?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
[INET]: Consolidate xxx_frag_alloc()
[linux-2.6/x86.git] / fs / ext3 / super.c
blob141573de7a9a4ca69416c24c0a84aa00e8e1674d
1 /*
2 * linux/fs/ext3/super.c
3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19 #include <linux/module.h>
20 #include <linux/string.h>
21 #include <linux/fs.h>
22 #include <linux/time.h>
23 #include <linux/jbd.h>
24 #include <linux/ext3_fs.h>
25 #include <linux/ext3_jbd.h>
26 #include <linux/slab.h>
27 #include <linux/init.h>
28 #include <linux/blkdev.h>
29 #include <linux/parser.h>
30 #include <linux/smp_lock.h>
31 #include <linux/buffer_head.h>
32 #include <linux/exportfs.h>
33 #include <linux/vfs.h>
34 #include <linux/random.h>
35 #include <linux/mount.h>
36 #include <linux/namei.h>
37 #include <linux/quotaops.h>
38 #include <linux/seq_file.h>
39 #include <linux/log2.h>
40
41 #include <asm/uaccess.h>
42
43 #include "xattr.h"
44 #include "acl.h"
45 #include "namei.h"
46
47 static int ext3_load_journal(struct super_block *, struct ext3_super_block *,
48 unsigned long journal_devnum);
49 static int ext3_create_journal(struct super_block *, struct ext3_super_block *,
50 unsigned int);
51 static void ext3_commit_super (struct super_block * sb,
52 struct ext3_super_block * es,
53 int sync);
54 static void ext3_mark_recovery_complete(struct super_block * sb,
55 struct ext3_super_block * es);
56 static void ext3_clear_journal_err(struct super_block * sb,
57 struct ext3_super_block * es);
58 static int ext3_sync_fs(struct super_block *sb, int wait);
59 static const char *ext3_decode_error(struct super_block * sb, int errno,
60 char nbuf[16]);
61 static int ext3_remount (struct super_block * sb, int * flags, char * data);
62 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf);
63 static void ext3_unlockfs(struct super_block *sb);
64 static void ext3_write_super (struct super_block * sb);
65 static void ext3_write_super_lockfs(struct super_block *sb);
66
67 /*
68 * Wrappers for journal_start/end.
69 *
70 * The only special thing we need to do here is to make sure that all
71 * journal_end calls result in the superblock being marked dirty, so
72 * that sync() will call the filesystem's write_super callback if
73 * appropriate.
74 */
75 handle_t *ext3_journal_start_sb(struct super_block *sb, int nblocks)
76 {
77 journal_t *journal;
78
79 if (sb->s_flags & MS_RDONLY)
80 return ERR_PTR(-EROFS);
81
82 /* Special case here: if the journal has aborted behind our
83 * backs (eg. EIO in the commit thread), then we still need to
84 * take the FS itself readonly cleanly. */
85 journal = EXT3_SB(sb)->s_journal;
86 if (is_journal_aborted(journal)) {
87 ext3_abort(sb, __FUNCTION__,
88 "Detected aborted journal");
89 return ERR_PTR(-EROFS);
90 }
91
92 return journal_start(journal, nblocks);
93 }
94
95 /*
96 * The only special thing we need to do here is to make sure that all
97 * journal_stop calls result in the superblock being marked dirty, so
98 * that sync() will call the filesystem's write_super callback if
99 * appropriate.
100 */
101 int __ext3_journal_stop(const char *where, handle_t *handle)
102 {
103 struct super_block *sb;
104 int err;
105 int rc;
106
107 sb = handle->h_transaction->t_journal->j_private;
108 err = handle->h_err;
109 rc = journal_stop(handle);
110
111 if (!err)
112 err = rc;
113 if (err)
114 __ext3_std_error(sb, where, err);
115 return err;
116 }
117
118 void ext3_journal_abort_handle(const char *caller, const char *err_fn,
119 struct buffer_head *bh, handle_t *handle, int err)
120 {
121 char nbuf[16];
122 const char *errstr = ext3_decode_error(NULL, err, nbuf);
123
124 if (bh)
125 BUFFER_TRACE(bh, "abort");
126
127 if (!handle->h_err)
128 handle->h_err = err;
129
130 if (is_handle_aborted(handle))
131 return;
132
133 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
134 caller, errstr, err_fn);
135
136 journal_abort_handle(handle);
137 }
138
139 /* Deal with the reporting of failure conditions on a filesystem such as
140 * inconsistencies detected or read IO failures.
141 *
142 * On ext2, we can store the error state of the filesystem in the
143 * superblock. That is not possible on ext3, because we may have other
144 * write ordering constraints on the superblock which prevent us from
145 * writing it out straight away; and given that the journal is about to
146 * be aborted, we can't rely on the current, or future, transactions to
147 * write out the superblock safely.
148 *
149 * We'll just use the journal_abort() error code to record an error in
150 * the journal instead. On recovery, the journal will compain about
151 * that error until we've noted it down and cleared it.
152 */
153
154 static void ext3_handle_error(struct super_block *sb)
155 {
156 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
157
158 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
159 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
160
161 if (sb->s_flags & MS_RDONLY)
162 return;
163
164 if (!test_opt (sb, ERRORS_CONT)) {
165 journal_t *journal = EXT3_SB(sb)->s_journal;
166
167 EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
168 if (journal)
169 journal_abort(journal, -EIO);
170 }
171 if (test_opt (sb, ERRORS_RO)) {
172 printk (KERN_CRIT "Remounting filesystem read-only\n");
173 sb->s_flags |= MS_RDONLY;
174 }
175 ext3_commit_super(sb, es, 1);
176 if (test_opt(sb, ERRORS_PANIC))
177 panic("EXT3-fs (device %s): panic forced after error\n",
178 sb->s_id);
179 }
180
181 void ext3_error (struct super_block * sb, const char * function,
182 const char * fmt, ...)
183 {
184 va_list args;
185
186 va_start(args, fmt);
187 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
188 vprintk(fmt, args);
189 printk("\n");
190 va_end(args);
191
192 ext3_handle_error(sb);
193 }
194
195 static const char *ext3_decode_error(struct super_block * sb, int errno,
196 char nbuf[16])
197 {
198 char *errstr = NULL;
199
200 switch (errno) {
201 case -EIO:
202 errstr = "IO failure";
203 break;
204 case -ENOMEM:
205 errstr = "Out of memory";
206 break;
207 case -EROFS:
208 if (!sb || EXT3_SB(sb)->s_journal->j_flags & JFS_ABORT)
209 errstr = "Journal has aborted";
210 else
211 errstr = "Readonly filesystem";
212 break;
213 default:
214 /* If the caller passed in an extra buffer for unknown
215 * errors, textualise them now. Else we just return
216 * NULL. */
217 if (nbuf) {
218 /* Check for truncated error codes... */
219 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
220 errstr = nbuf;
221 }
222 break;
223 }
224
225 return errstr;
226 }
227
228 /* __ext3_std_error decodes expected errors from journaling functions
229 * automatically and invokes the appropriate error response. */
230
231 void __ext3_std_error (struct super_block * sb, const char * function,
232 int errno)
233 {
234 char nbuf[16];
235 const char *errstr;
236
237 /* Special case: if the error is EROFS, and we're not already
238 * inside a transaction, then there's really no point in logging
239 * an error. */
240 if (errno == -EROFS && journal_current_handle() == NULL &&
241 (sb->s_flags & MS_RDONLY))
242 return;
243
244 errstr = ext3_decode_error(sb, errno, nbuf);
245 printk (KERN_CRIT "EXT3-fs error (device %s) in %s: %s\n",
246 sb->s_id, function, errstr);
247
248 ext3_handle_error(sb);
249 }
250
251 /*
252 * ext3_abort is a much stronger failure handler than ext3_error. The
253 * abort function may be used to deal with unrecoverable failures such
254 * as journal IO errors or ENOMEM at a critical moment in log management.
255 *
256 * We unconditionally force the filesystem into an ABORT|READONLY state,
257 * unless the error response on the fs has been set to panic in which
258 * case we take the easy way out and panic immediately.
259 */
260
261 void ext3_abort (struct super_block * sb, const char * function,
262 const char * fmt, ...)
263 {
264 va_list args;
265
266 printk (KERN_CRIT "ext3_abort called.\n");
267
268 va_start(args, fmt);
269 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
270 vprintk(fmt, args);
271 printk("\n");
272 va_end(args);
273
274 if (test_opt(sb, ERRORS_PANIC))
275 panic("EXT3-fs panic from previous error\n");
276
277 if (sb->s_flags & MS_RDONLY)
278 return;
279
280 printk(KERN_CRIT "Remounting filesystem read-only\n");
281 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
282 sb->s_flags |= MS_RDONLY;
283 EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
284 journal_abort(EXT3_SB(sb)->s_journal, -EIO);
285 }
286
287 void ext3_warning (struct super_block * sb, const char * function,
288 const char * fmt, ...)
289 {
290 va_list args;
291
292 va_start(args, fmt);
293 printk(KERN_WARNING "EXT3-fs warning (device %s): %s: ",
294 sb->s_id, function);
295 vprintk(fmt, args);
296 printk("\n");
297 va_end(args);
298 }
299
300 void ext3_update_dynamic_rev(struct super_block *sb)
301 {
302 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
303
304 if (le32_to_cpu(es->s_rev_level) > EXT3_GOOD_OLD_REV)
305 return;
306
307 ext3_warning(sb, __FUNCTION__,
308 "updating to rev %d because of new feature flag, "
309 "running e2fsck is recommended",
310 EXT3_DYNAMIC_REV);
311
312 es->s_first_ino = cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO);
313 es->s_inode_size = cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE);
314 es->s_rev_level = cpu_to_le32(EXT3_DYNAMIC_REV);
315 /* leave es->s_feature_*compat flags alone */
316 /* es->s_uuid will be set by e2fsck if empty */
317
318 /*
319 * The rest of the superblock fields should be zero, and if not it
320 * means they are likely already in use, so leave them alone. We
321 * can leave it up to e2fsck to clean up any inconsistencies there.
322 */
323 }
324
325 /*
326 * Open the external journal device
327 */
328 static struct block_device *ext3_blkdev_get(dev_t dev)
329 {
330 struct block_device *bdev;
331 char b[BDEVNAME_SIZE];
332
333 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
334 if (IS_ERR(bdev))
335 goto fail;
336 return bdev;
337
338 fail:
339 printk(KERN_ERR "EXT3: failed to open journal device %s: %ld\n",
340 __bdevname(dev, b), PTR_ERR(bdev));
341 return NULL;
342 }
343
344 /*
345 * Release the journal device
346 */
347 static int ext3_blkdev_put(struct block_device *bdev)
348 {
349 bd_release(bdev);
350 return blkdev_put(bdev);
351 }
352
353 static int ext3_blkdev_remove(struct ext3_sb_info *sbi)
354 {
355 struct block_device *bdev;
356 int ret = -ENODEV;
357
358 bdev = sbi->journal_bdev;
359 if (bdev) {
360 ret = ext3_blkdev_put(bdev);
361 sbi->journal_bdev = NULL;
362 }
363 return ret;
364 }
365
366 static inline struct inode *orphan_list_entry(struct list_head *l)
367 {
368 return &list_entry(l, struct ext3_inode_info, i_orphan)->vfs_inode;
369 }
370
371 static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi)
372 {
373 struct list_head *l;
374
375 printk(KERN_ERR "sb orphan head is %d\n",
376 le32_to_cpu(sbi->s_es->s_last_orphan));
377
378 printk(KERN_ERR "sb_info orphan list:\n");
379 list_for_each(l, &sbi->s_orphan) {
380 struct inode *inode = orphan_list_entry(l);
381 printk(KERN_ERR " "
382 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
383 inode->i_sb->s_id, inode->i_ino, inode,
384 inode->i_mode, inode->i_nlink,
385 NEXT_ORPHAN(inode));
386 }
387 }
388
389 static void ext3_put_super (struct super_block * sb)
390 {
391 struct ext3_sb_info *sbi = EXT3_SB(sb);
392 struct ext3_super_block *es = sbi->s_es;
393 int i;
394
395 ext3_xattr_put_super(sb);
396 journal_destroy(sbi->s_journal);
397 if (!(sb->s_flags & MS_RDONLY)) {
398 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
399 es->s_state = cpu_to_le16(sbi->s_mount_state);
400 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
401 mark_buffer_dirty(sbi->s_sbh);
402 ext3_commit_super(sb, es, 1);
403 }
404
405 for (i = 0; i < sbi->s_gdb_count; i++)
406 brelse(sbi->s_group_desc[i]);
407 kfree(sbi->s_group_desc);
408 percpu_counter_destroy(&sbi->s_freeblocks_counter);
409 percpu_counter_destroy(&sbi->s_freeinodes_counter);
410 percpu_counter_destroy(&sbi->s_dirs_counter);
411 brelse(sbi->s_sbh);
412 #ifdef CONFIG_QUOTA
413 for (i = 0; i < MAXQUOTAS; i++)
414 kfree(sbi->s_qf_names[i]);
415 #endif
416
417 /* Debugging code just in case the in-memory inode orphan list
418 * isn't empty. The on-disk one can be non-empty if we've
419 * detected an error and taken the fs readonly, but the
420 * in-memory list had better be clean by this point. */
421 if (!list_empty(&sbi->s_orphan))
422 dump_orphan_list(sb, sbi);
423 J_ASSERT(list_empty(&sbi->s_orphan));
424
425 invalidate_bdev(sb->s_bdev);
426 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
427 /*
428 * Invalidate the journal device's buffers. We don't want them
429 * floating about in memory - the physical journal device may
430 * hotswapped, and it breaks the `ro-after' testing code.
431 */
432 sync_blockdev(sbi->journal_bdev);
433 invalidate_bdev(sbi->journal_bdev);
434 ext3_blkdev_remove(sbi);
435 }
436 sb->s_fs_info = NULL;
437 kfree(sbi);
438 return;
439 }
440
441 static struct kmem_cache *ext3_inode_cachep;
442
443 /*
444 * Called inside transaction, so use GFP_NOFS
445 */
446 static struct inode *ext3_alloc_inode(struct super_block *sb)
447 {
448 struct ext3_inode_info *ei;
449
450 ei = kmem_cache_alloc(ext3_inode_cachep, GFP_NOFS);
451 if (!ei)
452 return NULL;
453 #ifdef CONFIG_EXT3_FS_POSIX_ACL
454 ei->i_acl = EXT3_ACL_NOT_CACHED;
455 ei->i_default_acl = EXT3_ACL_NOT_CACHED;
456 #endif
457 ei->i_block_alloc_info = NULL;
458 ei->vfs_inode.i_version = 1;
459 return &ei->vfs_inode;
460 }
461
462 static void ext3_destroy_inode(struct inode *inode)
463 {
464 if (!list_empty(&(EXT3_I(inode)->i_orphan))) {
465 printk("EXT3 Inode %p: orphan list check failed!\n",
466 EXT3_I(inode));
467 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
468 EXT3_I(inode), sizeof(struct ext3_inode_info),
469 false);
470 dump_stack();
471 }
472 kmem_cache_free(ext3_inode_cachep, EXT3_I(inode));
473 }
474
475 static void init_once(struct kmem_cache * cachep, void *foo)
476 {
477 struct ext3_inode_info *ei = (struct ext3_inode_info *) foo;
478
479 INIT_LIST_HEAD(&ei->i_orphan);
480 #ifdef CONFIG_EXT3_FS_XATTR
481 init_rwsem(&ei->xattr_sem);
482 #endif
483 mutex_init(&ei->truncate_mutex);
484 inode_init_once(&ei->vfs_inode);
485 }
486
487 static int init_inodecache(void)
488 {
489 ext3_inode_cachep = kmem_cache_create("ext3_inode_cache",
490 sizeof(struct ext3_inode_info),
491 0, (SLAB_RECLAIM_ACCOUNT|
492 SLAB_MEM_SPREAD),
493 init_once);
494 if (ext3_inode_cachep == NULL)
495 return -ENOMEM;
496 return 0;
497 }
498
499 static void destroy_inodecache(void)
500 {
501 kmem_cache_destroy(ext3_inode_cachep);
502 }
503
504 static void ext3_clear_inode(struct inode *inode)
505 {
506 struct ext3_block_alloc_info *rsv = EXT3_I(inode)->i_block_alloc_info;
507 #ifdef CONFIG_EXT3_FS_POSIX_ACL
508 if (EXT3_I(inode)->i_acl &&
509 EXT3_I(inode)->i_acl != EXT3_ACL_NOT_CACHED) {
510 posix_acl_release(EXT3_I(inode)->i_acl);
511 EXT3_I(inode)->i_acl = EXT3_ACL_NOT_CACHED;
512 }
513 if (EXT3_I(inode)->i_default_acl &&
514 EXT3_I(inode)->i_default_acl != EXT3_ACL_NOT_CACHED) {
515 posix_acl_release(EXT3_I(inode)->i_default_acl);
516 EXT3_I(inode)->i_default_acl = EXT3_ACL_NOT_CACHED;
517 }
518 #endif
519 ext3_discard_reservation(inode);
520 EXT3_I(inode)->i_block_alloc_info = NULL;
521 if (unlikely(rsv))
522 kfree(rsv);
523 }
524
525 static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb)
526 {
527 #if defined(CONFIG_QUOTA)
528 struct ext3_sb_info *sbi = EXT3_SB(sb);
529
530 if (sbi->s_jquota_fmt)
531 seq_printf(seq, ",jqfmt=%s",
532 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
533
534 if (sbi->s_qf_names[USRQUOTA])
535 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
536
537 if (sbi->s_qf_names[GRPQUOTA])
538 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
539
540 if (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA)
541 seq_puts(seq, ",usrquota");
542
543 if (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)
544 seq_puts(seq, ",grpquota");
545 #endif
546 }
547
548 /*
549 * Show an option if
550 * - it's set to a non-default value OR
551 * - if the per-sb default is different from the global default
552 */
553 static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs)
554 {
555 struct super_block *sb = vfs->mnt_sb;
556 struct ext3_sb_info *sbi = EXT3_SB(sb);
557 struct ext3_super_block *es = sbi->s_es;
558 unsigned long def_mount_opts;
559
560 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
561
562 if (sbi->s_sb_block != 1)
563 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
564 if (test_opt(sb, MINIX_DF))
565 seq_puts(seq, ",minixdf");
566 if (test_opt(sb, GRPID))
567 seq_puts(seq, ",grpid");
568 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT3_DEFM_BSDGROUPS))
569 seq_puts(seq, ",nogrpid");
570 if (sbi->s_resuid != EXT3_DEF_RESUID ||
571 le16_to_cpu(es->s_def_resuid) != EXT3_DEF_RESUID) {
572 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
573 }
574 if (sbi->s_resgid != EXT3_DEF_RESGID ||
575 le16_to_cpu(es->s_def_resgid) != EXT3_DEF_RESGID) {
576 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
577 }
578 if (test_opt(sb, ERRORS_CONT)) {
579 int def_errors = le16_to_cpu(es->s_errors);
580
581 if (def_errors == EXT3_ERRORS_PANIC ||
582 def_errors == EXT3_ERRORS_RO) {
583 seq_puts(seq, ",errors=continue");
584 }
585 }
586 if (test_opt(sb, ERRORS_RO))
587 seq_puts(seq, ",errors=remount-ro");
588 if (test_opt(sb, ERRORS_PANIC))
589 seq_puts(seq, ",errors=panic");
590 if (test_opt(sb, NO_UID32))
591 seq_puts(seq, ",nouid32");
592 if (test_opt(sb, DEBUG))
593 seq_puts(seq, ",debug");
594 if (test_opt(sb, OLDALLOC))
595 seq_puts(seq, ",oldalloc");
596 #ifdef CONFIG_EXT3_FS_XATTR
597 if (test_opt(sb, XATTR_USER))
598 seq_puts(seq, ",user_xattr");
599 if (!test_opt(sb, XATTR_USER) &&
600 (def_mount_opts & EXT3_DEFM_XATTR_USER)) {
601 seq_puts(seq, ",nouser_xattr");
602 }
603 #endif
604 #ifdef CONFIG_EXT3_FS_POSIX_ACL
605 if (test_opt(sb, POSIX_ACL))
606 seq_puts(seq, ",acl");
607 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT3_DEFM_ACL))
608 seq_puts(seq, ",noacl");
609 #endif
610 if (!test_opt(sb, RESERVATION))
611 seq_puts(seq, ",noreservation");
612 if (sbi->s_commit_interval) {
613 seq_printf(seq, ",commit=%u",
614 (unsigned) (sbi->s_commit_interval / HZ));
615 }
616 if (test_opt(sb, BARRIER))
617 seq_puts(seq, ",barrier=1");
618 if (test_opt(sb, NOBH))
619 seq_puts(seq, ",nobh");
620
621 if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA)
622 seq_puts(seq, ",data=journal");
623 else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA)
624 seq_puts(seq, ",data=ordered");
625 else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)
626 seq_puts(seq, ",data=writeback");
627
628 ext3_show_quota_options(seq, sb);
629
630 return 0;
631 }
632
633
634 static struct dentry *ext3_get_dentry(struct super_block *sb, void *vobjp)
635 {
636 __u32 *objp = vobjp;
637 unsigned long ino = objp[0];
638 __u32 generation = objp[1];
639 struct inode *inode;
640 struct dentry *result;
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 = iget(sb, ino);
656 if (inode == NULL)
657 return ERR_PTR(-ENOMEM);
658 if (is_bad_inode(inode) ||
659 (generation && inode->i_generation != generation)) {
660 iput(inode);
661 return ERR_PTR(-ESTALE);
662 }
663 /* now to find a dentry.
664 * If possible, get a well-connected one
665 */
666 result = d_alloc_anon(inode);
667 if (!result) {
668 iput(inode);
669 return ERR_PTR(-ENOMEM);
670 }
671 return result;
672 }
673
674 #ifdef CONFIG_QUOTA
675 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
676 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
677
678 static int ext3_dquot_initialize(struct inode *inode, int type);
679 static int ext3_dquot_drop(struct inode *inode);
680 static int ext3_write_dquot(struct dquot *dquot);
681 static int ext3_acquire_dquot(struct dquot *dquot);
682 static int ext3_release_dquot(struct dquot *dquot);
683 static int ext3_mark_dquot_dirty(struct dquot *dquot);
684 static int ext3_write_info(struct super_block *sb, int type);
685 static int ext3_quota_on(struct super_block *sb, int type, int format_id, char *path);
686 static int ext3_quota_on_mount(struct super_block *sb, int type);
687 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
688 size_t len, loff_t off);
689 static ssize_t ext3_quota_write(struct super_block *sb, int type,
690 const char *data, size_t len, loff_t off);
691
692 static struct dquot_operations ext3_quota_operations = {
693 .initialize = ext3_dquot_initialize,
694 .drop = ext3_dquot_drop,
695 .alloc_space = dquot_alloc_space,
696 .alloc_inode = dquot_alloc_inode,
697 .free_space = dquot_free_space,
698 .free_inode = dquot_free_inode,
699 .transfer = dquot_transfer,
700 .write_dquot = ext3_write_dquot,
701 .acquire_dquot = ext3_acquire_dquot,
702 .release_dquot = ext3_release_dquot,
703 .mark_dirty = ext3_mark_dquot_dirty,
704 .write_info = ext3_write_info
705 };
706
707 static struct quotactl_ops ext3_qctl_operations = {
708 .quota_on = ext3_quota_on,
709 .quota_off = vfs_quota_off,
710 .quota_sync = vfs_quota_sync,
711 .get_info = vfs_get_dqinfo,
712 .set_info = vfs_set_dqinfo,
713 .get_dqblk = vfs_get_dqblk,
714 .set_dqblk = vfs_set_dqblk
715 };
716 #endif
717
718 static const struct super_operations ext3_sops = {
719 .alloc_inode = ext3_alloc_inode,
720 .destroy_inode = ext3_destroy_inode,
721 .read_inode = ext3_read_inode,
722 .write_inode = ext3_write_inode,
723 .dirty_inode = ext3_dirty_inode,
724 .delete_inode = ext3_delete_inode,
725 .put_super = ext3_put_super,
726 .write_super = ext3_write_super,
727 .sync_fs = ext3_sync_fs,
728 .write_super_lockfs = ext3_write_super_lockfs,
729 .unlockfs = ext3_unlockfs,
730 .statfs = ext3_statfs,
731 .remount_fs = ext3_remount,
732 .clear_inode = ext3_clear_inode,
733 .show_options = ext3_show_options,
734 #ifdef CONFIG_QUOTA
735 .quota_read = ext3_quota_read,
736 .quota_write = ext3_quota_write,
737 #endif
738 };
739
740 static struct export_operations ext3_export_ops = {
741 .get_parent = ext3_get_parent,
742 .get_dentry = ext3_get_dentry,
743 };
744
745 enum {
746 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
747 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
748 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
749 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
750 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
751 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
752 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
753 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
754 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
755 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
756 Opt_grpquota
757 };
758
759 static match_table_t tokens = {
760 {Opt_bsd_df, "bsddf"},
761 {Opt_minix_df, "minixdf"},
762 {Opt_grpid, "grpid"},
763 {Opt_grpid, "bsdgroups"},
764 {Opt_nogrpid, "nogrpid"},
765 {Opt_nogrpid, "sysvgroups"},
766 {Opt_resgid, "resgid=%u"},
767 {Opt_resuid, "resuid=%u"},
768 {Opt_sb, "sb=%u"},
769 {Opt_err_cont, "errors=continue"},
770 {Opt_err_panic, "errors=panic"},
771 {Opt_err_ro, "errors=remount-ro"},
772 {Opt_nouid32, "nouid32"},
773 {Opt_nocheck, "nocheck"},
774 {Opt_nocheck, "check=none"},
775 {Opt_debug, "debug"},
776 {Opt_oldalloc, "oldalloc"},
777 {Opt_orlov, "orlov"},
778 {Opt_user_xattr, "user_xattr"},
779 {Opt_nouser_xattr, "nouser_xattr"},
780 {Opt_acl, "acl"},
781 {Opt_noacl, "noacl"},
782 {Opt_reservation, "reservation"},
783 {Opt_noreservation, "noreservation"},
784 {Opt_noload, "noload"},
785 {Opt_nobh, "nobh"},
786 {Opt_bh, "bh"},
787 {Opt_commit, "commit=%u"},
788 {Opt_journal_update, "journal=update"},
789 {Opt_journal_inum, "journal=%u"},
790 {Opt_journal_dev, "journal_dev=%u"},
791 {Opt_abort, "abort"},
792 {Opt_data_journal, "data=journal"},
793 {Opt_data_ordered, "data=ordered"},
794 {Opt_data_writeback, "data=writeback"},
795 {Opt_offusrjquota, "usrjquota="},
796 {Opt_usrjquota, "usrjquota=%s"},
797 {Opt_offgrpjquota, "grpjquota="},
798 {Opt_grpjquota, "grpjquota=%s"},
799 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
800 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
801 {Opt_grpquota, "grpquota"},
802 {Opt_noquota, "noquota"},
803 {Opt_quota, "quota"},
804 {Opt_usrquota, "usrquota"},
805 {Opt_barrier, "barrier=%u"},
806 {Opt_err, NULL},
807 {Opt_resize, "resize"},
808 };
809
810 static ext3_fsblk_t get_sb_block(void **data)
811 {
812 ext3_fsblk_t sb_block;
813 char *options = (char *) *data;
814
815 if (!options || strncmp(options, "sb=", 3) != 0)
816 return 1; /* Default location */
817 options += 3;
818 /*todo: use simple_strtoll with >32bit ext3 */
819 sb_block = simple_strtoul(options, &options, 0);
820 if (*options && *options != ',') {
821 printk("EXT3-fs: Invalid sb specification: %s\n",
822 (char *) *data);
823 return 1;
824 }
825 if (*options == ',')
826 options++;
827 *data = (void *) options;
828 return sb_block;
829 }
830
831 static int parse_options (char *options, struct super_block *sb,
832 unsigned int *inum, unsigned long *journal_devnum,
833 ext3_fsblk_t *n_blocks_count, int is_remount)
834 {
835 struct ext3_sb_info *sbi = EXT3_SB(sb);
836 char * p;
837 substring_t args[MAX_OPT_ARGS];
838 int data_opt = 0;
839 int option;
840 #ifdef CONFIG_QUOTA
841 int qtype;
842 char *qname;
843 #endif
844
845 if (!options)
846 return 1;
847
848 while ((p = strsep (&options, ",")) != NULL) {
849 int token;
850 if (!*p)
851 continue;
852
853 token = match_token(p, tokens, args);
854 switch (token) {
855 case Opt_bsd_df:
856 clear_opt (sbi->s_mount_opt, MINIX_DF);
857 break;
858 case Opt_minix_df:
859 set_opt (sbi->s_mount_opt, MINIX_DF);
860 break;
861 case Opt_grpid:
862 set_opt (sbi->s_mount_opt, GRPID);
863 break;
864 case Opt_nogrpid:
865 clear_opt (sbi->s_mount_opt, GRPID);
866 break;
867 case Opt_resuid:
868 if (match_int(&args[0], &option))
869 return 0;
870 sbi->s_resuid = option;
871 break;
872 case Opt_resgid:
873 if (match_int(&args[0], &option))
874 return 0;
875 sbi->s_resgid = option;
876 break;
877 case Opt_sb:
878 /* handled by get_sb_block() instead of here */
879 /* *sb_block = match_int(&args[0]); */
880 break;
881 case Opt_err_panic:
882 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
883 clear_opt (sbi->s_mount_opt, ERRORS_RO);
884 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
885 break;
886 case Opt_err_ro:
887 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
888 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
889 set_opt (sbi->s_mount_opt, ERRORS_RO);
890 break;
891 case Opt_err_cont:
892 clear_opt (sbi->s_mount_opt, ERRORS_RO);
893 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
894 set_opt (sbi->s_mount_opt, ERRORS_CONT);
895 break;
896 case Opt_nouid32:
897 set_opt (sbi->s_mount_opt, NO_UID32);
898 break;
899 case Opt_nocheck:
900 clear_opt (sbi->s_mount_opt, CHECK);
901 break;
902 case Opt_debug:
903 set_opt (sbi->s_mount_opt, DEBUG);
904 break;
905 case Opt_oldalloc:
906 set_opt (sbi->s_mount_opt, OLDALLOC);
907 break;
908 case Opt_orlov:
909 clear_opt (sbi->s_mount_opt, OLDALLOC);
910 break;
911 #ifdef CONFIG_EXT3_FS_XATTR
912 case Opt_user_xattr:
913 set_opt (sbi->s_mount_opt, XATTR_USER);
914 break;
915 case Opt_nouser_xattr:
916 clear_opt (sbi->s_mount_opt, XATTR_USER);
917 break;
918 #else
919 case Opt_user_xattr:
920 case Opt_nouser_xattr:
921 printk("EXT3 (no)user_xattr options not supported\n");
922 break;
923 #endif
924 #ifdef CONFIG_EXT3_FS_POSIX_ACL
925 case Opt_acl:
926 set_opt(sbi->s_mount_opt, POSIX_ACL);
927 break;
928 case Opt_noacl:
929 clear_opt(sbi->s_mount_opt, POSIX_ACL);
930 break;
931 #else
932 case Opt_acl:
933 case Opt_noacl:
934 printk("EXT3 (no)acl options not supported\n");
935 break;
936 #endif
937 case Opt_reservation:
938 set_opt(sbi->s_mount_opt, RESERVATION);
939 break;
940 case Opt_noreservation:
941 clear_opt(sbi->s_mount_opt, RESERVATION);
942 break;
943 case Opt_journal_update:
944 /* @@@ FIXME */
945 /* Eventually we will want to be able to create
946 a journal file here. For now, only allow the
947 user to specify an existing inode to be the
948 journal file. */
949 if (is_remount) {
950 printk(KERN_ERR "EXT3-fs: cannot specify "
951 "journal on remount\n");
952 return 0;
953 }
954 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
955 break;
956 case Opt_journal_inum:
957 if (is_remount) {
958 printk(KERN_ERR "EXT3-fs: cannot specify "
959 "journal on remount\n");
960 return 0;
961 }
962 if (match_int(&args[0], &option))
963 return 0;
964 *inum = option;
965 break;
966 case Opt_journal_dev:
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 *journal_devnum = option;
975 break;
976 case Opt_noload:
977 set_opt (sbi->s_mount_opt, NOLOAD);
978 break;
979 case Opt_commit:
980 if (match_int(&args[0], &option))
981 return 0;
982 if (option < 0)
983 return 0;
984 if (option == 0)
985 option = JBD_DEFAULT_MAX_COMMIT_AGE;
986 sbi->s_commit_interval = HZ * option;
987 break;
988 case Opt_data_journal:
989 data_opt = EXT3_MOUNT_JOURNAL_DATA;
990 goto datacheck;
991 case Opt_data_ordered:
992 data_opt = EXT3_MOUNT_ORDERED_DATA;
993 goto datacheck;
994 case Opt_data_writeback:
995 data_opt = EXT3_MOUNT_WRITEBACK_DATA;
996 datacheck:
997 if (is_remount) {
998 if ((sbi->s_mount_opt & EXT3_MOUNT_DATA_FLAGS)
999 != data_opt) {
1000 printk(KERN_ERR
1001 "EXT3-fs: cannot change data "
1002 "mode on remount\n");
1003 return 0;
1004 }
1005 } else {
1006 sbi->s_mount_opt &= ~EXT3_MOUNT_DATA_FLAGS;
1007 sbi->s_mount_opt |= data_opt;
1008 }
1009 break;
1010 #ifdef CONFIG_QUOTA
1011 case Opt_usrjquota:
1012 qtype = USRQUOTA;
1013 goto set_qf_name;
1014 case Opt_grpjquota:
1015 qtype = GRPQUOTA;
1016 set_qf_name:
1017 if (sb_any_quota_enabled(sb)) {
1018 printk(KERN_ERR
1019 "EXT3-fs: Cannot change journalled "
1020 "quota options when quota turned on.\n");
1021 return 0;
1022 }
1023 qname = match_strdup(&args[0]);
1024 if (!qname) {
1025 printk(KERN_ERR
1026 "EXT3-fs: not enough memory for "
1027 "storing quotafile name.\n");
1028 return 0;
1029 }
1030 if (sbi->s_qf_names[qtype] &&
1031 strcmp(sbi->s_qf_names[qtype], qname)) {
1032 printk(KERN_ERR
1033 "EXT3-fs: %s quota file already "
1034 "specified.\n", QTYPE2NAME(qtype));
1035 kfree(qname);
1036 return 0;
1037 }
1038 sbi->s_qf_names[qtype] = qname;
1039 if (strchr(sbi->s_qf_names[qtype], '/')) {
1040 printk(KERN_ERR
1041 "EXT3-fs: quotafile must be on "
1042 "filesystem root.\n");
1043 kfree(sbi->s_qf_names[qtype]);
1044 sbi->s_qf_names[qtype] = NULL;
1045 return 0;
1046 }
1047 set_opt(sbi->s_mount_opt, QUOTA);
1048 break;
1049 case Opt_offusrjquota:
1050 qtype = USRQUOTA;
1051 goto clear_qf_name;
1052 case Opt_offgrpjquota:
1053 qtype = GRPQUOTA;
1054 clear_qf_name:
1055 if (sb_any_quota_enabled(sb)) {
1056 printk(KERN_ERR "EXT3-fs: Cannot change "
1057 "journalled quota options when "
1058 "quota turned on.\n");
1059 return 0;
1060 }
1061 /*
1062 * The space will be released later when all options
1063 * are confirmed to be correct
1064 */
1065 sbi->s_qf_names[qtype] = NULL;
1066 break;
1067 case Opt_jqfmt_vfsold:
1068 sbi->s_jquota_fmt = QFMT_VFS_OLD;
1069 break;
1070 case Opt_jqfmt_vfsv0:
1071 sbi->s_jquota_fmt = QFMT_VFS_V0;
1072 break;
1073 case Opt_quota:
1074 case Opt_usrquota:
1075 set_opt(sbi->s_mount_opt, QUOTA);
1076 set_opt(sbi->s_mount_opt, USRQUOTA);
1077 break;
1078 case Opt_grpquota:
1079 set_opt(sbi->s_mount_opt, QUOTA);
1080 set_opt(sbi->s_mount_opt, GRPQUOTA);
1081 break;
1082 case Opt_noquota:
1083 if (sb_any_quota_enabled(sb)) {
1084 printk(KERN_ERR "EXT3-fs: Cannot change quota "
1085 "options when quota turned on.\n");
1086 return 0;
1087 }
1088 clear_opt(sbi->s_mount_opt, QUOTA);
1089 clear_opt(sbi->s_mount_opt, USRQUOTA);
1090 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1091 break;
1092 #else
1093 case Opt_quota:
1094 case Opt_usrquota:
1095 case Opt_grpquota:
1096 case Opt_usrjquota:
1097 case Opt_grpjquota:
1098 case Opt_offusrjquota:
1099 case Opt_offgrpjquota:
1100 case Opt_jqfmt_vfsold:
1101 case Opt_jqfmt_vfsv0:
1102 printk(KERN_ERR
1103 "EXT3-fs: journalled quota options not "
1104 "supported.\n");
1105 break;
1106 case Opt_noquota:
1107 break;
1108 #endif
1109 case Opt_abort:
1110 set_opt(sbi->s_mount_opt, ABORT);
1111 break;
1112 case Opt_barrier:
1113 if (match_int(&args[0], &option))
1114 return 0;
1115 if (option)
1116 set_opt(sbi->s_mount_opt, BARRIER);
1117 else
1118 clear_opt(sbi->s_mount_opt, BARRIER);
1119 break;
1120 case Opt_ignore:
1121 break;
1122 case Opt_resize:
1123 if (!is_remount) {
1124 printk("EXT3-fs: resize option only available "
1125 "for remount\n");
1126 return 0;
1127 }
1128 if (match_int(&args[0], &option) != 0)
1129 return 0;
1130 *n_blocks_count = option;
1131 break;
1132 case Opt_nobh:
1133 set_opt(sbi->s_mount_opt, NOBH);
1134 break;
1135 case Opt_bh:
1136 clear_opt(sbi->s_mount_opt, NOBH);
1137 break;
1138 default:
1139 printk (KERN_ERR
1140 "EXT3-fs: Unrecognized mount option \"%s\" "
1141 "or missing value\n", p);
1142 return 0;
1143 }
1144 }
1145 #ifdef CONFIG_QUOTA
1146 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1147 if ((sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA) &&
1148 sbi->s_qf_names[USRQUOTA])
1149 clear_opt(sbi->s_mount_opt, USRQUOTA);
1150
1151 if ((sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA) &&
1152 sbi->s_qf_names[GRPQUOTA])
1153 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1154
1155 if ((sbi->s_qf_names[USRQUOTA] &&
1156 (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)) ||
1157 (sbi->s_qf_names[GRPQUOTA] &&
1158 (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA))) {
1159 printk(KERN_ERR "EXT3-fs: old and new quota "
1160 "format mixing.\n");
1161 return 0;
1162 }
1163
1164 if (!sbi->s_jquota_fmt) {
1165 printk(KERN_ERR "EXT3-fs: journalled quota format "
1166 "not specified.\n");
1167 return 0;
1168 }
1169 } else {
1170 if (sbi->s_jquota_fmt) {
1171 printk(KERN_ERR "EXT3-fs: journalled quota format "
1172 "specified with no journalling "
1173 "enabled.\n");
1174 return 0;
1175 }
1176 }
1177 #endif
1178 return 1;
1179 }
1180
1181 static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
1182 int read_only)
1183 {
1184 struct ext3_sb_info *sbi = EXT3_SB(sb);
1185 int res = 0;
1186
1187 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
1188 printk (KERN_ERR "EXT3-fs warning: revision level too high, "
1189 "forcing read-only mode\n");
1190 res = MS_RDONLY;
1191 }
1192 if (read_only)
1193 return res;
1194 if (!(sbi->s_mount_state & EXT3_VALID_FS))
1195 printk (KERN_WARNING "EXT3-fs warning: mounting unchecked fs, "
1196 "running e2fsck is recommended\n");
1197 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
1198 printk (KERN_WARNING
1199 "EXT3-fs warning: mounting fs with errors, "
1200 "running e2fsck is recommended\n");
1201 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1202 le16_to_cpu(es->s_mnt_count) >=
1203 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1204 printk (KERN_WARNING
1205 "EXT3-fs warning: maximal mount count reached, "
1206 "running e2fsck is recommended\n");
1207 else if (le32_to_cpu(es->s_checkinterval) &&
1208 (le32_to_cpu(es->s_lastcheck) +
1209 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1210 printk (KERN_WARNING
1211 "EXT3-fs warning: checktime reached, "
1212 "running e2fsck is recommended\n");
1213 #if 0
1214 /* @@@ We _will_ want to clear the valid bit if we find
1215 inconsistencies, to force a fsck at reboot. But for
1216 a plain journaled filesystem we can keep it set as
1217 valid forever! :) */
1218 es->s_state = cpu_to_le16(le16_to_cpu(es->s_state) & ~EXT3_VALID_FS);
1219 #endif
1220 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1221 es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
1222 es->s_mnt_count=cpu_to_le16(le16_to_cpu(es->s_mnt_count) + 1);
1223 es->s_mtime = cpu_to_le32(get_seconds());
1224 ext3_update_dynamic_rev(sb);
1225 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1226
1227 ext3_commit_super(sb, es, 1);
1228 if (test_opt(sb, DEBUG))
1229 printk(KERN_INFO "[EXT3 FS bs=%lu, gc=%lu, "
1230 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1231 sb->s_blocksize,
1232 sbi->s_groups_count,
1233 EXT3_BLOCKS_PER_GROUP(sb),
1234 EXT3_INODES_PER_GROUP(sb),
1235 sbi->s_mount_opt);
1236
1237 printk(KERN_INFO "EXT3 FS on %s, ", sb->s_id);
1238 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
1239 char b[BDEVNAME_SIZE];
1240
1241 printk("external journal on %s\n",
1242 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1243 } else {
1244 printk("internal journal\n");
1245 }
1246 return res;
1247 }
1248
1249 /* Called at mount-time, super-block is locked */
1250 static int ext3_check_descriptors (struct super_block * sb)
1251 {
1252 struct ext3_sb_info *sbi = EXT3_SB(sb);
1253 ext3_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1254 ext3_fsblk_t last_block;
1255 struct ext3_group_desc * gdp = NULL;
1256 int desc_block = 0;
1257 int i;
1258
1259 ext3_debug ("Checking group descriptors");
1260
1261 for (i = 0; i < sbi->s_groups_count; i++)
1262 {
1263 if (i == sbi->s_groups_count - 1)
1264 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
1265 else
1266 last_block = first_block +
1267 (EXT3_BLOCKS_PER_GROUP(sb) - 1);
1268
1269 if ((i % EXT3_DESC_PER_BLOCK(sb)) == 0)
1270 gdp = (struct ext3_group_desc *)
1271 sbi->s_group_desc[desc_block++]->b_data;
1272 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
1273 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
1274 {
1275 ext3_error (sb, "ext3_check_descriptors",
1276 "Block bitmap for group %d"
1277 " not in group (block %lu)!",
1278 i, (unsigned long)
1279 le32_to_cpu(gdp->bg_block_bitmap));
1280 return 0;
1281 }
1282 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
1283 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
1284 {
1285 ext3_error (sb, "ext3_check_descriptors",
1286 "Inode bitmap for group %d"
1287 " not in group (block %lu)!",
1288 i, (unsigned long)
1289 le32_to_cpu(gdp->bg_inode_bitmap));
1290 return 0;
1291 }
1292 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
1293 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
1294 last_block)
1295 {
1296 ext3_error (sb, "ext3_check_descriptors",
1297 "Inode table for group %d"
1298 " not in group (block %lu)!",
1299 i, (unsigned long)
1300 le32_to_cpu(gdp->bg_inode_table));
1301 return 0;
1302 }
1303 first_block += EXT3_BLOCKS_PER_GROUP(sb);
1304 gdp++;
1305 }
1306
1307 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1308 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1309 return 1;
1310 }
1311
1312
1313 /* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1314 * the superblock) which were deleted from all directories, but held open by
1315 * a process at the time of a crash. We walk the list and try to delete these
1316 * inodes at recovery time (only with a read-write filesystem).
1317 *
1318 * In order to keep the orphan inode chain consistent during traversal (in
1319 * case of crash during recovery), we link each inode into the superblock
1320 * orphan list_head and handle it the same way as an inode deletion during
1321 * normal operation (which journals the operations for us).
1322 *
1323 * We only do an iget() and an iput() on each inode, which is very safe if we
1324 * accidentally point at an in-use or already deleted inode. The worst that
1325 * can happen in this case is that we get a "bit already cleared" message from
1326 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1327 * e2fsck was run on this filesystem, and it must have already done the orphan
1328 * inode cleanup for us, so we can safely abort without any further action.
1329 */
1330 static void ext3_orphan_cleanup (struct super_block * sb,
1331 struct ext3_super_block * es)
1332 {
1333 unsigned int s_flags = sb->s_flags;
1334 int nr_orphans = 0, nr_truncates = 0;
1335 #ifdef CONFIG_QUOTA
1336 int i;
1337 #endif
1338 if (!es->s_last_orphan) {
1339 jbd_debug(4, "no orphan inodes to clean up\n");
1340 return;
1341 }
1342
1343 if (bdev_read_only(sb->s_bdev)) {
1344 printk(KERN_ERR "EXT3-fs: write access "
1345 "unavailable, skipping orphan cleanup.\n");
1346 return;
1347 }
1348
1349 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1350 if (es->s_last_orphan)
1351 jbd_debug(1, "Errors on filesystem, "
1352 "clearing orphan list.\n");
1353 es->s_last_orphan = 0;
1354 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1355 return;
1356 }
1357
1358 if (s_flags & MS_RDONLY) {
1359 printk(KERN_INFO "EXT3-fs: %s: orphan cleanup on readonly fs\n",
1360 sb->s_id);
1361 sb->s_flags &= ~MS_RDONLY;
1362 }
1363 #ifdef CONFIG_QUOTA
1364 /* Needed for iput() to work correctly and not trash data */
1365 sb->s_flags |= MS_ACTIVE;
1366 /* Turn on quotas so that they are updated correctly */
1367 for (i = 0; i < MAXQUOTAS; i++) {
1368 if (EXT3_SB(sb)->s_qf_names[i]) {
1369 int ret = ext3_quota_on_mount(sb, i);
1370 if (ret < 0)
1371 printk(KERN_ERR
1372 "EXT3-fs: Cannot turn on journalled "
1373 "quota: error %d\n", ret);
1374 }
1375 }
1376 #endif
1377
1378 while (es->s_last_orphan) {
1379 struct inode *inode;
1380
1381 if (!(inode =
1382 ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan)))) {
1383 es->s_last_orphan = 0;
1384 break;
1385 }
1386
1387 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
1388 DQUOT_INIT(inode);
1389 if (inode->i_nlink) {
1390 printk(KERN_DEBUG
1391 "%s: truncating inode %lu to %Ld bytes\n",
1392 __FUNCTION__, inode->i_ino, inode->i_size);
1393 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1394 inode->i_ino, inode->i_size);
1395 ext3_truncate(inode);
1396 nr_truncates++;
1397 } else {
1398 printk(KERN_DEBUG
1399 "%s: deleting unreferenced inode %lu\n",
1400 __FUNCTION__, inode->i_ino);
1401 jbd_debug(2, "deleting unreferenced inode %lu\n",
1402 inode->i_ino);
1403 nr_orphans++;
1404 }
1405 iput(inode); /* The delete magic happens here! */
1406 }
1407
1408 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1409
1410 if (nr_orphans)
1411 printk(KERN_INFO "EXT3-fs: %s: %d orphan inode%s deleted\n",
1412 sb->s_id, PLURAL(nr_orphans));
1413 if (nr_truncates)
1414 printk(KERN_INFO "EXT3-fs: %s: %d truncate%s cleaned up\n",
1415 sb->s_id, PLURAL(nr_truncates));
1416 #ifdef CONFIG_QUOTA
1417 /* Turn quotas off */
1418 for (i = 0; i < MAXQUOTAS; i++) {
1419 if (sb_dqopt(sb)->files[i])
1420 vfs_quota_off(sb, i);
1421 }
1422 #endif
1423 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1424 }
1425
1426 /*
1427 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1428 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1429 * We need to be 1 filesystem block less than the 2^32 sector limit.
1430 */
1431 static loff_t ext3_max_size(int bits)
1432 {
1433 loff_t res = EXT3_NDIR_BLOCKS;
1434 /* This constant is calculated to be the largest file size for a
1435 * dense, 4k-blocksize file such that the total number of
1436 * sectors in the file, including data and all indirect blocks,
1437 * does not exceed 2^32. */
1438 const loff_t upper_limit = 0x1ff7fffd000LL;
1439
1440 res += 1LL << (bits-2);
1441 res += 1LL << (2*(bits-2));
1442 res += 1LL << (3*(bits-2));
1443 res <<= bits;
1444 if (res > upper_limit)
1445 res = upper_limit;
1446 return res;
1447 }
1448
1449 static ext3_fsblk_t descriptor_loc(struct super_block *sb,
1450 ext3_fsblk_t logic_sb_block,
1451 int nr)
1452 {
1453 struct ext3_sb_info *sbi = EXT3_SB(sb);
1454 unsigned long bg, first_meta_bg;
1455 int has_super = 0;
1456
1457 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1458
1459 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1460 nr < first_meta_bg)
1461 return (logic_sb_block + nr + 1);
1462 bg = sbi->s_desc_per_block * nr;
1463 if (ext3_bg_has_super(sb, bg))
1464 has_super = 1;
1465 return (has_super + ext3_group_first_block_no(sb, bg));
1466 }
1467
1468
1469 static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1470 {
1471 struct buffer_head * bh;
1472 struct ext3_super_block *es = NULL;
1473 struct ext3_sb_info *sbi;
1474 ext3_fsblk_t block;
1475 ext3_fsblk_t sb_block = get_sb_block(&data);
1476 ext3_fsblk_t logic_sb_block;
1477 unsigned long offset = 0;
1478 unsigned int journal_inum = 0;
1479 unsigned long journal_devnum = 0;
1480 unsigned long def_mount_opts;
1481 struct inode *root;
1482 int blocksize;
1483 int hblock;
1484 int db_count;
1485 int i;
1486 int needs_recovery;
1487 __le32 features;
1488 int err;
1489
1490 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1491 if (!sbi)
1492 return -ENOMEM;
1493 sb->s_fs_info = sbi;
1494 sbi->s_mount_opt = 0;
1495 sbi->s_resuid = EXT3_DEF_RESUID;
1496 sbi->s_resgid = EXT3_DEF_RESGID;
1497 sbi->s_sb_block = sb_block;
1498
1499 unlock_kernel();
1500
1501 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1502 if (!blocksize) {
1503 printk(KERN_ERR "EXT3-fs: unable to set blocksize\n");
1504 goto out_fail;
1505 }
1506
1507 /*
1508 * The ext3 superblock will not be buffer aligned for other than 1kB
1509 * block sizes. We need to calculate the offset from buffer start.
1510 */
1511 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1512 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1513 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1514 } else {
1515 logic_sb_block = sb_block;
1516 }
1517
1518 if (!(bh = sb_bread(sb, logic_sb_block))) {
1519 printk (KERN_ERR "EXT3-fs: unable to read superblock\n");
1520 goto out_fail;
1521 }
1522 /*
1523 * Note: s_es must be initialized as soon as possible because
1524 * some ext3 macro-instructions depend on its value
1525 */
1526 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1527 sbi->s_es = es;
1528 sb->s_magic = le16_to_cpu(es->s_magic);
1529 if (sb->s_magic != EXT3_SUPER_MAGIC)
1530 goto cantfind_ext3;
1531
1532 /* Set defaults before we parse the mount options */
1533 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1534 if (def_mount_opts & EXT3_DEFM_DEBUG)
1535 set_opt(sbi->s_mount_opt, DEBUG);
1536 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1537 set_opt(sbi->s_mount_opt, GRPID);
1538 if (def_mount_opts & EXT3_DEFM_UID16)
1539 set_opt(sbi->s_mount_opt, NO_UID32);
1540 #ifdef CONFIG_EXT3_FS_XATTR
1541 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1542 set_opt(sbi->s_mount_opt, XATTR_USER);
1543 #endif
1544 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1545 if (def_mount_opts & EXT3_DEFM_ACL)
1546 set_opt(sbi->s_mount_opt, POSIX_ACL);
1547 #endif
1548 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
1549 sbi->s_mount_opt |= EXT3_MOUNT_JOURNAL_DATA;
1550 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
1551 sbi->s_mount_opt |= EXT3_MOUNT_ORDERED_DATA;
1552 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
1553 sbi->s_mount_opt |= EXT3_MOUNT_WRITEBACK_DATA;
1554
1555 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1556 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1557 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_RO)
1558 set_opt(sbi->s_mount_opt, ERRORS_RO);
1559 else
1560 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1561
1562 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1563 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1564
1565 set_opt(sbi->s_mount_opt, RESERVATION);
1566
1567 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1568 NULL, 0))
1569 goto failed_mount;
1570
1571 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1572 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1573
1574 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1575 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1576 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1577 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1578 printk(KERN_WARNING
1579 "EXT3-fs warning: feature flags set on rev 0 fs, "
1580 "running e2fsck is recommended\n");
1581 /*
1582 * Check feature flags regardless of the revision level, since we
1583 * previously didn't change the revision level when setting the flags,
1584 * so there is a chance incompat flags are set on a rev 0 filesystem.
1585 */
1586 features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
1587 if (features) {
1588 printk(KERN_ERR "EXT3-fs: %s: couldn't mount because of "
1589 "unsupported optional features (%x).\n",
1590 sb->s_id, le32_to_cpu(features));
1591 goto failed_mount;
1592 }
1593 features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
1594 if (!(sb->s_flags & MS_RDONLY) && features) {
1595 printk(KERN_ERR "EXT3-fs: %s: couldn't mount RDWR because of "
1596 "unsupported optional features (%x).\n",
1597 sb->s_id, le32_to_cpu(features));
1598 goto failed_mount;
1599 }
1600 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1601
1602 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1603 blocksize > EXT3_MAX_BLOCK_SIZE) {
1604 printk(KERN_ERR
1605 "EXT3-fs: Unsupported filesystem blocksize %d on %s.\n",
1606 blocksize, sb->s_id);
1607 goto failed_mount;
1608 }
1609
1610 hblock = bdev_hardsect_size(sb->s_bdev);
1611 if (sb->s_blocksize != blocksize) {
1612 /*
1613 * Make sure the blocksize for the filesystem is larger
1614 * than the hardware sectorsize for the machine.
1615 */
1616 if (blocksize < hblock) {
1617 printk(KERN_ERR "EXT3-fs: blocksize %d too small for "
1618 "device blocksize %d.\n", blocksize, hblock);
1619 goto failed_mount;
1620 }
1621
1622 brelse (bh);
1623 sb_set_blocksize(sb, blocksize);
1624 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1625 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1626 bh = sb_bread(sb, logic_sb_block);
1627 if (!bh) {
1628 printk(KERN_ERR
1629 "EXT3-fs: Can't read superblock on 2nd try.\n");
1630 goto failed_mount;
1631 }
1632 es = (struct ext3_super_block *)(((char *)bh->b_data) + offset);
1633 sbi->s_es = es;
1634 if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
1635 printk (KERN_ERR
1636 "EXT3-fs: Magic mismatch, very weird !\n");
1637 goto failed_mount;
1638 }
1639 }
1640
1641 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1642
1643 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1644 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1645 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1646 } else {
1647 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1648 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1649 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
1650 (!is_power_of_2(sbi->s_inode_size)) ||
1651 (sbi->s_inode_size > blocksize)) {
1652 printk (KERN_ERR
1653 "EXT3-fs: unsupported inode size: %d\n",
1654 sbi->s_inode_size);
1655 goto failed_mount;
1656 }
1657 }
1658 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1659 le32_to_cpu(es->s_log_frag_size);
1660 if (blocksize != sbi->s_frag_size) {
1661 printk(KERN_ERR
1662 "EXT3-fs: fragsize %lu != blocksize %u (unsupported)\n",
1663 sbi->s_frag_size, blocksize);
1664 goto failed_mount;
1665 }
1666 sbi->s_frags_per_block = 1;
1667 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1668 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1669 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1670 if (EXT3_INODE_SIZE(sb) == 0)
1671 goto cantfind_ext3;
1672 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1673 if (sbi->s_inodes_per_block == 0)
1674 goto cantfind_ext3;
1675 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1676 sbi->s_inodes_per_block;
1677 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1678 sbi->s_sbh = bh;
1679 sbi->s_mount_state = le16_to_cpu(es->s_state);
1680 sbi->s_addr_per_block_bits = ilog2(EXT3_ADDR_PER_BLOCK(sb));
1681 sbi->s_desc_per_block_bits = ilog2(EXT3_DESC_PER_BLOCK(sb));
1682 for (i=0; i < 4; i++)
1683 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1684 sbi->s_def_hash_version = es->s_def_hash_version;
1685
1686 if (sbi->s_blocks_per_group > blocksize * 8) {
1687 printk (KERN_ERR
1688 "EXT3-fs: #blocks per group too big: %lu\n",
1689 sbi->s_blocks_per_group);
1690 goto failed_mount;
1691 }
1692 if (sbi->s_frags_per_group > blocksize * 8) {
1693 printk (KERN_ERR
1694 "EXT3-fs: #fragments per group too big: %lu\n",
1695 sbi->s_frags_per_group);
1696 goto failed_mount;
1697 }
1698 if (sbi->s_inodes_per_group > blocksize * 8) {
1699 printk (KERN_ERR
1700 "EXT3-fs: #inodes per group too big: %lu\n",
1701 sbi->s_inodes_per_group);
1702 goto failed_mount;
1703 }
1704
1705 if (le32_to_cpu(es->s_blocks_count) >
1706 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
1707 printk(KERN_ERR "EXT3-fs: filesystem on %s:"
1708 " too large to mount safely\n", sb->s_id);
1709 if (sizeof(sector_t) < 8)
1710 printk(KERN_WARNING "EXT3-fs: CONFIG_LBD not "
1711 "enabled\n");
1712 goto failed_mount;
1713 }
1714
1715 if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1716 goto cantfind_ext3;
1717 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1718 le32_to_cpu(es->s_first_data_block) - 1)
1719 / EXT3_BLOCKS_PER_GROUP(sb)) + 1;
1720 db_count = (sbi->s_groups_count + EXT3_DESC_PER_BLOCK(sb) - 1) /
1721 EXT3_DESC_PER_BLOCK(sb);
1722 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1723 GFP_KERNEL);
1724 if (sbi->s_group_desc == NULL) {
1725 printk (KERN_ERR "EXT3-fs: not enough memory\n");
1726 goto failed_mount;
1727 }
1728
1729 bgl_lock_init(&sbi->s_blockgroup_lock);
1730
1731 for (i = 0; i < db_count; i++) {
1732 block = descriptor_loc(sb, logic_sb_block, i);
1733 sbi->s_group_desc[i] = sb_bread(sb, block);
1734 if (!sbi->s_group_desc[i]) {
1735 printk (KERN_ERR "EXT3-fs: "
1736 "can't read group descriptor %d\n", i);
1737 db_count = i;
1738 goto failed_mount2;
1739 }
1740 }
1741 if (!ext3_check_descriptors (sb)) {
1742 printk(KERN_ERR "EXT3-fs: group descriptors corrupted!\n");
1743 goto failed_mount2;
1744 }
1745 sbi->s_gdb_count = db_count;
1746 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1747 spin_lock_init(&sbi->s_next_gen_lock);
1748
1749 err = percpu_counter_init(&sbi->s_freeblocks_counter,
1750 ext3_count_free_blocks(sb));
1751 if (!err) {
1752 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1753 ext3_count_free_inodes(sb));
1754 }
1755 if (!err) {
1756 err = percpu_counter_init(&sbi->s_dirs_counter,
1757 ext3_count_dirs(sb));
1758 }
1759 if (err) {
1760 printk(KERN_ERR "EXT3-fs: insufficient memory\n");
1761 goto failed_mount3;
1762 }
1763
1764 /* per fileystem reservation list head & lock */
1765 spin_lock_init(&sbi->s_rsv_window_lock);
1766 sbi->s_rsv_window_root = RB_ROOT;
1767 /* Add a single, static dummy reservation to the start of the
1768 * reservation window list --- it gives us a placeholder for
1769 * append-at-start-of-list which makes the allocation logic
1770 * _much_ simpler. */
1771 sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1772 sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1773 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1774 sbi->s_rsv_window_head.rsv_goal_size = 0;
1775 ext3_rsv_window_add(sb, &sbi->s_rsv_window_head);
1776
1777 /*
1778 * set up enough so that it can read an inode
1779 */
1780 sb->s_op = &ext3_sops;
1781 sb->s_export_op = &ext3_export_ops;
1782 sb->s_xattr = ext3_xattr_handlers;
1783 #ifdef CONFIG_QUOTA
1784 sb->s_qcop = &ext3_qctl_operations;
1785 sb->dq_op = &ext3_quota_operations;
1786 #endif
1787 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1788
1789 sb->s_root = NULL;
1790
1791 needs_recovery = (es->s_last_orphan != 0 ||
1792 EXT3_HAS_INCOMPAT_FEATURE(sb,
1793 EXT3_FEATURE_INCOMPAT_RECOVER));
1794
1795 /*
1796 * The first inode we look at is the journal inode. Don't try
1797 * root first: it may be modified in the journal!
1798 */
1799 if (!test_opt(sb, NOLOAD) &&
1800 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
1801 if (ext3_load_journal(sb, es, journal_devnum))
1802 goto failed_mount3;
1803 } else if (journal_inum) {
1804 if (ext3_create_journal(sb, es, journal_inum))
1805 goto failed_mount3;
1806 } else {
1807 if (!silent)
1808 printk (KERN_ERR
1809 "ext3: No journal on filesystem on %s\n",
1810 sb->s_id);
1811 goto failed_mount3;
1812 }
1813
1814 /* We have now updated the journal if required, so we can
1815 * validate the data journaling mode. */
1816 switch (test_opt(sb, DATA_FLAGS)) {
1817 case 0:
1818 /* No mode set, assume a default based on the journal
1819 capabilities: ORDERED_DATA if the journal can
1820 cope, else JOURNAL_DATA */
1821 if (journal_check_available_features
1822 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
1823 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1824 else
1825 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1826 break;
1827
1828 case EXT3_MOUNT_ORDERED_DATA:
1829 case EXT3_MOUNT_WRITEBACK_DATA:
1830 if (!journal_check_available_features
1831 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
1832 printk(KERN_ERR "EXT3-fs: Journal does not support "
1833 "requested data journaling mode\n");
1834 goto failed_mount4;
1835 }
1836 default:
1837 break;
1838 }
1839
1840 if (test_opt(sb, NOBH)) {
1841 if (!(test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)) {
1842 printk(KERN_WARNING "EXT3-fs: Ignoring nobh option - "
1843 "its supported only with writeback mode\n");
1844 clear_opt(sbi->s_mount_opt, NOBH);
1845 }
1846 }
1847 /*
1848 * The journal_load will have done any necessary log recovery,
1849 * so we can safely mount the rest of the filesystem now.
1850 */
1851
1852 root = iget(sb, EXT3_ROOT_INO);
1853 sb->s_root = d_alloc_root(root);
1854 if (!sb->s_root) {
1855 printk(KERN_ERR "EXT3-fs: get root inode failed\n");
1856 iput(root);
1857 goto failed_mount4;
1858 }
1859 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1860 dput(sb->s_root);
1861 sb->s_root = NULL;
1862 printk(KERN_ERR "EXT3-fs: corrupt root inode, run e2fsck\n");
1863 goto failed_mount4;
1864 }
1865
1866 ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1867 /*
1868 * akpm: core read_super() calls in here with the superblock locked.
1869 * That deadlocks, because orphan cleanup needs to lock the superblock
1870 * in numerous places. Here we just pop the lock - it's relatively
1871 * harmless, because we are now ready to accept write_super() requests,
1872 * and aviro says that's the only reason for hanging onto the
1873 * superblock lock.
1874 */
1875 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
1876 ext3_orphan_cleanup(sb, es);
1877 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
1878 if (needs_recovery)
1879 printk (KERN_INFO "EXT3-fs: recovery complete.\n");
1880 ext3_mark_recovery_complete(sb, es);
1881 printk (KERN_INFO "EXT3-fs: mounted filesystem with %s data mode.\n",
1882 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
1883 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
1884 "writeback");
1885
1886 lock_kernel();
1887 return 0;
1888
1889 cantfind_ext3:
1890 if (!silent)
1891 printk(KERN_ERR "VFS: Can't find ext3 filesystem on dev %s.\n",
1892 sb->s_id);
1893 goto failed_mount;
1894
1895 failed_mount4:
1896 journal_destroy(sbi->s_journal);
1897 failed_mount3:
1898 percpu_counter_destroy(&sbi->s_freeblocks_counter);
1899 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1900 percpu_counter_destroy(&sbi->s_dirs_counter);
1901 failed_mount2:
1902 for (i = 0; i < db_count; i++)
1903 brelse(sbi->s_group_desc[i]);
1904 kfree(sbi->s_group_desc);
1905 failed_mount:
1906 #ifdef CONFIG_QUOTA
1907 for (i = 0; i < MAXQUOTAS; i++)
1908 kfree(sbi->s_qf_names[i]);
1909 #endif
1910 ext3_blkdev_remove(sbi);
1911 brelse(bh);
1912 out_fail:
1913 sb->s_fs_info = NULL;
1914 kfree(sbi);
1915 lock_kernel();
1916 return -EINVAL;
1917 }
1918
1919 /*
1920 * Setup any per-fs journal parameters now. We'll do this both on
1921 * initial mount, once the journal has been initialised but before we've
1922 * done any recovery; and again on any subsequent remount.
1923 */
1924 static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
1925 {
1926 struct ext3_sb_info *sbi = EXT3_SB(sb);
1927
1928 if (sbi->s_commit_interval)
1929 journal->j_commit_interval = sbi->s_commit_interval;
1930 /* We could also set up an ext3-specific default for the commit
1931 * interval here, but for now we'll just fall back to the jbd
1932 * default. */
1933
1934 spin_lock(&journal->j_state_lock);
1935 if (test_opt(sb, BARRIER))
1936 journal->j_flags |= JFS_BARRIER;
1937 else
1938 journal->j_flags &= ~JFS_BARRIER;
1939 spin_unlock(&journal->j_state_lock);
1940 }
1941
1942 static journal_t *ext3_get_journal(struct super_block *sb,
1943 unsigned int journal_inum)
1944 {
1945 struct inode *journal_inode;
1946 journal_t *journal;
1947
1948 /* First, test for the existence of a valid inode on disk. Bad
1949 * things happen if we iget() an unused inode, as the subsequent
1950 * iput() will try to delete it. */
1951
1952 journal_inode = iget(sb, journal_inum);
1953 if (!journal_inode) {
1954 printk(KERN_ERR "EXT3-fs: no journal found.\n");
1955 return NULL;
1956 }
1957 if (!journal_inode->i_nlink) {
1958 make_bad_inode(journal_inode);
1959 iput(journal_inode);
1960 printk(KERN_ERR "EXT3-fs: journal inode is deleted.\n");
1961 return NULL;
1962 }
1963
1964 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
1965 journal_inode, journal_inode->i_size);
1966 if (is_bad_inode(journal_inode) || !S_ISREG(journal_inode->i_mode)) {
1967 printk(KERN_ERR "EXT3-fs: invalid journal inode.\n");
1968 iput(journal_inode);
1969 return NULL;
1970 }
1971
1972 journal = journal_init_inode(journal_inode);
1973 if (!journal) {
1974 printk(KERN_ERR "EXT3-fs: Could not load journal inode\n");
1975 iput(journal_inode);
1976 return NULL;
1977 }
1978 journal->j_private = sb;
1979 ext3_init_journal_params(sb, journal);
1980 return journal;
1981 }
1982
1983 static journal_t *ext3_get_dev_journal(struct super_block *sb,
1984 dev_t j_dev)
1985 {
1986 struct buffer_head * bh;
1987 journal_t *journal;
1988 ext3_fsblk_t start;
1989 ext3_fsblk_t len;
1990 int hblock, blocksize;
1991 ext3_fsblk_t sb_block;
1992 unsigned long offset;
1993 struct ext3_super_block * es;
1994 struct block_device *bdev;
1995
1996 bdev = ext3_blkdev_get(j_dev);
1997 if (bdev == NULL)
1998 return NULL;
1999
2000 if (bd_claim(bdev, sb)) {
2001 printk(KERN_ERR
2002 "EXT3: failed to claim external journal device.\n");
2003 blkdev_put(bdev);
2004 return NULL;
2005 }
2006
2007 blocksize = sb->s_blocksize;
2008 hblock = bdev_hardsect_size(bdev);
2009 if (blocksize < hblock) {
2010 printk(KERN_ERR
2011 "EXT3-fs: blocksize too small for journal device.\n");
2012 goto out_bdev;
2013 }
2014
2015 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
2016 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
2017 set_blocksize(bdev, blocksize);
2018 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2019 printk(KERN_ERR "EXT3-fs: couldn't read superblock of "
2020 "external journal\n");
2021 goto out_bdev;
2022 }
2023
2024 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
2025 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
2026 !(le32_to_cpu(es->s_feature_incompat) &
2027 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2028 printk(KERN_ERR "EXT3-fs: external journal has "
2029 "bad superblock\n");
2030 brelse(bh);
2031 goto out_bdev;
2032 }
2033
2034 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2035 printk(KERN_ERR "EXT3-fs: journal UUID does not match\n");
2036 brelse(bh);
2037 goto out_bdev;
2038 }
2039
2040 len = le32_to_cpu(es->s_blocks_count);
2041 start = sb_block + 1;
2042 brelse(bh); /* we're done with the superblock */
2043
2044 journal = journal_init_dev(bdev, sb->s_bdev,
2045 start, len, blocksize);
2046 if (!journal) {
2047 printk(KERN_ERR "EXT3-fs: failed to create device journal\n");
2048 goto out_bdev;
2049 }
2050 journal->j_private = sb;
2051 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2052 wait_on_buffer(journal->j_sb_buffer);
2053 if (!buffer_uptodate(journal->j_sb_buffer)) {
2054 printk(KERN_ERR "EXT3-fs: I/O error on journal device\n");
2055 goto out_journal;
2056 }
2057 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2058 printk(KERN_ERR "EXT3-fs: External journal has more than one "
2059 "user (unsupported) - %d\n",
2060 be32_to_cpu(journal->j_superblock->s_nr_users));
2061 goto out_journal;
2062 }
2063 EXT3_SB(sb)->journal_bdev = bdev;
2064 ext3_init_journal_params(sb, journal);
2065 return journal;
2066 out_journal:
2067 journal_destroy(journal);
2068 out_bdev:
2069 ext3_blkdev_put(bdev);
2070 return NULL;
2071 }
2072
2073 static int ext3_load_journal(struct super_block *sb,
2074 struct ext3_super_block *es,
2075 unsigned long journal_devnum)
2076 {
2077 journal_t *journal;
2078 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2079 dev_t journal_dev;
2080 int err = 0;
2081 int really_read_only;
2082
2083 if (journal_devnum &&
2084 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2085 printk(KERN_INFO "EXT3-fs: external journal device major/minor "
2086 "numbers have changed\n");
2087 journal_dev = new_decode_dev(journal_devnum);
2088 } else
2089 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2090
2091 really_read_only = bdev_read_only(sb->s_bdev);
2092
2093 /*
2094 * Are we loading a blank journal or performing recovery after a
2095 * crash? For recovery, we need to check in advance whether we
2096 * can get read-write access to the device.
2097 */
2098
2099 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
2100 if (sb->s_flags & MS_RDONLY) {
2101 printk(KERN_INFO "EXT3-fs: INFO: recovery "
2102 "required on readonly filesystem.\n");
2103 if (really_read_only) {
2104 printk(KERN_ERR "EXT3-fs: write access "
2105 "unavailable, cannot proceed.\n");
2106 return -EROFS;
2107 }
2108 printk (KERN_INFO "EXT3-fs: write access will "
2109 "be enabled during recovery.\n");
2110 }
2111 }
2112
2113 if (journal_inum && journal_dev) {
2114 printk(KERN_ERR "EXT3-fs: filesystem has both journal "
2115 "and inode journals!\n");
2116 return -EINVAL;
2117 }
2118
2119 if (journal_inum) {
2120 if (!(journal = ext3_get_journal(sb, journal_inum)))
2121 return -EINVAL;
2122 } else {
2123 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
2124 return -EINVAL;
2125 }
2126
2127 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2128 err = journal_update_format(journal);
2129 if (err) {
2130 printk(KERN_ERR "EXT3-fs: error updating journal.\n");
2131 journal_destroy(journal);
2132 return err;
2133 }
2134 }
2135
2136 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
2137 err = journal_wipe(journal, !really_read_only);
2138 if (!err)
2139 err = journal_load(journal);
2140
2141 if (err) {
2142 printk(KERN_ERR "EXT3-fs: error loading journal.\n");
2143 journal_destroy(journal);
2144 return err;
2145 }
2146
2147 EXT3_SB(sb)->s_journal = journal;
2148 ext3_clear_journal_err(sb, es);
2149
2150 if (journal_devnum &&
2151 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2152 es->s_journal_dev = cpu_to_le32(journal_devnum);
2153 sb->s_dirt = 1;
2154
2155 /* Make sure we flush the recovery flag to disk. */
2156 ext3_commit_super(sb, es, 1);
2157 }
2158
2159 return 0;
2160 }
2161
2162 static int ext3_create_journal(struct super_block * sb,
2163 struct ext3_super_block * es,
2164 unsigned int journal_inum)
2165 {
2166 journal_t *journal;
2167 int err;
2168
2169 if (sb->s_flags & MS_RDONLY) {
2170 printk(KERN_ERR "EXT3-fs: readonly filesystem when trying to "
2171 "create journal.\n");
2172 return -EROFS;
2173 }
2174
2175 journal = ext3_get_journal(sb, journal_inum);
2176 if (!journal)
2177 return -EINVAL;
2178
2179 printk(KERN_INFO "EXT3-fs: creating new journal on inode %u\n",
2180 journal_inum);
2181
2182 err = journal_create(journal);
2183 if (err) {
2184 printk(KERN_ERR "EXT3-fs: error creating journal.\n");
2185 journal_destroy(journal);
2186 return -EIO;
2187 }
2188
2189 EXT3_SB(sb)->s_journal = journal;
2190
2191 ext3_update_dynamic_rev(sb);
2192 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2193 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
2194
2195 es->s_journal_inum = cpu_to_le32(journal_inum);
2196 sb->s_dirt = 1;
2197
2198 /* Make sure we flush the recovery flag to disk. */
2199 ext3_commit_super(sb, es, 1);
2200
2201 return 0;
2202 }
2203
2204 static void ext3_commit_super (struct super_block * sb,
2205 struct ext3_super_block * es,
2206 int sync)
2207 {
2208 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
2209
2210 if (!sbh)
2211 return;
2212 es->s_wtime = cpu_to_le32(get_seconds());
2213 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
2214 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
2215 BUFFER_TRACE(sbh, "marking dirty");
2216 mark_buffer_dirty(sbh);
2217 if (sync)
2218 sync_dirty_buffer(sbh);
2219 }
2220
2221
2222 /*
2223 * Have we just finished recovery? If so, and if we are mounting (or
2224 * remounting) the filesystem readonly, then we will end up with a
2225 * consistent fs on disk. Record that fact.
2226 */
2227 static void ext3_mark_recovery_complete(struct super_block * sb,
2228 struct ext3_super_block * es)
2229 {
2230 journal_t *journal = EXT3_SB(sb)->s_journal;
2231
2232 journal_lock_updates(journal);
2233 journal_flush(journal);
2234 lock_super(sb);
2235 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
2236 sb->s_flags & MS_RDONLY) {
2237 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2238 sb->s_dirt = 0;
2239 ext3_commit_super(sb, es, 1);
2240 }
2241 unlock_super(sb);
2242 journal_unlock_updates(journal);
2243 }
2244
2245 /*
2246 * If we are mounting (or read-write remounting) a filesystem whose journal
2247 * has recorded an error from a previous lifetime, move that error to the
2248 * main filesystem now.
2249 */
2250 static void ext3_clear_journal_err(struct super_block * sb,
2251 struct ext3_super_block * es)
2252 {
2253 journal_t *journal;
2254 int j_errno;
2255 const char *errstr;
2256
2257 journal = EXT3_SB(sb)->s_journal;
2258
2259 /*
2260 * Now check for any error status which may have been recorded in the
2261 * journal by a prior ext3_error() or ext3_abort()
2262 */
2263
2264 j_errno = journal_errno(journal);
2265 if (j_errno) {
2266 char nbuf[16];
2267
2268 errstr = ext3_decode_error(sb, j_errno, nbuf);
2269 ext3_warning(sb, __FUNCTION__, "Filesystem error recorded "
2270 "from previous mount: %s", errstr);
2271 ext3_warning(sb, __FUNCTION__, "Marking fs in need of "
2272 "filesystem check.");
2273
2274 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
2275 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
2276 ext3_commit_super (sb, es, 1);
2277
2278 journal_clear_err(journal);
2279 }
2280 }
2281
2282 /*
2283 * Force the running and committing transactions to commit,
2284 * and wait on the commit.
2285 */
2286 int ext3_force_commit(struct super_block *sb)
2287 {
2288 journal_t *journal;
2289 int ret;
2290
2291 if (sb->s_flags & MS_RDONLY)
2292 return 0;
2293
2294 journal = EXT3_SB(sb)->s_journal;
2295 sb->s_dirt = 0;
2296 ret = ext3_journal_force_commit(journal);
2297 return ret;
2298 }
2299
2300 /*
2301 * Ext3 always journals updates to the superblock itself, so we don't
2302 * have to propagate any other updates to the superblock on disk at this
2303 * point. Just start an async writeback to get the buffers on their way
2304 * to the disk.
2305 *
2306 * This implicitly triggers the writebehind on sync().
2307 */
2308
2309 static void ext3_write_super (struct super_block * sb)
2310 {
2311 if (mutex_trylock(&sb->s_lock) != 0)
2312 BUG();
2313 sb->s_dirt = 0;
2314 }
2315
2316 static int ext3_sync_fs(struct super_block *sb, int wait)
2317 {
2318 tid_t target;
2319
2320 sb->s_dirt = 0;
2321 if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
2322 if (wait)
2323 log_wait_commit(EXT3_SB(sb)->s_journal, target);
2324 }
2325 return 0;
2326 }
2327
2328 /*
2329 * LVM calls this function before a (read-only) snapshot is created. This
2330 * gives us a chance to flush the journal completely and mark the fs clean.
2331 */
2332 static void ext3_write_super_lockfs(struct super_block *sb)
2333 {
2334 sb->s_dirt = 0;
2335
2336 if (!(sb->s_flags & MS_RDONLY)) {
2337 journal_t *journal = EXT3_SB(sb)->s_journal;
2338
2339 /* Now we set up the journal barrier. */
2340 journal_lock_updates(journal);
2341 journal_flush(journal);
2342
2343 /* Journal blocked and flushed, clear needs_recovery flag. */
2344 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2345 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2346 }
2347 }
2348
2349 /*
2350 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2351 * flag here, even though the filesystem is not technically dirty yet.
2352 */
2353 static void ext3_unlockfs(struct super_block *sb)
2354 {
2355 if (!(sb->s_flags & MS_RDONLY)) {
2356 lock_super(sb);
2357 /* Reser the needs_recovery flag before the fs is unlocked. */
2358 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2359 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2360 unlock_super(sb);
2361 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2362 }
2363 }
2364
2365 static int ext3_remount (struct super_block * sb, int * flags, char * data)
2366 {
2367 struct ext3_super_block * es;
2368 struct ext3_sb_info *sbi = EXT3_SB(sb);
2369 ext3_fsblk_t n_blocks_count = 0;
2370 unsigned long old_sb_flags;
2371 struct ext3_mount_options old_opts;
2372 int err;
2373 #ifdef CONFIG_QUOTA
2374 int i;
2375 #endif
2376
2377 /* Store the original options */
2378 old_sb_flags = sb->s_flags;
2379 old_opts.s_mount_opt = sbi->s_mount_opt;
2380 old_opts.s_resuid = sbi->s_resuid;
2381 old_opts.s_resgid = sbi->s_resgid;
2382 old_opts.s_commit_interval = sbi->s_commit_interval;
2383 #ifdef CONFIG_QUOTA
2384 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2385 for (i = 0; i < MAXQUOTAS; i++)
2386 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2387 #endif
2388
2389 /*
2390 * Allow the "check" option to be passed as a remount option.
2391 */
2392 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2393 err = -EINVAL;
2394 goto restore_opts;
2395 }
2396
2397 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT)
2398 ext3_abort(sb, __FUNCTION__, "Abort forced by user");
2399
2400 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2401 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2402
2403 es = sbi->s_es;
2404
2405 ext3_init_journal_params(sb, sbi->s_journal);
2406
2407 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2408 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
2409 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT) {
2410 err = -EROFS;
2411 goto restore_opts;
2412 }
2413
2414 if (*flags & MS_RDONLY) {
2415 /*
2416 * First of all, the unconditional stuff we have to do
2417 * to disable replay of the journal when we next remount
2418 */
2419 sb->s_flags |= MS_RDONLY;
2420
2421 /*
2422 * OK, test if we are remounting a valid rw partition
2423 * readonly, and if so set the rdonly flag and then
2424 * mark the partition as valid again.
2425 */
2426 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2427 (sbi->s_mount_state & EXT3_VALID_FS))
2428 es->s_state = cpu_to_le16(sbi->s_mount_state);
2429
2430 /*
2431 * We have to unlock super so that we can wait for
2432 * transactions.
2433 */
2434 unlock_super(sb);
2435 ext3_mark_recovery_complete(sb, es);
2436 lock_super(sb);
2437 } else {
2438 __le32 ret;
2439 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2440 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
2441 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2442 "remount RDWR because of unsupported "
2443 "optional features (%x).\n",
2444 sb->s_id, le32_to_cpu(ret));
2445 err = -EROFS;
2446 goto restore_opts;
2447 }
2448
2449 /*
2450 * If we have an unprocessed orphan list hanging
2451 * around from a previously readonly bdev mount,
2452 * require a full umount/remount for now.
2453 */
2454 if (es->s_last_orphan) {
2455 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2456 "remount RDWR because of unprocessed "
2457 "orphan inode list. Please "
2458 "umount/remount instead.\n",
2459 sb->s_id);
2460 err = -EINVAL;
2461 goto restore_opts;
2462 }
2463
2464 /*
2465 * Mounting a RDONLY partition read-write, so reread
2466 * and store the current valid flag. (It may have
2467 * been changed by e2fsck since we originally mounted
2468 * the partition.)
2469 */
2470 ext3_clear_journal_err(sb, es);
2471 sbi->s_mount_state = le16_to_cpu(es->s_state);
2472 if ((err = ext3_group_extend(sb, es, n_blocks_count)))
2473 goto restore_opts;
2474 if (!ext3_setup_super (sb, es, 0))
2475 sb->s_flags &= ~MS_RDONLY;
2476 }
2477 }
2478 #ifdef CONFIG_QUOTA
2479 /* Release old quota file names */
2480 for (i = 0; i < MAXQUOTAS; i++)
2481 if (old_opts.s_qf_names[i] &&
2482 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2483 kfree(old_opts.s_qf_names[i]);
2484 #endif
2485 return 0;
2486 restore_opts:
2487 sb->s_flags = old_sb_flags;
2488 sbi->s_mount_opt = old_opts.s_mount_opt;
2489 sbi->s_resuid = old_opts.s_resuid;
2490 sbi->s_resgid = old_opts.s_resgid;
2491 sbi->s_commit_interval = old_opts.s_commit_interval;
2492 #ifdef CONFIG_QUOTA
2493 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2494 for (i = 0; i < MAXQUOTAS; i++) {
2495 if (sbi->s_qf_names[i] &&
2496 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2497 kfree(sbi->s_qf_names[i]);
2498 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2499 }
2500 #endif
2501 return err;
2502 }
2503
2504 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf)
2505 {
2506 struct super_block *sb = dentry->d_sb;
2507 struct ext3_sb_info *sbi = EXT3_SB(sb);
2508 struct ext3_super_block *es = sbi->s_es;
2509 u64 fsid;
2510
2511 if (test_opt(sb, MINIX_DF)) {
2512 sbi->s_overhead_last = 0;
2513 } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
2514 unsigned long ngroups = sbi->s_groups_count, i;
2515 ext3_fsblk_t overhead = 0;
2516 smp_rmb();
2517
2518 /*
2519 * Compute the overhead (FS structures). This is constant
2520 * for a given filesystem unless the number of block groups
2521 * changes so we cache the previous value until it does.
2522 */
2523
2524 /*
2525 * All of the blocks before first_data_block are
2526 * overhead
2527 */
2528 overhead = le32_to_cpu(es->s_first_data_block);
2529
2530 /*
2531 * Add the overhead attributed to the superblock and
2532 * block group descriptors. If the sparse superblocks
2533 * feature is turned on, then not all groups have this.
2534 */
2535 for (i = 0; i < ngroups; i++) {
2536 overhead += ext3_bg_has_super(sb, i) +
2537 ext3_bg_num_gdb(sb, i);
2538 cond_resched();
2539 }
2540
2541 /*
2542 * Every block group has an inode bitmap, a block
2543 * bitmap, and an inode table.
2544 */
2545 overhead += ngroups * (2 + sbi->s_itb_per_group);
2546 sbi->s_overhead_last = overhead;
2547 smp_wmb();
2548 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
2549 }
2550
2551 buf->f_type = EXT3_SUPER_MAGIC;
2552 buf->f_bsize = sb->s_blocksize;
2553 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
2554 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
2555 es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
2556 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2557 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2558 buf->f_bavail = 0;
2559 buf->f_files = le32_to_cpu(es->s_inodes_count);
2560 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
2561 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
2562 buf->f_namelen = EXT3_NAME_LEN;
2563 fsid = le64_to_cpup((void *)es->s_uuid) ^
2564 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2565 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2566 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
2567 return 0;
2568 }
2569
2570 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2571 * is locked for write. Otherwise the are possible deadlocks:
2572 * Process 1 Process 2
2573 * ext3_create() quota_sync()
2574 * journal_start() write_dquot()
2575 * DQUOT_INIT() down(dqio_mutex)
2576 * down(dqio_mutex) journal_start()
2577 *
2578 */
2579
2580 #ifdef CONFIG_QUOTA
2581
2582 static inline struct inode *dquot_to_inode(struct dquot *dquot)
2583 {
2584 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2585 }
2586
2587 static int ext3_dquot_initialize(struct inode *inode, int type)
2588 {
2589 handle_t *handle;
2590 int ret, err;
2591
2592 /* We may create quota structure so we need to reserve enough blocks */
2593 handle = ext3_journal_start(inode, 2*EXT3_QUOTA_INIT_BLOCKS(inode->i_sb));
2594 if (IS_ERR(handle))
2595 return PTR_ERR(handle);
2596 ret = dquot_initialize(inode, type);
2597 err = ext3_journal_stop(handle);
2598 if (!ret)
2599 ret = err;
2600 return ret;
2601 }
2602
2603 static int ext3_dquot_drop(struct inode *inode)
2604 {
2605 handle_t *handle;
2606 int ret, err;
2607
2608 /* We may delete quota structure so we need to reserve enough blocks */
2609 handle = ext3_journal_start(inode, 2*EXT3_QUOTA_DEL_BLOCKS(inode->i_sb));
2610 if (IS_ERR(handle))
2611 return PTR_ERR(handle);
2612 ret = dquot_drop(inode);
2613 err = ext3_journal_stop(handle);
2614 if (!ret)
2615 ret = err;
2616 return ret;
2617 }
2618
2619 static int ext3_write_dquot(struct dquot *dquot)
2620 {
2621 int ret, err;
2622 handle_t *handle;
2623 struct inode *inode;
2624
2625 inode = dquot_to_inode(dquot);
2626 handle = ext3_journal_start(inode,
2627 EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
2628 if (IS_ERR(handle))
2629 return PTR_ERR(handle);
2630 ret = dquot_commit(dquot);
2631 err = ext3_journal_stop(handle);
2632 if (!ret)
2633 ret = err;
2634 return ret;
2635 }
2636
2637 static int ext3_acquire_dquot(struct dquot *dquot)
2638 {
2639 int ret, err;
2640 handle_t *handle;
2641
2642 handle = ext3_journal_start(dquot_to_inode(dquot),
2643 EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
2644 if (IS_ERR(handle))
2645 return PTR_ERR(handle);
2646 ret = dquot_acquire(dquot);
2647 err = ext3_journal_stop(handle);
2648 if (!ret)
2649 ret = err;
2650 return ret;
2651 }
2652
2653 static int ext3_release_dquot(struct dquot *dquot)
2654 {
2655 int ret, err;
2656 handle_t *handle;
2657
2658 handle = ext3_journal_start(dquot_to_inode(dquot),
2659 EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
2660 if (IS_ERR(handle)) {
2661 /* Release dquot anyway to avoid endless cycle in dqput() */
2662 dquot_release(dquot);
2663 return PTR_ERR(handle);
2664 }
2665 ret = dquot_release(dquot);
2666 err = ext3_journal_stop(handle);
2667 if (!ret)
2668 ret = err;
2669 return ret;
2670 }
2671
2672 static int ext3_mark_dquot_dirty(struct dquot *dquot)
2673 {
2674 /* Are we journalling quotas? */
2675 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2676 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2677 dquot_mark_dquot_dirty(dquot);
2678 return ext3_write_dquot(dquot);
2679 } else {
2680 return dquot_mark_dquot_dirty(dquot);
2681 }
2682 }
2683
2684 static int ext3_write_info(struct super_block *sb, int type)
2685 {
2686 int ret, err;
2687 handle_t *handle;
2688
2689 /* Data block + inode block */
2690 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2691 if (IS_ERR(handle))
2692 return PTR_ERR(handle);
2693 ret = dquot_commit_info(sb, type);
2694 err = ext3_journal_stop(handle);
2695 if (!ret)
2696 ret = err;
2697 return ret;
2698 }
2699
2700 /*
2701 * Turn on quotas during mount time - we need to find
2702 * the quota file and such...
2703 */
2704 static int ext3_quota_on_mount(struct super_block *sb, int type)
2705 {
2706 return vfs_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2707 EXT3_SB(sb)->s_jquota_fmt, type);
2708 }
2709
2710 /*
2711 * Standard function to be called on quota_on
2712 */
2713 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
2714 char *path)
2715 {
2716 int err;
2717 struct nameidata nd;
2718
2719 if (!test_opt(sb, QUOTA))
2720 return -EINVAL;
2721 /* Not journalling quota? */
2722 if (!EXT3_SB(sb)->s_qf_names[USRQUOTA] &&
2723 !EXT3_SB(sb)->s_qf_names[GRPQUOTA])
2724 return vfs_quota_on(sb, type, format_id, path);
2725 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
2726 if (err)
2727 return err;
2728 /* Quotafile not on the same filesystem? */
2729 if (nd.mnt->mnt_sb != sb) {
2730 path_release(&nd);
2731 return -EXDEV;
2732 }
2733 /* Quotafile not of fs root? */
2734 if (nd.dentry->d_parent->d_inode != sb->s_root->d_inode)
2735 printk(KERN_WARNING
2736 "EXT3-fs: Quota file not on filesystem root. "
2737 "Journalled quota will not work.\n");
2738 path_release(&nd);
2739 return vfs_quota_on(sb, type, format_id, path);
2740 }
2741
2742 /* Read data from quotafile - avoid pagecache and such because we cannot afford
2743 * acquiring the locks... As quota files are never truncated and quota code
2744 * itself serializes the operations (and noone else should touch the files)
2745 * we don't have to be afraid of races */
2746 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
2747 size_t len, loff_t off)
2748 {
2749 struct inode *inode = sb_dqopt(sb)->files[type];
2750 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2751 int err = 0;
2752 int offset = off & (sb->s_blocksize - 1);
2753 int tocopy;
2754 size_t toread;
2755 struct buffer_head *bh;
2756 loff_t i_size = i_size_read(inode);
2757
2758 if (off > i_size)
2759 return 0;
2760 if (off+len > i_size)
2761 len = i_size-off;
2762 toread = len;
2763 while (toread > 0) {
2764 tocopy = sb->s_blocksize - offset < toread ?
2765 sb->s_blocksize - offset : toread;
2766 bh = ext3_bread(NULL, inode, blk, 0, &err);
2767 if (err)
2768 return err;
2769 if (!bh) /* A hole? */
2770 memset(data, 0, tocopy);
2771 else
2772 memcpy(data, bh->b_data+offset, tocopy);
2773 brelse(bh);
2774 offset = 0;
2775 toread -= tocopy;
2776 data += tocopy;
2777 blk++;
2778 }
2779 return len;
2780 }
2781
2782 /* Write to quotafile (we know the transaction is already started and has
2783 * enough credits) */
2784 static ssize_t ext3_quota_write(struct super_block *sb, int type,
2785 const char *data, size_t len, loff_t off)
2786 {
2787 struct inode *inode = sb_dqopt(sb)->files[type];
2788 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2789 int err = 0;
2790 int offset = off & (sb->s_blocksize - 1);
2791 int tocopy;
2792 int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
2793 size_t towrite = len;
2794 struct buffer_head *bh;
2795 handle_t *handle = journal_current_handle();
2796
2797 if (!handle) {
2798 printk(KERN_WARNING "EXT3-fs: Quota write (off=%Lu, len=%Lu)"
2799 " cancelled because transaction is not started.\n",
2800 (unsigned long long)off, (unsigned long long)len);
2801 return -EIO;
2802 }
2803 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
2804 while (towrite > 0) {
2805 tocopy = sb->s_blocksize - offset < towrite ?
2806 sb->s_blocksize - offset : towrite;
2807 bh = ext3_bread(handle, inode, blk, 1, &err);
2808 if (!bh)
2809 goto out;
2810 if (journal_quota) {
2811 err = ext3_journal_get_write_access(handle, bh);
2812 if (err) {
2813 brelse(bh);
2814 goto out;
2815 }
2816 }
2817 lock_buffer(bh);
2818 memcpy(bh->b_data+offset, data, tocopy);
2819 flush_dcache_page(bh->b_page);
2820 unlock_buffer(bh);
2821 if (journal_quota)
2822 err = ext3_journal_dirty_metadata(handle, bh);
2823 else {
2824 /* Always do at least ordered writes for quotas */
2825 err = ext3_journal_dirty_data(handle, bh);
2826 mark_buffer_dirty(bh);
2827 }
2828 brelse(bh);
2829 if (err)
2830 goto out;
2831 offset = 0;
2832 towrite -= tocopy;
2833 data += tocopy;
2834 blk++;
2835 }
2836 out:
2837 if (len == towrite)
2838 return err;
2839 if (inode->i_size < off+len-towrite) {
2840 i_size_write(inode, off+len-towrite);
2841 EXT3_I(inode)->i_disksize = inode->i_size;
2842 }
2843 inode->i_version++;
2844 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2845 ext3_mark_inode_dirty(handle, inode);
2846 mutex_unlock(&inode->i_mutex);
2847 return len - towrite;
2848 }
2849
2850 #endif
2851
2852 static int ext3_get_sb(struct file_system_type *fs_type,
2853 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
2854 {
2855 return get_sb_bdev(fs_type, flags, dev_name, data, ext3_fill_super, mnt);
2856 }
2857
2858 static struct file_system_type ext3_fs_type = {
2859 .owner = THIS_MODULE,
2860 .name = "ext3",
2861 .get_sb = ext3_get_sb,
2862 .kill_sb = kill_block_super,
2863 .fs_flags = FS_REQUIRES_DEV,
2864 };
2865
2866 static int __init init_ext3_fs(void)
2867 {
2868 int err = init_ext3_xattr();
2869 if (err)
2870 return err;
2871 err = init_inodecache();
2872 if (err)
2873 goto out1;
2874 err = register_filesystem(&ext3_fs_type);
2875 if (err)
2876 goto out;
2877 return 0;
2878 out:
2879 destroy_inodecache();
2880 out1:
2881 exit_ext3_xattr();
2882 return err;
2883 }
2884
2885 static void __exit exit_ext3_fs(void)
2886 {
2887 unregister_filesystem(&ext3_fs_type);
2888 destroy_inodecache();
2889 exit_ext3_xattr();
2890 }
2891
2892 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
2893 MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
2894 MODULE_LICENSE("GPL");
2895 module_init(init_ext3_fs)
2896 module_exit(exit_ext3_fs)
x86 "subsystem" repository