?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Staging: usbip: fix build warning on 64bit kernels
[linux/fpc-iii.git] / fs / ext3 / super.c
blob3a260af5544d24dfc85f0dffc03b42771c322295
1 /*
2 * linux/fs/ext3/super.c
3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19 #include <linux/module.h>
20 #include <linux/string.h>
21 #include <linux/fs.h>
22 #include <linux/time.h>
23 #include <linux/jbd.h>
24 #include <linux/ext3_fs.h>
25 #include <linux/ext3_jbd.h>
26 #include <linux/slab.h>
27 #include <linux/init.h>
28 #include <linux/blkdev.h>
29 #include <linux/parser.h>
30 #include <linux/smp_lock.h>
31 #include <linux/buffer_head.h>
32 #include <linux/exportfs.h>
33 #include <linux/vfs.h>
34 #include <linux/random.h>
35 #include <linux/mount.h>
36 #include <linux/namei.h>
37 #include <linux/quotaops.h>
38 #include <linux/seq_file.h>
39 #include <linux/log2.h>
40
41 #include <asm/uaccess.h>
42
43 #include "xattr.h"
44 #include "acl.h"
45 #include "namei.h"
46
47 static int ext3_load_journal(struct super_block *, struct ext3_super_block *,
48 unsigned long journal_devnum);
49 static int ext3_create_journal(struct super_block *, struct ext3_super_block *,
50 unsigned int);
51 static void ext3_commit_super (struct super_block * sb,
52 struct ext3_super_block * es,
53 int sync);
54 static void ext3_mark_recovery_complete(struct super_block * sb,
55 struct ext3_super_block * es);
56 static void ext3_clear_journal_err(struct super_block * sb,
57 struct ext3_super_block * es);
58 static int ext3_sync_fs(struct super_block *sb, int wait);
59 static const char *ext3_decode_error(struct super_block * sb, int errno,
60 char nbuf[16]);
61 static int ext3_remount (struct super_block * sb, int * flags, char * data);
62 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf);
63 static void ext3_unlockfs(struct super_block *sb);
64 static void ext3_write_super (struct super_block * sb);
65 static void ext3_write_super_lockfs(struct super_block *sb);
66
67 /*
68 * Wrappers for journal_start/end.
69 *
70 * The only special thing we need to do here is to make sure that all
71 * journal_end calls result in the superblock being marked dirty, so
72 * that sync() will call the filesystem's write_super callback if
73 * appropriate.
74 */
75 handle_t *ext3_journal_start_sb(struct super_block *sb, int nblocks)
76 {
77 journal_t *journal;
78
79 if (sb->s_flags & MS_RDONLY)
80 return ERR_PTR(-EROFS);
81
82 /* Special case here: if the journal has aborted behind our
83 * backs (eg. EIO in the commit thread), then we still need to
84 * take the FS itself readonly cleanly. */
85 journal = EXT3_SB(sb)->s_journal;
86 if (is_journal_aborted(journal)) {
87 ext3_abort(sb, __func__,
88 "Detected aborted journal");
89 return ERR_PTR(-EROFS);
90 }
91
92 return journal_start(journal, nblocks);
93 }
94
95 /*
96 * The only special thing we need to do here is to make sure that all
97 * journal_stop calls result in the superblock being marked dirty, so
98 * that sync() will call the filesystem's write_super callback if
99 * appropriate.
100 */
101 int __ext3_journal_stop(const char *where, handle_t *handle)
102 {
103 struct super_block *sb;
104 int err;
105 int rc;
106
107 sb = handle->h_transaction->t_journal->j_private;
108 err = handle->h_err;
109 rc = journal_stop(handle);
110
111 if (!err)
112 err = rc;
113 if (err)
114 __ext3_std_error(sb, where, err);
115 return err;
116 }
117
118 void ext3_journal_abort_handle(const char *caller, const char *err_fn,
119 struct buffer_head *bh, handle_t *handle, int err)
120 {
121 char nbuf[16];
122 const char *errstr = ext3_decode_error(NULL, err, nbuf);
123
124 if (bh)
125 BUFFER_TRACE(bh, "abort");
126
127 if (!handle->h_err)
128 handle->h_err = err;
129
130 if (is_handle_aborted(handle))
131 return;
132
133 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
134 caller, errstr, err_fn);
135
136 journal_abort_handle(handle);
137 }
138
139 /* Deal with the reporting of failure conditions on a filesystem such as
140 * inconsistencies detected or read IO failures.
141 *
142 * On ext2, we can store the error state of the filesystem in the
143 * superblock. That is not possible on ext3, because we may have other
144 * write ordering constraints on the superblock which prevent us from
145 * writing it out straight away; and given that the journal is about to
146 * be aborted, we can't rely on the current, or future, transactions to
147 * write out the superblock safely.
148 *
149 * We'll just use the journal_abort() error code to record an error in
150 * the journal instead. On recovery, the journal will compain about
151 * that error until we've noted it down and cleared it.
152 */
153
154 static void ext3_handle_error(struct super_block *sb)
155 {
156 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
157
158 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
159 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
160
161 if (sb->s_flags & MS_RDONLY)
162 return;
163
164 if (!test_opt (sb, ERRORS_CONT)) {
165 journal_t *journal = EXT3_SB(sb)->s_journal;
166
167 EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
168 if (journal)
169 journal_abort(journal, -EIO);
170 }
171 if (test_opt (sb, ERRORS_RO)) {
172 printk (KERN_CRIT "Remounting filesystem read-only\n");
173 sb->s_flags |= MS_RDONLY;
174 }
175 ext3_commit_super(sb, es, 1);
176 if (test_opt(sb, ERRORS_PANIC))
177 panic("EXT3-fs (device %s): panic forced after error\n",
178 sb->s_id);
179 }
180
181 void ext3_error (struct super_block * sb, const char * function,
182 const char * fmt, ...)
183 {
184 va_list args;
185
186 va_start(args, fmt);
187 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
188 vprintk(fmt, args);
189 printk("\n");
190 va_end(args);
191
192 ext3_handle_error(sb);
193 }
194
195 static const char *ext3_decode_error(struct super_block * sb, int errno,
196 char nbuf[16])
197 {
198 char *errstr = NULL;
199
200 switch (errno) {
201 case -EIO:
202 errstr = "IO failure";
203 break;
204 case -ENOMEM:
205 errstr = "Out of memory";
206 break;
207 case -EROFS:
208 if (!sb || EXT3_SB(sb)->s_journal->j_flags & JFS_ABORT)
209 errstr = "Journal has aborted";
210 else
211 errstr = "Readonly filesystem";
212 break;
213 default:
214 /* If the caller passed in an extra buffer for unknown
215 * errors, textualise them now. Else we just return
216 * NULL. */
217 if (nbuf) {
218 /* Check for truncated error codes... */
219 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
220 errstr = nbuf;
221 }
222 break;
223 }
224
225 return errstr;
226 }
227
228 /* __ext3_std_error decodes expected errors from journaling functions
229 * automatically and invokes the appropriate error response. */
230
231 void __ext3_std_error (struct super_block * sb, const char * function,
232 int errno)
233 {
234 char nbuf[16];
235 const char *errstr;
236
237 /* Special case: if the error is EROFS, and we're not already
238 * inside a transaction, then there's really no point in logging
239 * an error. */
240 if (errno == -EROFS && journal_current_handle() == NULL &&
241 (sb->s_flags & MS_RDONLY))
242 return;
243
244 errstr = ext3_decode_error(sb, errno, nbuf);
245 printk (KERN_CRIT "EXT3-fs error (device %s) in %s: %s\n",
246 sb->s_id, function, errstr);
247
248 ext3_handle_error(sb);
249 }
250
251 /*
252 * ext3_abort is a much stronger failure handler than ext3_error. The
253 * abort function may be used to deal with unrecoverable failures such
254 * as journal IO errors or ENOMEM at a critical moment in log management.
255 *
256 * We unconditionally force the filesystem into an ABORT|READONLY state,
257 * unless the error response on the fs has been set to panic in which
258 * case we take the easy way out and panic immediately.
259 */
260
261 void ext3_abort (struct super_block * sb, const char * function,
262 const char * fmt, ...)
263 {
264 va_list args;
265
266 printk (KERN_CRIT "ext3_abort called.\n");
267
268 va_start(args, fmt);
269 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
270 vprintk(fmt, args);
271 printk("\n");
272 va_end(args);
273
274 if (test_opt(sb, ERRORS_PANIC))
275 panic("EXT3-fs panic from previous error\n");
276
277 if (sb->s_flags & MS_RDONLY)
278 return;
279
280 printk(KERN_CRIT "Remounting filesystem read-only\n");
281 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
282 sb->s_flags |= MS_RDONLY;
283 EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
284 journal_abort(EXT3_SB(sb)->s_journal, -EIO);
285 }
286
287 void ext3_warning (struct super_block * sb, const char * function,
288 const char * fmt, ...)
289 {
290 va_list args;
291
292 va_start(args, fmt);
293 printk(KERN_WARNING "EXT3-fs warning (device %s): %s: ",
294 sb->s_id, function);
295 vprintk(fmt, args);
296 printk("\n");
297 va_end(args);
298 }
299
300 void ext3_update_dynamic_rev(struct super_block *sb)
301 {
302 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
303
304 if (le32_to_cpu(es->s_rev_level) > EXT3_GOOD_OLD_REV)
305 return;
306
307 ext3_warning(sb, __func__,
308 "updating to rev %d because of new feature flag, "
309 "running e2fsck is recommended",
310 EXT3_DYNAMIC_REV);
311
312 es->s_first_ino = cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO);
313 es->s_inode_size = cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE);
314 es->s_rev_level = cpu_to_le32(EXT3_DYNAMIC_REV);
315 /* leave es->s_feature_*compat flags alone */
316 /* es->s_uuid will be set by e2fsck if empty */
317
318 /*
319 * The rest of the superblock fields should be zero, and if not it
320 * means they are likely already in use, so leave them alone. We
321 * can leave it up to e2fsck to clean up any inconsistencies there.
322 */
323 }
324
325 /*
326 * Open the external journal device
327 */
328 static struct block_device *ext3_blkdev_get(dev_t dev)
329 {
330 struct block_device *bdev;
331 char b[BDEVNAME_SIZE];
332
333 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
334 if (IS_ERR(bdev))
335 goto fail;
336 return bdev;
337
338 fail:
339 printk(KERN_ERR "EXT3: failed to open journal device %s: %ld\n",
340 __bdevname(dev, b), PTR_ERR(bdev));
341 return NULL;
342 }
343
344 /*
345 * Release the journal device
346 */
347 static int ext3_blkdev_put(struct block_device *bdev)
348 {
349 bd_release(bdev);
350 return blkdev_put(bdev);
351 }
352
353 static int ext3_blkdev_remove(struct ext3_sb_info *sbi)
354 {
355 struct block_device *bdev;
356 int ret = -ENODEV;
357
358 bdev = sbi->journal_bdev;
359 if (bdev) {
360 ret = ext3_blkdev_put(bdev);
361 sbi->journal_bdev = NULL;
362 }
363 return ret;
364 }
365
366 static inline struct inode *orphan_list_entry(struct list_head *l)
367 {
368 return &list_entry(l, struct ext3_inode_info, i_orphan)->vfs_inode;
369 }
370
371 static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi)
372 {
373 struct list_head *l;
374
375 printk(KERN_ERR "sb orphan head is %d\n",
376 le32_to_cpu(sbi->s_es->s_last_orphan));
377
378 printk(KERN_ERR "sb_info orphan list:\n");
379 list_for_each(l, &sbi->s_orphan) {
380 struct inode *inode = orphan_list_entry(l);
381 printk(KERN_ERR " "
382 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
383 inode->i_sb->s_id, inode->i_ino, inode,
384 inode->i_mode, inode->i_nlink,
385 NEXT_ORPHAN(inode));
386 }
387 }
388
389 static void ext3_put_super (struct super_block * sb)
390 {
391 struct ext3_sb_info *sbi = EXT3_SB(sb);
392 struct ext3_super_block *es = sbi->s_es;
393 int i;
394
395 ext3_xattr_put_super(sb);
396 journal_destroy(sbi->s_journal);
397 if (!(sb->s_flags & MS_RDONLY)) {
398 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
399 es->s_state = cpu_to_le16(sbi->s_mount_state);
400 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
401 mark_buffer_dirty(sbi->s_sbh);
402 ext3_commit_super(sb, es, 1);
403 }
404
405 for (i = 0; i < sbi->s_gdb_count; i++)
406 brelse(sbi->s_group_desc[i]);
407 kfree(sbi->s_group_desc);
408 percpu_counter_destroy(&sbi->s_freeblocks_counter);
409 percpu_counter_destroy(&sbi->s_freeinodes_counter);
410 percpu_counter_destroy(&sbi->s_dirs_counter);
411 brelse(sbi->s_sbh);
412 #ifdef CONFIG_QUOTA
413 for (i = 0; i < MAXQUOTAS; i++)
414 kfree(sbi->s_qf_names[i]);
415 #endif
416
417 /* Debugging code just in case the in-memory inode orphan list
418 * isn't empty. The on-disk one can be non-empty if we've
419 * detected an error and taken the fs readonly, but the
420 * in-memory list had better be clean by this point. */
421 if (!list_empty(&sbi->s_orphan))
422 dump_orphan_list(sb, sbi);
423 J_ASSERT(list_empty(&sbi->s_orphan));
424
425 invalidate_bdev(sb->s_bdev);
426 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
427 /*
428 * Invalidate the journal device's buffers. We don't want them
429 * floating about in memory - the physical journal device may
430 * hotswapped, and it breaks the `ro-after' testing code.
431 */
432 sync_blockdev(sbi->journal_bdev);
433 invalidate_bdev(sbi->journal_bdev);
434 ext3_blkdev_remove(sbi);
435 }
436 sb->s_fs_info = NULL;
437 kfree(sbi);
438 return;
439 }
440
441 static struct kmem_cache *ext3_inode_cachep;
442
443 /*
444 * Called inside transaction, so use GFP_NOFS
445 */
446 static struct inode *ext3_alloc_inode(struct super_block *sb)
447 {
448 struct ext3_inode_info *ei;
449
450 ei = kmem_cache_alloc(ext3_inode_cachep, GFP_NOFS);
451 if (!ei)
452 return NULL;
453 #ifdef CONFIG_EXT3_FS_POSIX_ACL
454 ei->i_acl = EXT3_ACL_NOT_CACHED;
455 ei->i_default_acl = EXT3_ACL_NOT_CACHED;
456 #endif
457 ei->i_block_alloc_info = NULL;
458 ei->vfs_inode.i_version = 1;
459 return &ei->vfs_inode;
460 }
461
462 static void ext3_destroy_inode(struct inode *inode)
463 {
464 if (!list_empty(&(EXT3_I(inode)->i_orphan))) {
465 printk("EXT3 Inode %p: orphan list check failed!\n",
466 EXT3_I(inode));
467 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
468 EXT3_I(inode), sizeof(struct ext3_inode_info),
469 false);
470 dump_stack();
471 }
472 kmem_cache_free(ext3_inode_cachep, EXT3_I(inode));
473 }
474
475 static void init_once(void *foo)
476 {
477 struct ext3_inode_info *ei = (struct ext3_inode_info *) foo;
478
479 INIT_LIST_HEAD(&ei->i_orphan);
480 #ifdef CONFIG_EXT3_FS_XATTR
481 init_rwsem(&ei->xattr_sem);
482 #endif
483 mutex_init(&ei->truncate_mutex);
484 inode_init_once(&ei->vfs_inode);
485 }
486
487 static int init_inodecache(void)
488 {
489 ext3_inode_cachep = kmem_cache_create("ext3_inode_cache",
490 sizeof(struct ext3_inode_info),
491 0, (SLAB_RECLAIM_ACCOUNT|
492 SLAB_MEM_SPREAD),
493 init_once);
494 if (ext3_inode_cachep == NULL)
495 return -ENOMEM;
496 return 0;
497 }
498
499 static void destroy_inodecache(void)
500 {
501 kmem_cache_destroy(ext3_inode_cachep);
502 }
503
504 static void ext3_clear_inode(struct inode *inode)
505 {
506 struct ext3_block_alloc_info *rsv = EXT3_I(inode)->i_block_alloc_info;
507 #ifdef CONFIG_EXT3_FS_POSIX_ACL
508 if (EXT3_I(inode)->i_acl &&
509 EXT3_I(inode)->i_acl != EXT3_ACL_NOT_CACHED) {
510 posix_acl_release(EXT3_I(inode)->i_acl);
511 EXT3_I(inode)->i_acl = EXT3_ACL_NOT_CACHED;
512 }
513 if (EXT3_I(inode)->i_default_acl &&
514 EXT3_I(inode)->i_default_acl != EXT3_ACL_NOT_CACHED) {
515 posix_acl_release(EXT3_I(inode)->i_default_acl);
516 EXT3_I(inode)->i_default_acl = EXT3_ACL_NOT_CACHED;
517 }
518 #endif
519 ext3_discard_reservation(inode);
520 EXT3_I(inode)->i_block_alloc_info = NULL;
521 if (unlikely(rsv))
522 kfree(rsv);
523 }
524
525 static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb)
526 {
527 #if defined(CONFIG_QUOTA)
528 struct ext3_sb_info *sbi = EXT3_SB(sb);
529
530 if (sbi->s_jquota_fmt)
531 seq_printf(seq, ",jqfmt=%s",
532 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
533
534 if (sbi->s_qf_names[USRQUOTA])
535 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
536
537 if (sbi->s_qf_names[GRPQUOTA])
538 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
539
540 if (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA)
541 seq_puts(seq, ",usrquota");
542
543 if (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)
544 seq_puts(seq, ",grpquota");
545 #endif
546 }
547
548 /*
549 * Show an option if
550 * - it's set to a non-default value OR
551 * - if the per-sb default is different from the global default
552 */
553 static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs)
554 {
555 struct super_block *sb = vfs->mnt_sb;
556 struct ext3_sb_info *sbi = EXT3_SB(sb);
557 struct ext3_super_block *es = sbi->s_es;
558 unsigned long def_mount_opts;
559
560 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
561
562 if (sbi->s_sb_block != 1)
563 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
564 if (test_opt(sb, MINIX_DF))
565 seq_puts(seq, ",minixdf");
566 if (test_opt(sb, GRPID))
567 seq_puts(seq, ",grpid");
568 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT3_DEFM_BSDGROUPS))
569 seq_puts(seq, ",nogrpid");
570 if (sbi->s_resuid != EXT3_DEF_RESUID ||
571 le16_to_cpu(es->s_def_resuid) != EXT3_DEF_RESUID) {
572 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
573 }
574 if (sbi->s_resgid != EXT3_DEF_RESGID ||
575 le16_to_cpu(es->s_def_resgid) != EXT3_DEF_RESGID) {
576 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
577 }
578 if (test_opt(sb, ERRORS_RO)) {
579 int def_errors = le16_to_cpu(es->s_errors);
580
581 if (def_errors == EXT3_ERRORS_PANIC ||
582 def_errors == EXT3_ERRORS_CONTINUE) {
583 seq_puts(seq, ",errors=remount-ro");
584 }
585 }
586 if (test_opt(sb, ERRORS_CONT))
587 seq_puts(seq, ",errors=continue");
588 if (test_opt(sb, ERRORS_PANIC))
589 seq_puts(seq, ",errors=panic");
590 if (test_opt(sb, NO_UID32))
591 seq_puts(seq, ",nouid32");
592 if (test_opt(sb, DEBUG))
593 seq_puts(seq, ",debug");
594 if (test_opt(sb, OLDALLOC))
595 seq_puts(seq, ",oldalloc");
596 #ifdef CONFIG_EXT3_FS_XATTR
597 if (test_opt(sb, XATTR_USER))
598 seq_puts(seq, ",user_xattr");
599 if (!test_opt(sb, XATTR_USER) &&
600 (def_mount_opts & EXT3_DEFM_XATTR_USER)) {
601 seq_puts(seq, ",nouser_xattr");
602 }
603 #endif
604 #ifdef CONFIG_EXT3_FS_POSIX_ACL
605 if (test_opt(sb, POSIX_ACL))
606 seq_puts(seq, ",acl");
607 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT3_DEFM_ACL))
608 seq_puts(seq, ",noacl");
609 #endif
610 if (!test_opt(sb, RESERVATION))
611 seq_puts(seq, ",noreservation");
612 if (sbi->s_commit_interval) {
613 seq_printf(seq, ",commit=%u",
614 (unsigned) (sbi->s_commit_interval / HZ));
615 }
616 if (test_opt(sb, BARRIER))
617 seq_puts(seq, ",barrier=1");
618 if (test_opt(sb, NOBH))
619 seq_puts(seq, ",nobh");
620
621 if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA)
622 seq_puts(seq, ",data=journal");
623 else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA)
624 seq_puts(seq, ",data=ordered");
625 else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)
626 seq_puts(seq, ",data=writeback");
627
628 if (test_opt(sb, DATA_ERR_ABORT))
629 seq_puts(seq, ",data_err=abort");
630
631 ext3_show_quota_options(seq, sb);
632
633 return 0;
634 }
635
636
637 static struct inode *ext3_nfs_get_inode(struct super_block *sb,
638 u64 ino, u32 generation)
639 {
640 struct inode *inode;
641
642 if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO)
643 return ERR_PTR(-ESTALE);
644 if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count))
645 return ERR_PTR(-ESTALE);
646
647 /* iget isn't really right if the inode is currently unallocated!!
648 *
649 * ext3_read_inode will return a bad_inode if the inode had been
650 * deleted, so we should be safe.
651 *
652 * Currently we don't know the generation for parent directory, so
653 * a generation of 0 means "accept any"
654 */
655 inode = ext3_iget(sb, ino);
656 if (IS_ERR(inode))
657 return ERR_CAST(inode);
658 if (generation && inode->i_generation != generation) {
659 iput(inode);
660 return ERR_PTR(-ESTALE);
661 }
662
663 return inode;
664 }
665
666 static struct dentry *ext3_fh_to_dentry(struct super_block *sb, struct fid *fid,
667 int fh_len, int fh_type)
668 {
669 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
670 ext3_nfs_get_inode);
671 }
672
673 static struct dentry *ext3_fh_to_parent(struct super_block *sb, struct fid *fid,
674 int fh_len, int fh_type)
675 {
676 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
677 ext3_nfs_get_inode);
678 }
679
680 #ifdef CONFIG_QUOTA
681 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
682 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
683
684 static int ext3_dquot_initialize(struct inode *inode, int type);
685 static int ext3_dquot_drop(struct inode *inode);
686 static int ext3_write_dquot(struct dquot *dquot);
687 static int ext3_acquire_dquot(struct dquot *dquot);
688 static int ext3_release_dquot(struct dquot *dquot);
689 static int ext3_mark_dquot_dirty(struct dquot *dquot);
690 static int ext3_write_info(struct super_block *sb, int type);
691 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
692 char *path, int remount);
693 static int ext3_quota_on_mount(struct super_block *sb, int type);
694 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
695 size_t len, loff_t off);
696 static ssize_t ext3_quota_write(struct super_block *sb, int type,
697 const char *data, size_t len, loff_t off);
698
699 static struct dquot_operations ext3_quota_operations = {
700 .initialize = ext3_dquot_initialize,
701 .drop = ext3_dquot_drop,
702 .alloc_space = dquot_alloc_space,
703 .alloc_inode = dquot_alloc_inode,
704 .free_space = dquot_free_space,
705 .free_inode = dquot_free_inode,
706 .transfer = dquot_transfer,
707 .write_dquot = ext3_write_dquot,
708 .acquire_dquot = ext3_acquire_dquot,
709 .release_dquot = ext3_release_dquot,
710 .mark_dirty = ext3_mark_dquot_dirty,
711 .write_info = ext3_write_info
712 };
713
714 static struct quotactl_ops ext3_qctl_operations = {
715 .quota_on = ext3_quota_on,
716 .quota_off = vfs_quota_off,
717 .quota_sync = vfs_quota_sync,
718 .get_info = vfs_get_dqinfo,
719 .set_info = vfs_set_dqinfo,
720 .get_dqblk = vfs_get_dqblk,
721 .set_dqblk = vfs_set_dqblk
722 };
723 #endif
724
725 static const struct super_operations ext3_sops = {
726 .alloc_inode = ext3_alloc_inode,
727 .destroy_inode = ext3_destroy_inode,
728 .write_inode = ext3_write_inode,
729 .dirty_inode = ext3_dirty_inode,
730 .delete_inode = ext3_delete_inode,
731 .put_super = ext3_put_super,
732 .write_super = ext3_write_super,
733 .sync_fs = ext3_sync_fs,
734 .write_super_lockfs = ext3_write_super_lockfs,
735 .unlockfs = ext3_unlockfs,
736 .statfs = ext3_statfs,
737 .remount_fs = ext3_remount,
738 .clear_inode = ext3_clear_inode,
739 .show_options = ext3_show_options,
740 #ifdef CONFIG_QUOTA
741 .quota_read = ext3_quota_read,
742 .quota_write = ext3_quota_write,
743 #endif
744 };
745
746 static const struct export_operations ext3_export_ops = {
747 .fh_to_dentry = ext3_fh_to_dentry,
748 .fh_to_parent = ext3_fh_to_parent,
749 .get_parent = ext3_get_parent,
750 };
751
752 enum {
753 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
754 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
755 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
756 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
757 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
758 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
759 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
760 Opt_data_err_abort, Opt_data_err_ignore,
761 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
762 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
763 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
764 Opt_grpquota
765 };
766
767 static const match_table_t tokens = {
768 {Opt_bsd_df, "bsddf"},
769 {Opt_minix_df, "minixdf"},
770 {Opt_grpid, "grpid"},
771 {Opt_grpid, "bsdgroups"},
772 {Opt_nogrpid, "nogrpid"},
773 {Opt_nogrpid, "sysvgroups"},
774 {Opt_resgid, "resgid=%u"},
775 {Opt_resuid, "resuid=%u"},
776 {Opt_sb, "sb=%u"},
777 {Opt_err_cont, "errors=continue"},
778 {Opt_err_panic, "errors=panic"},
779 {Opt_err_ro, "errors=remount-ro"},
780 {Opt_nouid32, "nouid32"},
781 {Opt_nocheck, "nocheck"},
782 {Opt_nocheck, "check=none"},
783 {Opt_debug, "debug"},
784 {Opt_oldalloc, "oldalloc"},
785 {Opt_orlov, "orlov"},
786 {Opt_user_xattr, "user_xattr"},
787 {Opt_nouser_xattr, "nouser_xattr"},
788 {Opt_acl, "acl"},
789 {Opt_noacl, "noacl"},
790 {Opt_reservation, "reservation"},
791 {Opt_noreservation, "noreservation"},
792 {Opt_noload, "noload"},
793 {Opt_nobh, "nobh"},
794 {Opt_bh, "bh"},
795 {Opt_commit, "commit=%u"},
796 {Opt_journal_update, "journal=update"},
797 {Opt_journal_inum, "journal=%u"},
798 {Opt_journal_dev, "journal_dev=%u"},
799 {Opt_abort, "abort"},
800 {Opt_data_journal, "data=journal"},
801 {Opt_data_ordered, "data=ordered"},
802 {Opt_data_writeback, "data=writeback"},
803 {Opt_data_err_abort, "data_err=abort"},
804 {Opt_data_err_ignore, "data_err=ignore"},
805 {Opt_offusrjquota, "usrjquota="},
806 {Opt_usrjquota, "usrjquota=%s"},
807 {Opt_offgrpjquota, "grpjquota="},
808 {Opt_grpjquota, "grpjquota=%s"},
809 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
810 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
811 {Opt_grpquota, "grpquota"},
812 {Opt_noquota, "noquota"},
813 {Opt_quota, "quota"},
814 {Opt_usrquota, "usrquota"},
815 {Opt_barrier, "barrier=%u"},
816 {Opt_resize, "resize"},
817 {Opt_err, NULL},
818 };
819
820 static ext3_fsblk_t get_sb_block(void **data)
821 {
822 ext3_fsblk_t sb_block;
823 char *options = (char *) *data;
824
825 if (!options || strncmp(options, "sb=", 3) != 0)
826 return 1; /* Default location */
827 options += 3;
828 /*todo: use simple_strtoll with >32bit ext3 */
829 sb_block = simple_strtoul(options, &options, 0);
830 if (*options && *options != ',') {
831 printk("EXT3-fs: Invalid sb specification: %s\n",
832 (char *) *data);
833 return 1;
834 }
835 if (*options == ',')
836 options++;
837 *data = (void *) options;
838 return sb_block;
839 }
840
841 static int parse_options (char *options, struct super_block *sb,
842 unsigned int *inum, unsigned long *journal_devnum,
843 ext3_fsblk_t *n_blocks_count, int is_remount)
844 {
845 struct ext3_sb_info *sbi = EXT3_SB(sb);
846 char * p;
847 substring_t args[MAX_OPT_ARGS];
848 int data_opt = 0;
849 int option;
850 #ifdef CONFIG_QUOTA
851 int qtype, qfmt;
852 char *qname;
853 #endif
854
855 if (!options)
856 return 1;
857
858 while ((p = strsep (&options, ",")) != NULL) {
859 int token;
860 if (!*p)
861 continue;
862
863 token = match_token(p, tokens, args);
864 switch (token) {
865 case Opt_bsd_df:
866 clear_opt (sbi->s_mount_opt, MINIX_DF);
867 break;
868 case Opt_minix_df:
869 set_opt (sbi->s_mount_opt, MINIX_DF);
870 break;
871 case Opt_grpid:
872 set_opt (sbi->s_mount_opt, GRPID);
873 break;
874 case Opt_nogrpid:
875 clear_opt (sbi->s_mount_opt, GRPID);
876 break;
877 case Opt_resuid:
878 if (match_int(&args[0], &option))
879 return 0;
880 sbi->s_resuid = option;
881 break;
882 case Opt_resgid:
883 if (match_int(&args[0], &option))
884 return 0;
885 sbi->s_resgid = option;
886 break;
887 case Opt_sb:
888 /* handled by get_sb_block() instead of here */
889 /* *sb_block = match_int(&args[0]); */
890 break;
891 case Opt_err_panic:
892 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
893 clear_opt (sbi->s_mount_opt, ERRORS_RO);
894 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
895 break;
896 case Opt_err_ro:
897 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
898 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
899 set_opt (sbi->s_mount_opt, ERRORS_RO);
900 break;
901 case Opt_err_cont:
902 clear_opt (sbi->s_mount_opt, ERRORS_RO);
903 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
904 set_opt (sbi->s_mount_opt, ERRORS_CONT);
905 break;
906 case Opt_nouid32:
907 set_opt (sbi->s_mount_opt, NO_UID32);
908 break;
909 case Opt_nocheck:
910 clear_opt (sbi->s_mount_opt, CHECK);
911 break;
912 case Opt_debug:
913 set_opt (sbi->s_mount_opt, DEBUG);
914 break;
915 case Opt_oldalloc:
916 set_opt (sbi->s_mount_opt, OLDALLOC);
917 break;
918 case Opt_orlov:
919 clear_opt (sbi->s_mount_opt, OLDALLOC);
920 break;
921 #ifdef CONFIG_EXT3_FS_XATTR
922 case Opt_user_xattr:
923 set_opt (sbi->s_mount_opt, XATTR_USER);
924 break;
925 case Opt_nouser_xattr:
926 clear_opt (sbi->s_mount_opt, XATTR_USER);
927 break;
928 #else
929 case Opt_user_xattr:
930 case Opt_nouser_xattr:
931 printk("EXT3 (no)user_xattr options not supported\n");
932 break;
933 #endif
934 #ifdef CONFIG_EXT3_FS_POSIX_ACL
935 case Opt_acl:
936 set_opt(sbi->s_mount_opt, POSIX_ACL);
937 break;
938 case Opt_noacl:
939 clear_opt(sbi->s_mount_opt, POSIX_ACL);
940 break;
941 #else
942 case Opt_acl:
943 case Opt_noacl:
944 printk("EXT3 (no)acl options not supported\n");
945 break;
946 #endif
947 case Opt_reservation:
948 set_opt(sbi->s_mount_opt, RESERVATION);
949 break;
950 case Opt_noreservation:
951 clear_opt(sbi->s_mount_opt, RESERVATION);
952 break;
953 case Opt_journal_update:
954 /* @@@ FIXME */
955 /* Eventually we will want to be able to create
956 a journal file here. For now, only allow the
957 user to specify an existing inode to be the
958 journal file. */
959 if (is_remount) {
960 printk(KERN_ERR "EXT3-fs: cannot specify "
961 "journal on remount\n");
962 return 0;
963 }
964 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
965 break;
966 case Opt_journal_inum:
967 if (is_remount) {
968 printk(KERN_ERR "EXT3-fs: cannot specify "
969 "journal on remount\n");
970 return 0;
971 }
972 if (match_int(&args[0], &option))
973 return 0;
974 *inum = option;
975 break;
976 case Opt_journal_dev:
977 if (is_remount) {
978 printk(KERN_ERR "EXT3-fs: cannot specify "
979 "journal on remount\n");
980 return 0;
981 }
982 if (match_int(&args[0], &option))
983 return 0;
984 *journal_devnum = option;
985 break;
986 case Opt_noload:
987 set_opt (sbi->s_mount_opt, NOLOAD);
988 break;
989 case Opt_commit:
990 if (match_int(&args[0], &option))
991 return 0;
992 if (option < 0)
993 return 0;
994 if (option == 0)
995 option = JBD_DEFAULT_MAX_COMMIT_AGE;
996 sbi->s_commit_interval = HZ * option;
997 break;
998 case Opt_data_journal:
999 data_opt = EXT3_MOUNT_JOURNAL_DATA;
1000 goto datacheck;
1001 case Opt_data_ordered:
1002 data_opt = EXT3_MOUNT_ORDERED_DATA;
1003 goto datacheck;
1004 case Opt_data_writeback:
1005 data_opt = EXT3_MOUNT_WRITEBACK_DATA;
1006 datacheck:
1007 if (is_remount) {
1008 if ((sbi->s_mount_opt & EXT3_MOUNT_DATA_FLAGS)
1009 != data_opt) {
1010 printk(KERN_ERR
1011 "EXT3-fs: cannot change data "
1012 "mode on remount\n");
1013 return 0;
1014 }
1015 } else {
1016 sbi->s_mount_opt &= ~EXT3_MOUNT_DATA_FLAGS;
1017 sbi->s_mount_opt |= data_opt;
1018 }
1019 break;
1020 case Opt_data_err_abort:
1021 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1022 break;
1023 case Opt_data_err_ignore:
1024 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1025 break;
1026 #ifdef CONFIG_QUOTA
1027 case Opt_usrjquota:
1028 qtype = USRQUOTA;
1029 goto set_qf_name;
1030 case Opt_grpjquota:
1031 qtype = GRPQUOTA;
1032 set_qf_name:
1033 if ((sb_any_quota_enabled(sb) ||
1034 sb_any_quota_suspended(sb)) &&
1035 !sbi->s_qf_names[qtype]) {
1036 printk(KERN_ERR
1037 "EXT3-fs: Cannot change journaled "
1038 "quota options when quota turned on.\n");
1039 return 0;
1040 }
1041 qname = match_strdup(&args[0]);
1042 if (!qname) {
1043 printk(KERN_ERR
1044 "EXT3-fs: not enough memory for "
1045 "storing quotafile name.\n");
1046 return 0;
1047 }
1048 if (sbi->s_qf_names[qtype] &&
1049 strcmp(sbi->s_qf_names[qtype], qname)) {
1050 printk(KERN_ERR
1051 "EXT3-fs: %s quota file already "
1052 "specified.\n", QTYPE2NAME(qtype));
1053 kfree(qname);
1054 return 0;
1055 }
1056 sbi->s_qf_names[qtype] = qname;
1057 if (strchr(sbi->s_qf_names[qtype], '/')) {
1058 printk(KERN_ERR
1059 "EXT3-fs: quotafile must be on "
1060 "filesystem root.\n");
1061 kfree(sbi->s_qf_names[qtype]);
1062 sbi->s_qf_names[qtype] = NULL;
1063 return 0;
1064 }
1065 set_opt(sbi->s_mount_opt, QUOTA);
1066 break;
1067 case Opt_offusrjquota:
1068 qtype = USRQUOTA;
1069 goto clear_qf_name;
1070 case Opt_offgrpjquota:
1071 qtype = GRPQUOTA;
1072 clear_qf_name:
1073 if ((sb_any_quota_enabled(sb) ||
1074 sb_any_quota_suspended(sb)) &&
1075 sbi->s_qf_names[qtype]) {
1076 printk(KERN_ERR "EXT3-fs: Cannot change "
1077 "journaled quota options when "
1078 "quota turned on.\n");
1079 return 0;
1080 }
1081 /*
1082 * The space will be released later when all options
1083 * are confirmed to be correct
1084 */
1085 sbi->s_qf_names[qtype] = NULL;
1086 break;
1087 case Opt_jqfmt_vfsold:
1088 qfmt = QFMT_VFS_OLD;
1089 goto set_qf_format;
1090 case Opt_jqfmt_vfsv0:
1091 qfmt = QFMT_VFS_V0;
1092 set_qf_format:
1093 if ((sb_any_quota_enabled(sb) ||
1094 sb_any_quota_suspended(sb)) &&
1095 sbi->s_jquota_fmt != qfmt) {
1096 printk(KERN_ERR "EXT3-fs: Cannot change "
1097 "journaled quota options when "
1098 "quota turned on.\n");
1099 return 0;
1100 }
1101 sbi->s_jquota_fmt = qfmt;
1102 break;
1103 case Opt_quota:
1104 case Opt_usrquota:
1105 set_opt(sbi->s_mount_opt, QUOTA);
1106 set_opt(sbi->s_mount_opt, USRQUOTA);
1107 break;
1108 case Opt_grpquota:
1109 set_opt(sbi->s_mount_opt, QUOTA);
1110 set_opt(sbi->s_mount_opt, GRPQUOTA);
1111 break;
1112 case Opt_noquota:
1113 if (sb_any_quota_enabled(sb) ||
1114 sb_any_quota_suspended(sb)) {
1115 printk(KERN_ERR "EXT3-fs: Cannot change quota "
1116 "options when quota turned on.\n");
1117 return 0;
1118 }
1119 clear_opt(sbi->s_mount_opt, QUOTA);
1120 clear_opt(sbi->s_mount_opt, USRQUOTA);
1121 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1122 break;
1123 #else
1124 case Opt_quota:
1125 case Opt_usrquota:
1126 case Opt_grpquota:
1127 printk(KERN_ERR
1128 "EXT3-fs: quota options not supported.\n");
1129 break;
1130 case Opt_usrjquota:
1131 case Opt_grpjquota:
1132 case Opt_offusrjquota:
1133 case Opt_offgrpjquota:
1134 case Opt_jqfmt_vfsold:
1135 case Opt_jqfmt_vfsv0:
1136 printk(KERN_ERR
1137 "EXT3-fs: journaled quota options not "
1138 "supported.\n");
1139 break;
1140 case Opt_noquota:
1141 break;
1142 #endif
1143 case Opt_abort:
1144 set_opt(sbi->s_mount_opt, ABORT);
1145 break;
1146 case Opt_barrier:
1147 if (match_int(&args[0], &option))
1148 return 0;
1149 if (option)
1150 set_opt(sbi->s_mount_opt, BARRIER);
1151 else
1152 clear_opt(sbi->s_mount_opt, BARRIER);
1153 break;
1154 case Opt_ignore:
1155 break;
1156 case Opt_resize:
1157 if (!is_remount) {
1158 printk("EXT3-fs: resize option only available "
1159 "for remount\n");
1160 return 0;
1161 }
1162 if (match_int(&args[0], &option) != 0)
1163 return 0;
1164 *n_blocks_count = option;
1165 break;
1166 case Opt_nobh:
1167 set_opt(sbi->s_mount_opt, NOBH);
1168 break;
1169 case Opt_bh:
1170 clear_opt(sbi->s_mount_opt, NOBH);
1171 break;
1172 default:
1173 printk (KERN_ERR
1174 "EXT3-fs: Unrecognized mount option \"%s\" "
1175 "or missing value\n", p);
1176 return 0;
1177 }
1178 }
1179 #ifdef CONFIG_QUOTA
1180 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1181 if ((sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA) &&
1182 sbi->s_qf_names[USRQUOTA])
1183 clear_opt(sbi->s_mount_opt, USRQUOTA);
1184
1185 if ((sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA) &&
1186 sbi->s_qf_names[GRPQUOTA])
1187 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1188
1189 if ((sbi->s_qf_names[USRQUOTA] &&
1190 (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)) ||
1191 (sbi->s_qf_names[GRPQUOTA] &&
1192 (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA))) {
1193 printk(KERN_ERR "EXT3-fs: old and new quota "
1194 "format mixing.\n");
1195 return 0;
1196 }
1197
1198 if (!sbi->s_jquota_fmt) {
1199 printk(KERN_ERR "EXT3-fs: journaled quota format "
1200 "not specified.\n");
1201 return 0;
1202 }
1203 } else {
1204 if (sbi->s_jquota_fmt) {
1205 printk(KERN_ERR "EXT3-fs: journaled quota format "
1206 "specified with no journaling "
1207 "enabled.\n");
1208 return 0;
1209 }
1210 }
1211 #endif
1212 return 1;
1213 }
1214
1215 static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
1216 int read_only)
1217 {
1218 struct ext3_sb_info *sbi = EXT3_SB(sb);
1219 int res = 0;
1220
1221 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
1222 printk (KERN_ERR "EXT3-fs warning: revision level too high, "
1223 "forcing read-only mode\n");
1224 res = MS_RDONLY;
1225 }
1226 if (read_only)
1227 return res;
1228 if (!(sbi->s_mount_state & EXT3_VALID_FS))
1229 printk (KERN_WARNING "EXT3-fs warning: mounting unchecked fs, "
1230 "running e2fsck is recommended\n");
1231 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
1232 printk (KERN_WARNING
1233 "EXT3-fs warning: mounting fs with errors, "
1234 "running e2fsck is recommended\n");
1235 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1236 le16_to_cpu(es->s_mnt_count) >=
1237 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1238 printk (KERN_WARNING
1239 "EXT3-fs warning: maximal mount count reached, "
1240 "running e2fsck is recommended\n");
1241 else if (le32_to_cpu(es->s_checkinterval) &&
1242 (le32_to_cpu(es->s_lastcheck) +
1243 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1244 printk (KERN_WARNING
1245 "EXT3-fs warning: checktime reached, "
1246 "running e2fsck is recommended\n");
1247 #if 0
1248 /* @@@ We _will_ want to clear the valid bit if we find
1249 inconsistencies, to force a fsck at reboot. But for
1250 a plain journaled filesystem we can keep it set as
1251 valid forever! :) */
1252 es->s_state &= cpu_to_le16(~EXT3_VALID_FS);
1253 #endif
1254 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1255 es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
1256 le16_add_cpu(&es->s_mnt_count, 1);
1257 es->s_mtime = cpu_to_le32(get_seconds());
1258 ext3_update_dynamic_rev(sb);
1259 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1260
1261 ext3_commit_super(sb, es, 1);
1262 if (test_opt(sb, DEBUG))
1263 printk(KERN_INFO "[EXT3 FS bs=%lu, gc=%lu, "
1264 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1265 sb->s_blocksize,
1266 sbi->s_groups_count,
1267 EXT3_BLOCKS_PER_GROUP(sb),
1268 EXT3_INODES_PER_GROUP(sb),
1269 sbi->s_mount_opt);
1270
1271 printk(KERN_INFO "EXT3 FS on %s, ", sb->s_id);
1272 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
1273 char b[BDEVNAME_SIZE];
1274
1275 printk("external journal on %s\n",
1276 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1277 } else {
1278 printk("internal journal\n");
1279 }
1280 return res;
1281 }
1282
1283 /* Called at mount-time, super-block is locked */
1284 static int ext3_check_descriptors(struct super_block *sb)
1285 {
1286 struct ext3_sb_info *sbi = EXT3_SB(sb);
1287 int i;
1288
1289 ext3_debug ("Checking group descriptors");
1290
1291 for (i = 0; i < sbi->s_groups_count; i++) {
1292 struct ext3_group_desc *gdp = ext3_get_group_desc(sb, i, NULL);
1293 ext3_fsblk_t first_block = ext3_group_first_block_no(sb, i);
1294 ext3_fsblk_t last_block;
1295
1296 if (i == sbi->s_groups_count - 1)
1297 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
1298 else
1299 last_block = first_block +
1300 (EXT3_BLOCKS_PER_GROUP(sb) - 1);
1301
1302 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
1303 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
1304 {
1305 ext3_error (sb, "ext3_check_descriptors",
1306 "Block bitmap for group %d"
1307 " not in group (block %lu)!",
1308 i, (unsigned long)
1309 le32_to_cpu(gdp->bg_block_bitmap));
1310 return 0;
1311 }
1312 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
1313 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
1314 {
1315 ext3_error (sb, "ext3_check_descriptors",
1316 "Inode bitmap for group %d"
1317 " not in group (block %lu)!",
1318 i, (unsigned long)
1319 le32_to_cpu(gdp->bg_inode_bitmap));
1320 return 0;
1321 }
1322 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
1323 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
1324 last_block)
1325 {
1326 ext3_error (sb, "ext3_check_descriptors",
1327 "Inode table for group %d"
1328 " not in group (block %lu)!",
1329 i, (unsigned long)
1330 le32_to_cpu(gdp->bg_inode_table));
1331 return 0;
1332 }
1333 }
1334
1335 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1336 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1337 return 1;
1338 }
1339
1340
1341 /* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1342 * the superblock) which were deleted from all directories, but held open by
1343 * a process at the time of a crash. We walk the list and try to delete these
1344 * inodes at recovery time (only with a read-write filesystem).
1345 *
1346 * In order to keep the orphan inode chain consistent during traversal (in
1347 * case of crash during recovery), we link each inode into the superblock
1348 * orphan list_head and handle it the same way as an inode deletion during
1349 * normal operation (which journals the operations for us).
1350 *
1351 * We only do an iget() and an iput() on each inode, which is very safe if we
1352 * accidentally point at an in-use or already deleted inode. The worst that
1353 * can happen in this case is that we get a "bit already cleared" message from
1354 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1355 * e2fsck was run on this filesystem, and it must have already done the orphan
1356 * inode cleanup for us, so we can safely abort without any further action.
1357 */
1358 static void ext3_orphan_cleanup (struct super_block * sb,
1359 struct ext3_super_block * es)
1360 {
1361 unsigned int s_flags = sb->s_flags;
1362 int nr_orphans = 0, nr_truncates = 0;
1363 #ifdef CONFIG_QUOTA
1364 int i;
1365 #endif
1366 if (!es->s_last_orphan) {
1367 jbd_debug(4, "no orphan inodes to clean up\n");
1368 return;
1369 }
1370
1371 if (bdev_read_only(sb->s_bdev)) {
1372 printk(KERN_ERR "EXT3-fs: write access "
1373 "unavailable, skipping orphan cleanup.\n");
1374 return;
1375 }
1376
1377 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1378 if (es->s_last_orphan)
1379 jbd_debug(1, "Errors on filesystem, "
1380 "clearing orphan list.\n");
1381 es->s_last_orphan = 0;
1382 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1383 return;
1384 }
1385
1386 if (s_flags & MS_RDONLY) {
1387 printk(KERN_INFO "EXT3-fs: %s: orphan cleanup on readonly fs\n",
1388 sb->s_id);
1389 sb->s_flags &= ~MS_RDONLY;
1390 }
1391 #ifdef CONFIG_QUOTA
1392 /* Needed for iput() to work correctly and not trash data */
1393 sb->s_flags |= MS_ACTIVE;
1394 /* Turn on quotas so that they are updated correctly */
1395 for (i = 0; i < MAXQUOTAS; i++) {
1396 if (EXT3_SB(sb)->s_qf_names[i]) {
1397 int ret = ext3_quota_on_mount(sb, i);
1398 if (ret < 0)
1399 printk(KERN_ERR
1400 "EXT3-fs: Cannot turn on journaled "
1401 "quota: error %d\n", ret);
1402 }
1403 }
1404 #endif
1405
1406 while (es->s_last_orphan) {
1407 struct inode *inode;
1408
1409 inode = ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1410 if (IS_ERR(inode)) {
1411 es->s_last_orphan = 0;
1412 break;
1413 }
1414
1415 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
1416 DQUOT_INIT(inode);
1417 if (inode->i_nlink) {
1418 printk(KERN_DEBUG
1419 "%s: truncating inode %lu to %Ld bytes\n",
1420 __func__, inode->i_ino, inode->i_size);
1421 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1422 inode->i_ino, inode->i_size);
1423 ext3_truncate(inode);
1424 nr_truncates++;
1425 } else {
1426 printk(KERN_DEBUG
1427 "%s: deleting unreferenced inode %lu\n",
1428 __func__, inode->i_ino);
1429 jbd_debug(2, "deleting unreferenced inode %lu\n",
1430 inode->i_ino);
1431 nr_orphans++;
1432 }
1433 iput(inode); /* The delete magic happens here! */
1434 }
1435
1436 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1437
1438 if (nr_orphans)
1439 printk(KERN_INFO "EXT3-fs: %s: %d orphan inode%s deleted\n",
1440 sb->s_id, PLURAL(nr_orphans));
1441 if (nr_truncates)
1442 printk(KERN_INFO "EXT3-fs: %s: %d truncate%s cleaned up\n",
1443 sb->s_id, PLURAL(nr_truncates));
1444 #ifdef CONFIG_QUOTA
1445 /* Turn quotas off */
1446 for (i = 0; i < MAXQUOTAS; i++) {
1447 if (sb_dqopt(sb)->files[i])
1448 vfs_quota_off(sb, i, 0);
1449 }
1450 #endif
1451 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1452 }
1453
1454 /*
1455 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1456 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1457 * We need to be 1 filesystem block less than the 2^32 sector limit.
1458 */
1459 static loff_t ext3_max_size(int bits)
1460 {
1461 loff_t res = EXT3_NDIR_BLOCKS;
1462 int meta_blocks;
1463 loff_t upper_limit;
1464
1465 /* This is calculated to be the largest file size for a
1466 * dense, file such that the total number of
1467 * sectors in the file, including data and all indirect blocks,
1468 * does not exceed 2^32 -1
1469 * __u32 i_blocks representing the total number of
1470 * 512 bytes blocks of the file
1471 */
1472 upper_limit = (1LL << 32) - 1;
1473
1474 /* total blocks in file system block size */
1475 upper_limit >>= (bits - 9);
1476
1477
1478 /* indirect blocks */
1479 meta_blocks = 1;
1480 /* double indirect blocks */
1481 meta_blocks += 1 + (1LL << (bits-2));
1482 /* tripple indirect blocks */
1483 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1484
1485 upper_limit -= meta_blocks;
1486 upper_limit <<= bits;
1487
1488 res += 1LL << (bits-2);
1489 res += 1LL << (2*(bits-2));
1490 res += 1LL << (3*(bits-2));
1491 res <<= bits;
1492 if (res > upper_limit)
1493 res = upper_limit;
1494
1495 if (res > MAX_LFS_FILESIZE)
1496 res = MAX_LFS_FILESIZE;
1497
1498 return res;
1499 }
1500
1501 static ext3_fsblk_t descriptor_loc(struct super_block *sb,
1502 ext3_fsblk_t logic_sb_block,
1503 int nr)
1504 {
1505 struct ext3_sb_info *sbi = EXT3_SB(sb);
1506 unsigned long bg, first_meta_bg;
1507 int has_super = 0;
1508
1509 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1510
1511 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1512 nr < first_meta_bg)
1513 return (logic_sb_block + nr + 1);
1514 bg = sbi->s_desc_per_block * nr;
1515 if (ext3_bg_has_super(sb, bg))
1516 has_super = 1;
1517 return (has_super + ext3_group_first_block_no(sb, bg));
1518 }
1519
1520
1521 static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1522 {
1523 struct buffer_head * bh;
1524 struct ext3_super_block *es = NULL;
1525 struct ext3_sb_info *sbi;
1526 ext3_fsblk_t block;
1527 ext3_fsblk_t sb_block = get_sb_block(&data);
1528 ext3_fsblk_t logic_sb_block;
1529 unsigned long offset = 0;
1530 unsigned int journal_inum = 0;
1531 unsigned long journal_devnum = 0;
1532 unsigned long def_mount_opts;
1533 struct inode *root;
1534 int blocksize;
1535 int hblock;
1536 int db_count;
1537 int i;
1538 int needs_recovery;
1539 int ret = -EINVAL;
1540 __le32 features;
1541 int err;
1542
1543 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1544 if (!sbi)
1545 return -ENOMEM;
1546 sb->s_fs_info = sbi;
1547 sbi->s_mount_opt = 0;
1548 sbi->s_resuid = EXT3_DEF_RESUID;
1549 sbi->s_resgid = EXT3_DEF_RESGID;
1550 sbi->s_sb_block = sb_block;
1551
1552 unlock_kernel();
1553
1554 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1555 if (!blocksize) {
1556 printk(KERN_ERR "EXT3-fs: unable to set blocksize\n");
1557 goto out_fail;
1558 }
1559
1560 /*
1561 * The ext3 superblock will not be buffer aligned for other than 1kB
1562 * block sizes. We need to calculate the offset from buffer start.
1563 */
1564 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1565 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1566 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1567 } else {
1568 logic_sb_block = sb_block;
1569 }
1570
1571 if (!(bh = sb_bread(sb, logic_sb_block))) {
1572 printk (KERN_ERR "EXT3-fs: unable to read superblock\n");
1573 goto out_fail;
1574 }
1575 /*
1576 * Note: s_es must be initialized as soon as possible because
1577 * some ext3 macro-instructions depend on its value
1578 */
1579 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1580 sbi->s_es = es;
1581 sb->s_magic = le16_to_cpu(es->s_magic);
1582 if (sb->s_magic != EXT3_SUPER_MAGIC)
1583 goto cantfind_ext3;
1584
1585 /* Set defaults before we parse the mount options */
1586 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1587 if (def_mount_opts & EXT3_DEFM_DEBUG)
1588 set_opt(sbi->s_mount_opt, DEBUG);
1589 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1590 set_opt(sbi->s_mount_opt, GRPID);
1591 if (def_mount_opts & EXT3_DEFM_UID16)
1592 set_opt(sbi->s_mount_opt, NO_UID32);
1593 #ifdef CONFIG_EXT3_FS_XATTR
1594 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1595 set_opt(sbi->s_mount_opt, XATTR_USER);
1596 #endif
1597 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1598 if (def_mount_opts & EXT3_DEFM_ACL)
1599 set_opt(sbi->s_mount_opt, POSIX_ACL);
1600 #endif
1601 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
1602 sbi->s_mount_opt |= EXT3_MOUNT_JOURNAL_DATA;
1603 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
1604 sbi->s_mount_opt |= EXT3_MOUNT_ORDERED_DATA;
1605 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
1606 sbi->s_mount_opt |= EXT3_MOUNT_WRITEBACK_DATA;
1607
1608 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1609 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1610 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_CONTINUE)
1611 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1612 else
1613 set_opt(sbi->s_mount_opt, ERRORS_RO);
1614
1615 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1616 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1617
1618 set_opt(sbi->s_mount_opt, RESERVATION);
1619
1620 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1621 NULL, 0))
1622 goto failed_mount;
1623
1624 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1625 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1626
1627 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1628 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1629 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1630 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1631 printk(KERN_WARNING
1632 "EXT3-fs warning: feature flags set on rev 0 fs, "
1633 "running e2fsck is recommended\n");
1634 /*
1635 * Check feature flags regardless of the revision level, since we
1636 * previously didn't change the revision level when setting the flags,
1637 * so there is a chance incompat flags are set on a rev 0 filesystem.
1638 */
1639 features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
1640 if (features) {
1641 printk(KERN_ERR "EXT3-fs: %s: couldn't mount because of "
1642 "unsupported optional features (%x).\n",
1643 sb->s_id, le32_to_cpu(features));
1644 goto failed_mount;
1645 }
1646 features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
1647 if (!(sb->s_flags & MS_RDONLY) && features) {
1648 printk(KERN_ERR "EXT3-fs: %s: couldn't mount RDWR because of "
1649 "unsupported optional features (%x).\n",
1650 sb->s_id, le32_to_cpu(features));
1651 goto failed_mount;
1652 }
1653 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1654
1655 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1656 blocksize > EXT3_MAX_BLOCK_SIZE) {
1657 printk(KERN_ERR
1658 "EXT3-fs: Unsupported filesystem blocksize %d on %s.\n",
1659 blocksize, sb->s_id);
1660 goto failed_mount;
1661 }
1662
1663 hblock = bdev_hardsect_size(sb->s_bdev);
1664 if (sb->s_blocksize != blocksize) {
1665 /*
1666 * Make sure the blocksize for the filesystem is larger
1667 * than the hardware sectorsize for the machine.
1668 */
1669 if (blocksize < hblock) {
1670 printk(KERN_ERR "EXT3-fs: blocksize %d too small for "
1671 "device blocksize %d.\n", blocksize, hblock);
1672 goto failed_mount;
1673 }
1674
1675 brelse (bh);
1676 if (!sb_set_blocksize(sb, blocksize)) {
1677 printk(KERN_ERR "EXT3-fs: bad blocksize %d.\n",
1678 blocksize);
1679 goto out_fail;
1680 }
1681 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1682 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1683 bh = sb_bread(sb, logic_sb_block);
1684 if (!bh) {
1685 printk(KERN_ERR
1686 "EXT3-fs: Can't read superblock on 2nd try.\n");
1687 goto failed_mount;
1688 }
1689 es = (struct ext3_super_block *)(((char *)bh->b_data) + offset);
1690 sbi->s_es = es;
1691 if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
1692 printk (KERN_ERR
1693 "EXT3-fs: Magic mismatch, very weird !\n");
1694 goto failed_mount;
1695 }
1696 }
1697
1698 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1699
1700 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1701 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1702 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1703 } else {
1704 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1705 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1706 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
1707 (!is_power_of_2(sbi->s_inode_size)) ||
1708 (sbi->s_inode_size > blocksize)) {
1709 printk (KERN_ERR
1710 "EXT3-fs: unsupported inode size: %d\n",
1711 sbi->s_inode_size);
1712 goto failed_mount;
1713 }
1714 }
1715 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1716 le32_to_cpu(es->s_log_frag_size);
1717 if (blocksize != sbi->s_frag_size) {
1718 printk(KERN_ERR
1719 "EXT3-fs: fragsize %lu != blocksize %u (unsupported)\n",
1720 sbi->s_frag_size, blocksize);
1721 goto failed_mount;
1722 }
1723 sbi->s_frags_per_block = 1;
1724 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1725 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1726 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1727 if (EXT3_INODE_SIZE(sb) == 0 || EXT3_INODES_PER_GROUP(sb) == 0)
1728 goto cantfind_ext3;
1729 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1730 if (sbi->s_inodes_per_block == 0)
1731 goto cantfind_ext3;
1732 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1733 sbi->s_inodes_per_block;
1734 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1735 sbi->s_sbh = bh;
1736 sbi->s_mount_state = le16_to_cpu(es->s_state);
1737 sbi->s_addr_per_block_bits = ilog2(EXT3_ADDR_PER_BLOCK(sb));
1738 sbi->s_desc_per_block_bits = ilog2(EXT3_DESC_PER_BLOCK(sb));
1739 for (i=0; i < 4; i++)
1740 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1741 sbi->s_def_hash_version = es->s_def_hash_version;
1742
1743 if (sbi->s_blocks_per_group > blocksize * 8) {
1744 printk (KERN_ERR
1745 "EXT3-fs: #blocks per group too big: %lu\n",
1746 sbi->s_blocks_per_group);
1747 goto failed_mount;
1748 }
1749 if (sbi->s_frags_per_group > blocksize * 8) {
1750 printk (KERN_ERR
1751 "EXT3-fs: #fragments per group too big: %lu\n",
1752 sbi->s_frags_per_group);
1753 goto failed_mount;
1754 }
1755 if (sbi->s_inodes_per_group > blocksize * 8) {
1756 printk (KERN_ERR
1757 "EXT3-fs: #inodes per group too big: %lu\n",
1758 sbi->s_inodes_per_group);
1759 goto failed_mount;
1760 }
1761
1762 if (le32_to_cpu(es->s_blocks_count) >
1763 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
1764 printk(KERN_ERR "EXT3-fs: filesystem on %s:"
1765 " too large to mount safely\n", sb->s_id);
1766 if (sizeof(sector_t) < 8)
1767 printk(KERN_WARNING "EXT3-fs: CONFIG_LBD not "
1768 "enabled\n");
1769 goto failed_mount;
1770 }
1771
1772 if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1773 goto cantfind_ext3;
1774 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1775 le32_to_cpu(es->s_first_data_block) - 1)
1776 / EXT3_BLOCKS_PER_GROUP(sb)) + 1;
1777 db_count = (sbi->s_groups_count + EXT3_DESC_PER_BLOCK(sb) - 1) /
1778 EXT3_DESC_PER_BLOCK(sb);
1779 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1780 GFP_KERNEL);
1781 if (sbi->s_group_desc == NULL) {
1782 printk (KERN_ERR "EXT3-fs: not enough memory\n");
1783 goto failed_mount;
1784 }
1785
1786 bgl_lock_init(&sbi->s_blockgroup_lock);
1787
1788 for (i = 0; i < db_count; i++) {
1789 block = descriptor_loc(sb, logic_sb_block, i);
1790 sbi->s_group_desc[i] = sb_bread(sb, block);
1791 if (!sbi->s_group_desc[i]) {
1792 printk (KERN_ERR "EXT3-fs: "
1793 "can't read group descriptor %d\n", i);
1794 db_count = i;
1795 goto failed_mount2;
1796 }
1797 }
1798 if (!ext3_check_descriptors (sb)) {
1799 printk(KERN_ERR "EXT3-fs: group descriptors corrupted!\n");
1800 goto failed_mount2;
1801 }
1802 sbi->s_gdb_count = db_count;
1803 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1804 spin_lock_init(&sbi->s_next_gen_lock);
1805
1806 err = percpu_counter_init(&sbi->s_freeblocks_counter,
1807 ext3_count_free_blocks(sb));
1808 if (!err) {
1809 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1810 ext3_count_free_inodes(sb));
1811 }
1812 if (!err) {
1813 err = percpu_counter_init(&sbi->s_dirs_counter,
1814 ext3_count_dirs(sb));
1815 }
1816 if (err) {
1817 printk(KERN_ERR "EXT3-fs: insufficient memory\n");
1818 goto failed_mount3;
1819 }
1820
1821 /* per fileystem reservation list head & lock */
1822 spin_lock_init(&sbi->s_rsv_window_lock);
1823 sbi->s_rsv_window_root = RB_ROOT;
1824 /* Add a single, static dummy reservation to the start of the
1825 * reservation window list --- it gives us a placeholder for
1826 * append-at-start-of-list which makes the allocation logic
1827 * _much_ simpler. */
1828 sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1829 sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1830 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1831 sbi->s_rsv_window_head.rsv_goal_size = 0;
1832 ext3_rsv_window_add(sb, &sbi->s_rsv_window_head);
1833
1834 /*
1835 * set up enough so that it can read an inode
1836 */
1837 sb->s_op = &ext3_sops;
1838 sb->s_export_op = &ext3_export_ops;
1839 sb->s_xattr = ext3_xattr_handlers;
1840 #ifdef CONFIG_QUOTA
1841 sb->s_qcop = &ext3_qctl_operations;
1842 sb->dq_op = &ext3_quota_operations;
1843 #endif
1844 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1845
1846 sb->s_root = NULL;
1847
1848 needs_recovery = (es->s_last_orphan != 0 ||
1849 EXT3_HAS_INCOMPAT_FEATURE(sb,
1850 EXT3_FEATURE_INCOMPAT_RECOVER));
1851
1852 /*
1853 * The first inode we look at is the journal inode. Don't try
1854 * root first: it may be modified in the journal!
1855 */
1856 if (!test_opt(sb, NOLOAD) &&
1857 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
1858 if (ext3_load_journal(sb, es, journal_devnum))
1859 goto failed_mount3;
1860 } else if (journal_inum) {
1861 if (ext3_create_journal(sb, es, journal_inum))
1862 goto failed_mount3;
1863 } else {
1864 if (!silent)
1865 printk (KERN_ERR
1866 "ext3: No journal on filesystem on %s\n",
1867 sb->s_id);
1868 goto failed_mount3;
1869 }
1870
1871 /* We have now updated the journal if required, so we can
1872 * validate the data journaling mode. */
1873 switch (test_opt(sb, DATA_FLAGS)) {
1874 case 0:
1875 /* No mode set, assume a default based on the journal
1876 capabilities: ORDERED_DATA if the journal can
1877 cope, else JOURNAL_DATA */
1878 if (journal_check_available_features
1879 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
1880 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1881 else
1882 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1883 break;
1884
1885 case EXT3_MOUNT_ORDERED_DATA:
1886 case EXT3_MOUNT_WRITEBACK_DATA:
1887 if (!journal_check_available_features
1888 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
1889 printk(KERN_ERR "EXT3-fs: Journal does not support "
1890 "requested data journaling mode\n");
1891 goto failed_mount4;
1892 }
1893 default:
1894 break;
1895 }
1896
1897 if (test_opt(sb, NOBH)) {
1898 if (!(test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)) {
1899 printk(KERN_WARNING "EXT3-fs: Ignoring nobh option - "
1900 "its supported only with writeback mode\n");
1901 clear_opt(sbi->s_mount_opt, NOBH);
1902 }
1903 }
1904 /*
1905 * The journal_load will have done any necessary log recovery,
1906 * so we can safely mount the rest of the filesystem now.
1907 */
1908
1909 root = ext3_iget(sb, EXT3_ROOT_INO);
1910 if (IS_ERR(root)) {
1911 printk(KERN_ERR "EXT3-fs: get root inode failed\n");
1912 ret = PTR_ERR(root);
1913 goto failed_mount4;
1914 }
1915 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1916 iput(root);
1917 printk(KERN_ERR "EXT3-fs: corrupt root inode, run e2fsck\n");
1918 goto failed_mount4;
1919 }
1920 sb->s_root = d_alloc_root(root);
1921 if (!sb->s_root) {
1922 printk(KERN_ERR "EXT3-fs: get root dentry failed\n");
1923 iput(root);
1924 ret = -ENOMEM;
1925 goto failed_mount4;
1926 }
1927
1928 ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1929 /*
1930 * akpm: core read_super() calls in here with the superblock locked.
1931 * That deadlocks, because orphan cleanup needs to lock the superblock
1932 * in numerous places. Here we just pop the lock - it's relatively
1933 * harmless, because we are now ready to accept write_super() requests,
1934 * and aviro says that's the only reason for hanging onto the
1935 * superblock lock.
1936 */
1937 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
1938 ext3_orphan_cleanup(sb, es);
1939 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
1940 if (needs_recovery)
1941 printk (KERN_INFO "EXT3-fs: recovery complete.\n");
1942 ext3_mark_recovery_complete(sb, es);
1943 printk (KERN_INFO "EXT3-fs: mounted filesystem with %s data mode.\n",
1944 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
1945 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
1946 "writeback");
1947
1948 lock_kernel();
1949 return 0;
1950
1951 cantfind_ext3:
1952 if (!silent)
1953 printk(KERN_ERR "VFS: Can't find ext3 filesystem on dev %s.\n",
1954 sb->s_id);
1955 goto failed_mount;
1956
1957 failed_mount4:
1958 journal_destroy(sbi->s_journal);
1959 failed_mount3:
1960 percpu_counter_destroy(&sbi->s_freeblocks_counter);
1961 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1962 percpu_counter_destroy(&sbi->s_dirs_counter);
1963 failed_mount2:
1964 for (i = 0; i < db_count; i++)
1965 brelse(sbi->s_group_desc[i]);
1966 kfree(sbi->s_group_desc);
1967 failed_mount:
1968 #ifdef CONFIG_QUOTA
1969 for (i = 0; i < MAXQUOTAS; i++)
1970 kfree(sbi->s_qf_names[i]);
1971 #endif
1972 ext3_blkdev_remove(sbi);
1973 brelse(bh);
1974 out_fail:
1975 sb->s_fs_info = NULL;
1976 kfree(sbi);
1977 lock_kernel();
1978 return ret;
1979 }
1980
1981 /*
1982 * Setup any per-fs journal parameters now. We'll do this both on
1983 * initial mount, once the journal has been initialised but before we've
1984 * done any recovery; and again on any subsequent remount.
1985 */
1986 static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
1987 {
1988 struct ext3_sb_info *sbi = EXT3_SB(sb);
1989
1990 if (sbi->s_commit_interval)
1991 journal->j_commit_interval = sbi->s_commit_interval;
1992 /* We could also set up an ext3-specific default for the commit
1993 * interval here, but for now we'll just fall back to the jbd
1994 * default. */
1995
1996 spin_lock(&journal->j_state_lock);
1997 if (test_opt(sb, BARRIER))
1998 journal->j_flags |= JFS_BARRIER;
1999 else
2000 journal->j_flags &= ~JFS_BARRIER;
2001 if (test_opt(sb, DATA_ERR_ABORT))
2002 journal->j_flags |= JFS_ABORT_ON_SYNCDATA_ERR;
2003 else
2004 journal->j_flags &= ~JFS_ABORT_ON_SYNCDATA_ERR;
2005 spin_unlock(&journal->j_state_lock);
2006 }
2007
2008 static journal_t *ext3_get_journal(struct super_block *sb,
2009 unsigned int journal_inum)
2010 {
2011 struct inode *journal_inode;
2012 journal_t *journal;
2013
2014 /* First, test for the existence of a valid inode on disk. Bad
2015 * things happen if we iget() an unused inode, as the subsequent
2016 * iput() will try to delete it. */
2017
2018 journal_inode = ext3_iget(sb, journal_inum);
2019 if (IS_ERR(journal_inode)) {
2020 printk(KERN_ERR "EXT3-fs: no journal found.\n");
2021 return NULL;
2022 }
2023 if (!journal_inode->i_nlink) {
2024 make_bad_inode(journal_inode);
2025 iput(journal_inode);
2026 printk(KERN_ERR "EXT3-fs: journal inode is deleted.\n");
2027 return NULL;
2028 }
2029
2030 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2031 journal_inode, journal_inode->i_size);
2032 if (!S_ISREG(journal_inode->i_mode)) {
2033 printk(KERN_ERR "EXT3-fs: invalid journal inode.\n");
2034 iput(journal_inode);
2035 return NULL;
2036 }
2037
2038 journal = journal_init_inode(journal_inode);
2039 if (!journal) {
2040 printk(KERN_ERR "EXT3-fs: Could not load journal inode\n");
2041 iput(journal_inode);
2042 return NULL;
2043 }
2044 journal->j_private = sb;
2045 ext3_init_journal_params(sb, journal);
2046 return journal;
2047 }
2048
2049 static journal_t *ext3_get_dev_journal(struct super_block *sb,
2050 dev_t j_dev)
2051 {
2052 struct buffer_head * bh;
2053 journal_t *journal;
2054 ext3_fsblk_t start;
2055 ext3_fsblk_t len;
2056 int hblock, blocksize;
2057 ext3_fsblk_t sb_block;
2058 unsigned long offset;
2059 struct ext3_super_block * es;
2060 struct block_device *bdev;
2061
2062 bdev = ext3_blkdev_get(j_dev);
2063 if (bdev == NULL)
2064 return NULL;
2065
2066 if (bd_claim(bdev, sb)) {
2067 printk(KERN_ERR
2068 "EXT3: failed to claim external journal device.\n");
2069 blkdev_put(bdev);
2070 return NULL;
2071 }
2072
2073 blocksize = sb->s_blocksize;
2074 hblock = bdev_hardsect_size(bdev);
2075 if (blocksize < hblock) {
2076 printk(KERN_ERR
2077 "EXT3-fs: blocksize too small for journal device.\n");
2078 goto out_bdev;
2079 }
2080
2081 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
2082 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
2083 set_blocksize(bdev, blocksize);
2084 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2085 printk(KERN_ERR "EXT3-fs: couldn't read superblock of "
2086 "external journal\n");
2087 goto out_bdev;
2088 }
2089
2090 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
2091 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
2092 !(le32_to_cpu(es->s_feature_incompat) &
2093 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2094 printk(KERN_ERR "EXT3-fs: external journal has "
2095 "bad superblock\n");
2096 brelse(bh);
2097 goto out_bdev;
2098 }
2099
2100 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2101 printk(KERN_ERR "EXT3-fs: journal UUID does not match\n");
2102 brelse(bh);
2103 goto out_bdev;
2104 }
2105
2106 len = le32_to_cpu(es->s_blocks_count);
2107 start = sb_block + 1;
2108 brelse(bh); /* we're done with the superblock */
2109
2110 journal = journal_init_dev(bdev, sb->s_bdev,
2111 start, len, blocksize);
2112 if (!journal) {
2113 printk(KERN_ERR "EXT3-fs: failed to create device journal\n");
2114 goto out_bdev;
2115 }
2116 journal->j_private = sb;
2117 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2118 wait_on_buffer(journal->j_sb_buffer);
2119 if (!buffer_uptodate(journal->j_sb_buffer)) {
2120 printk(KERN_ERR "EXT3-fs: I/O error on journal device\n");
2121 goto out_journal;
2122 }
2123 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2124 printk(KERN_ERR "EXT3-fs: External journal has more than one "
2125 "user (unsupported) - %d\n",
2126 be32_to_cpu(journal->j_superblock->s_nr_users));
2127 goto out_journal;
2128 }
2129 EXT3_SB(sb)->journal_bdev = bdev;
2130 ext3_init_journal_params(sb, journal);
2131 return journal;
2132 out_journal:
2133 journal_destroy(journal);
2134 out_bdev:
2135 ext3_blkdev_put(bdev);
2136 return NULL;
2137 }
2138
2139 static int ext3_load_journal(struct super_block *sb,
2140 struct ext3_super_block *es,
2141 unsigned long journal_devnum)
2142 {
2143 journal_t *journal;
2144 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2145 dev_t journal_dev;
2146 int err = 0;
2147 int really_read_only;
2148
2149 if (journal_devnum &&
2150 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2151 printk(KERN_INFO "EXT3-fs: external journal device major/minor "
2152 "numbers have changed\n");
2153 journal_dev = new_decode_dev(journal_devnum);
2154 } else
2155 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2156
2157 really_read_only = bdev_read_only(sb->s_bdev);
2158
2159 /*
2160 * Are we loading a blank journal or performing recovery after a
2161 * crash? For recovery, we need to check in advance whether we
2162 * can get read-write access to the device.
2163 */
2164
2165 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
2166 if (sb->s_flags & MS_RDONLY) {
2167 printk(KERN_INFO "EXT3-fs: INFO: recovery "
2168 "required on readonly filesystem.\n");
2169 if (really_read_only) {
2170 printk(KERN_ERR "EXT3-fs: write access "
2171 "unavailable, cannot proceed.\n");
2172 return -EROFS;
2173 }
2174 printk (KERN_INFO "EXT3-fs: write access will "
2175 "be enabled during recovery.\n");
2176 }
2177 }
2178
2179 if (journal_inum && journal_dev) {
2180 printk(KERN_ERR "EXT3-fs: filesystem has both journal "
2181 "and inode journals!\n");
2182 return -EINVAL;
2183 }
2184
2185 if (journal_inum) {
2186 if (!(journal = ext3_get_journal(sb, journal_inum)))
2187 return -EINVAL;
2188 } else {
2189 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
2190 return -EINVAL;
2191 }
2192
2193 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2194 err = journal_update_format(journal);
2195 if (err) {
2196 printk(KERN_ERR "EXT3-fs: error updating journal.\n");
2197 journal_destroy(journal);
2198 return err;
2199 }
2200 }
2201
2202 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
2203 err = journal_wipe(journal, !really_read_only);
2204 if (!err)
2205 err = journal_load(journal);
2206
2207 if (err) {
2208 printk(KERN_ERR "EXT3-fs: error loading journal.\n");
2209 journal_destroy(journal);
2210 return err;
2211 }
2212
2213 EXT3_SB(sb)->s_journal = journal;
2214 ext3_clear_journal_err(sb, es);
2215
2216 if (journal_devnum &&
2217 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2218 es->s_journal_dev = cpu_to_le32(journal_devnum);
2219 sb->s_dirt = 1;
2220
2221 /* Make sure we flush the recovery flag to disk. */
2222 ext3_commit_super(sb, es, 1);
2223 }
2224
2225 return 0;
2226 }
2227
2228 static int ext3_create_journal(struct super_block * sb,
2229 struct ext3_super_block * es,
2230 unsigned int journal_inum)
2231 {
2232 journal_t *journal;
2233 int err;
2234
2235 if (sb->s_flags & MS_RDONLY) {
2236 printk(KERN_ERR "EXT3-fs: readonly filesystem when trying to "
2237 "create journal.\n");
2238 return -EROFS;
2239 }
2240
2241 journal = ext3_get_journal(sb, journal_inum);
2242 if (!journal)
2243 return -EINVAL;
2244
2245 printk(KERN_INFO "EXT3-fs: creating new journal on inode %u\n",
2246 journal_inum);
2247
2248 err = journal_create(journal);
2249 if (err) {
2250 printk(KERN_ERR "EXT3-fs: error creating journal.\n");
2251 journal_destroy(journal);
2252 return -EIO;
2253 }
2254
2255 EXT3_SB(sb)->s_journal = journal;
2256
2257 ext3_update_dynamic_rev(sb);
2258 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2259 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
2260
2261 es->s_journal_inum = cpu_to_le32(journal_inum);
2262 sb->s_dirt = 1;
2263
2264 /* Make sure we flush the recovery flag to disk. */
2265 ext3_commit_super(sb, es, 1);
2266
2267 return 0;
2268 }
2269
2270 static void ext3_commit_super (struct super_block * sb,
2271 struct ext3_super_block * es,
2272 int sync)
2273 {
2274 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
2275
2276 if (!sbh)
2277 return;
2278 es->s_wtime = cpu_to_le32(get_seconds());
2279 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
2280 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
2281 BUFFER_TRACE(sbh, "marking dirty");
2282 mark_buffer_dirty(sbh);
2283 if (sync)
2284 sync_dirty_buffer(sbh);
2285 }
2286
2287
2288 /*
2289 * Have we just finished recovery? If so, and if we are mounting (or
2290 * remounting) the filesystem readonly, then we will end up with a
2291 * consistent fs on disk. Record that fact.
2292 */
2293 static void ext3_mark_recovery_complete(struct super_block * sb,
2294 struct ext3_super_block * es)
2295 {
2296 journal_t *journal = EXT3_SB(sb)->s_journal;
2297
2298 journal_lock_updates(journal);
2299 journal_flush(journal);
2300 lock_super(sb);
2301 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
2302 sb->s_flags & MS_RDONLY) {
2303 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2304 sb->s_dirt = 0;
2305 ext3_commit_super(sb, es, 1);
2306 }
2307 unlock_super(sb);
2308 journal_unlock_updates(journal);
2309 }
2310
2311 /*
2312 * If we are mounting (or read-write remounting) a filesystem whose journal
2313 * has recorded an error from a previous lifetime, move that error to the
2314 * main filesystem now.
2315 */
2316 static void ext3_clear_journal_err(struct super_block * sb,
2317 struct ext3_super_block * es)
2318 {
2319 journal_t *journal;
2320 int j_errno;
2321 const char *errstr;
2322
2323 journal = EXT3_SB(sb)->s_journal;
2324
2325 /*
2326 * Now check for any error status which may have been recorded in the
2327 * journal by a prior ext3_error() or ext3_abort()
2328 */
2329
2330 j_errno = journal_errno(journal);
2331 if (j_errno) {
2332 char nbuf[16];
2333
2334 errstr = ext3_decode_error(sb, j_errno, nbuf);
2335 ext3_warning(sb, __func__, "Filesystem error recorded "
2336 "from previous mount: %s", errstr);
2337 ext3_warning(sb, __func__, "Marking fs in need of "
2338 "filesystem check.");
2339
2340 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
2341 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
2342 ext3_commit_super (sb, es, 1);
2343
2344 journal_clear_err(journal);
2345 }
2346 }
2347
2348 /*
2349 * Force the running and committing transactions to commit,
2350 * and wait on the commit.
2351 */
2352 int ext3_force_commit(struct super_block *sb)
2353 {
2354 journal_t *journal;
2355 int ret;
2356
2357 if (sb->s_flags & MS_RDONLY)
2358 return 0;
2359
2360 journal = EXT3_SB(sb)->s_journal;
2361 sb->s_dirt = 0;
2362 ret = ext3_journal_force_commit(journal);
2363 return ret;
2364 }
2365
2366 /*
2367 * Ext3 always journals updates to the superblock itself, so we don't
2368 * have to propagate any other updates to the superblock on disk at this
2369 * point. Just start an async writeback to get the buffers on their way
2370 * to the disk.
2371 *
2372 * This implicitly triggers the writebehind on sync().
2373 */
2374
2375 static void ext3_write_super (struct super_block * sb)
2376 {
2377 if (mutex_trylock(&sb->s_lock) != 0)
2378 BUG();
2379 sb->s_dirt = 0;
2380 }
2381
2382 static int ext3_sync_fs(struct super_block *sb, int wait)
2383 {
2384 tid_t target;
2385
2386 sb->s_dirt = 0;
2387 if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
2388 if (wait)
2389 log_wait_commit(EXT3_SB(sb)->s_journal, target);
2390 }
2391 return 0;
2392 }
2393
2394 /*
2395 * LVM calls this function before a (read-only) snapshot is created. This
2396 * gives us a chance to flush the journal completely and mark the fs clean.
2397 */
2398 static void ext3_write_super_lockfs(struct super_block *sb)
2399 {
2400 sb->s_dirt = 0;
2401
2402 if (!(sb->s_flags & MS_RDONLY)) {
2403 journal_t *journal = EXT3_SB(sb)->s_journal;
2404
2405 /* Now we set up the journal barrier. */
2406 journal_lock_updates(journal);
2407 journal_flush(journal);
2408
2409 /* Journal blocked and flushed, clear needs_recovery flag. */
2410 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2411 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2412 }
2413 }
2414
2415 /*
2416 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2417 * flag here, even though the filesystem is not technically dirty yet.
2418 */
2419 static void ext3_unlockfs(struct super_block *sb)
2420 {
2421 if (!(sb->s_flags & MS_RDONLY)) {
2422 lock_super(sb);
2423 /* Reser the needs_recovery flag before the fs is unlocked. */
2424 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2425 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2426 unlock_super(sb);
2427 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2428 }
2429 }
2430
2431 static int ext3_remount (struct super_block * sb, int * flags, char * data)
2432 {
2433 struct ext3_super_block * es;
2434 struct ext3_sb_info *sbi = EXT3_SB(sb);
2435 ext3_fsblk_t n_blocks_count = 0;
2436 unsigned long old_sb_flags;
2437 struct ext3_mount_options old_opts;
2438 int err;
2439 #ifdef CONFIG_QUOTA
2440 int i;
2441 #endif
2442
2443 /* Store the original options */
2444 old_sb_flags = sb->s_flags;
2445 old_opts.s_mount_opt = sbi->s_mount_opt;
2446 old_opts.s_resuid = sbi->s_resuid;
2447 old_opts.s_resgid = sbi->s_resgid;
2448 old_opts.s_commit_interval = sbi->s_commit_interval;
2449 #ifdef CONFIG_QUOTA
2450 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2451 for (i = 0; i < MAXQUOTAS; i++)
2452 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2453 #endif
2454
2455 /*
2456 * Allow the "check" option to be passed as a remount option.
2457 */
2458 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2459 err = -EINVAL;
2460 goto restore_opts;
2461 }
2462
2463 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT)
2464 ext3_abort(sb, __func__, "Abort forced by user");
2465
2466 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2467 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2468
2469 es = sbi->s_es;
2470
2471 ext3_init_journal_params(sb, sbi->s_journal);
2472
2473 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2474 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
2475 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT) {
2476 err = -EROFS;
2477 goto restore_opts;
2478 }
2479
2480 if (*flags & MS_RDONLY) {
2481 /*
2482 * First of all, the unconditional stuff we have to do
2483 * to disable replay of the journal when we next remount
2484 */
2485 sb->s_flags |= MS_RDONLY;
2486
2487 /*
2488 * OK, test if we are remounting a valid rw partition
2489 * readonly, and if so set the rdonly flag and then
2490 * mark the partition as valid again.
2491 */
2492 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2493 (sbi->s_mount_state & EXT3_VALID_FS))
2494 es->s_state = cpu_to_le16(sbi->s_mount_state);
2495
2496 /*
2497 * We have to unlock super so that we can wait for
2498 * transactions.
2499 */
2500 unlock_super(sb);
2501 ext3_mark_recovery_complete(sb, es);
2502 lock_super(sb);
2503 } else {
2504 __le32 ret;
2505 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2506 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
2507 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2508 "remount RDWR because of unsupported "
2509 "optional features (%x).\n",
2510 sb->s_id, le32_to_cpu(ret));
2511 err = -EROFS;
2512 goto restore_opts;
2513 }
2514
2515 /*
2516 * If we have an unprocessed orphan list hanging
2517 * around from a previously readonly bdev mount,
2518 * require a full umount/remount for now.
2519 */
2520 if (es->s_last_orphan) {
2521 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2522 "remount RDWR because of unprocessed "
2523 "orphan inode list. Please "
2524 "umount/remount instead.\n",
2525 sb->s_id);
2526 err = -EINVAL;
2527 goto restore_opts;
2528 }
2529
2530 /*
2531 * Mounting a RDONLY partition read-write, so reread
2532 * and store the current valid flag. (It may have
2533 * been changed by e2fsck since we originally mounted
2534 * the partition.)
2535 */
2536 ext3_clear_journal_err(sb, es);
2537 sbi->s_mount_state = le16_to_cpu(es->s_state);
2538 if ((err = ext3_group_extend(sb, es, n_blocks_count)))
2539 goto restore_opts;
2540 if (!ext3_setup_super (sb, es, 0))
2541 sb->s_flags &= ~MS_RDONLY;
2542 }
2543 }
2544 #ifdef CONFIG_QUOTA
2545 /* Release old quota file names */
2546 for (i = 0; i < MAXQUOTAS; i++)
2547 if (old_opts.s_qf_names[i] &&
2548 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2549 kfree(old_opts.s_qf_names[i]);
2550 #endif
2551 return 0;
2552 restore_opts:
2553 sb->s_flags = old_sb_flags;
2554 sbi->s_mount_opt = old_opts.s_mount_opt;
2555 sbi->s_resuid = old_opts.s_resuid;
2556 sbi->s_resgid = old_opts.s_resgid;
2557 sbi->s_commit_interval = old_opts.s_commit_interval;
2558 #ifdef CONFIG_QUOTA
2559 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2560 for (i = 0; i < MAXQUOTAS; i++) {
2561 if (sbi->s_qf_names[i] &&
2562 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2563 kfree(sbi->s_qf_names[i]);
2564 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2565 }
2566 #endif
2567 return err;
2568 }
2569
2570 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf)
2571 {
2572 struct super_block *sb = dentry->d_sb;
2573 struct ext3_sb_info *sbi = EXT3_SB(sb);
2574 struct ext3_super_block *es = sbi->s_es;
2575 u64 fsid;
2576
2577 if (test_opt(sb, MINIX_DF)) {
2578 sbi->s_overhead_last = 0;
2579 } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
2580 unsigned long ngroups = sbi->s_groups_count, i;
2581 ext3_fsblk_t overhead = 0;
2582 smp_rmb();
2583
2584 /*
2585 * Compute the overhead (FS structures). This is constant
2586 * for a given filesystem unless the number of block groups
2587 * changes so we cache the previous value until it does.
2588 */
2589
2590 /*
2591 * All of the blocks before first_data_block are
2592 * overhead
2593 */
2594 overhead = le32_to_cpu(es->s_first_data_block);
2595
2596 /*
2597 * Add the overhead attributed to the superblock and
2598 * block group descriptors. If the sparse superblocks
2599 * feature is turned on, then not all groups have this.
2600 */
2601 for (i = 0; i < ngroups; i++) {
2602 overhead += ext3_bg_has_super(sb, i) +
2603 ext3_bg_num_gdb(sb, i);
2604 cond_resched();
2605 }
2606
2607 /*
2608 * Every block group has an inode bitmap, a block
2609 * bitmap, and an inode table.
2610 */
2611 overhead += ngroups * (2 + sbi->s_itb_per_group);
2612 sbi->s_overhead_last = overhead;
2613 smp_wmb();
2614 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
2615 }
2616
2617 buf->f_type = EXT3_SUPER_MAGIC;
2618 buf->f_bsize = sb->s_blocksize;
2619 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
2620 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
2621 es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
2622 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2623 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2624 buf->f_bavail = 0;
2625 buf->f_files = le32_to_cpu(es->s_inodes_count);
2626 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
2627 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
2628 buf->f_namelen = EXT3_NAME_LEN;
2629 fsid = le64_to_cpup((void *)es->s_uuid) ^
2630 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2631 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2632 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
2633 return 0;
2634 }
2635
2636 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2637 * is locked for write. Otherwise the are possible deadlocks:
2638 * Process 1 Process 2
2639 * ext3_create() quota_sync()
2640 * journal_start() write_dquot()
2641 * DQUOT_INIT() down(dqio_mutex)
2642 * down(dqio_mutex) journal_start()
2643 *
2644 */
2645
2646 #ifdef CONFIG_QUOTA
2647
2648 static inline struct inode *dquot_to_inode(struct dquot *dquot)
2649 {
2650 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2651 }
2652
2653 static int ext3_dquot_initialize(struct inode *inode, int type)
2654 {
2655 handle_t *handle;
2656 int ret, err;
2657
2658 /* We may create quota structure so we need to reserve enough blocks */
2659 handle = ext3_journal_start(inode, 2*EXT3_QUOTA_INIT_BLOCKS(inode->i_sb));
2660 if (IS_ERR(handle))
2661 return PTR_ERR(handle);
2662 ret = dquot_initialize(inode, type);
2663 err = ext3_journal_stop(handle);
2664 if (!ret)
2665 ret = err;
2666 return ret;
2667 }
2668
2669 static int ext3_dquot_drop(struct inode *inode)
2670 {
2671 handle_t *handle;
2672 int ret, err;
2673
2674 /* We may delete quota structure so we need to reserve enough blocks */
2675 handle = ext3_journal_start(inode, 2*EXT3_QUOTA_DEL_BLOCKS(inode->i_sb));
2676 if (IS_ERR(handle)) {
2677 /*
2678 * We call dquot_drop() anyway to at least release references
2679 * to quota structures so that umount does not hang.
2680 */
2681 dquot_drop(inode);
2682 return PTR_ERR(handle);
2683 }
2684 ret = dquot_drop(inode);
2685 err = ext3_journal_stop(handle);
2686 if (!ret)
2687 ret = err;
2688 return ret;
2689 }
2690
2691 static int ext3_write_dquot(struct dquot *dquot)
2692 {
2693 int ret, err;
2694 handle_t *handle;
2695 struct inode *inode;
2696
2697 inode = dquot_to_inode(dquot);
2698 handle = ext3_journal_start(inode,
2699 EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
2700 if (IS_ERR(handle))
2701 return PTR_ERR(handle);
2702 ret = dquot_commit(dquot);
2703 err = ext3_journal_stop(handle);
2704 if (!ret)
2705 ret = err;
2706 return ret;
2707 }
2708
2709 static int ext3_acquire_dquot(struct dquot *dquot)
2710 {
2711 int ret, err;
2712 handle_t *handle;
2713
2714 handle = ext3_journal_start(dquot_to_inode(dquot),
2715 EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
2716 if (IS_ERR(handle))
2717 return PTR_ERR(handle);
2718 ret = dquot_acquire(dquot);
2719 err = ext3_journal_stop(handle);
2720 if (!ret)
2721 ret = err;
2722 return ret;
2723 }
2724
2725 static int ext3_release_dquot(struct dquot *dquot)
2726 {
2727 int ret, err;
2728 handle_t *handle;
2729
2730 handle = ext3_journal_start(dquot_to_inode(dquot),
2731 EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
2732 if (IS_ERR(handle)) {
2733 /* Release dquot anyway to avoid endless cycle in dqput() */
2734 dquot_release(dquot);
2735 return PTR_ERR(handle);
2736 }
2737 ret = dquot_release(dquot);
2738 err = ext3_journal_stop(handle);
2739 if (!ret)
2740 ret = err;
2741 return ret;
2742 }
2743
2744 static int ext3_mark_dquot_dirty(struct dquot *dquot)
2745 {
2746 /* Are we journaling quotas? */
2747 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2748 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2749 dquot_mark_dquot_dirty(dquot);
2750 return ext3_write_dquot(dquot);
2751 } else {
2752 return dquot_mark_dquot_dirty(dquot);
2753 }
2754 }
2755
2756 static int ext3_write_info(struct super_block *sb, int type)
2757 {
2758 int ret, err;
2759 handle_t *handle;
2760
2761 /* Data block + inode block */
2762 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2763 if (IS_ERR(handle))
2764 return PTR_ERR(handle);
2765 ret = dquot_commit_info(sb, type);
2766 err = ext3_journal_stop(handle);
2767 if (!ret)
2768 ret = err;
2769 return ret;
2770 }
2771
2772 /*
2773 * Turn on quotas during mount time - we need to find
2774 * the quota file and such...
2775 */
2776 static int ext3_quota_on_mount(struct super_block *sb, int type)
2777 {
2778 return vfs_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2779 EXT3_SB(sb)->s_jquota_fmt, type);
2780 }
2781
2782 /*
2783 * Standard function to be called on quota_on
2784 */
2785 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
2786 char *path, int remount)
2787 {
2788 int err;
2789 struct nameidata nd;
2790
2791 if (!test_opt(sb, QUOTA))
2792 return -EINVAL;
2793 /* When remounting, no checks are needed and in fact, path is NULL */
2794 if (remount)
2795 return vfs_quota_on(sb, type, format_id, path, remount);
2796
2797 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
2798 if (err)
2799 return err;
2800
2801 /* Quotafile not on the same filesystem? */
2802 if (nd.path.mnt->mnt_sb != sb) {
2803 path_put(&nd.path);
2804 return -EXDEV;
2805 }
2806 /* Journaling quota? */
2807 if (EXT3_SB(sb)->s_qf_names[type]) {
2808 /* Quotafile not of fs root? */
2809 if (nd.path.dentry->d_parent->d_inode != sb->s_root->d_inode)
2810 printk(KERN_WARNING
2811 "EXT3-fs: Quota file not on filesystem root. "
2812 "Journaled quota will not work.\n");
2813 }
2814
2815 /*
2816 * When we journal data on quota file, we have to flush journal to see
2817 * all updates to the file when we bypass pagecache...
2818 */
2819 if (ext3_should_journal_data(nd.path.dentry->d_inode)) {
2820 /*
2821 * We don't need to lock updates but journal_flush() could
2822 * otherwise be livelocked...
2823 */
2824 journal_lock_updates(EXT3_SB(sb)->s_journal);
2825 journal_flush(EXT3_SB(sb)->s_journal);
2826 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2827 }
2828
2829 err = vfs_quota_on_path(sb, type, format_id, &nd.path);
2830 path_put(&nd.path);
2831 return err;
2832 }
2833
2834 /* Read data from quotafile - avoid pagecache and such because we cannot afford
2835 * acquiring the locks... As quota files are never truncated and quota code
2836 * itself serializes the operations (and noone else should touch the files)
2837 * we don't have to be afraid of races */
2838 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
2839 size_t len, loff_t off)
2840 {
2841 struct inode *inode = sb_dqopt(sb)->files[type];
2842 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2843 int err = 0;
2844 int offset = off & (sb->s_blocksize - 1);
2845 int tocopy;
2846 size_t toread;
2847 struct buffer_head *bh;
2848 loff_t i_size = i_size_read(inode);
2849
2850 if (off > i_size)
2851 return 0;
2852 if (off+len > i_size)
2853 len = i_size-off;
2854 toread = len;
2855 while (toread > 0) {
2856 tocopy = sb->s_blocksize - offset < toread ?
2857 sb->s_blocksize - offset : toread;
2858 bh = ext3_bread(NULL, inode, blk, 0, &err);
2859 if (err)
2860 return err;
2861 if (!bh) /* A hole? */
2862 memset(data, 0, tocopy);
2863 else
2864 memcpy(data, bh->b_data+offset, tocopy);
2865 brelse(bh);
2866 offset = 0;
2867 toread -= tocopy;
2868 data += tocopy;
2869 blk++;
2870 }
2871 return len;
2872 }
2873
2874 /* Write to quotafile (we know the transaction is already started and has
2875 * enough credits) */
2876 static ssize_t ext3_quota_write(struct super_block *sb, int type,
2877 const char *data, size_t len, loff_t off)
2878 {
2879 struct inode *inode = sb_dqopt(sb)->files[type];
2880 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2881 int err = 0;
2882 int offset = off & (sb->s_blocksize - 1);
2883 int tocopy;
2884 int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
2885 size_t towrite = len;
2886 struct buffer_head *bh;
2887 handle_t *handle = journal_current_handle();
2888
2889 if (!handle) {
2890 printk(KERN_WARNING "EXT3-fs: Quota write (off=%Lu, len=%Lu)"
2891 " cancelled because transaction is not started.\n",
2892 (unsigned long long)off, (unsigned long long)len);
2893 return -EIO;
2894 }
2895 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
2896 while (towrite > 0) {
2897 tocopy = sb->s_blocksize - offset < towrite ?
2898 sb->s_blocksize - offset : towrite;
2899 bh = ext3_bread(handle, inode, blk, 1, &err);
2900 if (!bh)
2901 goto out;
2902 if (journal_quota) {
2903 err = ext3_journal_get_write_access(handle, bh);
2904 if (err) {
2905 brelse(bh);
2906 goto out;
2907 }
2908 }
2909 lock_buffer(bh);
2910 memcpy(bh->b_data+offset, data, tocopy);
2911 flush_dcache_page(bh->b_page);
2912 unlock_buffer(bh);
2913 if (journal_quota)
2914 err = ext3_journal_dirty_metadata(handle, bh);
2915 else {
2916 /* Always do at least ordered writes for quotas */
2917 err = ext3_journal_dirty_data(handle, bh);
2918 mark_buffer_dirty(bh);
2919 }
2920 brelse(bh);
2921 if (err)
2922 goto out;
2923 offset = 0;
2924 towrite -= tocopy;
2925 data += tocopy;
2926 blk++;
2927 }
2928 out:
2929 if (len == towrite) {
2930 mutex_unlock(&inode->i_mutex);
2931 return err;
2932 }
2933 if (inode->i_size < off+len-towrite) {
2934 i_size_write(inode, off+len-towrite);
2935 EXT3_I(inode)->i_disksize = inode->i_size;
2936 }
2937 inode->i_version++;
2938 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2939 ext3_mark_inode_dirty(handle, inode);
2940 mutex_unlock(&inode->i_mutex);
2941 return len - towrite;
2942 }
2943
2944 #endif
2945
2946 static int ext3_get_sb(struct file_system_type *fs_type,
2947 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
2948 {
2949 return get_sb_bdev(fs_type, flags, dev_name, data, ext3_fill_super, mnt);
2950 }
2951
2952 static struct file_system_type ext3_fs_type = {
2953 .owner = THIS_MODULE,
2954 .name = "ext3",
2955 .get_sb = ext3_get_sb,
2956 .kill_sb = kill_block_super,
2957 .fs_flags = FS_REQUIRES_DEV,
2958 };
2959
2960 static int __init init_ext3_fs(void)
2961 {
2962 int err = init_ext3_xattr();
2963 if (err)
2964 return err;
2965 err = init_inodecache();
2966 if (err)
2967 goto out1;
2968 err = register_filesystem(&ext3_fs_type);
2969 if (err)
2970 goto out;
2971 return 0;
2972 out:
2973 destroy_inodecache();
2974 out1:
2975 exit_ext3_xattr();
2976 return err;
2977 }
2978
2979 static void __exit exit_ext3_fs(void)
2980 {
2981 unregister_filesystem(&ext3_fs_type);
2982 destroy_inodecache();
2983 exit_ext3_xattr();
2984 }
2985
2986 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
2987 MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
2988 MODULE_LICENSE("GPL");
2989 module_init(init_ext3_fs)
2990 module_exit(exit_ext3_fs)
Linux with added FPC-III support