?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
[ALSA] wavefront: only declare isapnp on CONFIG_PNP
[linux-2.6/mini2440.git] / fs / ext3 / super.c
blob54d3c90412598f3c13bb0c4c76dc8ad5545819ee
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/vfs.h>
33 #include <linux/random.h>
34 #include <linux/mount.h>
35 #include <linux/namei.h>
36 #include <linux/quotaops.h>
37 #include <linux/seq_file.h>
38
39 #include <asm/uaccess.h>
40
41 #include "xattr.h"
42 #include "acl.h"
43 #include "namei.h"
44
45 static int ext3_load_journal(struct super_block *, struct ext3_super_block *,
46 unsigned long journal_devnum);
47 static int ext3_create_journal(struct super_block *, struct ext3_super_block *,
48 unsigned int);
49 static void ext3_commit_super (struct super_block * sb,
50 struct ext3_super_block * es,
51 int sync);
52 static void ext3_mark_recovery_complete(struct super_block * sb,
53 struct ext3_super_block * es);
54 static void ext3_clear_journal_err(struct super_block * sb,
55 struct ext3_super_block * es);
56 static int ext3_sync_fs(struct super_block *sb, int wait);
57 static const char *ext3_decode_error(struct super_block * sb, int errno,
58 char nbuf[16]);
59 static int ext3_remount (struct super_block * sb, int * flags, char * data);
60 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf);
61 static void ext3_unlockfs(struct super_block *sb);
62 static void ext3_write_super (struct super_block * sb);
63 static void ext3_write_super_lockfs(struct super_block *sb);
64
65 /*
66 * Wrappers for journal_start/end.
67 *
68 * The only special thing we need to do here is to make sure that all
69 * journal_end calls result in the superblock being marked dirty, so
70 * that sync() will call the filesystem's write_super callback if
71 * appropriate.
72 */
73 handle_t *ext3_journal_start_sb(struct super_block *sb, int nblocks)
74 {
75 journal_t *journal;
76
77 if (sb->s_flags & MS_RDONLY)
78 return ERR_PTR(-EROFS);
79
80 /* Special case here: if the journal has aborted behind our
81 * backs (eg. EIO in the commit thread), then we still need to
82 * take the FS itself readonly cleanly. */
83 journal = EXT3_SB(sb)->s_journal;
84 if (is_journal_aborted(journal)) {
85 ext3_abort(sb, __FUNCTION__,
86 "Detected aborted journal");
87 return ERR_PTR(-EROFS);
88 }
89
90 return journal_start(journal, nblocks);
91 }
92
93 /*
94 * The only special thing we need to do here is to make sure that all
95 * journal_stop calls result in the superblock being marked dirty, so
96 * that sync() will call the filesystem's write_super callback if
97 * appropriate.
98 */
99 int __ext3_journal_stop(const char *where, handle_t *handle)
100 {
101 struct super_block *sb;
102 int err;
103 int rc;
104
105 sb = handle->h_transaction->t_journal->j_private;
106 err = handle->h_err;
107 rc = journal_stop(handle);
108
109 if (!err)
110 err = rc;
111 if (err)
112 __ext3_std_error(sb, where, err);
113 return err;
114 }
115
116 void ext3_journal_abort_handle(const char *caller, const char *err_fn,
117 struct buffer_head *bh, handle_t *handle, int err)
118 {
119 char nbuf[16];
120 const char *errstr = ext3_decode_error(NULL, err, nbuf);
121
122 if (bh)
123 BUFFER_TRACE(bh, "abort");
124
125 if (!handle->h_err)
126 handle->h_err = err;
127
128 if (is_handle_aborted(handle))
129 return;
130
131 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
132 caller, errstr, err_fn);
133
134 journal_abort_handle(handle);
135 }
136
137 /* Deal with the reporting of failure conditions on a filesystem such as
138 * inconsistencies detected or read IO failures.
139 *
140 * On ext2, we can store the error state of the filesystem in the
141 * superblock. That is not possible on ext3, because we may have other
142 * write ordering constraints on the superblock which prevent us from
143 * writing it out straight away; and given that the journal is about to
144 * be aborted, we can't rely on the current, or future, transactions to
145 * write out the superblock safely.
146 *
147 * We'll just use the journal_abort() error code to record an error in
148 * the journal instead. On recovery, the journal will compain about
149 * that error until we've noted it down and cleared it.
150 */
151
152 static void ext3_handle_error(struct super_block *sb)
153 {
154 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
155
156 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
157 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
158
159 if (sb->s_flags & MS_RDONLY)
160 return;
161
162 if (!test_opt (sb, ERRORS_CONT)) {
163 journal_t *journal = EXT3_SB(sb)->s_journal;
164
165 EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
166 if (journal)
167 journal_abort(journal, -EIO);
168 }
169 if (test_opt (sb, ERRORS_RO)) {
170 printk (KERN_CRIT "Remounting filesystem read-only\n");
171 sb->s_flags |= MS_RDONLY;
172 }
173 ext3_commit_super(sb, es, 1);
174 if (test_opt(sb, ERRORS_PANIC))
175 panic("EXT3-fs (device %s): panic forced after error\n",
176 sb->s_id);
177 }
178
179 void ext3_error (struct super_block * sb, const char * function,
180 const char * fmt, ...)
181 {
182 va_list args;
183
184 va_start(args, fmt);
185 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
186 vprintk(fmt, args);
187 printk("\n");
188 va_end(args);
189
190 ext3_handle_error(sb);
191 }
192
193 static const char *ext3_decode_error(struct super_block * sb, int errno,
194 char nbuf[16])
195 {
196 char *errstr = NULL;
197
198 switch (errno) {
199 case -EIO:
200 errstr = "IO failure";
201 break;
202 case -ENOMEM:
203 errstr = "Out of memory";
204 break;
205 case -EROFS:
206 if (!sb || EXT3_SB(sb)->s_journal->j_flags & JFS_ABORT)
207 errstr = "Journal has aborted";
208 else
209 errstr = "Readonly filesystem";
210 break;
211 default:
212 /* If the caller passed in an extra buffer for unknown
213 * errors, textualise them now. Else we just return
214 * NULL. */
215 if (nbuf) {
216 /* Check for truncated error codes... */
217 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
218 errstr = nbuf;
219 }
220 break;
221 }
222
223 return errstr;
224 }
225
226 /* __ext3_std_error decodes expected errors from journaling functions
227 * automatically and invokes the appropriate error response. */
228
229 void __ext3_std_error (struct super_block * sb, const char * function,
230 int errno)
231 {
232 char nbuf[16];
233 const char *errstr;
234
235 /* Special case: if the error is EROFS, and we're not already
236 * inside a transaction, then there's really no point in logging
237 * an error. */
238 if (errno == -EROFS && journal_current_handle() == NULL &&
239 (sb->s_flags & MS_RDONLY))
240 return;
241
242 errstr = ext3_decode_error(sb, errno, nbuf);
243 printk (KERN_CRIT "EXT3-fs error (device %s) in %s: %s\n",
244 sb->s_id, function, errstr);
245
246 ext3_handle_error(sb);
247 }
248
249 /*
250 * ext3_abort is a much stronger failure handler than ext3_error. The
251 * abort function may be used to deal with unrecoverable failures such
252 * as journal IO errors or ENOMEM at a critical moment in log management.
253 *
254 * We unconditionally force the filesystem into an ABORT|READONLY state,
255 * unless the error response on the fs has been set to panic in which
256 * case we take the easy way out and panic immediately.
257 */
258
259 void ext3_abort (struct super_block * sb, const char * function,
260 const char * fmt, ...)
261 {
262 va_list args;
263
264 printk (KERN_CRIT "ext3_abort called.\n");
265
266 va_start(args, fmt);
267 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
268 vprintk(fmt, args);
269 printk("\n");
270 va_end(args);
271
272 if (test_opt(sb, ERRORS_PANIC))
273 panic("EXT3-fs panic from previous error\n");
274
275 if (sb->s_flags & MS_RDONLY)
276 return;
277
278 printk(KERN_CRIT "Remounting filesystem read-only\n");
279 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
280 sb->s_flags |= MS_RDONLY;
281 EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
282 journal_abort(EXT3_SB(sb)->s_journal, -EIO);
283 }
284
285 void ext3_warning (struct super_block * sb, const char * function,
286 const char * fmt, ...)
287 {
288 va_list args;
289
290 va_start(args, fmt);
291 printk(KERN_WARNING "EXT3-fs warning (device %s): %s: ",
292 sb->s_id, function);
293 vprintk(fmt, args);
294 printk("\n");
295 va_end(args);
296 }
297
298 void ext3_update_dynamic_rev(struct super_block *sb)
299 {
300 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
301
302 if (le32_to_cpu(es->s_rev_level) > EXT3_GOOD_OLD_REV)
303 return;
304
305 ext3_warning(sb, __FUNCTION__,
306 "updating to rev %d because of new feature flag, "
307 "running e2fsck is recommended",
308 EXT3_DYNAMIC_REV);
309
310 es->s_first_ino = cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO);
311 es->s_inode_size = cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE);
312 es->s_rev_level = cpu_to_le32(EXT3_DYNAMIC_REV);
313 /* leave es->s_feature_*compat flags alone */
314 /* es->s_uuid will be set by e2fsck if empty */
315
316 /*
317 * The rest of the superblock fields should be zero, and if not it
318 * means they are likely already in use, so leave them alone. We
319 * can leave it up to e2fsck to clean up any inconsistencies there.
320 */
321 }
322
323 /*
324 * Open the external journal device
325 */
326 static struct block_device *ext3_blkdev_get(dev_t dev)
327 {
328 struct block_device *bdev;
329 char b[BDEVNAME_SIZE];
330
331 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
332 if (IS_ERR(bdev))
333 goto fail;
334 return bdev;
335
336 fail:
337 printk(KERN_ERR "EXT3: failed to open journal device %s: %ld\n",
338 __bdevname(dev, b), PTR_ERR(bdev));
339 return NULL;
340 }
341
342 /*
343 * Release the journal device
344 */
345 static int ext3_blkdev_put(struct block_device *bdev)
346 {
347 bd_release(bdev);
348 return blkdev_put(bdev);
349 }
350
351 static int ext3_blkdev_remove(struct ext3_sb_info *sbi)
352 {
353 struct block_device *bdev;
354 int ret = -ENODEV;
355
356 bdev = sbi->journal_bdev;
357 if (bdev) {
358 ret = ext3_blkdev_put(bdev);
359 sbi->journal_bdev = NULL;
360 }
361 return ret;
362 }
363
364 static inline struct inode *orphan_list_entry(struct list_head *l)
365 {
366 return &list_entry(l, struct ext3_inode_info, i_orphan)->vfs_inode;
367 }
368
369 static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi)
370 {
371 struct list_head *l;
372
373 printk(KERN_ERR "sb orphan head is %d\n",
374 le32_to_cpu(sbi->s_es->s_last_orphan));
375
376 printk(KERN_ERR "sb_info orphan list:\n");
377 list_for_each(l, &sbi->s_orphan) {
378 struct inode *inode = orphan_list_entry(l);
379 printk(KERN_ERR " "
380 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
381 inode->i_sb->s_id, inode->i_ino, inode,
382 inode->i_mode, inode->i_nlink,
383 NEXT_ORPHAN(inode));
384 }
385 }
386
387 static void ext3_put_super (struct super_block * sb)
388 {
389 struct ext3_sb_info *sbi = EXT3_SB(sb);
390 struct ext3_super_block *es = sbi->s_es;
391 int i;
392
393 ext3_xattr_put_super(sb);
394 journal_destroy(sbi->s_journal);
395 if (!(sb->s_flags & MS_RDONLY)) {
396 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
397 es->s_state = cpu_to_le16(sbi->s_mount_state);
398 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
399 mark_buffer_dirty(sbi->s_sbh);
400 ext3_commit_super(sb, es, 1);
401 }
402
403 for (i = 0; i < sbi->s_gdb_count; i++)
404 brelse(sbi->s_group_desc[i]);
405 kfree(sbi->s_group_desc);
406 percpu_counter_destroy(&sbi->s_freeblocks_counter);
407 percpu_counter_destroy(&sbi->s_freeinodes_counter);
408 percpu_counter_destroy(&sbi->s_dirs_counter);
409 brelse(sbi->s_sbh);
410 #ifdef CONFIG_QUOTA
411 for (i = 0; i < MAXQUOTAS; i++)
412 kfree(sbi->s_qf_names[i]);
413 #endif
414
415 /* Debugging code just in case the in-memory inode orphan list
416 * isn't empty. The on-disk one can be non-empty if we've
417 * detected an error and taken the fs readonly, but the
418 * in-memory list had better be clean by this point. */
419 if (!list_empty(&sbi->s_orphan))
420 dump_orphan_list(sb, sbi);
421 J_ASSERT(list_empty(&sbi->s_orphan));
422
423 invalidate_bdev(sb->s_bdev);
424 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
425 /*
426 * Invalidate the journal device's buffers. We don't want them
427 * floating about in memory - the physical journal device may
428 * hotswapped, and it breaks the `ro-after' testing code.
429 */
430 sync_blockdev(sbi->journal_bdev);
431 invalidate_bdev(sbi->journal_bdev);
432 ext3_blkdev_remove(sbi);
433 }
434 sb->s_fs_info = NULL;
435 kfree(sbi);
436 return;
437 }
438
439 static struct kmem_cache *ext3_inode_cachep;
440
441 /*
442 * Called inside transaction, so use GFP_NOFS
443 */
444 static struct inode *ext3_alloc_inode(struct super_block *sb)
445 {
446 struct ext3_inode_info *ei;
447
448 ei = kmem_cache_alloc(ext3_inode_cachep, GFP_NOFS);
449 if (!ei)
450 return NULL;
451 #ifdef CONFIG_EXT3_FS_POSIX_ACL
452 ei->i_acl = EXT3_ACL_NOT_CACHED;
453 ei->i_default_acl = EXT3_ACL_NOT_CACHED;
454 #endif
455 ei->i_block_alloc_info = NULL;
456 ei->vfs_inode.i_version = 1;
457 return &ei->vfs_inode;
458 }
459
460 static void ext3_destroy_inode(struct inode *inode)
461 {
462 kmem_cache_free(ext3_inode_cachep, EXT3_I(inode));
463 }
464
465 static void init_once(void * foo, struct kmem_cache * cachep, unsigned long flags)
466 {
467 struct ext3_inode_info *ei = (struct ext3_inode_info *) foo;
468
469 if (flags & SLAB_CTOR_CONSTRUCTOR) {
470 INIT_LIST_HEAD(&ei->i_orphan);
471 #ifdef CONFIG_EXT3_FS_XATTR
472 init_rwsem(&ei->xattr_sem);
473 #endif
474 mutex_init(&ei->truncate_mutex);
475 inode_init_once(&ei->vfs_inode);
476 }
477 }
478
479 static int init_inodecache(void)
480 {
481 ext3_inode_cachep = kmem_cache_create("ext3_inode_cache",
482 sizeof(struct ext3_inode_info),
483 0, (SLAB_RECLAIM_ACCOUNT|
484 SLAB_MEM_SPREAD),
485 init_once, NULL);
486 if (ext3_inode_cachep == NULL)
487 return -ENOMEM;
488 return 0;
489 }
490
491 static void destroy_inodecache(void)
492 {
493 kmem_cache_destroy(ext3_inode_cachep);
494 }
495
496 static void ext3_clear_inode(struct inode *inode)
497 {
498 struct ext3_block_alloc_info *rsv = EXT3_I(inode)->i_block_alloc_info;
499 #ifdef CONFIG_EXT3_FS_POSIX_ACL
500 if (EXT3_I(inode)->i_acl &&
501 EXT3_I(inode)->i_acl != EXT3_ACL_NOT_CACHED) {
502 posix_acl_release(EXT3_I(inode)->i_acl);
503 EXT3_I(inode)->i_acl = EXT3_ACL_NOT_CACHED;
504 }
505 if (EXT3_I(inode)->i_default_acl &&
506 EXT3_I(inode)->i_default_acl != EXT3_ACL_NOT_CACHED) {
507 posix_acl_release(EXT3_I(inode)->i_default_acl);
508 EXT3_I(inode)->i_default_acl = EXT3_ACL_NOT_CACHED;
509 }
510 #endif
511 ext3_discard_reservation(inode);
512 EXT3_I(inode)->i_block_alloc_info = NULL;
513 if (unlikely(rsv))
514 kfree(rsv);
515 }
516
517 static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb)
518 {
519 #if defined(CONFIG_QUOTA)
520 struct ext3_sb_info *sbi = EXT3_SB(sb);
521
522 if (sbi->s_jquota_fmt)
523 seq_printf(seq, ",jqfmt=%s",
524 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
525
526 if (sbi->s_qf_names[USRQUOTA])
527 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
528
529 if (sbi->s_qf_names[GRPQUOTA])
530 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
531
532 if (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA)
533 seq_puts(seq, ",usrquota");
534
535 if (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)
536 seq_puts(seq, ",grpquota");
537 #endif
538 }
539
540 static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs)
541 {
542 struct super_block *sb = vfs->mnt_sb;
543
544 if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA)
545 seq_puts(seq, ",data=journal");
546 else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA)
547 seq_puts(seq, ",data=ordered");
548 else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)
549 seq_puts(seq, ",data=writeback");
550
551 ext3_show_quota_options(seq, sb);
552
553 return 0;
554 }
555
556
557 static struct dentry *ext3_get_dentry(struct super_block *sb, void *vobjp)
558 {
559 __u32 *objp = vobjp;
560 unsigned long ino = objp[0];
561 __u32 generation = objp[1];
562 struct inode *inode;
563 struct dentry *result;
564
565 if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO)
566 return ERR_PTR(-ESTALE);
567 if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count))
568 return ERR_PTR(-ESTALE);
569
570 /* iget isn't really right if the inode is currently unallocated!!
571 *
572 * ext3_read_inode will return a bad_inode if the inode had been
573 * deleted, so we should be safe.
574 *
575 * Currently we don't know the generation for parent directory, so
576 * a generation of 0 means "accept any"
577 */
578 inode = iget(sb, ino);
579 if (inode == NULL)
580 return ERR_PTR(-ENOMEM);
581 if (is_bad_inode(inode) ||
582 (generation && inode->i_generation != generation)) {
583 iput(inode);
584 return ERR_PTR(-ESTALE);
585 }
586 /* now to find a dentry.
587 * If possible, get a well-connected one
588 */
589 result = d_alloc_anon(inode);
590 if (!result) {
591 iput(inode);
592 return ERR_PTR(-ENOMEM);
593 }
594 return result;
595 }
596
597 #ifdef CONFIG_QUOTA
598 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
599 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
600
601 static int ext3_dquot_initialize(struct inode *inode, int type);
602 static int ext3_dquot_drop(struct inode *inode);
603 static int ext3_write_dquot(struct dquot *dquot);
604 static int ext3_acquire_dquot(struct dquot *dquot);
605 static int ext3_release_dquot(struct dquot *dquot);
606 static int ext3_mark_dquot_dirty(struct dquot *dquot);
607 static int ext3_write_info(struct super_block *sb, int type);
608 static int ext3_quota_on(struct super_block *sb, int type, int format_id, char *path);
609 static int ext3_quota_on_mount(struct super_block *sb, int type);
610 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
611 size_t len, loff_t off);
612 static ssize_t ext3_quota_write(struct super_block *sb, int type,
613 const char *data, size_t len, loff_t off);
614
615 static struct dquot_operations ext3_quota_operations = {
616 .initialize = ext3_dquot_initialize,
617 .drop = ext3_dquot_drop,
618 .alloc_space = dquot_alloc_space,
619 .alloc_inode = dquot_alloc_inode,
620 .free_space = dquot_free_space,
621 .free_inode = dquot_free_inode,
622 .transfer = dquot_transfer,
623 .write_dquot = ext3_write_dquot,
624 .acquire_dquot = ext3_acquire_dquot,
625 .release_dquot = ext3_release_dquot,
626 .mark_dirty = ext3_mark_dquot_dirty,
627 .write_info = ext3_write_info
628 };
629
630 static struct quotactl_ops ext3_qctl_operations = {
631 .quota_on = ext3_quota_on,
632 .quota_off = vfs_quota_off,
633 .quota_sync = vfs_quota_sync,
634 .get_info = vfs_get_dqinfo,
635 .set_info = vfs_set_dqinfo,
636 .get_dqblk = vfs_get_dqblk,
637 .set_dqblk = vfs_set_dqblk
638 };
639 #endif
640
641 static const struct super_operations ext3_sops = {
642 .alloc_inode = ext3_alloc_inode,
643 .destroy_inode = ext3_destroy_inode,
644 .read_inode = ext3_read_inode,
645 .write_inode = ext3_write_inode,
646 .dirty_inode = ext3_dirty_inode,
647 .delete_inode = ext3_delete_inode,
648 .put_super = ext3_put_super,
649 .write_super = ext3_write_super,
650 .sync_fs = ext3_sync_fs,
651 .write_super_lockfs = ext3_write_super_lockfs,
652 .unlockfs = ext3_unlockfs,
653 .statfs = ext3_statfs,
654 .remount_fs = ext3_remount,
655 .clear_inode = ext3_clear_inode,
656 .show_options = ext3_show_options,
657 #ifdef CONFIG_QUOTA
658 .quota_read = ext3_quota_read,
659 .quota_write = ext3_quota_write,
660 #endif
661 };
662
663 static struct export_operations ext3_export_ops = {
664 .get_parent = ext3_get_parent,
665 .get_dentry = ext3_get_dentry,
666 };
667
668 enum {
669 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
670 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
671 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
672 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
673 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
674 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
675 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
676 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
677 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
678 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
679 Opt_grpquota
680 };
681
682 static match_table_t tokens = {
683 {Opt_bsd_df, "bsddf"},
684 {Opt_minix_df, "minixdf"},
685 {Opt_grpid, "grpid"},
686 {Opt_grpid, "bsdgroups"},
687 {Opt_nogrpid, "nogrpid"},
688 {Opt_nogrpid, "sysvgroups"},
689 {Opt_resgid, "resgid=%u"},
690 {Opt_resuid, "resuid=%u"},
691 {Opt_sb, "sb=%u"},
692 {Opt_err_cont, "errors=continue"},
693 {Opt_err_panic, "errors=panic"},
694 {Opt_err_ro, "errors=remount-ro"},
695 {Opt_nouid32, "nouid32"},
696 {Opt_nocheck, "nocheck"},
697 {Opt_nocheck, "check=none"},
698 {Opt_debug, "debug"},
699 {Opt_oldalloc, "oldalloc"},
700 {Opt_orlov, "orlov"},
701 {Opt_user_xattr, "user_xattr"},
702 {Opt_nouser_xattr, "nouser_xattr"},
703 {Opt_acl, "acl"},
704 {Opt_noacl, "noacl"},
705 {Opt_reservation, "reservation"},
706 {Opt_noreservation, "noreservation"},
707 {Opt_noload, "noload"},
708 {Opt_nobh, "nobh"},
709 {Opt_bh, "bh"},
710 {Opt_commit, "commit=%u"},
711 {Opt_journal_update, "journal=update"},
712 {Opt_journal_inum, "journal=%u"},
713 {Opt_journal_dev, "journal_dev=%u"},
714 {Opt_abort, "abort"},
715 {Opt_data_journal, "data=journal"},
716 {Opt_data_ordered, "data=ordered"},
717 {Opt_data_writeback, "data=writeback"},
718 {Opt_offusrjquota, "usrjquota="},
719 {Opt_usrjquota, "usrjquota=%s"},
720 {Opt_offgrpjquota, "grpjquota="},
721 {Opt_grpjquota, "grpjquota=%s"},
722 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
723 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
724 {Opt_grpquota, "grpquota"},
725 {Opt_noquota, "noquota"},
726 {Opt_quota, "quota"},
727 {Opt_usrquota, "usrquota"},
728 {Opt_barrier, "barrier=%u"},
729 {Opt_err, NULL},
730 {Opt_resize, "resize"},
731 };
732
733 static ext3_fsblk_t get_sb_block(void **data)
734 {
735 ext3_fsblk_t sb_block;
736 char *options = (char *) *data;
737
738 if (!options || strncmp(options, "sb=", 3) != 0)
739 return 1; /* Default location */
740 options += 3;
741 /*todo: use simple_strtoll with >32bit ext3 */
742 sb_block = simple_strtoul(options, &options, 0);
743 if (*options && *options != ',') {
744 printk("EXT3-fs: Invalid sb specification: %s\n",
745 (char *) *data);
746 return 1;
747 }
748 if (*options == ',')
749 options++;
750 *data = (void *) options;
751 return sb_block;
752 }
753
754 static int parse_options (char *options, struct super_block *sb,
755 unsigned int *inum, unsigned long *journal_devnum,
756 ext3_fsblk_t *n_blocks_count, int is_remount)
757 {
758 struct ext3_sb_info *sbi = EXT3_SB(sb);
759 char * p;
760 substring_t args[MAX_OPT_ARGS];
761 int data_opt = 0;
762 int option;
763 #ifdef CONFIG_QUOTA
764 int qtype;
765 char *qname;
766 #endif
767
768 if (!options)
769 return 1;
770
771 while ((p = strsep (&options, ",")) != NULL) {
772 int token;
773 if (!*p)
774 continue;
775
776 token = match_token(p, tokens, args);
777 switch (token) {
778 case Opt_bsd_df:
779 clear_opt (sbi->s_mount_opt, MINIX_DF);
780 break;
781 case Opt_minix_df:
782 set_opt (sbi->s_mount_opt, MINIX_DF);
783 break;
784 case Opt_grpid:
785 set_opt (sbi->s_mount_opt, GRPID);
786 break;
787 case Opt_nogrpid:
788 clear_opt (sbi->s_mount_opt, GRPID);
789 break;
790 case Opt_resuid:
791 if (match_int(&args[0], &option))
792 return 0;
793 sbi->s_resuid = option;
794 break;
795 case Opt_resgid:
796 if (match_int(&args[0], &option))
797 return 0;
798 sbi->s_resgid = option;
799 break;
800 case Opt_sb:
801 /* handled by get_sb_block() instead of here */
802 /* *sb_block = match_int(&args[0]); */
803 break;
804 case Opt_err_panic:
805 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
806 clear_opt (sbi->s_mount_opt, ERRORS_RO);
807 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
808 break;
809 case Opt_err_ro:
810 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
811 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
812 set_opt (sbi->s_mount_opt, ERRORS_RO);
813 break;
814 case Opt_err_cont:
815 clear_opt (sbi->s_mount_opt, ERRORS_RO);
816 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
817 set_opt (sbi->s_mount_opt, ERRORS_CONT);
818 break;
819 case Opt_nouid32:
820 set_opt (sbi->s_mount_opt, NO_UID32);
821 break;
822 case Opt_nocheck:
823 clear_opt (sbi->s_mount_opt, CHECK);
824 break;
825 case Opt_debug:
826 set_opt (sbi->s_mount_opt, DEBUG);
827 break;
828 case Opt_oldalloc:
829 set_opt (sbi->s_mount_opt, OLDALLOC);
830 break;
831 case Opt_orlov:
832 clear_opt (sbi->s_mount_opt, OLDALLOC);
833 break;
834 #ifdef CONFIG_EXT3_FS_XATTR
835 case Opt_user_xattr:
836 set_opt (sbi->s_mount_opt, XATTR_USER);
837 break;
838 case Opt_nouser_xattr:
839 clear_opt (sbi->s_mount_opt, XATTR_USER);
840 break;
841 #else
842 case Opt_user_xattr:
843 case Opt_nouser_xattr:
844 printk("EXT3 (no)user_xattr options not supported\n");
845 break;
846 #endif
847 #ifdef CONFIG_EXT3_FS_POSIX_ACL
848 case Opt_acl:
849 set_opt(sbi->s_mount_opt, POSIX_ACL);
850 break;
851 case Opt_noacl:
852 clear_opt(sbi->s_mount_opt, POSIX_ACL);
853 break;
854 #else
855 case Opt_acl:
856 case Opt_noacl:
857 printk("EXT3 (no)acl options not supported\n");
858 break;
859 #endif
860 case Opt_reservation:
861 set_opt(sbi->s_mount_opt, RESERVATION);
862 break;
863 case Opt_noreservation:
864 clear_opt(sbi->s_mount_opt, RESERVATION);
865 break;
866 case Opt_journal_update:
867 /* @@@ FIXME */
868 /* Eventually we will want to be able to create
869 a journal file here. For now, only allow the
870 user to specify an existing inode to be the
871 journal file. */
872 if (is_remount) {
873 printk(KERN_ERR "EXT3-fs: cannot specify "
874 "journal on remount\n");
875 return 0;
876 }
877 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
878 break;
879 case Opt_journal_inum:
880 if (is_remount) {
881 printk(KERN_ERR "EXT3-fs: cannot specify "
882 "journal on remount\n");
883 return 0;
884 }
885 if (match_int(&args[0], &option))
886 return 0;
887 *inum = option;
888 break;
889 case Opt_journal_dev:
890 if (is_remount) {
891 printk(KERN_ERR "EXT3-fs: cannot specify "
892 "journal on remount\n");
893 return 0;
894 }
895 if (match_int(&args[0], &option))
896 return 0;
897 *journal_devnum = option;
898 break;
899 case Opt_noload:
900 set_opt (sbi->s_mount_opt, NOLOAD);
901 break;
902 case Opt_commit:
903 if (match_int(&args[0], &option))
904 return 0;
905 if (option < 0)
906 return 0;
907 if (option == 0)
908 option = JBD_DEFAULT_MAX_COMMIT_AGE;
909 sbi->s_commit_interval = HZ * option;
910 break;
911 case Opt_data_journal:
912 data_opt = EXT3_MOUNT_JOURNAL_DATA;
913 goto datacheck;
914 case Opt_data_ordered:
915 data_opt = EXT3_MOUNT_ORDERED_DATA;
916 goto datacheck;
917 case Opt_data_writeback:
918 data_opt = EXT3_MOUNT_WRITEBACK_DATA;
919 datacheck:
920 if (is_remount) {
921 if ((sbi->s_mount_opt & EXT3_MOUNT_DATA_FLAGS)
922 != data_opt) {
923 printk(KERN_ERR
924 "EXT3-fs: cannot change data "
925 "mode on remount\n");
926 return 0;
927 }
928 } else {
929 sbi->s_mount_opt &= ~EXT3_MOUNT_DATA_FLAGS;
930 sbi->s_mount_opt |= data_opt;
931 }
932 break;
933 #ifdef CONFIG_QUOTA
934 case Opt_usrjquota:
935 qtype = USRQUOTA;
936 goto set_qf_name;
937 case Opt_grpjquota:
938 qtype = GRPQUOTA;
939 set_qf_name:
940 if (sb_any_quota_enabled(sb)) {
941 printk(KERN_ERR
942 "EXT3-fs: Cannot change journalled "
943 "quota options when quota turned on.\n");
944 return 0;
945 }
946 qname = match_strdup(&args[0]);
947 if (!qname) {
948 printk(KERN_ERR
949 "EXT3-fs: not enough memory for "
950 "storing quotafile name.\n");
951 return 0;
952 }
953 if (sbi->s_qf_names[qtype] &&
954 strcmp(sbi->s_qf_names[qtype], qname)) {
955 printk(KERN_ERR
956 "EXT3-fs: %s quota file already "
957 "specified.\n", QTYPE2NAME(qtype));
958 kfree(qname);
959 return 0;
960 }
961 sbi->s_qf_names[qtype] = qname;
962 if (strchr(sbi->s_qf_names[qtype], '/')) {
963 printk(KERN_ERR
964 "EXT3-fs: quotafile must be on "
965 "filesystem root.\n");
966 kfree(sbi->s_qf_names[qtype]);
967 sbi->s_qf_names[qtype] = NULL;
968 return 0;
969 }
970 set_opt(sbi->s_mount_opt, QUOTA);
971 break;
972 case Opt_offusrjquota:
973 qtype = USRQUOTA;
974 goto clear_qf_name;
975 case Opt_offgrpjquota:
976 qtype = GRPQUOTA;
977 clear_qf_name:
978 if (sb_any_quota_enabled(sb)) {
979 printk(KERN_ERR "EXT3-fs: Cannot change "
980 "journalled quota options when "
981 "quota turned on.\n");
982 return 0;
983 }
984 /*
985 * The space will be released later when all options
986 * are confirmed to be correct
987 */
988 sbi->s_qf_names[qtype] = NULL;
989 break;
990 case Opt_jqfmt_vfsold:
991 sbi->s_jquota_fmt = QFMT_VFS_OLD;
992 break;
993 case Opt_jqfmt_vfsv0:
994 sbi->s_jquota_fmt = QFMT_VFS_V0;
995 break;
996 case Opt_quota:
997 case Opt_usrquota:
998 set_opt(sbi->s_mount_opt, QUOTA);
999 set_opt(sbi->s_mount_opt, USRQUOTA);
1000 break;
1001 case Opt_grpquota:
1002 set_opt(sbi->s_mount_opt, QUOTA);
1003 set_opt(sbi->s_mount_opt, GRPQUOTA);
1004 break;
1005 case Opt_noquota:
1006 if (sb_any_quota_enabled(sb)) {
1007 printk(KERN_ERR "EXT3-fs: Cannot change quota "
1008 "options when quota turned on.\n");
1009 return 0;
1010 }
1011 clear_opt(sbi->s_mount_opt, QUOTA);
1012 clear_opt(sbi->s_mount_opt, USRQUOTA);
1013 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1014 break;
1015 #else
1016 case Opt_quota:
1017 case Opt_usrquota:
1018 case Opt_grpquota:
1019 case Opt_usrjquota:
1020 case Opt_grpjquota:
1021 case Opt_offusrjquota:
1022 case Opt_offgrpjquota:
1023 case Opt_jqfmt_vfsold:
1024 case Opt_jqfmt_vfsv0:
1025 printk(KERN_ERR
1026 "EXT3-fs: journalled quota options not "
1027 "supported.\n");
1028 break;
1029 case Opt_noquota:
1030 break;
1031 #endif
1032 case Opt_abort:
1033 set_opt(sbi->s_mount_opt, ABORT);
1034 break;
1035 case Opt_barrier:
1036 if (match_int(&args[0], &option))
1037 return 0;
1038 if (option)
1039 set_opt(sbi->s_mount_opt, BARRIER);
1040 else
1041 clear_opt(sbi->s_mount_opt, BARRIER);
1042 break;
1043 case Opt_ignore:
1044 break;
1045 case Opt_resize:
1046 if (!is_remount) {
1047 printk("EXT3-fs: resize option only available "
1048 "for remount\n");
1049 return 0;
1050 }
1051 if (match_int(&args[0], &option) != 0)
1052 return 0;
1053 *n_blocks_count = option;
1054 break;
1055 case Opt_nobh:
1056 set_opt(sbi->s_mount_opt, NOBH);
1057 break;
1058 case Opt_bh:
1059 clear_opt(sbi->s_mount_opt, NOBH);
1060 break;
1061 default:
1062 printk (KERN_ERR
1063 "EXT3-fs: Unrecognized mount option \"%s\" "
1064 "or missing value\n", p);
1065 return 0;
1066 }
1067 }
1068 #ifdef CONFIG_QUOTA
1069 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1070 if ((sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA) &&
1071 sbi->s_qf_names[USRQUOTA])
1072 clear_opt(sbi->s_mount_opt, USRQUOTA);
1073
1074 if ((sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA) &&
1075 sbi->s_qf_names[GRPQUOTA])
1076 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1077
1078 if ((sbi->s_qf_names[USRQUOTA] &&
1079 (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)) ||
1080 (sbi->s_qf_names[GRPQUOTA] &&
1081 (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA))) {
1082 printk(KERN_ERR "EXT3-fs: old and new quota "
1083 "format mixing.\n");
1084 return 0;
1085 }
1086
1087 if (!sbi->s_jquota_fmt) {
1088 printk(KERN_ERR "EXT3-fs: journalled quota format "
1089 "not specified.\n");
1090 return 0;
1091 }
1092 } else {
1093 if (sbi->s_jquota_fmt) {
1094 printk(KERN_ERR "EXT3-fs: journalled quota format "
1095 "specified with no journalling "
1096 "enabled.\n");
1097 return 0;
1098 }
1099 }
1100 #endif
1101 return 1;
1102 }
1103
1104 static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
1105 int read_only)
1106 {
1107 struct ext3_sb_info *sbi = EXT3_SB(sb);
1108 int res = 0;
1109
1110 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
1111 printk (KERN_ERR "EXT3-fs warning: revision level too high, "
1112 "forcing read-only mode\n");
1113 res = MS_RDONLY;
1114 }
1115 if (read_only)
1116 return res;
1117 if (!(sbi->s_mount_state & EXT3_VALID_FS))
1118 printk (KERN_WARNING "EXT3-fs warning: mounting unchecked fs, "
1119 "running e2fsck is recommended\n");
1120 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
1121 printk (KERN_WARNING
1122 "EXT3-fs warning: mounting fs with errors, "
1123 "running e2fsck is recommended\n");
1124 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1125 le16_to_cpu(es->s_mnt_count) >=
1126 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1127 printk (KERN_WARNING
1128 "EXT3-fs warning: maximal mount count reached, "
1129 "running e2fsck is recommended\n");
1130 else if (le32_to_cpu(es->s_checkinterval) &&
1131 (le32_to_cpu(es->s_lastcheck) +
1132 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1133 printk (KERN_WARNING
1134 "EXT3-fs warning: checktime reached, "
1135 "running e2fsck is recommended\n");
1136 #if 0
1137 /* @@@ We _will_ want to clear the valid bit if we find
1138 inconsistencies, to force a fsck at reboot. But for
1139 a plain journaled filesystem we can keep it set as
1140 valid forever! :) */
1141 es->s_state = cpu_to_le16(le16_to_cpu(es->s_state) & ~EXT3_VALID_FS);
1142 #endif
1143 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1144 es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
1145 es->s_mnt_count=cpu_to_le16(le16_to_cpu(es->s_mnt_count) + 1);
1146 es->s_mtime = cpu_to_le32(get_seconds());
1147 ext3_update_dynamic_rev(sb);
1148 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1149
1150 ext3_commit_super(sb, es, 1);
1151 if (test_opt(sb, DEBUG))
1152 printk(KERN_INFO "[EXT3 FS bs=%lu, gc=%lu, "
1153 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1154 sb->s_blocksize,
1155 sbi->s_groups_count,
1156 EXT3_BLOCKS_PER_GROUP(sb),
1157 EXT3_INODES_PER_GROUP(sb),
1158 sbi->s_mount_opt);
1159
1160 printk(KERN_INFO "EXT3 FS on %s, ", sb->s_id);
1161 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
1162 char b[BDEVNAME_SIZE];
1163
1164 printk("external journal on %s\n",
1165 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1166 } else {
1167 printk("internal journal\n");
1168 }
1169 return res;
1170 }
1171
1172 /* Called at mount-time, super-block is locked */
1173 static int ext3_check_descriptors (struct super_block * sb)
1174 {
1175 struct ext3_sb_info *sbi = EXT3_SB(sb);
1176 ext3_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1177 ext3_fsblk_t last_block;
1178 struct ext3_group_desc * gdp = NULL;
1179 int desc_block = 0;
1180 int i;
1181
1182 ext3_debug ("Checking group descriptors");
1183
1184 for (i = 0; i < sbi->s_groups_count; i++)
1185 {
1186 if (i == sbi->s_groups_count - 1)
1187 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
1188 else
1189 last_block = first_block +
1190 (EXT3_BLOCKS_PER_GROUP(sb) - 1);
1191
1192 if ((i % EXT3_DESC_PER_BLOCK(sb)) == 0)
1193 gdp = (struct ext3_group_desc *)
1194 sbi->s_group_desc[desc_block++]->b_data;
1195 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
1196 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
1197 {
1198 ext3_error (sb, "ext3_check_descriptors",
1199 "Block bitmap for group %d"
1200 " not in group (block %lu)!",
1201 i, (unsigned long)
1202 le32_to_cpu(gdp->bg_block_bitmap));
1203 return 0;
1204 }
1205 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
1206 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
1207 {
1208 ext3_error (sb, "ext3_check_descriptors",
1209 "Inode bitmap for group %d"
1210 " not in group (block %lu)!",
1211 i, (unsigned long)
1212 le32_to_cpu(gdp->bg_inode_bitmap));
1213 return 0;
1214 }
1215 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
1216 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group >
1217 last_block)
1218 {
1219 ext3_error (sb, "ext3_check_descriptors",
1220 "Inode table for group %d"
1221 " not in group (block %lu)!",
1222 i, (unsigned long)
1223 le32_to_cpu(gdp->bg_inode_table));
1224 return 0;
1225 }
1226 first_block += EXT3_BLOCKS_PER_GROUP(sb);
1227 gdp++;
1228 }
1229
1230 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1231 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1232 return 1;
1233 }
1234
1235
1236 /* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1237 * the superblock) which were deleted from all directories, but held open by
1238 * a process at the time of a crash. We walk the list and try to delete these
1239 * inodes at recovery time (only with a read-write filesystem).
1240 *
1241 * In order to keep the orphan inode chain consistent during traversal (in
1242 * case of crash during recovery), we link each inode into the superblock
1243 * orphan list_head and handle it the same way as an inode deletion during
1244 * normal operation (which journals the operations for us).
1245 *
1246 * We only do an iget() and an iput() on each inode, which is very safe if we
1247 * accidentally point at an in-use or already deleted inode. The worst that
1248 * can happen in this case is that we get a "bit already cleared" message from
1249 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1250 * e2fsck was run on this filesystem, and it must have already done the orphan
1251 * inode cleanup for us, so we can safely abort without any further action.
1252 */
1253 static void ext3_orphan_cleanup (struct super_block * sb,
1254 struct ext3_super_block * es)
1255 {
1256 unsigned int s_flags = sb->s_flags;
1257 int nr_orphans = 0, nr_truncates = 0;
1258 #ifdef CONFIG_QUOTA
1259 int i;
1260 #endif
1261 if (!es->s_last_orphan) {
1262 jbd_debug(4, "no orphan inodes to clean up\n");
1263 return;
1264 }
1265
1266 if (bdev_read_only(sb->s_bdev)) {
1267 printk(KERN_ERR "EXT3-fs: write access "
1268 "unavailable, skipping orphan cleanup.\n");
1269 return;
1270 }
1271
1272 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1273 if (es->s_last_orphan)
1274 jbd_debug(1, "Errors on filesystem, "
1275 "clearing orphan list.\n");
1276 es->s_last_orphan = 0;
1277 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1278 return;
1279 }
1280
1281 if (s_flags & MS_RDONLY) {
1282 printk(KERN_INFO "EXT3-fs: %s: orphan cleanup on readonly fs\n",
1283 sb->s_id);
1284 sb->s_flags &= ~MS_RDONLY;
1285 }
1286 #ifdef CONFIG_QUOTA
1287 /* Needed for iput() to work correctly and not trash data */
1288 sb->s_flags |= MS_ACTIVE;
1289 /* Turn on quotas so that they are updated correctly */
1290 for (i = 0; i < MAXQUOTAS; i++) {
1291 if (EXT3_SB(sb)->s_qf_names[i]) {
1292 int ret = ext3_quota_on_mount(sb, i);
1293 if (ret < 0)
1294 printk(KERN_ERR
1295 "EXT3-fs: Cannot turn on journalled "
1296 "quota: error %d\n", ret);
1297 }
1298 }
1299 #endif
1300
1301 while (es->s_last_orphan) {
1302 struct inode *inode;
1303
1304 if (!(inode =
1305 ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan)))) {
1306 es->s_last_orphan = 0;
1307 break;
1308 }
1309
1310 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
1311 DQUOT_INIT(inode);
1312 if (inode->i_nlink) {
1313 printk(KERN_DEBUG
1314 "%s: truncating inode %lu to %Ld bytes\n",
1315 __FUNCTION__, inode->i_ino, inode->i_size);
1316 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1317 inode->i_ino, inode->i_size);
1318 ext3_truncate(inode);
1319 nr_truncates++;
1320 } else {
1321 printk(KERN_DEBUG
1322 "%s: deleting unreferenced inode %lu\n",
1323 __FUNCTION__, inode->i_ino);
1324 jbd_debug(2, "deleting unreferenced inode %lu\n",
1325 inode->i_ino);
1326 nr_orphans++;
1327 }
1328 iput(inode); /* The delete magic happens here! */
1329 }
1330
1331 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1332
1333 if (nr_orphans)
1334 printk(KERN_INFO "EXT3-fs: %s: %d orphan inode%s deleted\n",
1335 sb->s_id, PLURAL(nr_orphans));
1336 if (nr_truncates)
1337 printk(KERN_INFO "EXT3-fs: %s: %d truncate%s cleaned up\n",
1338 sb->s_id, PLURAL(nr_truncates));
1339 #ifdef CONFIG_QUOTA
1340 /* Turn quotas off */
1341 for (i = 0; i < MAXQUOTAS; i++) {
1342 if (sb_dqopt(sb)->files[i])
1343 vfs_quota_off(sb, i);
1344 }
1345 #endif
1346 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1347 }
1348
1349 /*
1350 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1351 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1352 * We need to be 1 filesystem block less than the 2^32 sector limit.
1353 */
1354 static loff_t ext3_max_size(int bits)
1355 {
1356 loff_t res = EXT3_NDIR_BLOCKS;
1357 /* This constant is calculated to be the largest file size for a
1358 * dense, 4k-blocksize file such that the total number of
1359 * sectors in the file, including data and all indirect blocks,
1360 * does not exceed 2^32. */
1361 const loff_t upper_limit = 0x1ff7fffd000LL;
1362
1363 res += 1LL << (bits-2);
1364 res += 1LL << (2*(bits-2));
1365 res += 1LL << (3*(bits-2));
1366 res <<= bits;
1367 if (res > upper_limit)
1368 res = upper_limit;
1369 return res;
1370 }
1371
1372 static ext3_fsblk_t descriptor_loc(struct super_block *sb,
1373 ext3_fsblk_t logic_sb_block,
1374 int nr)
1375 {
1376 struct ext3_sb_info *sbi = EXT3_SB(sb);
1377 unsigned long bg, first_meta_bg;
1378 int has_super = 0;
1379
1380 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1381
1382 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1383 nr < first_meta_bg)
1384 return (logic_sb_block + nr + 1);
1385 bg = sbi->s_desc_per_block * nr;
1386 if (ext3_bg_has_super(sb, bg))
1387 has_super = 1;
1388 return (has_super + ext3_group_first_block_no(sb, bg));
1389 }
1390
1391
1392 static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1393 {
1394 struct buffer_head * bh;
1395 struct ext3_super_block *es = NULL;
1396 struct ext3_sb_info *sbi;
1397 ext3_fsblk_t block;
1398 ext3_fsblk_t sb_block = get_sb_block(&data);
1399 ext3_fsblk_t logic_sb_block;
1400 unsigned long offset = 0;
1401 unsigned int journal_inum = 0;
1402 unsigned long journal_devnum = 0;
1403 unsigned long def_mount_opts;
1404 struct inode *root;
1405 int blocksize;
1406 int hblock;
1407 int db_count;
1408 int i;
1409 int needs_recovery;
1410 __le32 features;
1411
1412 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1413 if (!sbi)
1414 return -ENOMEM;
1415 sb->s_fs_info = sbi;
1416 sbi->s_mount_opt = 0;
1417 sbi->s_resuid = EXT3_DEF_RESUID;
1418 sbi->s_resgid = EXT3_DEF_RESGID;
1419
1420 unlock_kernel();
1421
1422 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1423 if (!blocksize) {
1424 printk(KERN_ERR "EXT3-fs: unable to set blocksize\n");
1425 goto out_fail;
1426 }
1427
1428 /*
1429 * The ext3 superblock will not be buffer aligned for other than 1kB
1430 * block sizes. We need to calculate the offset from buffer start.
1431 */
1432 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1433 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1434 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1435 } else {
1436 logic_sb_block = sb_block;
1437 }
1438
1439 if (!(bh = sb_bread(sb, logic_sb_block))) {
1440 printk (KERN_ERR "EXT3-fs: unable to read superblock\n");
1441 goto out_fail;
1442 }
1443 /*
1444 * Note: s_es must be initialized as soon as possible because
1445 * some ext3 macro-instructions depend on its value
1446 */
1447 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1448 sbi->s_es = es;
1449 sb->s_magic = le16_to_cpu(es->s_magic);
1450 if (sb->s_magic != EXT3_SUPER_MAGIC)
1451 goto cantfind_ext3;
1452
1453 /* Set defaults before we parse the mount options */
1454 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1455 if (def_mount_opts & EXT3_DEFM_DEBUG)
1456 set_opt(sbi->s_mount_opt, DEBUG);
1457 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1458 set_opt(sbi->s_mount_opt, GRPID);
1459 if (def_mount_opts & EXT3_DEFM_UID16)
1460 set_opt(sbi->s_mount_opt, NO_UID32);
1461 #ifdef CONFIG_EXT3_FS_XATTR
1462 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1463 set_opt(sbi->s_mount_opt, XATTR_USER);
1464 #endif
1465 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1466 if (def_mount_opts & EXT3_DEFM_ACL)
1467 set_opt(sbi->s_mount_opt, POSIX_ACL);
1468 #endif
1469 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
1470 sbi->s_mount_opt |= EXT3_MOUNT_JOURNAL_DATA;
1471 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
1472 sbi->s_mount_opt |= EXT3_MOUNT_ORDERED_DATA;
1473 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
1474 sbi->s_mount_opt |= EXT3_MOUNT_WRITEBACK_DATA;
1475
1476 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1477 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1478 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_RO)
1479 set_opt(sbi->s_mount_opt, ERRORS_RO);
1480 else
1481 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1482
1483 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1484 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1485
1486 set_opt(sbi->s_mount_opt, RESERVATION);
1487
1488 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1489 NULL, 0))
1490 goto failed_mount;
1491
1492 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1493 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1494
1495 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1496 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1497 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1498 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1499 printk(KERN_WARNING
1500 "EXT3-fs warning: feature flags set on rev 0 fs, "
1501 "running e2fsck is recommended\n");
1502 /*
1503 * Check feature flags regardless of the revision level, since we
1504 * previously didn't change the revision level when setting the flags,
1505 * so there is a chance incompat flags are set on a rev 0 filesystem.
1506 */
1507 features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
1508 if (features) {
1509 printk(KERN_ERR "EXT3-fs: %s: couldn't mount because of "
1510 "unsupported optional features (%x).\n",
1511 sb->s_id, le32_to_cpu(features));
1512 goto failed_mount;
1513 }
1514 features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
1515 if (!(sb->s_flags & MS_RDONLY) && features) {
1516 printk(KERN_ERR "EXT3-fs: %s: couldn't mount RDWR because of "
1517 "unsupported optional features (%x).\n",
1518 sb->s_id, le32_to_cpu(features));
1519 goto failed_mount;
1520 }
1521 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1522
1523 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1524 blocksize > EXT3_MAX_BLOCK_SIZE) {
1525 printk(KERN_ERR
1526 "EXT3-fs: Unsupported filesystem blocksize %d on %s.\n",
1527 blocksize, sb->s_id);
1528 goto failed_mount;
1529 }
1530
1531 hblock = bdev_hardsect_size(sb->s_bdev);
1532 if (sb->s_blocksize != blocksize) {
1533 /*
1534 * Make sure the blocksize for the filesystem is larger
1535 * than the hardware sectorsize for the machine.
1536 */
1537 if (blocksize < hblock) {
1538 printk(KERN_ERR "EXT3-fs: blocksize %d too small for "
1539 "device blocksize %d.\n", blocksize, hblock);
1540 goto failed_mount;
1541 }
1542
1543 brelse (bh);
1544 sb_set_blocksize(sb, blocksize);
1545 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1546 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1547 bh = sb_bread(sb, logic_sb_block);
1548 if (!bh) {
1549 printk(KERN_ERR
1550 "EXT3-fs: Can't read superblock on 2nd try.\n");
1551 goto failed_mount;
1552 }
1553 es = (struct ext3_super_block *)(((char *)bh->b_data) + offset);
1554 sbi->s_es = es;
1555 if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
1556 printk (KERN_ERR
1557 "EXT3-fs: Magic mismatch, very weird !\n");
1558 goto failed_mount;
1559 }
1560 }
1561
1562 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1563
1564 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1565 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1566 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1567 } else {
1568 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1569 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1570 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
1571 (sbi->s_inode_size & (sbi->s_inode_size - 1)) ||
1572 (sbi->s_inode_size > blocksize)) {
1573 printk (KERN_ERR
1574 "EXT3-fs: unsupported inode size: %d\n",
1575 sbi->s_inode_size);
1576 goto failed_mount;
1577 }
1578 }
1579 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1580 le32_to_cpu(es->s_log_frag_size);
1581 if (blocksize != sbi->s_frag_size) {
1582 printk(KERN_ERR
1583 "EXT3-fs: fragsize %lu != blocksize %u (unsupported)\n",
1584 sbi->s_frag_size, blocksize);
1585 goto failed_mount;
1586 }
1587 sbi->s_frags_per_block = 1;
1588 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1589 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1590 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1591 if (EXT3_INODE_SIZE(sb) == 0)
1592 goto cantfind_ext3;
1593 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1594 if (sbi->s_inodes_per_block == 0)
1595 goto cantfind_ext3;
1596 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1597 sbi->s_inodes_per_block;
1598 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1599 sbi->s_sbh = bh;
1600 sbi->s_mount_state = le16_to_cpu(es->s_state);
1601 sbi->s_addr_per_block_bits = ilog2(EXT3_ADDR_PER_BLOCK(sb));
1602 sbi->s_desc_per_block_bits = ilog2(EXT3_DESC_PER_BLOCK(sb));
1603 for (i=0; i < 4; i++)
1604 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1605 sbi->s_def_hash_version = es->s_def_hash_version;
1606
1607 if (sbi->s_blocks_per_group > blocksize * 8) {
1608 printk (KERN_ERR
1609 "EXT3-fs: #blocks per group too big: %lu\n",
1610 sbi->s_blocks_per_group);
1611 goto failed_mount;
1612 }
1613 if (sbi->s_frags_per_group > blocksize * 8) {
1614 printk (KERN_ERR
1615 "EXT3-fs: #fragments per group too big: %lu\n",
1616 sbi->s_frags_per_group);
1617 goto failed_mount;
1618 }
1619 if (sbi->s_inodes_per_group > blocksize * 8) {
1620 printk (KERN_ERR
1621 "EXT3-fs: #inodes per group too big: %lu\n",
1622 sbi->s_inodes_per_group);
1623 goto failed_mount;
1624 }
1625
1626 if (le32_to_cpu(es->s_blocks_count) >
1627 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
1628 printk(KERN_ERR "EXT3-fs: filesystem on %s:"
1629 " too large to mount safely\n", sb->s_id);
1630 if (sizeof(sector_t) < 8)
1631 printk(KERN_WARNING "EXT3-fs: CONFIG_LBD not "
1632 "enabled\n");
1633 goto failed_mount;
1634 }
1635
1636 if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1637 goto cantfind_ext3;
1638 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1639 le32_to_cpu(es->s_first_data_block) - 1)
1640 / EXT3_BLOCKS_PER_GROUP(sb)) + 1;
1641 db_count = (sbi->s_groups_count + EXT3_DESC_PER_BLOCK(sb) - 1) /
1642 EXT3_DESC_PER_BLOCK(sb);
1643 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1644 GFP_KERNEL);
1645 if (sbi->s_group_desc == NULL) {
1646 printk (KERN_ERR "EXT3-fs: not enough memory\n");
1647 goto failed_mount;
1648 }
1649
1650 bgl_lock_init(&sbi->s_blockgroup_lock);
1651
1652 for (i = 0; i < db_count; i++) {
1653 block = descriptor_loc(sb, logic_sb_block, i);
1654 sbi->s_group_desc[i] = sb_bread(sb, block);
1655 if (!sbi->s_group_desc[i]) {
1656 printk (KERN_ERR "EXT3-fs: "
1657 "can't read group descriptor %d\n", i);
1658 db_count = i;
1659 goto failed_mount2;
1660 }
1661 }
1662 if (!ext3_check_descriptors (sb)) {
1663 printk(KERN_ERR "EXT3-fs: group descriptors corrupted!\n");
1664 goto failed_mount2;
1665 }
1666 sbi->s_gdb_count = db_count;
1667 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1668 spin_lock_init(&sbi->s_next_gen_lock);
1669
1670 percpu_counter_init(&sbi->s_freeblocks_counter,
1671 ext3_count_free_blocks(sb));
1672 percpu_counter_init(&sbi->s_freeinodes_counter,
1673 ext3_count_free_inodes(sb));
1674 percpu_counter_init(&sbi->s_dirs_counter,
1675 ext3_count_dirs(sb));
1676
1677 /* per fileystem reservation list head & lock */
1678 spin_lock_init(&sbi->s_rsv_window_lock);
1679 sbi->s_rsv_window_root = RB_ROOT;
1680 /* Add a single, static dummy reservation to the start of the
1681 * reservation window list --- it gives us a placeholder for
1682 * append-at-start-of-list which makes the allocation logic
1683 * _much_ simpler. */
1684 sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1685 sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1686 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1687 sbi->s_rsv_window_head.rsv_goal_size = 0;
1688 ext3_rsv_window_add(sb, &sbi->s_rsv_window_head);
1689
1690 /*
1691 * set up enough so that it can read an inode
1692 */
1693 sb->s_op = &ext3_sops;
1694 sb->s_export_op = &ext3_export_ops;
1695 sb->s_xattr = ext3_xattr_handlers;
1696 #ifdef CONFIG_QUOTA
1697 sb->s_qcop = &ext3_qctl_operations;
1698 sb->dq_op = &ext3_quota_operations;
1699 #endif
1700 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1701
1702 sb->s_root = NULL;
1703
1704 needs_recovery = (es->s_last_orphan != 0 ||
1705 EXT3_HAS_INCOMPAT_FEATURE(sb,
1706 EXT3_FEATURE_INCOMPAT_RECOVER));
1707
1708 /*
1709 * The first inode we look at is the journal inode. Don't try
1710 * root first: it may be modified in the journal!
1711 */
1712 if (!test_opt(sb, NOLOAD) &&
1713 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
1714 if (ext3_load_journal(sb, es, journal_devnum))
1715 goto failed_mount3;
1716 } else if (journal_inum) {
1717 if (ext3_create_journal(sb, es, journal_inum))
1718 goto failed_mount3;
1719 } else {
1720 if (!silent)
1721 printk (KERN_ERR
1722 "ext3: No journal on filesystem on %s\n",
1723 sb->s_id);
1724 goto failed_mount3;
1725 }
1726
1727 /* We have now updated the journal if required, so we can
1728 * validate the data journaling mode. */
1729 switch (test_opt(sb, DATA_FLAGS)) {
1730 case 0:
1731 /* No mode set, assume a default based on the journal
1732 capabilities: ORDERED_DATA if the journal can
1733 cope, else JOURNAL_DATA */
1734 if (journal_check_available_features
1735 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
1736 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1737 else
1738 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1739 break;
1740
1741 case EXT3_MOUNT_ORDERED_DATA:
1742 case EXT3_MOUNT_WRITEBACK_DATA:
1743 if (!journal_check_available_features
1744 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
1745 printk(KERN_ERR "EXT3-fs: Journal does not support "
1746 "requested data journaling mode\n");
1747 goto failed_mount4;
1748 }
1749 default:
1750 break;
1751 }
1752
1753 if (test_opt(sb, NOBH)) {
1754 if (!(test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)) {
1755 printk(KERN_WARNING "EXT3-fs: Ignoring nobh option - "
1756 "its supported only with writeback mode\n");
1757 clear_opt(sbi->s_mount_opt, NOBH);
1758 }
1759 }
1760 /*
1761 * The journal_load will have done any necessary log recovery,
1762 * so we can safely mount the rest of the filesystem now.
1763 */
1764
1765 root = iget(sb, EXT3_ROOT_INO);
1766 sb->s_root = d_alloc_root(root);
1767 if (!sb->s_root) {
1768 printk(KERN_ERR "EXT3-fs: get root inode failed\n");
1769 iput(root);
1770 goto failed_mount4;
1771 }
1772 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1773 dput(sb->s_root);
1774 sb->s_root = NULL;
1775 printk(KERN_ERR "EXT3-fs: corrupt root inode, run e2fsck\n");
1776 goto failed_mount4;
1777 }
1778
1779 ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1780 /*
1781 * akpm: core read_super() calls in here with the superblock locked.
1782 * That deadlocks, because orphan cleanup needs to lock the superblock
1783 * in numerous places. Here we just pop the lock - it's relatively
1784 * harmless, because we are now ready to accept write_super() requests,
1785 * and aviro says that's the only reason for hanging onto the
1786 * superblock lock.
1787 */
1788 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
1789 ext3_orphan_cleanup(sb, es);
1790 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
1791 if (needs_recovery)
1792 printk (KERN_INFO "EXT3-fs: recovery complete.\n");
1793 ext3_mark_recovery_complete(sb, es);
1794 printk (KERN_INFO "EXT3-fs: mounted filesystem with %s data mode.\n",
1795 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
1796 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
1797 "writeback");
1798
1799 lock_kernel();
1800 return 0;
1801
1802 cantfind_ext3:
1803 if (!silent)
1804 printk(KERN_ERR "VFS: Can't find ext3 filesystem on dev %s.\n",
1805 sb->s_id);
1806 goto failed_mount;
1807
1808 failed_mount4:
1809 journal_destroy(sbi->s_journal);
1810 failed_mount3:
1811 percpu_counter_destroy(&sbi->s_freeblocks_counter);
1812 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1813 percpu_counter_destroy(&sbi->s_dirs_counter);
1814 failed_mount2:
1815 for (i = 0; i < db_count; i++)
1816 brelse(sbi->s_group_desc[i]);
1817 kfree(sbi->s_group_desc);
1818 failed_mount:
1819 #ifdef CONFIG_QUOTA
1820 for (i = 0; i < MAXQUOTAS; i++)
1821 kfree(sbi->s_qf_names[i]);
1822 #endif
1823 ext3_blkdev_remove(sbi);
1824 brelse(bh);
1825 out_fail:
1826 sb->s_fs_info = NULL;
1827 kfree(sbi);
1828 lock_kernel();
1829 return -EINVAL;
1830 }
1831
1832 /*
1833 * Setup any per-fs journal parameters now. We'll do this both on
1834 * initial mount, once the journal has been initialised but before we've
1835 * done any recovery; and again on any subsequent remount.
1836 */
1837 static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
1838 {
1839 struct ext3_sb_info *sbi = EXT3_SB(sb);
1840
1841 if (sbi->s_commit_interval)
1842 journal->j_commit_interval = sbi->s_commit_interval;
1843 /* We could also set up an ext3-specific default for the commit
1844 * interval here, but for now we'll just fall back to the jbd
1845 * default. */
1846
1847 spin_lock(&journal->j_state_lock);
1848 if (test_opt(sb, BARRIER))
1849 journal->j_flags |= JFS_BARRIER;
1850 else
1851 journal->j_flags &= ~JFS_BARRIER;
1852 spin_unlock(&journal->j_state_lock);
1853 }
1854
1855 static journal_t *ext3_get_journal(struct super_block *sb,
1856 unsigned int journal_inum)
1857 {
1858 struct inode *journal_inode;
1859 journal_t *journal;
1860
1861 /* First, test for the existence of a valid inode on disk. Bad
1862 * things happen if we iget() an unused inode, as the subsequent
1863 * iput() will try to delete it. */
1864
1865 journal_inode = iget(sb, journal_inum);
1866 if (!journal_inode) {
1867 printk(KERN_ERR "EXT3-fs: no journal found.\n");
1868 return NULL;
1869 }
1870 if (!journal_inode->i_nlink) {
1871 make_bad_inode(journal_inode);
1872 iput(journal_inode);
1873 printk(KERN_ERR "EXT3-fs: journal inode is deleted.\n");
1874 return NULL;
1875 }
1876
1877 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
1878 journal_inode, journal_inode->i_size);
1879 if (is_bad_inode(journal_inode) || !S_ISREG(journal_inode->i_mode)) {
1880 printk(KERN_ERR "EXT3-fs: invalid journal inode.\n");
1881 iput(journal_inode);
1882 return NULL;
1883 }
1884
1885 journal = journal_init_inode(journal_inode);
1886 if (!journal) {
1887 printk(KERN_ERR "EXT3-fs: Could not load journal inode\n");
1888 iput(journal_inode);
1889 return NULL;
1890 }
1891 journal->j_private = sb;
1892 ext3_init_journal_params(sb, journal);
1893 return journal;
1894 }
1895
1896 static journal_t *ext3_get_dev_journal(struct super_block *sb,
1897 dev_t j_dev)
1898 {
1899 struct buffer_head * bh;
1900 journal_t *journal;
1901 ext3_fsblk_t start;
1902 ext3_fsblk_t len;
1903 int hblock, blocksize;
1904 ext3_fsblk_t sb_block;
1905 unsigned long offset;
1906 struct ext3_super_block * es;
1907 struct block_device *bdev;
1908
1909 bdev = ext3_blkdev_get(j_dev);
1910 if (bdev == NULL)
1911 return NULL;
1912
1913 if (bd_claim(bdev, sb)) {
1914 printk(KERN_ERR
1915 "EXT3: failed to claim external journal device.\n");
1916 blkdev_put(bdev);
1917 return NULL;
1918 }
1919
1920 blocksize = sb->s_blocksize;
1921 hblock = bdev_hardsect_size(bdev);
1922 if (blocksize < hblock) {
1923 printk(KERN_ERR
1924 "EXT3-fs: blocksize too small for journal device.\n");
1925 goto out_bdev;
1926 }
1927
1928 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
1929 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
1930 set_blocksize(bdev, blocksize);
1931 if (!(bh = __bread(bdev, sb_block, blocksize))) {
1932 printk(KERN_ERR "EXT3-fs: couldn't read superblock of "
1933 "external journal\n");
1934 goto out_bdev;
1935 }
1936
1937 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1938 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
1939 !(le32_to_cpu(es->s_feature_incompat) &
1940 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
1941 printk(KERN_ERR "EXT3-fs: external journal has "
1942 "bad superblock\n");
1943 brelse(bh);
1944 goto out_bdev;
1945 }
1946
1947 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
1948 printk(KERN_ERR "EXT3-fs: journal UUID does not match\n");
1949 brelse(bh);
1950 goto out_bdev;
1951 }
1952
1953 len = le32_to_cpu(es->s_blocks_count);
1954 start = sb_block + 1;
1955 brelse(bh); /* we're done with the superblock */
1956
1957 journal = journal_init_dev(bdev, sb->s_bdev,
1958 start, len, blocksize);
1959 if (!journal) {
1960 printk(KERN_ERR "EXT3-fs: failed to create device journal\n");
1961 goto out_bdev;
1962 }
1963 journal->j_private = sb;
1964 ll_rw_block(READ, 1, &journal->j_sb_buffer);
1965 wait_on_buffer(journal->j_sb_buffer);
1966 if (!buffer_uptodate(journal->j_sb_buffer)) {
1967 printk(KERN_ERR "EXT3-fs: I/O error on journal device\n");
1968 goto out_journal;
1969 }
1970 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
1971 printk(KERN_ERR "EXT3-fs: External journal has more than one "
1972 "user (unsupported) - %d\n",
1973 be32_to_cpu(journal->j_superblock->s_nr_users));
1974 goto out_journal;
1975 }
1976 EXT3_SB(sb)->journal_bdev = bdev;
1977 ext3_init_journal_params(sb, journal);
1978 return journal;
1979 out_journal:
1980 journal_destroy(journal);
1981 out_bdev:
1982 ext3_blkdev_put(bdev);
1983 return NULL;
1984 }
1985
1986 static int ext3_load_journal(struct super_block *sb,
1987 struct ext3_super_block *es,
1988 unsigned long journal_devnum)
1989 {
1990 journal_t *journal;
1991 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
1992 dev_t journal_dev;
1993 int err = 0;
1994 int really_read_only;
1995
1996 if (journal_devnum &&
1997 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
1998 printk(KERN_INFO "EXT3-fs: external journal device major/minor "
1999 "numbers have changed\n");
2000 journal_dev = new_decode_dev(journal_devnum);
2001 } else
2002 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2003
2004 really_read_only = bdev_read_only(sb->s_bdev);
2005
2006 /*
2007 * Are we loading a blank journal or performing recovery after a
2008 * crash? For recovery, we need to check in advance whether we
2009 * can get read-write access to the device.
2010 */
2011
2012 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
2013 if (sb->s_flags & MS_RDONLY) {
2014 printk(KERN_INFO "EXT3-fs: INFO: recovery "
2015 "required on readonly filesystem.\n");
2016 if (really_read_only) {
2017 printk(KERN_ERR "EXT3-fs: write access "
2018 "unavailable, cannot proceed.\n");
2019 return -EROFS;
2020 }
2021 printk (KERN_INFO "EXT3-fs: write access will "
2022 "be enabled during recovery.\n");
2023 }
2024 }
2025
2026 if (journal_inum && journal_dev) {
2027 printk(KERN_ERR "EXT3-fs: filesystem has both journal "
2028 "and inode journals!\n");
2029 return -EINVAL;
2030 }
2031
2032 if (journal_inum) {
2033 if (!(journal = ext3_get_journal(sb, journal_inum)))
2034 return -EINVAL;
2035 } else {
2036 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
2037 return -EINVAL;
2038 }
2039
2040 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2041 err = journal_update_format(journal);
2042 if (err) {
2043 printk(KERN_ERR "EXT3-fs: error updating journal.\n");
2044 journal_destroy(journal);
2045 return err;
2046 }
2047 }
2048
2049 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
2050 err = journal_wipe(journal, !really_read_only);
2051 if (!err)
2052 err = journal_load(journal);
2053
2054 if (err) {
2055 printk(KERN_ERR "EXT3-fs: error loading journal.\n");
2056 journal_destroy(journal);
2057 return err;
2058 }
2059
2060 EXT3_SB(sb)->s_journal = journal;
2061 ext3_clear_journal_err(sb, es);
2062
2063 if (journal_devnum &&
2064 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2065 es->s_journal_dev = cpu_to_le32(journal_devnum);
2066 sb->s_dirt = 1;
2067
2068 /* Make sure we flush the recovery flag to disk. */
2069 ext3_commit_super(sb, es, 1);
2070 }
2071
2072 return 0;
2073 }
2074
2075 static int ext3_create_journal(struct super_block * sb,
2076 struct ext3_super_block * es,
2077 unsigned int journal_inum)
2078 {
2079 journal_t *journal;
2080
2081 if (sb->s_flags & MS_RDONLY) {
2082 printk(KERN_ERR "EXT3-fs: readonly filesystem when trying to "
2083 "create journal.\n");
2084 return -EROFS;
2085 }
2086
2087 if (!(journal = ext3_get_journal(sb, journal_inum)))
2088 return -EINVAL;
2089
2090 printk(KERN_INFO "EXT3-fs: creating new journal on inode %u\n",
2091 journal_inum);
2092
2093 if (journal_create(journal)) {
2094 printk(KERN_ERR "EXT3-fs: error creating journal.\n");
2095 journal_destroy(journal);
2096 return -EIO;
2097 }
2098
2099 EXT3_SB(sb)->s_journal = journal;
2100
2101 ext3_update_dynamic_rev(sb);
2102 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2103 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
2104
2105 es->s_journal_inum = cpu_to_le32(journal_inum);
2106 sb->s_dirt = 1;
2107
2108 /* Make sure we flush the recovery flag to disk. */
2109 ext3_commit_super(sb, es, 1);
2110
2111 return 0;
2112 }
2113
2114 static void ext3_commit_super (struct super_block * sb,
2115 struct ext3_super_block * es,
2116 int sync)
2117 {
2118 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
2119
2120 if (!sbh)
2121 return;
2122 es->s_wtime = cpu_to_le32(get_seconds());
2123 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
2124 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
2125 BUFFER_TRACE(sbh, "marking dirty");
2126 mark_buffer_dirty(sbh);
2127 if (sync)
2128 sync_dirty_buffer(sbh);
2129 }
2130
2131
2132 /*
2133 * Have we just finished recovery? If so, and if we are mounting (or
2134 * remounting) the filesystem readonly, then we will end up with a
2135 * consistent fs on disk. Record that fact.
2136 */
2137 static void ext3_mark_recovery_complete(struct super_block * sb,
2138 struct ext3_super_block * es)
2139 {
2140 journal_t *journal = EXT3_SB(sb)->s_journal;
2141
2142 journal_lock_updates(journal);
2143 journal_flush(journal);
2144 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
2145 sb->s_flags & MS_RDONLY) {
2146 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2147 sb->s_dirt = 0;
2148 ext3_commit_super(sb, es, 1);
2149 }
2150 journal_unlock_updates(journal);
2151 }
2152
2153 /*
2154 * If we are mounting (or read-write remounting) a filesystem whose journal
2155 * has recorded an error from a previous lifetime, move that error to the
2156 * main filesystem now.
2157 */
2158 static void ext3_clear_journal_err(struct super_block * sb,
2159 struct ext3_super_block * es)
2160 {
2161 journal_t *journal;
2162 int j_errno;
2163 const char *errstr;
2164
2165 journal = EXT3_SB(sb)->s_journal;
2166
2167 /*
2168 * Now check for any error status which may have been recorded in the
2169 * journal by a prior ext3_error() or ext3_abort()
2170 */
2171
2172 j_errno = journal_errno(journal);
2173 if (j_errno) {
2174 char nbuf[16];
2175
2176 errstr = ext3_decode_error(sb, j_errno, nbuf);
2177 ext3_warning(sb, __FUNCTION__, "Filesystem error recorded "
2178 "from previous mount: %s", errstr);
2179 ext3_warning(sb, __FUNCTION__, "Marking fs in need of "
2180 "filesystem check.");
2181
2182 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
2183 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
2184 ext3_commit_super (sb, es, 1);
2185
2186 journal_clear_err(journal);
2187 }
2188 }
2189
2190 /*
2191 * Force the running and committing transactions to commit,
2192 * and wait on the commit.
2193 */
2194 int ext3_force_commit(struct super_block *sb)
2195 {
2196 journal_t *journal;
2197 int ret;
2198
2199 if (sb->s_flags & MS_RDONLY)
2200 return 0;
2201
2202 journal = EXT3_SB(sb)->s_journal;
2203 sb->s_dirt = 0;
2204 ret = ext3_journal_force_commit(journal);
2205 return ret;
2206 }
2207
2208 /*
2209 * Ext3 always journals updates to the superblock itself, so we don't
2210 * have to propagate any other updates to the superblock on disk at this
2211 * point. Just start an async writeback to get the buffers on their way
2212 * to the disk.
2213 *
2214 * This implicitly triggers the writebehind on sync().
2215 */
2216
2217 static void ext3_write_super (struct super_block * sb)
2218 {
2219 if (mutex_trylock(&sb->s_lock) != 0)
2220 BUG();
2221 sb->s_dirt = 0;
2222 }
2223
2224 static int ext3_sync_fs(struct super_block *sb, int wait)
2225 {
2226 tid_t target;
2227
2228 sb->s_dirt = 0;
2229 if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
2230 if (wait)
2231 log_wait_commit(EXT3_SB(sb)->s_journal, target);
2232 }
2233 return 0;
2234 }
2235
2236 /*
2237 * LVM calls this function before a (read-only) snapshot is created. This
2238 * gives us a chance to flush the journal completely and mark the fs clean.
2239 */
2240 static void ext3_write_super_lockfs(struct super_block *sb)
2241 {
2242 sb->s_dirt = 0;
2243
2244 if (!(sb->s_flags & MS_RDONLY)) {
2245 journal_t *journal = EXT3_SB(sb)->s_journal;
2246
2247 /* Now we set up the journal barrier. */
2248 journal_lock_updates(journal);
2249 journal_flush(journal);
2250
2251 /* Journal blocked and flushed, clear needs_recovery flag. */
2252 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2253 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2254 }
2255 }
2256
2257 /*
2258 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2259 * flag here, even though the filesystem is not technically dirty yet.
2260 */
2261 static void ext3_unlockfs(struct super_block *sb)
2262 {
2263 if (!(sb->s_flags & MS_RDONLY)) {
2264 lock_super(sb);
2265 /* Reser the needs_recovery flag before the fs is unlocked. */
2266 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2267 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2268 unlock_super(sb);
2269 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2270 }
2271 }
2272
2273 static int ext3_remount (struct super_block * sb, int * flags, char * data)
2274 {
2275 struct ext3_super_block * es;
2276 struct ext3_sb_info *sbi = EXT3_SB(sb);
2277 ext3_fsblk_t n_blocks_count = 0;
2278 unsigned long old_sb_flags;
2279 struct ext3_mount_options old_opts;
2280 int err;
2281 #ifdef CONFIG_QUOTA
2282 int i;
2283 #endif
2284
2285 /* Store the original options */
2286 old_sb_flags = sb->s_flags;
2287 old_opts.s_mount_opt = sbi->s_mount_opt;
2288 old_opts.s_resuid = sbi->s_resuid;
2289 old_opts.s_resgid = sbi->s_resgid;
2290 old_opts.s_commit_interval = sbi->s_commit_interval;
2291 #ifdef CONFIG_QUOTA
2292 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2293 for (i = 0; i < MAXQUOTAS; i++)
2294 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2295 #endif
2296
2297 /*
2298 * Allow the "check" option to be passed as a remount option.
2299 */
2300 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2301 err = -EINVAL;
2302 goto restore_opts;
2303 }
2304
2305 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT)
2306 ext3_abort(sb, __FUNCTION__, "Abort forced by user");
2307
2308 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2309 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2310
2311 es = sbi->s_es;
2312
2313 ext3_init_journal_params(sb, sbi->s_journal);
2314
2315 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2316 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
2317 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT) {
2318 err = -EROFS;
2319 goto restore_opts;
2320 }
2321
2322 if (*flags & MS_RDONLY) {
2323 /*
2324 * First of all, the unconditional stuff we have to do
2325 * to disable replay of the journal when we next remount
2326 */
2327 sb->s_flags |= MS_RDONLY;
2328
2329 /*
2330 * OK, test if we are remounting a valid rw partition
2331 * readonly, and if so set the rdonly flag and then
2332 * mark the partition as valid again.
2333 */
2334 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2335 (sbi->s_mount_state & EXT3_VALID_FS))
2336 es->s_state = cpu_to_le16(sbi->s_mount_state);
2337
2338 ext3_mark_recovery_complete(sb, es);
2339 } else {
2340 __le32 ret;
2341 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2342 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
2343 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2344 "remount RDWR because of unsupported "
2345 "optional features (%x).\n",
2346 sb->s_id, le32_to_cpu(ret));
2347 err = -EROFS;
2348 goto restore_opts;
2349 }
2350
2351 /*
2352 * If we have an unprocessed orphan list hanging
2353 * around from a previously readonly bdev mount,
2354 * require a full umount/remount for now.
2355 */
2356 if (es->s_last_orphan) {
2357 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2358 "remount RDWR because of unprocessed "
2359 "orphan inode list. Please "
2360 "umount/remount instead.\n",
2361 sb->s_id);
2362 err = -EINVAL;
2363 goto restore_opts;
2364 }
2365
2366 /*
2367 * Mounting a RDONLY partition read-write, so reread
2368 * and store the current valid flag. (It may have
2369 * been changed by e2fsck since we originally mounted
2370 * the partition.)
2371 */
2372 ext3_clear_journal_err(sb, es);
2373 sbi->s_mount_state = le16_to_cpu(es->s_state);
2374 if ((err = ext3_group_extend(sb, es, n_blocks_count)))
2375 goto restore_opts;
2376 if (!ext3_setup_super (sb, es, 0))
2377 sb->s_flags &= ~MS_RDONLY;
2378 }
2379 }
2380 #ifdef CONFIG_QUOTA
2381 /* Release old quota file names */
2382 for (i = 0; i < MAXQUOTAS; i++)
2383 if (old_opts.s_qf_names[i] &&
2384 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2385 kfree(old_opts.s_qf_names[i]);
2386 #endif
2387 return 0;
2388 restore_opts:
2389 sb->s_flags = old_sb_flags;
2390 sbi->s_mount_opt = old_opts.s_mount_opt;
2391 sbi->s_resuid = old_opts.s_resuid;
2392 sbi->s_resgid = old_opts.s_resgid;
2393 sbi->s_commit_interval = old_opts.s_commit_interval;
2394 #ifdef CONFIG_QUOTA
2395 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2396 for (i = 0; i < MAXQUOTAS; i++) {
2397 if (sbi->s_qf_names[i] &&
2398 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2399 kfree(sbi->s_qf_names[i]);
2400 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2401 }
2402 #endif
2403 return err;
2404 }
2405
2406 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf)
2407 {
2408 struct super_block *sb = dentry->d_sb;
2409 struct ext3_sb_info *sbi = EXT3_SB(sb);
2410 struct ext3_super_block *es = sbi->s_es;
2411 ext3_fsblk_t overhead;
2412 int i;
2413 u64 fsid;
2414
2415 if (test_opt (sb, MINIX_DF))
2416 overhead = 0;
2417 else {
2418 unsigned long ngroups;
2419 ngroups = EXT3_SB(sb)->s_groups_count;
2420 smp_rmb();
2421
2422 /*
2423 * Compute the overhead (FS structures)
2424 */
2425
2426 /*
2427 * All of the blocks before first_data_block are
2428 * overhead
2429 */
2430 overhead = le32_to_cpu(es->s_first_data_block);
2431
2432 /*
2433 * Add the overhead attributed to the superblock and
2434 * block group descriptors. If the sparse superblocks
2435 * feature is turned on, then not all groups have this.
2436 */
2437 for (i = 0; i < ngroups; i++) {
2438 overhead += ext3_bg_has_super(sb, i) +
2439 ext3_bg_num_gdb(sb, i);
2440 cond_resched();
2441 }
2442
2443 /*
2444 * Every block group has an inode bitmap, a block
2445 * bitmap, and an inode table.
2446 */
2447 overhead += (ngroups * (2 + EXT3_SB(sb)->s_itb_per_group));
2448 }
2449
2450 buf->f_type = EXT3_SUPER_MAGIC;
2451 buf->f_bsize = sb->s_blocksize;
2452 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - overhead;
2453 buf->f_bfree = percpu_counter_sum(&sbi->s_freeblocks_counter);
2454 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2455 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2456 buf->f_bavail = 0;
2457 buf->f_files = le32_to_cpu(es->s_inodes_count);
2458 buf->f_ffree = percpu_counter_sum(&sbi->s_freeinodes_counter);
2459 buf->f_namelen = EXT3_NAME_LEN;
2460 fsid = le64_to_cpup((void *)es->s_uuid) ^
2461 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2462 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2463 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
2464 return 0;
2465 }
2466
2467 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2468 * is locked for write. Otherwise the are possible deadlocks:
2469 * Process 1 Process 2
2470 * ext3_create() quota_sync()
2471 * journal_start() write_dquot()
2472 * DQUOT_INIT() down(dqio_mutex)
2473 * down(dqio_mutex) journal_start()
2474 *
2475 */
2476
2477 #ifdef CONFIG_QUOTA
2478
2479 static inline struct inode *dquot_to_inode(struct dquot *dquot)
2480 {
2481 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2482 }
2483
2484 static int ext3_dquot_initialize(struct inode *inode, int type)
2485 {
2486 handle_t *handle;
2487 int ret, err;
2488
2489 /* We may create quota structure so we need to reserve enough blocks */
2490 handle = ext3_journal_start(inode, 2*EXT3_QUOTA_INIT_BLOCKS(inode->i_sb));
2491 if (IS_ERR(handle))
2492 return PTR_ERR(handle);
2493 ret = dquot_initialize(inode, type);
2494 err = ext3_journal_stop(handle);
2495 if (!ret)
2496 ret = err;
2497 return ret;
2498 }
2499
2500 static int ext3_dquot_drop(struct inode *inode)
2501 {
2502 handle_t *handle;
2503 int ret, err;
2504
2505 /* We may delete quota structure so we need to reserve enough blocks */
2506 handle = ext3_journal_start(inode, 2*EXT3_QUOTA_DEL_BLOCKS(inode->i_sb));
2507 if (IS_ERR(handle))
2508 return PTR_ERR(handle);
2509 ret = dquot_drop(inode);
2510 err = ext3_journal_stop(handle);
2511 if (!ret)
2512 ret = err;
2513 return ret;
2514 }
2515
2516 static int ext3_write_dquot(struct dquot *dquot)
2517 {
2518 int ret, err;
2519 handle_t *handle;
2520 struct inode *inode;
2521
2522 inode = dquot_to_inode(dquot);
2523 handle = ext3_journal_start(inode,
2524 EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
2525 if (IS_ERR(handle))
2526 return PTR_ERR(handle);
2527 ret = dquot_commit(dquot);
2528 err = ext3_journal_stop(handle);
2529 if (!ret)
2530 ret = err;
2531 return ret;
2532 }
2533
2534 static int ext3_acquire_dquot(struct dquot *dquot)
2535 {
2536 int ret, err;
2537 handle_t *handle;
2538
2539 handle = ext3_journal_start(dquot_to_inode(dquot),
2540 EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
2541 if (IS_ERR(handle))
2542 return PTR_ERR(handle);
2543 ret = dquot_acquire(dquot);
2544 err = ext3_journal_stop(handle);
2545 if (!ret)
2546 ret = err;
2547 return ret;
2548 }
2549
2550 static int ext3_release_dquot(struct dquot *dquot)
2551 {
2552 int ret, err;
2553 handle_t *handle;
2554
2555 handle = ext3_journal_start(dquot_to_inode(dquot),
2556 EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
2557 if (IS_ERR(handle))
2558 return PTR_ERR(handle);
2559 ret = dquot_release(dquot);
2560 err = ext3_journal_stop(handle);
2561 if (!ret)
2562 ret = err;
2563 return ret;
2564 }
2565
2566 static int ext3_mark_dquot_dirty(struct dquot *dquot)
2567 {
2568 /* Are we journalling quotas? */
2569 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2570 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2571 dquot_mark_dquot_dirty(dquot);
2572 return ext3_write_dquot(dquot);
2573 } else {
2574 return dquot_mark_dquot_dirty(dquot);
2575 }
2576 }
2577
2578 static int ext3_write_info(struct super_block *sb, int type)
2579 {
2580 int ret, err;
2581 handle_t *handle;
2582
2583 /* Data block + inode block */
2584 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2585 if (IS_ERR(handle))
2586 return PTR_ERR(handle);
2587 ret = dquot_commit_info(sb, type);
2588 err = ext3_journal_stop(handle);
2589 if (!ret)
2590 ret = err;
2591 return ret;
2592 }
2593
2594 /*
2595 * Turn on quotas during mount time - we need to find
2596 * the quota file and such...
2597 */
2598 static int ext3_quota_on_mount(struct super_block *sb, int type)
2599 {
2600 return vfs_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2601 EXT3_SB(sb)->s_jquota_fmt, type);
2602 }
2603
2604 /*
2605 * Standard function to be called on quota_on
2606 */
2607 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
2608 char *path)
2609 {
2610 int err;
2611 struct nameidata nd;
2612
2613 if (!test_opt(sb, QUOTA))
2614 return -EINVAL;
2615 /* Not journalling quota? */
2616 if (!EXT3_SB(sb)->s_qf_names[USRQUOTA] &&
2617 !EXT3_SB(sb)->s_qf_names[GRPQUOTA])
2618 return vfs_quota_on(sb, type, format_id, path);
2619 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
2620 if (err)
2621 return err;
2622 /* Quotafile not on the same filesystem? */
2623 if (nd.mnt->mnt_sb != sb) {
2624 path_release(&nd);
2625 return -EXDEV;
2626 }
2627 /* Quotafile not of fs root? */
2628 if (nd.dentry->d_parent->d_inode != sb->s_root->d_inode)
2629 printk(KERN_WARNING
2630 "EXT3-fs: Quota file not on filesystem root. "
2631 "Journalled quota will not work.\n");
2632 path_release(&nd);
2633 return vfs_quota_on(sb, type, format_id, path);
2634 }
2635
2636 /* Read data from quotafile - avoid pagecache and such because we cannot afford
2637 * acquiring the locks... As quota files are never truncated and quota code
2638 * itself serializes the operations (and noone else should touch the files)
2639 * we don't have to be afraid of races */
2640 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
2641 size_t len, loff_t off)
2642 {
2643 struct inode *inode = sb_dqopt(sb)->files[type];
2644 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2645 int err = 0;
2646 int offset = off & (sb->s_blocksize - 1);
2647 int tocopy;
2648 size_t toread;
2649 struct buffer_head *bh;
2650 loff_t i_size = i_size_read(inode);
2651
2652 if (off > i_size)
2653 return 0;
2654 if (off+len > i_size)
2655 len = i_size-off;
2656 toread = len;
2657 while (toread > 0) {
2658 tocopy = sb->s_blocksize - offset < toread ?
2659 sb->s_blocksize - offset : toread;
2660 bh = ext3_bread(NULL, inode, blk, 0, &err);
2661 if (err)
2662 return err;
2663 if (!bh) /* A hole? */
2664 memset(data, 0, tocopy);
2665 else
2666 memcpy(data, bh->b_data+offset, tocopy);
2667 brelse(bh);
2668 offset = 0;
2669 toread -= tocopy;
2670 data += tocopy;
2671 blk++;
2672 }
2673 return len;
2674 }
2675
2676 /* Write to quotafile (we know the transaction is already started and has
2677 * enough credits) */
2678 static ssize_t ext3_quota_write(struct super_block *sb, int type,
2679 const char *data, size_t len, loff_t off)
2680 {
2681 struct inode *inode = sb_dqopt(sb)->files[type];
2682 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2683 int err = 0;
2684 int offset = off & (sb->s_blocksize - 1);
2685 int tocopy;
2686 int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
2687 size_t towrite = len;
2688 struct buffer_head *bh;
2689 handle_t *handle = journal_current_handle();
2690
2691 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
2692 while (towrite > 0) {
2693 tocopy = sb->s_blocksize - offset < towrite ?
2694 sb->s_blocksize - offset : towrite;
2695 bh = ext3_bread(handle, inode, blk, 1, &err);
2696 if (!bh)
2697 goto out;
2698 if (journal_quota) {
2699 err = ext3_journal_get_write_access(handle, bh);
2700 if (err) {
2701 brelse(bh);
2702 goto out;
2703 }
2704 }
2705 lock_buffer(bh);
2706 memcpy(bh->b_data+offset, data, tocopy);
2707 flush_dcache_page(bh->b_page);
2708 unlock_buffer(bh);
2709 if (journal_quota)
2710 err = ext3_journal_dirty_metadata(handle, bh);
2711 else {
2712 /* Always do at least ordered writes for quotas */
2713 err = ext3_journal_dirty_data(handle, bh);
2714 mark_buffer_dirty(bh);
2715 }
2716 brelse(bh);
2717 if (err)
2718 goto out;
2719 offset = 0;
2720 towrite -= tocopy;
2721 data += tocopy;
2722 blk++;
2723 }
2724 out:
2725 if (len == towrite)
2726 return err;
2727 if (inode->i_size < off+len-towrite) {
2728 i_size_write(inode, off+len-towrite);
2729 EXT3_I(inode)->i_disksize = inode->i_size;
2730 }
2731 inode->i_version++;
2732 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2733 ext3_mark_inode_dirty(handle, inode);
2734 mutex_unlock(&inode->i_mutex);
2735 return len - towrite;
2736 }
2737
2738 #endif
2739
2740 static int ext3_get_sb(struct file_system_type *fs_type,
2741 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
2742 {
2743 return get_sb_bdev(fs_type, flags, dev_name, data, ext3_fill_super, mnt);
2744 }
2745
2746 static struct file_system_type ext3_fs_type = {
2747 .owner = THIS_MODULE,
2748 .name = "ext3",
2749 .get_sb = ext3_get_sb,
2750 .kill_sb = kill_block_super,
2751 .fs_flags = FS_REQUIRES_DEV,
2752 };
2753
2754 static int __init init_ext3_fs(void)
2755 {
2756 int err = init_ext3_xattr();
2757 if (err)
2758 return err;
2759 err = init_inodecache();
2760 if (err)
2761 goto out1;
2762 err = register_filesystem(&ext3_fs_type);
2763 if (err)
2764 goto out;
2765 return 0;
2766 out:
2767 destroy_inodecache();
2768 out1:
2769 exit_ext3_xattr();
2770 return err;
2771 }
2772
2773 static void __exit exit_ext3_fs(void)
2774 {
2775 unregister_filesystem(&ext3_fs_type);
2776 destroy_inodecache();
2777 exit_ext3_xattr();
2778 }
2779
2780 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
2781 MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
2782 MODULE_LICENSE("GPL");
2783 module_init(init_ext3_fs)
2784 module_exit(exit_ext3_fs)
Support for the Chinese Samsung S3C2440 based development boards