?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
x86: cpu_init(): fix memory leak when using CPU hotplug
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / fs / ext3 / super.c
blobf38a5afc39a160c15b5a9e518ba41dd8bd4da2af
1 /*
2 * linux/fs/ext3/super.c
3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19 #include <linux/module.h>
20 #include <linux/string.h>
21 #include <linux/fs.h>
22 #include <linux/time.h>
23 #include <linux/jbd.h>
24 #include <linux/ext3_fs.h>
25 #include <linux/ext3_jbd.h>
26 #include <linux/slab.h>
27 #include <linux/init.h>
28 #include <linux/blkdev.h>
29 #include <linux/parser.h>
30 #include <linux/smp_lock.h>
31 #include <linux/buffer_head.h>
32 #include <linux/exportfs.h>
33 #include <linux/vfs.h>
34 #include <linux/random.h>
35 #include <linux/mount.h>
36 #include <linux/namei.h>
37 #include <linux/quotaops.h>
38 #include <linux/seq_file.h>
39 #include <linux/log2.h>
40
41 #include <asm/uaccess.h>
42
43 #include "xattr.h"
44 #include "acl.h"
45 #include "namei.h"
46
47 static int ext3_load_journal(struct super_block *, struct ext3_super_block *,
48 unsigned long journal_devnum);
49 static int ext3_create_journal(struct super_block *, struct ext3_super_block *,
50 unsigned int);
51 static void ext3_commit_super (struct super_block * sb,
52 struct ext3_super_block * es,
53 int sync);
54 static void ext3_mark_recovery_complete(struct super_block * sb,
55 struct ext3_super_block * es);
56 static void ext3_clear_journal_err(struct super_block * sb,
57 struct ext3_super_block * es);
58 static int ext3_sync_fs(struct super_block *sb, int wait);
59 static const char *ext3_decode_error(struct super_block * sb, int errno,
60 char nbuf[16]);
61 static int ext3_remount (struct super_block * sb, int * flags, char * data);
62 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf);
63 static void ext3_unlockfs(struct super_block *sb);
64 static void ext3_write_super (struct super_block * sb);
65 static void ext3_write_super_lockfs(struct super_block *sb);
66
67 /*
68 * Wrappers for journal_start/end.
69 *
70 * The only special thing we need to do here is to make sure that all
71 * journal_end calls result in the superblock being marked dirty, so
72 * that sync() will call the filesystem's write_super callback if
73 * appropriate.
74 */
75 handle_t *ext3_journal_start_sb(struct super_block *sb, int nblocks)
76 {
77 journal_t *journal;
78
79 if (sb->s_flags & MS_RDONLY)
80 return ERR_PTR(-EROFS);
81
82 /* Special case here: if the journal has aborted behind our
83 * backs (eg. EIO in the commit thread), then we still need to
84 * take the FS itself readonly cleanly. */
85 journal = EXT3_SB(sb)->s_journal;
86 if (is_journal_aborted(journal)) {
87 ext3_abort(sb, __func__,
88 "Detected aborted journal");
89 return ERR_PTR(-EROFS);
90 }
91
92 return journal_start(journal, nblocks);
93 }
94
95 /*
96 * The only special thing we need to do here is to make sure that all
97 * journal_stop calls result in the superblock being marked dirty, so
98 * that sync() will call the filesystem's write_super callback if
99 * appropriate.
100 */
101 int __ext3_journal_stop(const char *where, handle_t *handle)
102 {
103 struct super_block *sb;
104 int err;
105 int rc;
106
107 sb = handle->h_transaction->t_journal->j_private;
108 err = handle->h_err;
109 rc = journal_stop(handle);
110
111 if (!err)
112 err = rc;
113 if (err)
114 __ext3_std_error(sb, where, err);
115 return err;
116 }
117
118 void ext3_journal_abort_handle(const char *caller, const char *err_fn,
119 struct buffer_head *bh, handle_t *handle, int err)
120 {
121 char nbuf[16];
122 const char *errstr = ext3_decode_error(NULL, err, nbuf);
123
124 if (bh)
125 BUFFER_TRACE(bh, "abort");
126
127 if (!handle->h_err)
128 handle->h_err = err;
129
130 if (is_handle_aborted(handle))
131 return;
132
133 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
134 caller, errstr, err_fn);
135
136 journal_abort_handle(handle);
137 }
138
139 /* Deal with the reporting of failure conditions on a filesystem such as
140 * inconsistencies detected or read IO failures.
141 *
142 * On ext2, we can store the error state of the filesystem in the
143 * superblock. That is not possible on ext3, because we may have other
144 * write ordering constraints on the superblock which prevent us from
145 * writing it out straight away; and given that the journal is about to
146 * be aborted, we can't rely on the current, or future, transactions to
147 * write out the superblock safely.
148 *
149 * We'll just use the journal_abort() error code to record an error in
150 * the journal instead. On recovery, the journal will compain about
151 * that error until we've noted it down and cleared it.
152 */
153
154 static void ext3_handle_error(struct super_block *sb)
155 {
156 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
157
158 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
159 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
160
161 if (sb->s_flags & MS_RDONLY)
162 return;
163
164 if (!test_opt (sb, ERRORS_CONT)) {
165 journal_t *journal = EXT3_SB(sb)->s_journal;
166
167 EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
168 if (journal)
169 journal_abort(journal, -EIO);
170 }
171 if (test_opt (sb, ERRORS_RO)) {
172 printk (KERN_CRIT "Remounting filesystem read-only\n");
173 sb->s_flags |= MS_RDONLY;
174 }
175 ext3_commit_super(sb, es, 1);
176 if (test_opt(sb, ERRORS_PANIC))
177 panic("EXT3-fs (device %s): panic forced after error\n",
178 sb->s_id);
179 }
180
181 void ext3_error (struct super_block * sb, const char * function,
182 const char * fmt, ...)
183 {
184 va_list args;
185
186 va_start(args, fmt);
187 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
188 vprintk(fmt, args);
189 printk("\n");
190 va_end(args);
191
192 ext3_handle_error(sb);
193 }
194
195 static const char *ext3_decode_error(struct super_block * sb, int errno,
196 char nbuf[16])
197 {
198 char *errstr = NULL;
199
200 switch (errno) {
201 case -EIO:
202 errstr = "IO failure";
203 break;
204 case -ENOMEM:
205 errstr = "Out of memory";
206 break;
207 case -EROFS:
208 if (!sb || EXT3_SB(sb)->s_journal->j_flags & JFS_ABORT)
209 errstr = "Journal has aborted";
210 else
211 errstr = "Readonly filesystem";
212 break;
213 default:
214 /* If the caller passed in an extra buffer for unknown
215 * errors, textualise them now. Else we just return
216 * NULL. */
217 if (nbuf) {
218 /* Check for truncated error codes... */
219 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
220 errstr = nbuf;
221 }
222 break;
223 }
224
225 return errstr;
226 }
227
228 /* __ext3_std_error decodes expected errors from journaling functions
229 * automatically and invokes the appropriate error response. */
230
231 void __ext3_std_error (struct super_block * sb, const char * function,
232 int errno)
233 {
234 char nbuf[16];
235 const char *errstr;
236
237 /* Special case: if the error is EROFS, and we're not already
238 * inside a transaction, then there's really no point in logging
239 * an error. */
240 if (errno == -EROFS && journal_current_handle() == NULL &&
241 (sb->s_flags & MS_RDONLY))
242 return;
243
244 errstr = ext3_decode_error(sb, errno, nbuf);
245 printk (KERN_CRIT "EXT3-fs error (device %s) in %s: %s\n",
246 sb->s_id, function, errstr);
247
248 ext3_handle_error(sb);
249 }
250
251 /*
252 * ext3_abort is a much stronger failure handler than ext3_error. The
253 * abort function may be used to deal with unrecoverable failures such
254 * as journal IO errors or ENOMEM at a critical moment in log management.
255 *
256 * We unconditionally force the filesystem into an ABORT|READONLY state,
257 * unless the error response on the fs has been set to panic in which
258 * case we take the easy way out and panic immediately.
259 */
260
261 void ext3_abort (struct super_block * sb, const char * function,
262 const char * fmt, ...)
263 {
264 va_list args;
265
266 printk (KERN_CRIT "ext3_abort called.\n");
267
268 va_start(args, fmt);
269 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
270 vprintk(fmt, args);
271 printk("\n");
272 va_end(args);
273
274 if (test_opt(sb, ERRORS_PANIC))
275 panic("EXT3-fs panic from previous error\n");
276
277 if (sb->s_flags & MS_RDONLY)
278 return;
279
280 printk(KERN_CRIT "Remounting filesystem read-only\n");
281 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
282 sb->s_flags |= MS_RDONLY;
283 EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
284 journal_abort(EXT3_SB(sb)->s_journal, -EIO);
285 }
286
287 void ext3_warning (struct super_block * sb, const char * function,
288 const char * fmt, ...)
289 {
290 va_list args;
291
292 va_start(args, fmt);
293 printk(KERN_WARNING "EXT3-fs warning (device %s): %s: ",
294 sb->s_id, function);
295 vprintk(fmt, args);
296 printk("\n");
297 va_end(args);
298 }
299
300 void ext3_update_dynamic_rev(struct super_block *sb)
301 {
302 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
303
304 if (le32_to_cpu(es->s_rev_level) > EXT3_GOOD_OLD_REV)
305 return;
306
307 ext3_warning(sb, __func__,
308 "updating to rev %d because of new feature flag, "
309 "running e2fsck is recommended",
310 EXT3_DYNAMIC_REV);
311
312 es->s_first_ino = cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO);
313 es->s_inode_size = cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE);
314 es->s_rev_level = cpu_to_le32(EXT3_DYNAMIC_REV);
315 /* leave es->s_feature_*compat flags alone */
316 /* es->s_uuid will be set by e2fsck if empty */
317
318 /*
319 * The rest of the superblock fields should be zero, and if not it
320 * means they are likely already in use, so leave them alone. We
321 * can leave it up to e2fsck to clean up any inconsistencies there.
322 */
323 }
324
325 /*
326 * Open the external journal device
327 */
328 static struct block_device *ext3_blkdev_get(dev_t dev)
329 {
330 struct block_device *bdev;
331 char b[BDEVNAME_SIZE];
332
333 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
334 if (IS_ERR(bdev))
335 goto fail;
336 return bdev;
337
338 fail:
339 printk(KERN_ERR "EXT3: failed to open journal device %s: %ld\n",
340 __bdevname(dev, b), PTR_ERR(bdev));
341 return NULL;
342 }
343
344 /*
345 * Release the journal device
346 */
347 static int ext3_blkdev_put(struct block_device *bdev)
348 {
349 bd_release(bdev);
350 return blkdev_put(bdev);
351 }
352
353 static int ext3_blkdev_remove(struct ext3_sb_info *sbi)
354 {
355 struct block_device *bdev;
356 int ret = -ENODEV;
357
358 bdev = sbi->journal_bdev;
359 if (bdev) {
360 ret = ext3_blkdev_put(bdev);
361 sbi->journal_bdev = NULL;
362 }
363 return ret;
364 }
365
366 static inline struct inode *orphan_list_entry(struct list_head *l)
367 {
368 return &list_entry(l, struct ext3_inode_info, i_orphan)->vfs_inode;
369 }
370
371 static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi)
372 {
373 struct list_head *l;
374
375 printk(KERN_ERR "sb orphan head is %d\n",
376 le32_to_cpu(sbi->s_es->s_last_orphan));
377
378 printk(KERN_ERR "sb_info orphan list:\n");
379 list_for_each(l, &sbi->s_orphan) {
380 struct inode *inode = orphan_list_entry(l);
381 printk(KERN_ERR " "
382 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
383 inode->i_sb->s_id, inode->i_ino, inode,
384 inode->i_mode, inode->i_nlink,
385 NEXT_ORPHAN(inode));
386 }
387 }
388
389 static void ext3_put_super (struct super_block * sb)
390 {
391 struct ext3_sb_info *sbi = EXT3_SB(sb);
392 struct ext3_super_block *es = sbi->s_es;
393 int i;
394
395 ext3_xattr_put_super(sb);
396 journal_destroy(sbi->s_journal);
397 if (!(sb->s_flags & MS_RDONLY)) {
398 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
399 es->s_state = cpu_to_le16(sbi->s_mount_state);
400 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
401 mark_buffer_dirty(sbi->s_sbh);
402 ext3_commit_super(sb, es, 1);
403 }
404
405 for (i = 0; i < sbi->s_gdb_count; i++)
406 brelse(sbi->s_group_desc[i]);
407 kfree(sbi->s_group_desc);
408 percpu_counter_destroy(&sbi->s_freeblocks_counter);
409 percpu_counter_destroy(&sbi->s_freeinodes_counter);
410 percpu_counter_destroy(&sbi->s_dirs_counter);
411 brelse(sbi->s_sbh);
412 #ifdef CONFIG_QUOTA
413 for (i = 0; i < MAXQUOTAS; i++)
414 kfree(sbi->s_qf_names[i]);
415 #endif
416
417 /* Debugging code just in case the in-memory inode orphan list
418 * isn't empty. The on-disk one can be non-empty if we've
419 * detected an error and taken the fs readonly, but the
420 * in-memory list had better be clean by this point. */
421 if (!list_empty(&sbi->s_orphan))
422 dump_orphan_list(sb, sbi);
423 J_ASSERT(list_empty(&sbi->s_orphan));
424
425 invalidate_bdev(sb->s_bdev);
426 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
427 /*
428 * Invalidate the journal device's buffers. We don't want them
429 * floating about in memory - the physical journal device may
430 * hotswapped, and it breaks the `ro-after' testing code.
431 */
432 sync_blockdev(sbi->journal_bdev);
433 invalidate_bdev(sbi->journal_bdev);
434 ext3_blkdev_remove(sbi);
435 }
436 sb->s_fs_info = NULL;
437 kfree(sbi);
438 return;
439 }
440
441 static struct kmem_cache *ext3_inode_cachep;
442
443 /*
444 * Called inside transaction, so use GFP_NOFS
445 */
446 static struct inode *ext3_alloc_inode(struct super_block *sb)
447 {
448 struct ext3_inode_info *ei;
449
450 ei = kmem_cache_alloc(ext3_inode_cachep, GFP_NOFS);
451 if (!ei)
452 return NULL;
453 #ifdef CONFIG_EXT3_FS_POSIX_ACL
454 ei->i_acl = EXT3_ACL_NOT_CACHED;
455 ei->i_default_acl = EXT3_ACL_NOT_CACHED;
456 #endif
457 ei->i_block_alloc_info = NULL;
458 ei->vfs_inode.i_version = 1;
459 return &ei->vfs_inode;
460 }
461
462 static void ext3_destroy_inode(struct inode *inode)
463 {
464 if (!list_empty(&(EXT3_I(inode)->i_orphan))) {
465 printk("EXT3 Inode %p: orphan list check failed!\n",
466 EXT3_I(inode));
467 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
468 EXT3_I(inode), sizeof(struct ext3_inode_info),
469 false);
470 dump_stack();
471 }
472 kmem_cache_free(ext3_inode_cachep, EXT3_I(inode));
473 }
474
475 static void init_once(void *foo)
476 {
477 struct ext3_inode_info *ei = (struct ext3_inode_info *) foo;
478
479 INIT_LIST_HEAD(&ei->i_orphan);
480 #ifdef CONFIG_EXT3_FS_XATTR
481 init_rwsem(&ei->xattr_sem);
482 #endif
483 mutex_init(&ei->truncate_mutex);
484 inode_init_once(&ei->vfs_inode);
485 }
486
487 static int init_inodecache(void)
488 {
489 ext3_inode_cachep = kmem_cache_create("ext3_inode_cache",
490 sizeof(struct ext3_inode_info),
491 0, (SLAB_RECLAIM_ACCOUNT|
492 SLAB_MEM_SPREAD),
493 init_once);
494 if (ext3_inode_cachep == NULL)
495 return -ENOMEM;
496 return 0;
497 }
498
499 static void destroy_inodecache(void)
500 {
501 kmem_cache_destroy(ext3_inode_cachep);
502 }
503
504 static void ext3_clear_inode(struct inode *inode)
505 {
506 struct ext3_block_alloc_info *rsv = EXT3_I(inode)->i_block_alloc_info;
507 #ifdef CONFIG_EXT3_FS_POSIX_ACL
508 if (EXT3_I(inode)->i_acl &&
509 EXT3_I(inode)->i_acl != EXT3_ACL_NOT_CACHED) {
510 posix_acl_release(EXT3_I(inode)->i_acl);
511 EXT3_I(inode)->i_acl = EXT3_ACL_NOT_CACHED;
512 }
513 if (EXT3_I(inode)->i_default_acl &&
514 EXT3_I(inode)->i_default_acl != EXT3_ACL_NOT_CACHED) {
515 posix_acl_release(EXT3_I(inode)->i_default_acl);
516 EXT3_I(inode)->i_default_acl = EXT3_ACL_NOT_CACHED;
517 }
518 #endif
519 ext3_discard_reservation(inode);
520 EXT3_I(inode)->i_block_alloc_info = NULL;
521 if (unlikely(rsv))
522 kfree(rsv);
523 }
524
525 static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb)
526 {
527 #if defined(CONFIG_QUOTA)
528 struct ext3_sb_info *sbi = EXT3_SB(sb);
529
530 if (sbi->s_jquota_fmt)
531 seq_printf(seq, ",jqfmt=%s",
532 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
533
534 if (sbi->s_qf_names[USRQUOTA])
535 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
536
537 if (sbi->s_qf_names[GRPQUOTA])
538 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
539
540 if (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA)
541 seq_puts(seq, ",usrquota");
542
543 if (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)
544 seq_puts(seq, ",grpquota");
545 #endif
546 }
547
548 /*
549 * Show an option if
550 * - it's set to a non-default value OR
551 * - if the per-sb default is different from the global default
552 */
553 static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs)
554 {
555 struct super_block *sb = vfs->mnt_sb;
556 struct ext3_sb_info *sbi = EXT3_SB(sb);
557 struct ext3_super_block *es = sbi->s_es;
558 unsigned long def_mount_opts;
559
560 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
561
562 if (sbi->s_sb_block != 1)
563 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
564 if (test_opt(sb, MINIX_DF))
565 seq_puts(seq, ",minixdf");
566 if (test_opt(sb, GRPID))
567 seq_puts(seq, ",grpid");
568 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT3_DEFM_BSDGROUPS))
569 seq_puts(seq, ",nogrpid");
570 if (sbi->s_resuid != EXT3_DEF_RESUID ||
571 le16_to_cpu(es->s_def_resuid) != EXT3_DEF_RESUID) {
572 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
573 }
574 if (sbi->s_resgid != EXT3_DEF_RESGID ||
575 le16_to_cpu(es->s_def_resgid) != EXT3_DEF_RESGID) {
576 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
577 }
578 if (test_opt(sb, ERRORS_RO)) {
579 int def_errors = le16_to_cpu(es->s_errors);
580
581 if (def_errors == EXT3_ERRORS_PANIC ||
582 def_errors == EXT3_ERRORS_CONTINUE) {
583 seq_puts(seq, ",errors=remount-ro");
584 }
585 }
586 if (test_opt(sb, ERRORS_CONT))
587 seq_puts(seq, ",errors=continue");
588 if (test_opt(sb, ERRORS_PANIC))
589 seq_puts(seq, ",errors=panic");
590 if (test_opt(sb, NO_UID32))
591 seq_puts(seq, ",nouid32");
592 if (test_opt(sb, DEBUG))
593 seq_puts(seq, ",debug");
594 if (test_opt(sb, OLDALLOC))
595 seq_puts(seq, ",oldalloc");
596 #ifdef CONFIG_EXT3_FS_XATTR
597 if (test_opt(sb, XATTR_USER))
598 seq_puts(seq, ",user_xattr");
599 if (!test_opt(sb, XATTR_USER) &&
600 (def_mount_opts & EXT3_DEFM_XATTR_USER)) {
601 seq_puts(seq, ",nouser_xattr");
602 }
603 #endif
604 #ifdef CONFIG_EXT3_FS_POSIX_ACL
605 if (test_opt(sb, POSIX_ACL))
606 seq_puts(seq, ",acl");
607 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT3_DEFM_ACL))
608 seq_puts(seq, ",noacl");
609 #endif
610 if (!test_opt(sb, RESERVATION))
611 seq_puts(seq, ",noreservation");
612 if (sbi->s_commit_interval) {
613 seq_printf(seq, ",commit=%u",
614 (unsigned) (sbi->s_commit_interval / HZ));
615 }
616 if (test_opt(sb, BARRIER))
617 seq_puts(seq, ",barrier=1");
618 if (test_opt(sb, NOBH))
619 seq_puts(seq, ",nobh");
620
621 if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA)
622 seq_puts(seq, ",data=journal");
623 else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA)
624 seq_puts(seq, ",data=ordered");
625 else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)
626 seq_puts(seq, ",data=writeback");
627
628 ext3_show_quota_options(seq, sb);
629
630 return 0;
631 }
632
633
634 static struct inode *ext3_nfs_get_inode(struct super_block *sb,
635 u64 ino, u32 generation)
636 {
637 struct inode *inode;
638
639 if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO)
640 return ERR_PTR(-ESTALE);
641 if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count))
642 return ERR_PTR(-ESTALE);
643
644 /* iget isn't really right if the inode is currently unallocated!!
645 *
646 * ext3_read_inode will return a bad_inode if the inode had been
647 * deleted, so we should be safe.
648 *
649 * Currently we don't know the generation for parent directory, so
650 * a generation of 0 means "accept any"
651 */
652 inode = ext3_iget(sb, ino);
653 if (IS_ERR(inode))
654 return ERR_CAST(inode);
655 if (generation && inode->i_generation != generation) {
656 iput(inode);
657 return ERR_PTR(-ESTALE);
658 }
659
660 return inode;
661 }
662
663 static struct dentry *ext3_fh_to_dentry(struct super_block *sb, struct fid *fid,
664 int fh_len, int fh_type)
665 {
666 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
667 ext3_nfs_get_inode);
668 }
669
670 static struct dentry *ext3_fh_to_parent(struct super_block *sb, struct fid *fid,
671 int fh_len, int fh_type)
672 {
673 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
674 ext3_nfs_get_inode);
675 }
676
677 #ifdef CONFIG_QUOTA
678 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
679 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
680
681 static int ext3_dquot_initialize(struct inode *inode, int type);
682 static int ext3_dquot_drop(struct inode *inode);
683 static int ext3_write_dquot(struct dquot *dquot);
684 static int ext3_acquire_dquot(struct dquot *dquot);
685 static int ext3_release_dquot(struct dquot *dquot);
686 static int ext3_mark_dquot_dirty(struct dquot *dquot);
687 static int ext3_write_info(struct super_block *sb, int type);
688 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
689 char *path, int remount);
690 static int ext3_quota_on_mount(struct super_block *sb, int type);
691 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
692 size_t len, loff_t off);
693 static ssize_t ext3_quota_write(struct super_block *sb, int type,
694 const char *data, size_t len, loff_t off);
695
696 static struct dquot_operations ext3_quota_operations = {
697 .initialize = ext3_dquot_initialize,
698 .drop = ext3_dquot_drop,
699 .alloc_space = dquot_alloc_space,
700 .alloc_inode = dquot_alloc_inode,
701 .free_space = dquot_free_space,
702 .free_inode = dquot_free_inode,
703 .transfer = dquot_transfer,
704 .write_dquot = ext3_write_dquot,
705 .acquire_dquot = ext3_acquire_dquot,
706 .release_dquot = ext3_release_dquot,
707 .mark_dirty = ext3_mark_dquot_dirty,
708 .write_info = ext3_write_info
709 };
710
711 static struct quotactl_ops ext3_qctl_operations = {
712 .quota_on = ext3_quota_on,
713 .quota_off = vfs_quota_off,
714 .quota_sync = vfs_quota_sync,
715 .get_info = vfs_get_dqinfo,
716 .set_info = vfs_set_dqinfo,
717 .get_dqblk = vfs_get_dqblk,
718 .set_dqblk = vfs_set_dqblk
719 };
720 #endif
721
722 static const struct super_operations ext3_sops = {
723 .alloc_inode = ext3_alloc_inode,
724 .destroy_inode = ext3_destroy_inode,
725 .write_inode = ext3_write_inode,
726 .dirty_inode = ext3_dirty_inode,
727 .delete_inode = ext3_delete_inode,
728 .put_super = ext3_put_super,
729 .write_super = ext3_write_super,
730 .sync_fs = ext3_sync_fs,
731 .write_super_lockfs = ext3_write_super_lockfs,
732 .unlockfs = ext3_unlockfs,
733 .statfs = ext3_statfs,
734 .remount_fs = ext3_remount,
735 .clear_inode = ext3_clear_inode,
736 .show_options = ext3_show_options,
737 #ifdef CONFIG_QUOTA
738 .quota_read = ext3_quota_read,
739 .quota_write = ext3_quota_write,
740 #endif
741 };
742
743 static const struct export_operations ext3_export_ops = {
744 .fh_to_dentry = ext3_fh_to_dentry,
745 .fh_to_parent = ext3_fh_to_parent,
746 .get_parent = ext3_get_parent,
747 };
748
749 enum {
750 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
751 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
752 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
753 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
754 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
755 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
756 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
757 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
758 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
759 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
760 Opt_grpquota
761 };
762
763 static match_table_t tokens = {
764 {Opt_bsd_df, "bsddf"},
765 {Opt_minix_df, "minixdf"},
766 {Opt_grpid, "grpid"},
767 {Opt_grpid, "bsdgroups"},
768 {Opt_nogrpid, "nogrpid"},
769 {Opt_nogrpid, "sysvgroups"},
770 {Opt_resgid, "resgid=%u"},
771 {Opt_resuid, "resuid=%u"},
772 {Opt_sb, "sb=%u"},
773 {Opt_err_cont, "errors=continue"},
774 {Opt_err_panic, "errors=panic"},
775 {Opt_err_ro, "errors=remount-ro"},
776 {Opt_nouid32, "nouid32"},
777 {Opt_nocheck, "nocheck"},
778 {Opt_nocheck, "check=none"},
779 {Opt_debug, "debug"},
780 {Opt_oldalloc, "oldalloc"},
781 {Opt_orlov, "orlov"},
782 {Opt_user_xattr, "user_xattr"},
783 {Opt_nouser_xattr, "nouser_xattr"},
784 {Opt_acl, "acl"},
785 {Opt_noacl, "noacl"},
786 {Opt_reservation, "reservation"},
787 {Opt_noreservation, "noreservation"},
788 {Opt_noload, "noload"},
789 {Opt_nobh, "nobh"},
790 {Opt_bh, "bh"},
791 {Opt_commit, "commit=%u"},
792 {Opt_journal_update, "journal=update"},
793 {Opt_journal_inum, "journal=%u"},
794 {Opt_journal_dev, "journal_dev=%u"},
795 {Opt_abort, "abort"},
796 {Opt_data_journal, "data=journal"},
797 {Opt_data_ordered, "data=ordered"},
798 {Opt_data_writeback, "data=writeback"},
799 {Opt_offusrjquota, "usrjquota="},
800 {Opt_usrjquota, "usrjquota=%s"},
801 {Opt_offgrpjquota, "grpjquota="},
802 {Opt_grpjquota, "grpjquota=%s"},
803 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
804 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
805 {Opt_grpquota, "grpquota"},
806 {Opt_noquota, "noquota"},
807 {Opt_quota, "quota"},
808 {Opt_usrquota, "usrquota"},
809 {Opt_barrier, "barrier=%u"},
810 {Opt_resize, "resize"},
811 {Opt_err, NULL},
812 };
813
814 static ext3_fsblk_t get_sb_block(void **data)
815 {
816 ext3_fsblk_t sb_block;
817 char *options = (char *) *data;
818
819 if (!options || strncmp(options, "sb=", 3) != 0)
820 return 1; /* Default location */
821 options += 3;
822 /*todo: use simple_strtoll with >32bit ext3 */
823 sb_block = simple_strtoul(options, &options, 0);
824 if (*options && *options != ',') {
825 printk("EXT3-fs: Invalid sb specification: %s\n",
826 (char *) *data);
827 return 1;
828 }
829 if (*options == ',')
830 options++;
831 *data = (void *) options;
832 return sb_block;
833 }
834
835 static int parse_options (char *options, struct super_block *sb,
836 unsigned int *inum, unsigned long *journal_devnum,
837 ext3_fsblk_t *n_blocks_count, int is_remount)
838 {
839 struct ext3_sb_info *sbi = EXT3_SB(sb);
840 char * p;
841 substring_t args[MAX_OPT_ARGS];
842 int data_opt = 0;
843 int option;
844 #ifdef CONFIG_QUOTA
845 int qtype, qfmt;
846 char *qname;
847 #endif
848
849 if (!options)
850 return 1;
851
852 while ((p = strsep (&options, ",")) != NULL) {
853 int token;
854 if (!*p)
855 continue;
856
857 token = match_token(p, tokens, args);
858 switch (token) {
859 case Opt_bsd_df:
860 clear_opt (sbi->s_mount_opt, MINIX_DF);
861 break;
862 case Opt_minix_df:
863 set_opt (sbi->s_mount_opt, MINIX_DF);
864 break;
865 case Opt_grpid:
866 set_opt (sbi->s_mount_opt, GRPID);
867 break;
868 case Opt_nogrpid:
869 clear_opt (sbi->s_mount_opt, GRPID);
870 break;
871 case Opt_resuid:
872 if (match_int(&args[0], &option))
873 return 0;
874 sbi->s_resuid = option;
875 break;
876 case Opt_resgid:
877 if (match_int(&args[0], &option))
878 return 0;
879 sbi->s_resgid = option;
880 break;
881 case Opt_sb:
882 /* handled by get_sb_block() instead of here */
883 /* *sb_block = match_int(&args[0]); */
884 break;
885 case Opt_err_panic:
886 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
887 clear_opt (sbi->s_mount_opt, ERRORS_RO);
888 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
889 break;
890 case Opt_err_ro:
891 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
892 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
893 set_opt (sbi->s_mount_opt, ERRORS_RO);
894 break;
895 case Opt_err_cont:
896 clear_opt (sbi->s_mount_opt, ERRORS_RO);
897 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
898 set_opt (sbi->s_mount_opt, ERRORS_CONT);
899 break;
900 case Opt_nouid32:
901 set_opt (sbi->s_mount_opt, NO_UID32);
902 break;
903 case Opt_nocheck:
904 clear_opt (sbi->s_mount_opt, CHECK);
905 break;
906 case Opt_debug:
907 set_opt (sbi->s_mount_opt, DEBUG);
908 break;
909 case Opt_oldalloc:
910 set_opt (sbi->s_mount_opt, OLDALLOC);
911 break;
912 case Opt_orlov:
913 clear_opt (sbi->s_mount_opt, OLDALLOC);
914 break;
915 #ifdef CONFIG_EXT3_FS_XATTR
916 case Opt_user_xattr:
917 set_opt (sbi->s_mount_opt, XATTR_USER);
918 break;
919 case Opt_nouser_xattr:
920 clear_opt (sbi->s_mount_opt, XATTR_USER);
921 break;
922 #else
923 case Opt_user_xattr:
924 case Opt_nouser_xattr:
925 printk("EXT3 (no)user_xattr options not supported\n");
926 break;
927 #endif
928 #ifdef CONFIG_EXT3_FS_POSIX_ACL
929 case Opt_acl:
930 set_opt(sbi->s_mount_opt, POSIX_ACL);
931 break;
932 case Opt_noacl:
933 clear_opt(sbi->s_mount_opt, POSIX_ACL);
934 break;
935 #else
936 case Opt_acl:
937 case Opt_noacl:
938 printk("EXT3 (no)acl options not supported\n");
939 break;
940 #endif
941 case Opt_reservation:
942 set_opt(sbi->s_mount_opt, RESERVATION);
943 break;
944 case Opt_noreservation:
945 clear_opt(sbi->s_mount_opt, RESERVATION);
946 break;
947 case Opt_journal_update:
948 /* @@@ FIXME */
949 /* Eventually we will want to be able to create
950 a journal file here. For now, only allow the
951 user to specify an existing inode to be the
952 journal file. */
953 if (is_remount) {
954 printk(KERN_ERR "EXT3-fs: cannot specify "
955 "journal on remount\n");
956 return 0;
957 }
958 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
959 break;
960 case Opt_journal_inum:
961 if (is_remount) {
962 printk(KERN_ERR "EXT3-fs: cannot specify "
963 "journal on remount\n");
964 return 0;
965 }
966 if (match_int(&args[0], &option))
967 return 0;
968 *inum = option;
969 break;
970 case Opt_journal_dev:
971 if (is_remount) {
972 printk(KERN_ERR "EXT3-fs: cannot specify "
973 "journal on remount\n");
974 return 0;
975 }
976 if (match_int(&args[0], &option))
977 return 0;
978 *journal_devnum = option;
979 break;
980 case Opt_noload:
981 set_opt (sbi->s_mount_opt, NOLOAD);
982 break;
983 case Opt_commit:
984 if (match_int(&args[0], &option))
985 return 0;
986 if (option < 0)
987 return 0;
988 if (option == 0)
989 option = JBD_DEFAULT_MAX_COMMIT_AGE;
990 sbi->s_commit_interval = HZ * option;
991 break;
992 case Opt_data_journal:
993 data_opt = EXT3_MOUNT_JOURNAL_DATA;
994 goto datacheck;
995 case Opt_data_ordered:
996 data_opt = EXT3_MOUNT_ORDERED_DATA;
997 goto datacheck;
998 case Opt_data_writeback:
999 data_opt = EXT3_MOUNT_WRITEBACK_DATA;
1000 datacheck:
1001 if (is_remount) {
1002 if ((sbi->s_mount_opt & EXT3_MOUNT_DATA_FLAGS)
1003 != data_opt) {
1004 printk(KERN_ERR
1005 "EXT3-fs: cannot change data "
1006 "mode on remount\n");
1007 return 0;
1008 }
1009 } else {
1010 sbi->s_mount_opt &= ~EXT3_MOUNT_DATA_FLAGS;
1011 sbi->s_mount_opt |= data_opt;
1012 }
1013 break;
1014 #ifdef CONFIG_QUOTA
1015 case Opt_usrjquota:
1016 qtype = USRQUOTA;
1017 goto set_qf_name;
1018 case Opt_grpjquota:
1019 qtype = GRPQUOTA;
1020 set_qf_name:
1021 if ((sb_any_quota_enabled(sb) ||
1022 sb_any_quota_suspended(sb)) &&
1023 !sbi->s_qf_names[qtype]) {
1024 printk(KERN_ERR
1025 "EXT3-fs: Cannot change journaled "
1026 "quota options when quota turned on.\n");
1027 return 0;
1028 }
1029 qname = match_strdup(&args[0]);
1030 if (!qname) {
1031 printk(KERN_ERR
1032 "EXT3-fs: not enough memory for "
1033 "storing quotafile name.\n");
1034 return 0;
1035 }
1036 if (sbi->s_qf_names[qtype] &&
1037 strcmp(sbi->s_qf_names[qtype], qname)) {
1038 printk(KERN_ERR
1039 "EXT3-fs: %s quota file already "
1040 "specified.\n", QTYPE2NAME(qtype));
1041 kfree(qname);
1042 return 0;
1043 }
1044 sbi->s_qf_names[qtype] = qname;
1045 if (strchr(sbi->s_qf_names[qtype], '/')) {
1046 printk(KERN_ERR
1047 "EXT3-fs: quotafile must be on "
1048 "filesystem root.\n");
1049 kfree(sbi->s_qf_names[qtype]);
1050 sbi->s_qf_names[qtype] = NULL;
1051 return 0;
1052 }
1053 set_opt(sbi->s_mount_opt, QUOTA);
1054 break;
1055 case Opt_offusrjquota:
1056 qtype = USRQUOTA;
1057 goto clear_qf_name;
1058 case Opt_offgrpjquota:
1059 qtype = GRPQUOTA;
1060 clear_qf_name:
1061 if ((sb_any_quota_enabled(sb) ||
1062 sb_any_quota_suspended(sb)) &&
1063 sbi->s_qf_names[qtype]) {
1064 printk(KERN_ERR "EXT3-fs: Cannot change "
1065 "journaled quota options when "
1066 "quota turned on.\n");
1067 return 0;
1068 }
1069 /*
1070 * The space will be released later when all options
1071 * are confirmed to be correct
1072 */
1073 sbi->s_qf_names[qtype] = NULL;
1074 break;
1075 case Opt_jqfmt_vfsold:
1076 qfmt = QFMT_VFS_OLD;
1077 goto set_qf_format;
1078 case Opt_jqfmt_vfsv0:
1079 qfmt = QFMT_VFS_V0;
1080 set_qf_format:
1081 if ((sb_any_quota_enabled(sb) ||
1082 sb_any_quota_suspended(sb)) &&
1083 sbi->s_jquota_fmt != qfmt) {
1084 printk(KERN_ERR "EXT3-fs: Cannot change "
1085 "journaled quota options when "
1086 "quota turned on.\n");
1087 return 0;
1088 }
1089 sbi->s_jquota_fmt = qfmt;
1090 break;
1091 case Opt_quota:
1092 case Opt_usrquota:
1093 set_opt(sbi->s_mount_opt, QUOTA);
1094 set_opt(sbi->s_mount_opt, USRQUOTA);
1095 break;
1096 case Opt_grpquota:
1097 set_opt(sbi->s_mount_opt, QUOTA);
1098 set_opt(sbi->s_mount_opt, GRPQUOTA);
1099 break;
1100 case Opt_noquota:
1101 if (sb_any_quota_enabled(sb) ||
1102 sb_any_quota_suspended(sb)) {
1103 printk(KERN_ERR "EXT3-fs: Cannot change quota "
1104 "options when quota turned on.\n");
1105 return 0;
1106 }
1107 clear_opt(sbi->s_mount_opt, QUOTA);
1108 clear_opt(sbi->s_mount_opt, USRQUOTA);
1109 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1110 break;
1111 #else
1112 case Opt_quota:
1113 case Opt_usrquota:
1114 case Opt_grpquota:
1115 printk(KERN_ERR
1116 "EXT3-fs: quota options not supported.\n");
1117 break;
1118 case Opt_usrjquota:
1119 case Opt_grpjquota:
1120 case Opt_offusrjquota:
1121 case Opt_offgrpjquota:
1122 case Opt_jqfmt_vfsold:
1123 case Opt_jqfmt_vfsv0:
1124 printk(KERN_ERR
1125 "EXT3-fs: journaled quota options not "
1126 "supported.\n");
1127 break;
1128 case Opt_noquota:
1129 break;
1130 #endif
1131 case Opt_abort:
1132 set_opt(sbi->s_mount_opt, ABORT);
1133 break;
1134 case Opt_barrier:
1135 if (match_int(&args[0], &option))
1136 return 0;
1137 if (option)
1138 set_opt(sbi->s_mount_opt, BARRIER);
1139 else
1140 clear_opt(sbi->s_mount_opt, BARRIER);
1141 break;
1142 case Opt_ignore:
1143 break;
1144 case Opt_resize:
1145 if (!is_remount) {
1146 printk("EXT3-fs: resize option only available "
1147 "for remount\n");
1148 return 0;
1149 }
1150 if (match_int(&args[0], &option) != 0)
1151 return 0;
1152 *n_blocks_count = option;
1153 break;
1154 case Opt_nobh:
1155 set_opt(sbi->s_mount_opt, NOBH);
1156 break;
1157 case Opt_bh:
1158 clear_opt(sbi->s_mount_opt, NOBH);
1159 break;
1160 default:
1161 printk (KERN_ERR
1162 "EXT3-fs: Unrecognized mount option \"%s\" "
1163 "or missing value\n", p);
1164 return 0;
1165 }
1166 }
1167 #ifdef CONFIG_QUOTA
1168 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1169 if ((sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA) &&
1170 sbi->s_qf_names[USRQUOTA])
1171 clear_opt(sbi->s_mount_opt, USRQUOTA);
1172
1173 if ((sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA) &&
1174 sbi->s_qf_names[GRPQUOTA])
1175 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1176
1177 if ((sbi->s_qf_names[USRQUOTA] &&
1178 (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)) ||
1179 (sbi->s_qf_names[GRPQUOTA] &&
1180 (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA))) {
1181 printk(KERN_ERR "EXT3-fs: old and new quota "
1182 "format mixing.\n");
1183 return 0;
1184 }
1185
1186 if (!sbi->s_jquota_fmt) {
1187 printk(KERN_ERR "EXT3-fs: journaled quota format "
1188 "not specified.\n");
1189 return 0;
1190 }
1191 } else {
1192 if (sbi->s_jquota_fmt) {
1193 printk(KERN_ERR "EXT3-fs: journaled quota format "
1194 "specified with no journaling "
1195 "enabled.\n");
1196 return 0;
1197 }
1198 }
1199 #endif
1200 return 1;
1201 }
1202
1203 static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
1204 int read_only)
1205 {
1206 struct ext3_sb_info *sbi = EXT3_SB(sb);
1207 int res = 0;
1208
1209 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
1210 printk (KERN_ERR "EXT3-fs warning: revision level too high, "
1211 "forcing read-only mode\n");
1212 res = MS_RDONLY;
1213 }
1214 if (read_only)
1215 return res;
1216 if (!(sbi->s_mount_state & EXT3_VALID_FS))
1217 printk (KERN_WARNING "EXT3-fs warning: mounting unchecked fs, "
1218 "running e2fsck is recommended\n");
1219 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
1220 printk (KERN_WARNING
1221 "EXT3-fs warning: mounting fs with errors, "
1222 "running e2fsck is recommended\n");
1223 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1224 le16_to_cpu(es->s_mnt_count) >=
1225 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1226 printk (KERN_WARNING
1227 "EXT3-fs warning: maximal mount count reached, "
1228 "running e2fsck is recommended\n");
1229 else if (le32_to_cpu(es->s_checkinterval) &&
1230 (le32_to_cpu(es->s_lastcheck) +
1231 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1232 printk (KERN_WARNING
1233 "EXT3-fs warning: checktime reached, "
1234 "running e2fsck is recommended\n");
1235 #if 0
1236 /* @@@ We _will_ want to clear the valid bit if we find
1237 inconsistencies, to force a fsck at reboot. But for
1238 a plain journaled filesystem we can keep it set as
1239 valid forever! :) */
1240 es->s_state &= cpu_to_le16(~EXT3_VALID_FS);
1241 #endif
1242 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1243 es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
1244 le16_add_cpu(&es->s_mnt_count, 1);
1245 es->s_mtime = cpu_to_le32(get_seconds());
1246 ext3_update_dynamic_rev(sb);
1247 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1248
1249 ext3_commit_super(sb, es, 1);
1250 if (test_opt(sb, DEBUG))
1251 printk(KERN_INFO "[EXT3 FS bs=%lu, gc=%lu, "
1252 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1253 sb->s_blocksize,
1254 sbi->s_groups_count,
1255 EXT3_BLOCKS_PER_GROUP(sb),
1256 EXT3_INODES_PER_GROUP(sb),
1257 sbi->s_mount_opt);
1258
1259 printk(KERN_INFO "EXT3 FS on %s, ", sb->s_id);
1260 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
1261 char b[BDEVNAME_SIZE];
1262
1263 printk("external journal on %s\n",
1264 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1265 } else {
1266 printk("internal journal\n");
1267 }
1268 return res;
1269 }
1270
1271 /* Called at mount-time, super-block is locked */
1272 static int ext3_check_descriptors(struct super_block *sb)
1273 {
1274 struct ext3_sb_info *sbi = EXT3_SB(sb);
1275 int i;
1276
1277 ext3_debug ("Checking group descriptors");
1278
1279 for (i = 0; i < sbi->s_groups_count; i++) {
1280 struct ext3_group_desc *gdp = ext3_get_group_desc(sb, i, NULL);
1281 ext3_fsblk_t first_block = ext3_group_first_block_no(sb, i);
1282 ext3_fsblk_t last_block;
1283
1284 if (i == sbi->s_groups_count - 1)
1285 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
1286 else
1287 last_block = first_block +
1288 (EXT3_BLOCKS_PER_GROUP(sb) - 1);
1289
1290 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
1291 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
1292 {
1293 ext3_error (sb, "ext3_check_descriptors",
1294 "Block bitmap for group %d"
1295 " not in group (block %lu)!",
1296 i, (unsigned long)
1297 le32_to_cpu(gdp->bg_block_bitmap));
1298 return 0;
1299 }
1300 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
1301 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
1302 {
1303 ext3_error (sb, "ext3_check_descriptors",
1304 "Inode bitmap for group %d"
1305 " not in group (block %lu)!",
1306 i, (unsigned long)
1307 le32_to_cpu(gdp->bg_inode_bitmap));
1308 return 0;
1309 }
1310 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
1311 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
1312 last_block)
1313 {
1314 ext3_error (sb, "ext3_check_descriptors",
1315 "Inode table for group %d"
1316 " not in group (block %lu)!",
1317 i, (unsigned long)
1318 le32_to_cpu(gdp->bg_inode_table));
1319 return 0;
1320 }
1321 }
1322
1323 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1324 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1325 return 1;
1326 }
1327
1328
1329 /* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1330 * the superblock) which were deleted from all directories, but held open by
1331 * a process at the time of a crash. We walk the list and try to delete these
1332 * inodes at recovery time (only with a read-write filesystem).
1333 *
1334 * In order to keep the orphan inode chain consistent during traversal (in
1335 * case of crash during recovery), we link each inode into the superblock
1336 * orphan list_head and handle it the same way as an inode deletion during
1337 * normal operation (which journals the operations for us).
1338 *
1339 * We only do an iget() and an iput() on each inode, which is very safe if we
1340 * accidentally point at an in-use or already deleted inode. The worst that
1341 * can happen in this case is that we get a "bit already cleared" message from
1342 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1343 * e2fsck was run on this filesystem, and it must have already done the orphan
1344 * inode cleanup for us, so we can safely abort without any further action.
1345 */
1346 static void ext3_orphan_cleanup (struct super_block * sb,
1347 struct ext3_super_block * es)
1348 {
1349 unsigned int s_flags = sb->s_flags;
1350 int nr_orphans = 0, nr_truncates = 0;
1351 #ifdef CONFIG_QUOTA
1352 int i;
1353 #endif
1354 if (!es->s_last_orphan) {
1355 jbd_debug(4, "no orphan inodes to clean up\n");
1356 return;
1357 }
1358
1359 if (bdev_read_only(sb->s_bdev)) {
1360 printk(KERN_ERR "EXT3-fs: write access "
1361 "unavailable, skipping orphan cleanup.\n");
1362 return;
1363 }
1364
1365 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1366 if (es->s_last_orphan)
1367 jbd_debug(1, "Errors on filesystem, "
1368 "clearing orphan list.\n");
1369 es->s_last_orphan = 0;
1370 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1371 return;
1372 }
1373
1374 if (s_flags & MS_RDONLY) {
1375 printk(KERN_INFO "EXT3-fs: %s: orphan cleanup on readonly fs\n",
1376 sb->s_id);
1377 sb->s_flags &= ~MS_RDONLY;
1378 }
1379 #ifdef CONFIG_QUOTA
1380 /* Needed for iput() to work correctly and not trash data */
1381 sb->s_flags |= MS_ACTIVE;
1382 /* Turn on quotas so that they are updated correctly */
1383 for (i = 0; i < MAXQUOTAS; i++) {
1384 if (EXT3_SB(sb)->s_qf_names[i]) {
1385 int ret = ext3_quota_on_mount(sb, i);
1386 if (ret < 0)
1387 printk(KERN_ERR
1388 "EXT3-fs: Cannot turn on journaled "
1389 "quota: error %d\n", ret);
1390 }
1391 }
1392 #endif
1393
1394 while (es->s_last_orphan) {
1395 struct inode *inode;
1396
1397 inode = ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1398 if (IS_ERR(inode)) {
1399 es->s_last_orphan = 0;
1400 break;
1401 }
1402
1403 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
1404 DQUOT_INIT(inode);
1405 if (inode->i_nlink) {
1406 printk(KERN_DEBUG
1407 "%s: truncating inode %lu to %Ld bytes\n",
1408 __func__, inode->i_ino, inode->i_size);
1409 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1410 inode->i_ino, inode->i_size);
1411 ext3_truncate(inode);
1412 nr_truncates++;
1413 } else {
1414 printk(KERN_DEBUG
1415 "%s: deleting unreferenced inode %lu\n",
1416 __func__, inode->i_ino);
1417 jbd_debug(2, "deleting unreferenced inode %lu\n",
1418 inode->i_ino);
1419 nr_orphans++;
1420 }
1421 iput(inode); /* The delete magic happens here! */
1422 }
1423
1424 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1425
1426 if (nr_orphans)
1427 printk(KERN_INFO "EXT3-fs: %s: %d orphan inode%s deleted\n",
1428 sb->s_id, PLURAL(nr_orphans));
1429 if (nr_truncates)
1430 printk(KERN_INFO "EXT3-fs: %s: %d truncate%s cleaned up\n",
1431 sb->s_id, PLURAL(nr_truncates));
1432 #ifdef CONFIG_QUOTA
1433 /* Turn quotas off */
1434 for (i = 0; i < MAXQUOTAS; i++) {
1435 if (sb_dqopt(sb)->files[i])
1436 vfs_quota_off(sb, i, 0);
1437 }
1438 #endif
1439 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1440 }
1441
1442 /*
1443 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1444 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1445 * We need to be 1 filesystem block less than the 2^32 sector limit.
1446 */
1447 static loff_t ext3_max_size(int bits)
1448 {
1449 loff_t res = EXT3_NDIR_BLOCKS;
1450 int meta_blocks;
1451 loff_t upper_limit;
1452
1453 /* This is calculated to be the largest file size for a
1454 * dense, file such that the total number of
1455 * sectors in the file, including data and all indirect blocks,
1456 * does not exceed 2^32 -1
1457 * __u32 i_blocks representing the total number of
1458 * 512 bytes blocks of the file
1459 */
1460 upper_limit = (1LL << 32) - 1;
1461
1462 /* total blocks in file system block size */
1463 upper_limit >>= (bits - 9);
1464
1465
1466 /* indirect blocks */
1467 meta_blocks = 1;
1468 /* double indirect blocks */
1469 meta_blocks += 1 + (1LL << (bits-2));
1470 /* tripple indirect blocks */
1471 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1472
1473 upper_limit -= meta_blocks;
1474 upper_limit <<= bits;
1475
1476 res += 1LL << (bits-2);
1477 res += 1LL << (2*(bits-2));
1478 res += 1LL << (3*(bits-2));
1479 res <<= bits;
1480 if (res > upper_limit)
1481 res = upper_limit;
1482
1483 if (res > MAX_LFS_FILESIZE)
1484 res = MAX_LFS_FILESIZE;
1485
1486 return res;
1487 }
1488
1489 static ext3_fsblk_t descriptor_loc(struct super_block *sb,
1490 ext3_fsblk_t logic_sb_block,
1491 int nr)
1492 {
1493 struct ext3_sb_info *sbi = EXT3_SB(sb);
1494 unsigned long bg, first_meta_bg;
1495 int has_super = 0;
1496
1497 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1498
1499 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1500 nr < first_meta_bg)
1501 return (logic_sb_block + nr + 1);
1502 bg = sbi->s_desc_per_block * nr;
1503 if (ext3_bg_has_super(sb, bg))
1504 has_super = 1;
1505 return (has_super + ext3_group_first_block_no(sb, bg));
1506 }
1507
1508
1509 static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1510 {
1511 struct buffer_head * bh;
1512 struct ext3_super_block *es = NULL;
1513 struct ext3_sb_info *sbi;
1514 ext3_fsblk_t block;
1515 ext3_fsblk_t sb_block = get_sb_block(&data);
1516 ext3_fsblk_t logic_sb_block;
1517 unsigned long offset = 0;
1518 unsigned int journal_inum = 0;
1519 unsigned long journal_devnum = 0;
1520 unsigned long def_mount_opts;
1521 struct inode *root;
1522 int blocksize;
1523 int hblock;
1524 int db_count;
1525 int i;
1526 int needs_recovery;
1527 int ret = -EINVAL;
1528 __le32 features;
1529 int err;
1530
1531 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1532 if (!sbi)
1533 return -ENOMEM;
1534 sb->s_fs_info = sbi;
1535 sbi->s_mount_opt = 0;
1536 sbi->s_resuid = EXT3_DEF_RESUID;
1537 sbi->s_resgid = EXT3_DEF_RESGID;
1538 sbi->s_sb_block = sb_block;
1539
1540 unlock_kernel();
1541
1542 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1543 if (!blocksize) {
1544 printk(KERN_ERR "EXT3-fs: unable to set blocksize\n");
1545 goto out_fail;
1546 }
1547
1548 /*
1549 * The ext3 superblock will not be buffer aligned for other than 1kB
1550 * block sizes. We need to calculate the offset from buffer start.
1551 */
1552 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1553 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1554 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1555 } else {
1556 logic_sb_block = sb_block;
1557 }
1558
1559 if (!(bh = sb_bread(sb, logic_sb_block))) {
1560 printk (KERN_ERR "EXT3-fs: unable to read superblock\n");
1561 goto out_fail;
1562 }
1563 /*
1564 * Note: s_es must be initialized as soon as possible because
1565 * some ext3 macro-instructions depend on its value
1566 */
1567 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1568 sbi->s_es = es;
1569 sb->s_magic = le16_to_cpu(es->s_magic);
1570 if (sb->s_magic != EXT3_SUPER_MAGIC)
1571 goto cantfind_ext3;
1572
1573 /* Set defaults before we parse the mount options */
1574 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1575 if (def_mount_opts & EXT3_DEFM_DEBUG)
1576 set_opt(sbi->s_mount_opt, DEBUG);
1577 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1578 set_opt(sbi->s_mount_opt, GRPID);
1579 if (def_mount_opts & EXT3_DEFM_UID16)
1580 set_opt(sbi->s_mount_opt, NO_UID32);
1581 #ifdef CONFIG_EXT3_FS_XATTR
1582 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1583 set_opt(sbi->s_mount_opt, XATTR_USER);
1584 #endif
1585 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1586 if (def_mount_opts & EXT3_DEFM_ACL)
1587 set_opt(sbi->s_mount_opt, POSIX_ACL);
1588 #endif
1589 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
1590 sbi->s_mount_opt |= EXT3_MOUNT_JOURNAL_DATA;
1591 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
1592 sbi->s_mount_opt |= EXT3_MOUNT_ORDERED_DATA;
1593 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
1594 sbi->s_mount_opt |= EXT3_MOUNT_WRITEBACK_DATA;
1595
1596 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1597 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1598 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_CONTINUE)
1599 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1600 else
1601 set_opt(sbi->s_mount_opt, ERRORS_RO);
1602
1603 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1604 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1605
1606 set_opt(sbi->s_mount_opt, RESERVATION);
1607
1608 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1609 NULL, 0))
1610 goto failed_mount;
1611
1612 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1613 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1614
1615 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1616 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1617 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1618 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1619 printk(KERN_WARNING
1620 "EXT3-fs warning: feature flags set on rev 0 fs, "
1621 "running e2fsck is recommended\n");
1622 /*
1623 * Check feature flags regardless of the revision level, since we
1624 * previously didn't change the revision level when setting the flags,
1625 * so there is a chance incompat flags are set on a rev 0 filesystem.
1626 */
1627 features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
1628 if (features) {
1629 printk(KERN_ERR "EXT3-fs: %s: couldn't mount because of "
1630 "unsupported optional features (%x).\n",
1631 sb->s_id, le32_to_cpu(features));
1632 goto failed_mount;
1633 }
1634 features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
1635 if (!(sb->s_flags & MS_RDONLY) && features) {
1636 printk(KERN_ERR "EXT3-fs: %s: couldn't mount RDWR because of "
1637 "unsupported optional features (%x).\n",
1638 sb->s_id, le32_to_cpu(features));
1639 goto failed_mount;
1640 }
1641 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1642
1643 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1644 blocksize > EXT3_MAX_BLOCK_SIZE) {
1645 printk(KERN_ERR
1646 "EXT3-fs: Unsupported filesystem blocksize %d on %s.\n",
1647 blocksize, sb->s_id);
1648 goto failed_mount;
1649 }
1650
1651 hblock = bdev_hardsect_size(sb->s_bdev);
1652 if (sb->s_blocksize != blocksize) {
1653 /*
1654 * Make sure the blocksize for the filesystem is larger
1655 * than the hardware sectorsize for the machine.
1656 */
1657 if (blocksize < hblock) {
1658 printk(KERN_ERR "EXT3-fs: blocksize %d too small for "
1659 "device blocksize %d.\n", blocksize, hblock);
1660 goto failed_mount;
1661 }
1662
1663 brelse (bh);
1664 if (!sb_set_blocksize(sb, blocksize)) {
1665 printk(KERN_ERR "EXT3-fs: bad blocksize %d.\n",
1666 blocksize);
1667 goto out_fail;
1668 }
1669 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1670 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1671 bh = sb_bread(sb, logic_sb_block);
1672 if (!bh) {
1673 printk(KERN_ERR
1674 "EXT3-fs: Can't read superblock on 2nd try.\n");
1675 goto failed_mount;
1676 }
1677 es = (struct ext3_super_block *)(((char *)bh->b_data) + offset);
1678 sbi->s_es = es;
1679 if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
1680 printk (KERN_ERR
1681 "EXT3-fs: Magic mismatch, very weird !\n");
1682 goto failed_mount;
1683 }
1684 }
1685
1686 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1687
1688 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1689 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1690 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1691 } else {
1692 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1693 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1694 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
1695 (!is_power_of_2(sbi->s_inode_size)) ||
1696 (sbi->s_inode_size > blocksize)) {
1697 printk (KERN_ERR
1698 "EXT3-fs: unsupported inode size: %d\n",
1699 sbi->s_inode_size);
1700 goto failed_mount;
1701 }
1702 }
1703 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1704 le32_to_cpu(es->s_log_frag_size);
1705 if (blocksize != sbi->s_frag_size) {
1706 printk(KERN_ERR
1707 "EXT3-fs: fragsize %lu != blocksize %u (unsupported)\n",
1708 sbi->s_frag_size, blocksize);
1709 goto failed_mount;
1710 }
1711 sbi->s_frags_per_block = 1;
1712 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1713 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1714 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1715 if (EXT3_INODE_SIZE(sb) == 0 || EXT3_INODES_PER_GROUP(sb) == 0)
1716 goto cantfind_ext3;
1717 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1718 if (sbi->s_inodes_per_block == 0)
1719 goto cantfind_ext3;
1720 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1721 sbi->s_inodes_per_block;
1722 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1723 sbi->s_sbh = bh;
1724 sbi->s_mount_state = le16_to_cpu(es->s_state);
1725 sbi->s_addr_per_block_bits = ilog2(EXT3_ADDR_PER_BLOCK(sb));
1726 sbi->s_desc_per_block_bits = ilog2(EXT3_DESC_PER_BLOCK(sb));
1727 for (i=0; i < 4; i++)
1728 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1729 sbi->s_def_hash_version = es->s_def_hash_version;
1730
1731 if (sbi->s_blocks_per_group > blocksize * 8) {
1732 printk (KERN_ERR
1733 "EXT3-fs: #blocks per group too big: %lu\n",
1734 sbi->s_blocks_per_group);
1735 goto failed_mount;
1736 }
1737 if (sbi->s_frags_per_group > blocksize * 8) {
1738 printk (KERN_ERR
1739 "EXT3-fs: #fragments per group too big: %lu\n",
1740 sbi->s_frags_per_group);
1741 goto failed_mount;
1742 }
1743 if (sbi->s_inodes_per_group > blocksize * 8) {
1744 printk (KERN_ERR
1745 "EXT3-fs: #inodes per group too big: %lu\n",
1746 sbi->s_inodes_per_group);
1747 goto failed_mount;
1748 }
1749
1750 if (le32_to_cpu(es->s_blocks_count) >
1751 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
1752 printk(KERN_ERR "EXT3-fs: filesystem on %s:"
1753 " too large to mount safely\n", sb->s_id);
1754 if (sizeof(sector_t) < 8)
1755 printk(KERN_WARNING "EXT3-fs: CONFIG_LBD not "
1756 "enabled\n");
1757 goto failed_mount;
1758 }
1759
1760 if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1761 goto cantfind_ext3;
1762 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1763 le32_to_cpu(es->s_first_data_block) - 1)
1764 / EXT3_BLOCKS_PER_GROUP(sb)) + 1;
1765 db_count = (sbi->s_groups_count + EXT3_DESC_PER_BLOCK(sb) - 1) /
1766 EXT3_DESC_PER_BLOCK(sb);
1767 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1768 GFP_KERNEL);
1769 if (sbi->s_group_desc == NULL) {
1770 printk (KERN_ERR "EXT3-fs: not enough memory\n");
1771 goto failed_mount;
1772 }
1773
1774 bgl_lock_init(&sbi->s_blockgroup_lock);
1775
1776 for (i = 0; i < db_count; i++) {
1777 block = descriptor_loc(sb, logic_sb_block, i);
1778 sbi->s_group_desc[i] = sb_bread(sb, block);
1779 if (!sbi->s_group_desc[i]) {
1780 printk (KERN_ERR "EXT3-fs: "
1781 "can't read group descriptor %d\n", i);
1782 db_count = i;
1783 goto failed_mount2;
1784 }
1785 }
1786 if (!ext3_check_descriptors (sb)) {
1787 printk(KERN_ERR "EXT3-fs: group descriptors corrupted!\n");
1788 goto failed_mount2;
1789 }
1790 sbi->s_gdb_count = db_count;
1791 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1792 spin_lock_init(&sbi->s_next_gen_lock);
1793
1794 err = percpu_counter_init(&sbi->s_freeblocks_counter,
1795 ext3_count_free_blocks(sb));
1796 if (!err) {
1797 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1798 ext3_count_free_inodes(sb));
1799 }
1800 if (!err) {
1801 err = percpu_counter_init(&sbi->s_dirs_counter,
1802 ext3_count_dirs(sb));
1803 }
1804 if (err) {
1805 printk(KERN_ERR "EXT3-fs: insufficient memory\n");
1806 goto failed_mount3;
1807 }
1808
1809 /* per fileystem reservation list head & lock */
1810 spin_lock_init(&sbi->s_rsv_window_lock);
1811 sbi->s_rsv_window_root = RB_ROOT;
1812 /* Add a single, static dummy reservation to the start of the
1813 * reservation window list --- it gives us a placeholder for
1814 * append-at-start-of-list which makes the allocation logic
1815 * _much_ simpler. */
1816 sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1817 sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1818 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1819 sbi->s_rsv_window_head.rsv_goal_size = 0;
1820 ext3_rsv_window_add(sb, &sbi->s_rsv_window_head);
1821
1822 /*
1823 * set up enough so that it can read an inode
1824 */
1825 sb->s_op = &ext3_sops;
1826 sb->s_export_op = &ext3_export_ops;
1827 sb->s_xattr = ext3_xattr_handlers;
1828 #ifdef CONFIG_QUOTA
1829 sb->s_qcop = &ext3_qctl_operations;
1830 sb->dq_op = &ext3_quota_operations;
1831 #endif
1832 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1833
1834 sb->s_root = NULL;
1835
1836 needs_recovery = (es->s_last_orphan != 0 ||
1837 EXT3_HAS_INCOMPAT_FEATURE(sb,
1838 EXT3_FEATURE_INCOMPAT_RECOVER));
1839
1840 /*
1841 * The first inode we look at is the journal inode. Don't try
1842 * root first: it may be modified in the journal!
1843 */
1844 if (!test_opt(sb, NOLOAD) &&
1845 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
1846 if (ext3_load_journal(sb, es, journal_devnum))
1847 goto failed_mount3;
1848 } else if (journal_inum) {
1849 if (ext3_create_journal(sb, es, journal_inum))
1850 goto failed_mount3;
1851 } else {
1852 if (!silent)
1853 printk (KERN_ERR
1854 "ext3: No journal on filesystem on %s\n",
1855 sb->s_id);
1856 goto failed_mount3;
1857 }
1858
1859 /* We have now updated the journal if required, so we can
1860 * validate the data journaling mode. */
1861 switch (test_opt(sb, DATA_FLAGS)) {
1862 case 0:
1863 /* No mode set, assume a default based on the journal
1864 capabilities: ORDERED_DATA if the journal can
1865 cope, else JOURNAL_DATA */
1866 if (journal_check_available_features
1867 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
1868 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1869 else
1870 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1871 break;
1872
1873 case EXT3_MOUNT_ORDERED_DATA:
1874 case EXT3_MOUNT_WRITEBACK_DATA:
1875 if (!journal_check_available_features
1876 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
1877 printk(KERN_ERR "EXT3-fs: Journal does not support "
1878 "requested data journaling mode\n");
1879 goto failed_mount4;
1880 }
1881 default:
1882 break;
1883 }
1884
1885 if (test_opt(sb, NOBH)) {
1886 if (!(test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)) {
1887 printk(KERN_WARNING "EXT3-fs: Ignoring nobh option - "
1888 "its supported only with writeback mode\n");
1889 clear_opt(sbi->s_mount_opt, NOBH);
1890 }
1891 }
1892 /*
1893 * The journal_load will have done any necessary log recovery,
1894 * so we can safely mount the rest of the filesystem now.
1895 */
1896
1897 root = ext3_iget(sb, EXT3_ROOT_INO);
1898 if (IS_ERR(root)) {
1899 printk(KERN_ERR "EXT3-fs: get root inode failed\n");
1900 ret = PTR_ERR(root);
1901 goto failed_mount4;
1902 }
1903 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1904 iput(root);
1905 printk(KERN_ERR "EXT3-fs: corrupt root inode, run e2fsck\n");
1906 goto failed_mount4;
1907 }
1908 sb->s_root = d_alloc_root(root);
1909 if (!sb->s_root) {
1910 printk(KERN_ERR "EXT3-fs: get root dentry failed\n");
1911 iput(root);
1912 ret = -ENOMEM;
1913 goto failed_mount4;
1914 }
1915
1916 ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1917 /*
1918 * akpm: core read_super() calls in here with the superblock locked.
1919 * That deadlocks, because orphan cleanup needs to lock the superblock
1920 * in numerous places. Here we just pop the lock - it's relatively
1921 * harmless, because we are now ready to accept write_super() requests,
1922 * and aviro says that's the only reason for hanging onto the
1923 * superblock lock.
1924 */
1925 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
1926 ext3_orphan_cleanup(sb, es);
1927 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
1928 if (needs_recovery)
1929 printk (KERN_INFO "EXT3-fs: recovery complete.\n");
1930 ext3_mark_recovery_complete(sb, es);
1931 printk (KERN_INFO "EXT3-fs: mounted filesystem with %s data mode.\n",
1932 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
1933 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
1934 "writeback");
1935
1936 lock_kernel();
1937 return 0;
1938
1939 cantfind_ext3:
1940 if (!silent)
1941 printk(KERN_ERR "VFS: Can't find ext3 filesystem on dev %s.\n",
1942 sb->s_id);
1943 goto failed_mount;
1944
1945 failed_mount4:
1946 journal_destroy(sbi->s_journal);
1947 failed_mount3:
1948 percpu_counter_destroy(&sbi->s_freeblocks_counter);
1949 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1950 percpu_counter_destroy(&sbi->s_dirs_counter);
1951 failed_mount2:
1952 for (i = 0; i < db_count; i++)
1953 brelse(sbi->s_group_desc[i]);
1954 kfree(sbi->s_group_desc);
1955 failed_mount:
1956 #ifdef CONFIG_QUOTA
1957 for (i = 0; i < MAXQUOTAS; i++)
1958 kfree(sbi->s_qf_names[i]);
1959 #endif
1960 ext3_blkdev_remove(sbi);
1961 brelse(bh);
1962 out_fail:
1963 sb->s_fs_info = NULL;
1964 kfree(sbi);
1965 lock_kernel();
1966 return ret;
1967 }
1968
1969 /*
1970 * Setup any per-fs journal parameters now. We'll do this both on
1971 * initial mount, once the journal has been initialised but before we've
1972 * done any recovery; and again on any subsequent remount.
1973 */
1974 static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
1975 {
1976 struct ext3_sb_info *sbi = EXT3_SB(sb);
1977
1978 if (sbi->s_commit_interval)
1979 journal->j_commit_interval = sbi->s_commit_interval;
1980 /* We could also set up an ext3-specific default for the commit
1981 * interval here, but for now we'll just fall back to the jbd
1982 * default. */
1983
1984 spin_lock(&journal->j_state_lock);
1985 if (test_opt(sb, BARRIER))
1986 journal->j_flags |= JFS_BARRIER;
1987 else
1988 journal->j_flags &= ~JFS_BARRIER;
1989 spin_unlock(&journal->j_state_lock);
1990 }
1991
1992 static journal_t *ext3_get_journal(struct super_block *sb,
1993 unsigned int journal_inum)
1994 {
1995 struct inode *journal_inode;
1996 journal_t *journal;
1997
1998 /* First, test for the existence of a valid inode on disk. Bad
1999 * things happen if we iget() an unused inode, as the subsequent
2000 * iput() will try to delete it. */
2001
2002 journal_inode = ext3_iget(sb, journal_inum);
2003 if (IS_ERR(journal_inode)) {
2004 printk(KERN_ERR "EXT3-fs: no journal found.\n");
2005 return NULL;
2006 }
2007 if (!journal_inode->i_nlink) {
2008 make_bad_inode(journal_inode);
2009 iput(journal_inode);
2010 printk(KERN_ERR "EXT3-fs: journal inode is deleted.\n");
2011 return NULL;
2012 }
2013
2014 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2015 journal_inode, journal_inode->i_size);
2016 if (!S_ISREG(journal_inode->i_mode)) {
2017 printk(KERN_ERR "EXT3-fs: invalid journal inode.\n");
2018 iput(journal_inode);
2019 return NULL;
2020 }
2021
2022 journal = journal_init_inode(journal_inode);
2023 if (!journal) {
2024 printk(KERN_ERR "EXT3-fs: Could not load journal inode\n");
2025 iput(journal_inode);
2026 return NULL;
2027 }
2028 journal->j_private = sb;
2029 ext3_init_journal_params(sb, journal);
2030 return journal;
2031 }
2032
2033 static journal_t *ext3_get_dev_journal(struct super_block *sb,
2034 dev_t j_dev)
2035 {
2036 struct buffer_head * bh;
2037 journal_t *journal;
2038 ext3_fsblk_t start;
2039 ext3_fsblk_t len;
2040 int hblock, blocksize;
2041 ext3_fsblk_t sb_block;
2042 unsigned long offset;
2043 struct ext3_super_block * es;
2044 struct block_device *bdev;
2045
2046 bdev = ext3_blkdev_get(j_dev);
2047 if (bdev == NULL)
2048 return NULL;
2049
2050 if (bd_claim(bdev, sb)) {
2051 printk(KERN_ERR
2052 "EXT3: failed to claim external journal device.\n");
2053 blkdev_put(bdev);
2054 return NULL;
2055 }
2056
2057 blocksize = sb->s_blocksize;
2058 hblock = bdev_hardsect_size(bdev);
2059 if (blocksize < hblock) {
2060 printk(KERN_ERR
2061 "EXT3-fs: blocksize too small for journal device.\n");
2062 goto out_bdev;
2063 }
2064
2065 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
2066 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
2067 set_blocksize(bdev, blocksize);
2068 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2069 printk(KERN_ERR "EXT3-fs: couldn't read superblock of "
2070 "external journal\n");
2071 goto out_bdev;
2072 }
2073
2074 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
2075 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
2076 !(le32_to_cpu(es->s_feature_incompat) &
2077 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2078 printk(KERN_ERR "EXT3-fs: external journal has "
2079 "bad superblock\n");
2080 brelse(bh);
2081 goto out_bdev;
2082 }
2083
2084 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2085 printk(KERN_ERR "EXT3-fs: journal UUID does not match\n");
2086 brelse(bh);
2087 goto out_bdev;
2088 }
2089
2090 len = le32_to_cpu(es->s_blocks_count);
2091 start = sb_block + 1;
2092 brelse(bh); /* we're done with the superblock */
2093
2094 journal = journal_init_dev(bdev, sb->s_bdev,
2095 start, len, blocksize);
2096 if (!journal) {
2097 printk(KERN_ERR "EXT3-fs: failed to create device journal\n");
2098 goto out_bdev;
2099 }
2100 journal->j_private = sb;
2101 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2102 wait_on_buffer(journal->j_sb_buffer);
2103 if (!buffer_uptodate(journal->j_sb_buffer)) {
2104 printk(KERN_ERR "EXT3-fs: I/O error on journal device\n");
2105 goto out_journal;
2106 }
2107 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2108 printk(KERN_ERR "EXT3-fs: External journal has more than one "
2109 "user (unsupported) - %d\n",
2110 be32_to_cpu(journal->j_superblock->s_nr_users));
2111 goto out_journal;
2112 }
2113 EXT3_SB(sb)->journal_bdev = bdev;
2114 ext3_init_journal_params(sb, journal);
2115 return journal;
2116 out_journal:
2117 journal_destroy(journal);
2118 out_bdev:
2119 ext3_blkdev_put(bdev);
2120 return NULL;
2121 }
2122
2123 static int ext3_load_journal(struct super_block *sb,
2124 struct ext3_super_block *es,
2125 unsigned long journal_devnum)
2126 {
2127 journal_t *journal;
2128 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2129 dev_t journal_dev;
2130 int err = 0;
2131 int really_read_only;
2132
2133 if (journal_devnum &&
2134 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2135 printk(KERN_INFO "EXT3-fs: external journal device major/minor "
2136 "numbers have changed\n");
2137 journal_dev = new_decode_dev(journal_devnum);
2138 } else
2139 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2140
2141 really_read_only = bdev_read_only(sb->s_bdev);
2142
2143 /*
2144 * Are we loading a blank journal or performing recovery after a
2145 * crash? For recovery, we need to check in advance whether we
2146 * can get read-write access to the device.
2147 */
2148
2149 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
2150 if (sb->s_flags & MS_RDONLY) {
2151 printk(KERN_INFO "EXT3-fs: INFO: recovery "
2152 "required on readonly filesystem.\n");
2153 if (really_read_only) {
2154 printk(KERN_ERR "EXT3-fs: write access "
2155 "unavailable, cannot proceed.\n");
2156 return -EROFS;
2157 }
2158 printk (KERN_INFO "EXT3-fs: write access will "
2159 "be enabled during recovery.\n");
2160 }
2161 }
2162
2163 if (journal_inum && journal_dev) {
2164 printk(KERN_ERR "EXT3-fs: filesystem has both journal "
2165 "and inode journals!\n");
2166 return -EINVAL;
2167 }
2168
2169 if (journal_inum) {
2170 if (!(journal = ext3_get_journal(sb, journal_inum)))
2171 return -EINVAL;
2172 } else {
2173 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
2174 return -EINVAL;
2175 }
2176
2177 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2178 err = journal_update_format(journal);
2179 if (err) {
2180 printk(KERN_ERR "EXT3-fs: error updating journal.\n");
2181 journal_destroy(journal);
2182 return err;
2183 }
2184 }
2185
2186 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
2187 err = journal_wipe(journal, !really_read_only);
2188 if (!err)
2189 err = journal_load(journal);
2190
2191 if (err) {
2192 printk(KERN_ERR "EXT3-fs: error loading journal.\n");
2193 journal_destroy(journal);
2194 return err;
2195 }
2196
2197 EXT3_SB(sb)->s_journal = journal;
2198 ext3_clear_journal_err(sb, es);
2199
2200 if (journal_devnum &&
2201 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2202 es->s_journal_dev = cpu_to_le32(journal_devnum);
2203 sb->s_dirt = 1;
2204
2205 /* Make sure we flush the recovery flag to disk. */
2206 ext3_commit_super(sb, es, 1);
2207 }
2208
2209 return 0;
2210 }
2211
2212 static int ext3_create_journal(struct super_block * sb,
2213 struct ext3_super_block * es,
2214 unsigned int journal_inum)
2215 {
2216 journal_t *journal;
2217 int err;
2218
2219 if (sb->s_flags & MS_RDONLY) {
2220 printk(KERN_ERR "EXT3-fs: readonly filesystem when trying to "
2221 "create journal.\n");
2222 return -EROFS;
2223 }
2224
2225 journal = ext3_get_journal(sb, journal_inum);
2226 if (!journal)
2227 return -EINVAL;
2228
2229 printk(KERN_INFO "EXT3-fs: creating new journal on inode %u\n",
2230 journal_inum);
2231
2232 err = journal_create(journal);
2233 if (err) {
2234 printk(KERN_ERR "EXT3-fs: error creating journal.\n");
2235 journal_destroy(journal);
2236 return -EIO;
2237 }
2238
2239 EXT3_SB(sb)->s_journal = journal;
2240
2241 ext3_update_dynamic_rev(sb);
2242 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2243 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
2244
2245 es->s_journal_inum = cpu_to_le32(journal_inum);
2246 sb->s_dirt = 1;
2247
2248 /* Make sure we flush the recovery flag to disk. */
2249 ext3_commit_super(sb, es, 1);
2250
2251 return 0;
2252 }
2253
2254 static void ext3_commit_super (struct super_block * sb,
2255 struct ext3_super_block * es,
2256 int sync)
2257 {
2258 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
2259
2260 if (!sbh)
2261 return;
2262 es->s_wtime = cpu_to_le32(get_seconds());
2263 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
2264 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
2265 BUFFER_TRACE(sbh, "marking dirty");
2266 mark_buffer_dirty(sbh);
2267 if (sync)
2268 sync_dirty_buffer(sbh);
2269 }
2270
2271
2272 /*
2273 * Have we just finished recovery? If so, and if we are mounting (or
2274 * remounting) the filesystem readonly, then we will end up with a
2275 * consistent fs on disk. Record that fact.
2276 */
2277 static void ext3_mark_recovery_complete(struct super_block * sb,
2278 struct ext3_super_block * es)
2279 {
2280 journal_t *journal = EXT3_SB(sb)->s_journal;
2281
2282 journal_lock_updates(journal);
2283 journal_flush(journal);
2284 lock_super(sb);
2285 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
2286 sb->s_flags & MS_RDONLY) {
2287 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2288 sb->s_dirt = 0;
2289 ext3_commit_super(sb, es, 1);
2290 }
2291 unlock_super(sb);
2292 journal_unlock_updates(journal);
2293 }
2294
2295 /*
2296 * If we are mounting (or read-write remounting) a filesystem whose journal
2297 * has recorded an error from a previous lifetime, move that error to the
2298 * main filesystem now.
2299 */
2300 static void ext3_clear_journal_err(struct super_block * sb,
2301 struct ext3_super_block * es)
2302 {
2303 journal_t *journal;
2304 int j_errno;
2305 const char *errstr;
2306
2307 journal = EXT3_SB(sb)->s_journal;
2308
2309 /*
2310 * Now check for any error status which may have been recorded in the
2311 * journal by a prior ext3_error() or ext3_abort()
2312 */
2313
2314 j_errno = journal_errno(journal);
2315 if (j_errno) {
2316 char nbuf[16];
2317
2318 errstr = ext3_decode_error(sb, j_errno, nbuf);
2319 ext3_warning(sb, __func__, "Filesystem error recorded "
2320 "from previous mount: %s", errstr);
2321 ext3_warning(sb, __func__, "Marking fs in need of "
2322 "filesystem check.");
2323
2324 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
2325 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
2326 ext3_commit_super (sb, es, 1);
2327
2328 journal_clear_err(journal);
2329 }
2330 }
2331
2332 /*
2333 * Force the running and committing transactions to commit,
2334 * and wait on the commit.
2335 */
2336 int ext3_force_commit(struct super_block *sb)
2337 {
2338 journal_t *journal;
2339 int ret;
2340
2341 if (sb->s_flags & MS_RDONLY)
2342 return 0;
2343
2344 journal = EXT3_SB(sb)->s_journal;
2345 sb->s_dirt = 0;
2346 ret = ext3_journal_force_commit(journal);
2347 return ret;
2348 }
2349
2350 /*
2351 * Ext3 always journals updates to the superblock itself, so we don't
2352 * have to propagate any other updates to the superblock on disk at this
2353 * point. Just start an async writeback to get the buffers on their way
2354 * to the disk.
2355 *
2356 * This implicitly triggers the writebehind on sync().
2357 */
2358
2359 static void ext3_write_super (struct super_block * sb)
2360 {
2361 if (mutex_trylock(&sb->s_lock) != 0)
2362 BUG();
2363 sb->s_dirt = 0;
2364 }
2365
2366 static int ext3_sync_fs(struct super_block *sb, int wait)
2367 {
2368 tid_t target;
2369
2370 sb->s_dirt = 0;
2371 if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
2372 if (wait)
2373 log_wait_commit(EXT3_SB(sb)->s_journal, target);
2374 }
2375 return 0;
2376 }
2377
2378 /*
2379 * LVM calls this function before a (read-only) snapshot is created. This
2380 * gives us a chance to flush the journal completely and mark the fs clean.
2381 */
2382 static void ext3_write_super_lockfs(struct super_block *sb)
2383 {
2384 sb->s_dirt = 0;
2385
2386 if (!(sb->s_flags & MS_RDONLY)) {
2387 journal_t *journal = EXT3_SB(sb)->s_journal;
2388
2389 /* Now we set up the journal barrier. */
2390 journal_lock_updates(journal);
2391 journal_flush(journal);
2392
2393 /* Journal blocked and flushed, clear needs_recovery flag. */
2394 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2395 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2396 }
2397 }
2398
2399 /*
2400 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2401 * flag here, even though the filesystem is not technically dirty yet.
2402 */
2403 static void ext3_unlockfs(struct super_block *sb)
2404 {
2405 if (!(sb->s_flags & MS_RDONLY)) {
2406 lock_super(sb);
2407 /* Reser the needs_recovery flag before the fs is unlocked. */
2408 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2409 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2410 unlock_super(sb);
2411 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2412 }
2413 }
2414
2415 static int ext3_remount (struct super_block * sb, int * flags, char * data)
2416 {
2417 struct ext3_super_block * es;
2418 struct ext3_sb_info *sbi = EXT3_SB(sb);
2419 ext3_fsblk_t n_blocks_count = 0;
2420 unsigned long old_sb_flags;
2421 struct ext3_mount_options old_opts;
2422 int err;
2423 #ifdef CONFIG_QUOTA
2424 int i;
2425 #endif
2426
2427 /* Store the original options */
2428 old_sb_flags = sb->s_flags;
2429 old_opts.s_mount_opt = sbi->s_mount_opt;
2430 old_opts.s_resuid = sbi->s_resuid;
2431 old_opts.s_resgid = sbi->s_resgid;
2432 old_opts.s_commit_interval = sbi->s_commit_interval;
2433 #ifdef CONFIG_QUOTA
2434 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2435 for (i = 0; i < MAXQUOTAS; i++)
2436 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2437 #endif
2438
2439 /*
2440 * Allow the "check" option to be passed as a remount option.
2441 */
2442 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2443 err = -EINVAL;
2444 goto restore_opts;
2445 }
2446
2447 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT)
2448 ext3_abort(sb, __func__, "Abort forced by user");
2449
2450 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2451 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2452
2453 es = sbi->s_es;
2454
2455 ext3_init_journal_params(sb, sbi->s_journal);
2456
2457 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2458 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
2459 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT) {
2460 err = -EROFS;
2461 goto restore_opts;
2462 }
2463
2464 if (*flags & MS_RDONLY) {
2465 /*
2466 * First of all, the unconditional stuff we have to do
2467 * to disable replay of the journal when we next remount
2468 */
2469 sb->s_flags |= MS_RDONLY;
2470
2471 /*
2472 * OK, test if we are remounting a valid rw partition
2473 * readonly, and if so set the rdonly flag and then
2474 * mark the partition as valid again.
2475 */
2476 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2477 (sbi->s_mount_state & EXT3_VALID_FS))
2478 es->s_state = cpu_to_le16(sbi->s_mount_state);
2479
2480 /*
2481 * We have to unlock super so that we can wait for
2482 * transactions.
2483 */
2484 unlock_super(sb);
2485 ext3_mark_recovery_complete(sb, es);
2486 lock_super(sb);
2487 } else {
2488 __le32 ret;
2489 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2490 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
2491 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2492 "remount RDWR because of unsupported "
2493 "optional features (%x).\n",
2494 sb->s_id, le32_to_cpu(ret));
2495 err = -EROFS;
2496 goto restore_opts;
2497 }
2498
2499 /*
2500 * If we have an unprocessed orphan list hanging
2501 * around from a previously readonly bdev mount,
2502 * require a full umount/remount for now.
2503 */
2504 if (es->s_last_orphan) {
2505 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2506 "remount RDWR because of unprocessed "
2507 "orphan inode list. Please "
2508 "umount/remount instead.\n",
2509 sb->s_id);
2510 err = -EINVAL;
2511 goto restore_opts;
2512 }
2513
2514 /*
2515 * Mounting a RDONLY partition read-write, so reread
2516 * and store the current valid flag. (It may have
2517 * been changed by e2fsck since we originally mounted
2518 * the partition.)
2519 */
2520 ext3_clear_journal_err(sb, es);
2521 sbi->s_mount_state = le16_to_cpu(es->s_state);
2522 if ((err = ext3_group_extend(sb, es, n_blocks_count)))
2523 goto restore_opts;
2524 if (!ext3_setup_super (sb, es, 0))
2525 sb->s_flags &= ~MS_RDONLY;
2526 }
2527 }
2528 #ifdef CONFIG_QUOTA
2529 /* Release old quota file names */
2530 for (i = 0; i < MAXQUOTAS; i++)
2531 if (old_opts.s_qf_names[i] &&
2532 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2533 kfree(old_opts.s_qf_names[i]);
2534 #endif
2535 return 0;
2536 restore_opts:
2537 sb->s_flags = old_sb_flags;
2538 sbi->s_mount_opt = old_opts.s_mount_opt;
2539 sbi->s_resuid = old_opts.s_resuid;
2540 sbi->s_resgid = old_opts.s_resgid;
2541 sbi->s_commit_interval = old_opts.s_commit_interval;
2542 #ifdef CONFIG_QUOTA
2543 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2544 for (i = 0; i < MAXQUOTAS; i++) {
2545 if (sbi->s_qf_names[i] &&
2546 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2547 kfree(sbi->s_qf_names[i]);
2548 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2549 }
2550 #endif
2551 return err;
2552 }
2553
2554 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf)
2555 {
2556 struct super_block *sb = dentry->d_sb;
2557 struct ext3_sb_info *sbi = EXT3_SB(sb);
2558 struct ext3_super_block *es = sbi->s_es;
2559 u64 fsid;
2560
2561 if (test_opt(sb, MINIX_DF)) {
2562 sbi->s_overhead_last = 0;
2563 } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
2564 unsigned long ngroups = sbi->s_groups_count, i;
2565 ext3_fsblk_t overhead = 0;
2566 smp_rmb();
2567
2568 /*
2569 * Compute the overhead (FS structures). This is constant
2570 * for a given filesystem unless the number of block groups
2571 * changes so we cache the previous value until it does.
2572 */
2573
2574 /*
2575 * All of the blocks before first_data_block are
2576 * overhead
2577 */
2578 overhead = le32_to_cpu(es->s_first_data_block);
2579
2580 /*
2581 * Add the overhead attributed to the superblock and
2582 * block group descriptors. If the sparse superblocks
2583 * feature is turned on, then not all groups have this.
2584 */
2585 for (i = 0; i < ngroups; i++) {
2586 overhead += ext3_bg_has_super(sb, i) +
2587 ext3_bg_num_gdb(sb, i);
2588 cond_resched();
2589 }
2590
2591 /*
2592 * Every block group has an inode bitmap, a block
2593 * bitmap, and an inode table.
2594 */
2595 overhead += ngroups * (2 + sbi->s_itb_per_group);
2596 sbi->s_overhead_last = overhead;
2597 smp_wmb();
2598 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
2599 }
2600
2601 buf->f_type = EXT3_SUPER_MAGIC;
2602 buf->f_bsize = sb->s_blocksize;
2603 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
2604 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
2605 es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
2606 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2607 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2608 buf->f_bavail = 0;
2609 buf->f_files = le32_to_cpu(es->s_inodes_count);
2610 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
2611 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
2612 buf->f_namelen = EXT3_NAME_LEN;
2613 fsid = le64_to_cpup((void *)es->s_uuid) ^
2614 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2615 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2616 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
2617 return 0;
2618 }
2619
2620 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2621 * is locked for write. Otherwise the are possible deadlocks:
2622 * Process 1 Process 2
2623 * ext3_create() quota_sync()
2624 * journal_start() write_dquot()
2625 * DQUOT_INIT() down(dqio_mutex)
2626 * down(dqio_mutex) journal_start()
2627 *
2628 */
2629
2630 #ifdef CONFIG_QUOTA
2631
2632 static inline struct inode *dquot_to_inode(struct dquot *dquot)
2633 {
2634 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2635 }
2636
2637 static int ext3_dquot_initialize(struct inode *inode, int type)
2638 {
2639 handle_t *handle;
2640 int ret, err;
2641
2642 /* We may create quota structure so we need to reserve enough blocks */
2643 handle = ext3_journal_start(inode, 2*EXT3_QUOTA_INIT_BLOCKS(inode->i_sb));
2644 if (IS_ERR(handle))
2645 return PTR_ERR(handle);
2646 ret = dquot_initialize(inode, type);
2647 err = ext3_journal_stop(handle);
2648 if (!ret)
2649 ret = err;
2650 return ret;
2651 }
2652
2653 static int ext3_dquot_drop(struct inode *inode)
2654 {
2655 handle_t *handle;
2656 int ret, err;
2657
2658 /* We may delete quota structure so we need to reserve enough blocks */
2659 handle = ext3_journal_start(inode, 2*EXT3_QUOTA_DEL_BLOCKS(inode->i_sb));
2660 if (IS_ERR(handle)) {
2661 /*
2662 * We call dquot_drop() anyway to at least release references
2663 * to quota structures so that umount does not hang.
2664 */
2665 dquot_drop(inode);
2666 return PTR_ERR(handle);
2667 }
2668 ret = dquot_drop(inode);
2669 err = ext3_journal_stop(handle);
2670 if (!ret)
2671 ret = err;
2672 return ret;
2673 }
2674
2675 static int ext3_write_dquot(struct dquot *dquot)
2676 {
2677 int ret, err;
2678 handle_t *handle;
2679 struct inode *inode;
2680
2681 inode = dquot_to_inode(dquot);
2682 handle = ext3_journal_start(inode,
2683 EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
2684 if (IS_ERR(handle))
2685 return PTR_ERR(handle);
2686 ret = dquot_commit(dquot);
2687 err = ext3_journal_stop(handle);
2688 if (!ret)
2689 ret = err;
2690 return ret;
2691 }
2692
2693 static int ext3_acquire_dquot(struct dquot *dquot)
2694 {
2695 int ret, err;
2696 handle_t *handle;
2697
2698 handle = ext3_journal_start(dquot_to_inode(dquot),
2699 EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
2700 if (IS_ERR(handle))
2701 return PTR_ERR(handle);
2702 ret = dquot_acquire(dquot);
2703 err = ext3_journal_stop(handle);
2704 if (!ret)
2705 ret = err;
2706 return ret;
2707 }
2708
2709 static int ext3_release_dquot(struct dquot *dquot)
2710 {
2711 int ret, err;
2712 handle_t *handle;
2713
2714 handle = ext3_journal_start(dquot_to_inode(dquot),
2715 EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
2716 if (IS_ERR(handle)) {
2717 /* Release dquot anyway to avoid endless cycle in dqput() */
2718 dquot_release(dquot);
2719 return PTR_ERR(handle);
2720 }
2721 ret = dquot_release(dquot);
2722 err = ext3_journal_stop(handle);
2723 if (!ret)
2724 ret = err;
2725 return ret;
2726 }
2727
2728 static int ext3_mark_dquot_dirty(struct dquot *dquot)
2729 {
2730 /* Are we journaling quotas? */
2731 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2732 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2733 dquot_mark_dquot_dirty(dquot);
2734 return ext3_write_dquot(dquot);
2735 } else {
2736 return dquot_mark_dquot_dirty(dquot);
2737 }
2738 }
2739
2740 static int ext3_write_info(struct super_block *sb, int type)
2741 {
2742 int ret, err;
2743 handle_t *handle;
2744
2745 /* Data block + inode block */
2746 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2747 if (IS_ERR(handle))
2748 return PTR_ERR(handle);
2749 ret = dquot_commit_info(sb, type);
2750 err = ext3_journal_stop(handle);
2751 if (!ret)
2752 ret = err;
2753 return ret;
2754 }
2755
2756 /*
2757 * Turn on quotas during mount time - we need to find
2758 * the quota file and such...
2759 */
2760 static int ext3_quota_on_mount(struct super_block *sb, int type)
2761 {
2762 return vfs_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2763 EXT3_SB(sb)->s_jquota_fmt, type);
2764 }
2765
2766 /*
2767 * Standard function to be called on quota_on
2768 */
2769 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
2770 char *path, int remount)
2771 {
2772 int err;
2773 struct nameidata nd;
2774
2775 if (!test_opt(sb, QUOTA))
2776 return -EINVAL;
2777 /* When remounting, no checks are needed and in fact, path is NULL */
2778 if (remount)
2779 return vfs_quota_on(sb, type, format_id, path, remount);
2780
2781 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
2782 if (err)
2783 return err;
2784
2785 /* Quotafile not on the same filesystem? */
2786 if (nd.path.mnt->mnt_sb != sb) {
2787 path_put(&nd.path);
2788 return -EXDEV;
2789 }
2790 /* Journaling quota? */
2791 if (EXT3_SB(sb)->s_qf_names[type]) {
2792 /* Quotafile not of fs root? */
2793 if (nd.path.dentry->d_parent->d_inode != sb->s_root->d_inode)
2794 printk(KERN_WARNING
2795 "EXT3-fs: Quota file not on filesystem root. "
2796 "Journaled quota will not work.\n");
2797 }
2798
2799 /*
2800 * When we journal data on quota file, we have to flush journal to see
2801 * all updates to the file when we bypass pagecache...
2802 */
2803 if (ext3_should_journal_data(nd.path.dentry->d_inode)) {
2804 /*
2805 * We don't need to lock updates but journal_flush() could
2806 * otherwise be livelocked...
2807 */
2808 journal_lock_updates(EXT3_SB(sb)->s_journal);
2809 journal_flush(EXT3_SB(sb)->s_journal);
2810 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2811 }
2812
2813 err = vfs_quota_on_path(sb, type, format_id, &nd.path);
2814 path_put(&nd.path);
2815 return err;
2816 }
2817
2818 /* Read data from quotafile - avoid pagecache and such because we cannot afford
2819 * acquiring the locks... As quota files are never truncated and quota code
2820 * itself serializes the operations (and noone else should touch the files)
2821 * we don't have to be afraid of races */
2822 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
2823 size_t len, loff_t off)
2824 {
2825 struct inode *inode = sb_dqopt(sb)->files[type];
2826 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2827 int err = 0;
2828 int offset = off & (sb->s_blocksize - 1);
2829 int tocopy;
2830 size_t toread;
2831 struct buffer_head *bh;
2832 loff_t i_size = i_size_read(inode);
2833
2834 if (off > i_size)
2835 return 0;
2836 if (off+len > i_size)
2837 len = i_size-off;
2838 toread = len;
2839 while (toread > 0) {
2840 tocopy = sb->s_blocksize - offset < toread ?
2841 sb->s_blocksize - offset : toread;
2842 bh = ext3_bread(NULL, inode, blk, 0, &err);
2843 if (err)
2844 return err;
2845 if (!bh) /* A hole? */
2846 memset(data, 0, tocopy);
2847 else
2848 memcpy(data, bh->b_data+offset, tocopy);
2849 brelse(bh);
2850 offset = 0;
2851 toread -= tocopy;
2852 data += tocopy;
2853 blk++;
2854 }
2855 return len;
2856 }
2857
2858 /* Write to quotafile (we know the transaction is already started and has
2859 * enough credits) */
2860 static ssize_t ext3_quota_write(struct super_block *sb, int type,
2861 const char *data, size_t len, loff_t off)
2862 {
2863 struct inode *inode = sb_dqopt(sb)->files[type];
2864 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2865 int err = 0;
2866 int offset = off & (sb->s_blocksize - 1);
2867 int tocopy;
2868 int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
2869 size_t towrite = len;
2870 struct buffer_head *bh;
2871 handle_t *handle = journal_current_handle();
2872
2873 if (!handle) {
2874 printk(KERN_WARNING "EXT3-fs: Quota write (off=%Lu, len=%Lu)"
2875 " cancelled because transaction is not started.\n",
2876 (unsigned long long)off, (unsigned long long)len);
2877 return -EIO;
2878 }
2879 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
2880 while (towrite > 0) {
2881 tocopy = sb->s_blocksize - offset < towrite ?
2882 sb->s_blocksize - offset : towrite;
2883 bh = ext3_bread(handle, inode, blk, 1, &err);
2884 if (!bh)
2885 goto out;
2886 if (journal_quota) {
2887 err = ext3_journal_get_write_access(handle, bh);
2888 if (err) {
2889 brelse(bh);
2890 goto out;
2891 }
2892 }
2893 lock_buffer(bh);
2894 memcpy(bh->b_data+offset, data, tocopy);
2895 flush_dcache_page(bh->b_page);
2896 unlock_buffer(bh);
2897 if (journal_quota)
2898 err = ext3_journal_dirty_metadata(handle, bh);
2899 else {
2900 /* Always do at least ordered writes for quotas */
2901 err = ext3_journal_dirty_data(handle, bh);
2902 mark_buffer_dirty(bh);
2903 }
2904 brelse(bh);
2905 if (err)
2906 goto out;
2907 offset = 0;
2908 towrite -= tocopy;
2909 data += tocopy;
2910 blk++;
2911 }
2912 out:
2913 if (len == towrite) {
2914 mutex_unlock(&inode->i_mutex);
2915 return err;
2916 }
2917 if (inode->i_size < off+len-towrite) {
2918 i_size_write(inode, off+len-towrite);
2919 EXT3_I(inode)->i_disksize = inode->i_size;
2920 }
2921 inode->i_version++;
2922 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2923 ext3_mark_inode_dirty(handle, inode);
2924 mutex_unlock(&inode->i_mutex);
2925 return len - towrite;
2926 }
2927
2928 #endif
2929
2930 static int ext3_get_sb(struct file_system_type *fs_type,
2931 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
2932 {
2933 return get_sb_bdev(fs_type, flags, dev_name, data, ext3_fill_super, mnt);
2934 }
2935
2936 static struct file_system_type ext3_fs_type = {
2937 .owner = THIS_MODULE,
2938 .name = "ext3",
2939 .get_sb = ext3_get_sb,
2940 .kill_sb = kill_block_super,
2941 .fs_flags = FS_REQUIRES_DEV,
2942 };
2943
2944 static int __init init_ext3_fs(void)
2945 {
2946 int err = init_ext3_xattr();
2947 if (err)
2948 return err;
2949 err = init_inodecache();
2950 if (err)
2951 goto out1;
2952 err = register_filesystem(&ext3_fs_type);
2953 if (err)
2954 goto out;
2955 return 0;
2956 out:
2957 destroy_inodecache();
2958 out1:
2959 exit_ext3_xattr();
2960 return err;
2961 }
2962
2963 static void __exit exit_ext3_fs(void)
2964 {
2965 unregister_filesystem(&ext3_fs_type);
2966 destroy_inodecache();
2967 exit_ext3_xattr();
2968 }
2969
2970 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
2971 MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
2972 MODULE_LICENSE("GPL");
2973 module_init(init_ext3_fs)
2974 module_exit(exit_ext3_fs)
Linux 2.6 ibm-acpi/thinkpad-acpi driver git tree