?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
x86/amd-iommu: Add per IOMMU reference counting
[linux/fpc-iii.git] / fs / ext3 / super.c
blob427496c4767cda4c542f43cb97b8a3089e77bcce
1 /*
2 * linux/fs/ext3/super.c
3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19 #include <linux/module.h>
20 #include <linux/string.h>
21 #include <linux/fs.h>
22 #include <linux/time.h>
23 #include <linux/jbd.h>
24 #include <linux/ext3_fs.h>
25 #include <linux/ext3_jbd.h>
26 #include <linux/slab.h>
27 #include <linux/init.h>
28 #include <linux/blkdev.h>
29 #include <linux/parser.h>
30 #include <linux/smp_lock.h>
31 #include <linux/buffer_head.h>
32 #include <linux/exportfs.h>
33 #include <linux/vfs.h>
34 #include <linux/random.h>
35 #include <linux/mount.h>
36 #include <linux/namei.h>
37 #include <linux/quotaops.h>
38 #include <linux/seq_file.h>
39 #include <linux/log2.h>
40
41 #include <asm/uaccess.h>
42
43 #include "xattr.h"
44 #include "acl.h"
45 #include "namei.h"
46
47 #ifdef CONFIG_EXT3_DEFAULTS_TO_ORDERED
48 #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_ORDERED_DATA
49 #else
50 #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_WRITEBACK_DATA
51 #endif
52
53 static int ext3_load_journal(struct super_block *, struct ext3_super_block *,
54 unsigned long journal_devnum);
55 static int ext3_create_journal(struct super_block *, struct ext3_super_block *,
56 unsigned int);
57 static int ext3_commit_super(struct super_block *sb,
58 struct ext3_super_block *es,
59 int sync);
60 static void ext3_mark_recovery_complete(struct super_block * sb,
61 struct ext3_super_block * es);
62 static void ext3_clear_journal_err(struct super_block * sb,
63 struct ext3_super_block * es);
64 static int ext3_sync_fs(struct super_block *sb, int wait);
65 static const char *ext3_decode_error(struct super_block * sb, int errno,
66 char nbuf[16]);
67 static int ext3_remount (struct super_block * sb, int * flags, char * data);
68 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf);
69 static int ext3_unfreeze(struct super_block *sb);
70 static int ext3_freeze(struct super_block *sb);
71
72 /*
73 * Wrappers for journal_start/end.
74 *
75 * The only special thing we need to do here is to make sure that all
76 * journal_end calls result in the superblock being marked dirty, so
77 * that sync() will call the filesystem's write_super callback if
78 * appropriate.
79 */
80 handle_t *ext3_journal_start_sb(struct super_block *sb, int nblocks)
81 {
82 journal_t *journal;
83
84 if (sb->s_flags & MS_RDONLY)
85 return ERR_PTR(-EROFS);
86
87 /* Special case here: if the journal has aborted behind our
88 * backs (eg. EIO in the commit thread), then we still need to
89 * take the FS itself readonly cleanly. */
90 journal = EXT3_SB(sb)->s_journal;
91 if (is_journal_aborted(journal)) {
92 ext3_abort(sb, __func__,
93 "Detected aborted journal");
94 return ERR_PTR(-EROFS);
95 }
96
97 return journal_start(journal, nblocks);
98 }
99
100 /*
101 * The only special thing we need to do here is to make sure that all
102 * journal_stop calls result in the superblock being marked dirty, so
103 * that sync() will call the filesystem's write_super callback if
104 * appropriate.
105 */
106 int __ext3_journal_stop(const char *where, handle_t *handle)
107 {
108 struct super_block *sb;
109 int err;
110 int rc;
111
112 sb = handle->h_transaction->t_journal->j_private;
113 err = handle->h_err;
114 rc = journal_stop(handle);
115
116 if (!err)
117 err = rc;
118 if (err)
119 __ext3_std_error(sb, where, err);
120 return err;
121 }
122
123 void ext3_journal_abort_handle(const char *caller, const char *err_fn,
124 struct buffer_head *bh, handle_t *handle, int err)
125 {
126 char nbuf[16];
127 const char *errstr = ext3_decode_error(NULL, err, nbuf);
128
129 if (bh)
130 BUFFER_TRACE(bh, "abort");
131
132 if (!handle->h_err)
133 handle->h_err = err;
134
135 if (is_handle_aborted(handle))
136 return;
137
138 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
139 caller, errstr, err_fn);
140
141 journal_abort_handle(handle);
142 }
143
144 /* Deal with the reporting of failure conditions on a filesystem such as
145 * inconsistencies detected or read IO failures.
146 *
147 * On ext2, we can store the error state of the filesystem in the
148 * superblock. That is not possible on ext3, because we may have other
149 * write ordering constraints on the superblock which prevent us from
150 * writing it out straight away; and given that the journal is about to
151 * be aborted, we can't rely on the current, or future, transactions to
152 * write out the superblock safely.
153 *
154 * We'll just use the journal_abort() error code to record an error in
155 * the journal instead. On recovery, the journal will compain about
156 * that error until we've noted it down and cleared it.
157 */
158
159 static void ext3_handle_error(struct super_block *sb)
160 {
161 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
162
163 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
164 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
165
166 if (sb->s_flags & MS_RDONLY)
167 return;
168
169 if (!test_opt (sb, ERRORS_CONT)) {
170 journal_t *journal = EXT3_SB(sb)->s_journal;
171
172 EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
173 if (journal)
174 journal_abort(journal, -EIO);
175 }
176 if (test_opt (sb, ERRORS_RO)) {
177 printk (KERN_CRIT "Remounting filesystem read-only\n");
178 sb->s_flags |= MS_RDONLY;
179 }
180 ext3_commit_super(sb, es, 1);
181 if (test_opt(sb, ERRORS_PANIC))
182 panic("EXT3-fs (device %s): panic forced after error\n",
183 sb->s_id);
184 }
185
186 void ext3_error (struct super_block * sb, const char * function,
187 const char * fmt, ...)
188 {
189 va_list args;
190
191 va_start(args, fmt);
192 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
193 vprintk(fmt, args);
194 printk("\n");
195 va_end(args);
196
197 ext3_handle_error(sb);
198 }
199
200 static const char *ext3_decode_error(struct super_block * sb, int errno,
201 char nbuf[16])
202 {
203 char *errstr = NULL;
204
205 switch (errno) {
206 case -EIO:
207 errstr = "IO failure";
208 break;
209 case -ENOMEM:
210 errstr = "Out of memory";
211 break;
212 case -EROFS:
213 if (!sb || EXT3_SB(sb)->s_journal->j_flags & JFS_ABORT)
214 errstr = "Journal has aborted";
215 else
216 errstr = "Readonly filesystem";
217 break;
218 default:
219 /* If the caller passed in an extra buffer for unknown
220 * errors, textualise them now. Else we just return
221 * NULL. */
222 if (nbuf) {
223 /* Check for truncated error codes... */
224 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
225 errstr = nbuf;
226 }
227 break;
228 }
229
230 return errstr;
231 }
232
233 /* __ext3_std_error decodes expected errors from journaling functions
234 * automatically and invokes the appropriate error response. */
235
236 void __ext3_std_error (struct super_block * sb, const char * function,
237 int errno)
238 {
239 char nbuf[16];
240 const char *errstr;
241
242 /* Special case: if the error is EROFS, and we're not already
243 * inside a transaction, then there's really no point in logging
244 * an error. */
245 if (errno == -EROFS && journal_current_handle() == NULL &&
246 (sb->s_flags & MS_RDONLY))
247 return;
248
249 errstr = ext3_decode_error(sb, errno, nbuf);
250 printk (KERN_CRIT "EXT3-fs error (device %s) in %s: %s\n",
251 sb->s_id, function, errstr);
252
253 ext3_handle_error(sb);
254 }
255
256 /*
257 * ext3_abort is a much stronger failure handler than ext3_error. The
258 * abort function may be used to deal with unrecoverable failures such
259 * as journal IO errors or ENOMEM at a critical moment in log management.
260 *
261 * We unconditionally force the filesystem into an ABORT|READONLY state,
262 * unless the error response on the fs has been set to panic in which
263 * case we take the easy way out and panic immediately.
264 */
265
266 void ext3_abort (struct super_block * sb, const char * function,
267 const char * fmt, ...)
268 {
269 va_list args;
270
271 printk (KERN_CRIT "ext3_abort called.\n");
272
273 va_start(args, fmt);
274 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
275 vprintk(fmt, args);
276 printk("\n");
277 va_end(args);
278
279 if (test_opt(sb, ERRORS_PANIC))
280 panic("EXT3-fs panic from previous error\n");
281
282 if (sb->s_flags & MS_RDONLY)
283 return;
284
285 printk(KERN_CRIT "Remounting filesystem read-only\n");
286 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
287 sb->s_flags |= MS_RDONLY;
288 EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
289 if (EXT3_SB(sb)->s_journal)
290 journal_abort(EXT3_SB(sb)->s_journal, -EIO);
291 }
292
293 void ext3_warning (struct super_block * sb, const char * function,
294 const char * fmt, ...)
295 {
296 va_list args;
297
298 va_start(args, fmt);
299 printk(KERN_WARNING "EXT3-fs warning (device %s): %s: ",
300 sb->s_id, function);
301 vprintk(fmt, args);
302 printk("\n");
303 va_end(args);
304 }
305
306 void ext3_update_dynamic_rev(struct super_block *sb)
307 {
308 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
309
310 if (le32_to_cpu(es->s_rev_level) > EXT3_GOOD_OLD_REV)
311 return;
312
313 ext3_warning(sb, __func__,
314 "updating to rev %d because of new feature flag, "
315 "running e2fsck is recommended",
316 EXT3_DYNAMIC_REV);
317
318 es->s_first_ino = cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO);
319 es->s_inode_size = cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE);
320 es->s_rev_level = cpu_to_le32(EXT3_DYNAMIC_REV);
321 /* leave es->s_feature_*compat flags alone */
322 /* es->s_uuid will be set by e2fsck if empty */
323
324 /*
325 * The rest of the superblock fields should be zero, and if not it
326 * means they are likely already in use, so leave them alone. We
327 * can leave it up to e2fsck to clean up any inconsistencies there.
328 */
329 }
330
331 /*
332 * Open the external journal device
333 */
334 static struct block_device *ext3_blkdev_get(dev_t dev)
335 {
336 struct block_device *bdev;
337 char b[BDEVNAME_SIZE];
338
339 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
340 if (IS_ERR(bdev))
341 goto fail;
342 return bdev;
343
344 fail:
345 printk(KERN_ERR "EXT3: failed to open journal device %s: %ld\n",
346 __bdevname(dev, b), PTR_ERR(bdev));
347 return NULL;
348 }
349
350 /*
351 * Release the journal device
352 */
353 static int ext3_blkdev_put(struct block_device *bdev)
354 {
355 bd_release(bdev);
356 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
357 }
358
359 static int ext3_blkdev_remove(struct ext3_sb_info *sbi)
360 {
361 struct block_device *bdev;
362 int ret = -ENODEV;
363
364 bdev = sbi->journal_bdev;
365 if (bdev) {
366 ret = ext3_blkdev_put(bdev);
367 sbi->journal_bdev = NULL;
368 }
369 return ret;
370 }
371
372 static inline struct inode *orphan_list_entry(struct list_head *l)
373 {
374 return &list_entry(l, struct ext3_inode_info, i_orphan)->vfs_inode;
375 }
376
377 static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi)
378 {
379 struct list_head *l;
380
381 printk(KERN_ERR "sb orphan head is %d\n",
382 le32_to_cpu(sbi->s_es->s_last_orphan));
383
384 printk(KERN_ERR "sb_info orphan list:\n");
385 list_for_each(l, &sbi->s_orphan) {
386 struct inode *inode = orphan_list_entry(l);
387 printk(KERN_ERR " "
388 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
389 inode->i_sb->s_id, inode->i_ino, inode,
390 inode->i_mode, inode->i_nlink,
391 NEXT_ORPHAN(inode));
392 }
393 }
394
395 static void ext3_put_super (struct super_block * sb)
396 {
397 struct ext3_sb_info *sbi = EXT3_SB(sb);
398 struct ext3_super_block *es = sbi->s_es;
399 int i, err;
400
401 lock_kernel();
402
403 ext3_xattr_put_super(sb);
404 err = journal_destroy(sbi->s_journal);
405 sbi->s_journal = NULL;
406 if (err < 0)
407 ext3_abort(sb, __func__, "Couldn't clean up the journal");
408
409 if (!(sb->s_flags & MS_RDONLY)) {
410 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
411 es->s_state = cpu_to_le16(sbi->s_mount_state);
412 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
413 mark_buffer_dirty(sbi->s_sbh);
414 ext3_commit_super(sb, es, 1);
415 }
416
417 for (i = 0; i < sbi->s_gdb_count; i++)
418 brelse(sbi->s_group_desc[i]);
419 kfree(sbi->s_group_desc);
420 percpu_counter_destroy(&sbi->s_freeblocks_counter);
421 percpu_counter_destroy(&sbi->s_freeinodes_counter);
422 percpu_counter_destroy(&sbi->s_dirs_counter);
423 brelse(sbi->s_sbh);
424 #ifdef CONFIG_QUOTA
425 for (i = 0; i < MAXQUOTAS; i++)
426 kfree(sbi->s_qf_names[i]);
427 #endif
428
429 /* Debugging code just in case the in-memory inode orphan list
430 * isn't empty. The on-disk one can be non-empty if we've
431 * detected an error and taken the fs readonly, but the
432 * in-memory list had better be clean by this point. */
433 if (!list_empty(&sbi->s_orphan))
434 dump_orphan_list(sb, sbi);
435 J_ASSERT(list_empty(&sbi->s_orphan));
436
437 invalidate_bdev(sb->s_bdev);
438 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
439 /*
440 * Invalidate the journal device's buffers. We don't want them
441 * floating about in memory - the physical journal device may
442 * hotswapped, and it breaks the `ro-after' testing code.
443 */
444 sync_blockdev(sbi->journal_bdev);
445 invalidate_bdev(sbi->journal_bdev);
446 ext3_blkdev_remove(sbi);
447 }
448 sb->s_fs_info = NULL;
449 kfree(sbi->s_blockgroup_lock);
450 kfree(sbi);
451
452 unlock_kernel();
453 }
454
455 static struct kmem_cache *ext3_inode_cachep;
456
457 /*
458 * Called inside transaction, so use GFP_NOFS
459 */
460 static struct inode *ext3_alloc_inode(struct super_block *sb)
461 {
462 struct ext3_inode_info *ei;
463
464 ei = kmem_cache_alloc(ext3_inode_cachep, GFP_NOFS);
465 if (!ei)
466 return NULL;
467 ei->i_block_alloc_info = NULL;
468 ei->vfs_inode.i_version = 1;
469 atomic_set(&ei->i_datasync_tid, 0);
470 atomic_set(&ei->i_sync_tid, 0);
471 return &ei->vfs_inode;
472 }
473
474 static void ext3_destroy_inode(struct inode *inode)
475 {
476 if (!list_empty(&(EXT3_I(inode)->i_orphan))) {
477 printk("EXT3 Inode %p: orphan list check failed!\n",
478 EXT3_I(inode));
479 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
480 EXT3_I(inode), sizeof(struct ext3_inode_info),
481 false);
482 dump_stack();
483 }
484 kmem_cache_free(ext3_inode_cachep, EXT3_I(inode));
485 }
486
487 static void init_once(void *foo)
488 {
489 struct ext3_inode_info *ei = (struct ext3_inode_info *) foo;
490
491 INIT_LIST_HEAD(&ei->i_orphan);
492 #ifdef CONFIG_EXT3_FS_XATTR
493 init_rwsem(&ei->xattr_sem);
494 #endif
495 mutex_init(&ei->truncate_mutex);
496 inode_init_once(&ei->vfs_inode);
497 }
498
499 static int init_inodecache(void)
500 {
501 ext3_inode_cachep = kmem_cache_create("ext3_inode_cache",
502 sizeof(struct ext3_inode_info),
503 0, (SLAB_RECLAIM_ACCOUNT|
504 SLAB_MEM_SPREAD),
505 init_once);
506 if (ext3_inode_cachep == NULL)
507 return -ENOMEM;
508 return 0;
509 }
510
511 static void destroy_inodecache(void)
512 {
513 kmem_cache_destroy(ext3_inode_cachep);
514 }
515
516 static void ext3_clear_inode(struct inode *inode)
517 {
518 struct ext3_block_alloc_info *rsv = EXT3_I(inode)->i_block_alloc_info;
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 static char *data_mode_string(unsigned long mode)
549 {
550 switch (mode) {
551 case EXT3_MOUNT_JOURNAL_DATA:
552 return "journal";
553 case EXT3_MOUNT_ORDERED_DATA:
554 return "ordered";
555 case EXT3_MOUNT_WRITEBACK_DATA:
556 return "writeback";
557 }
558 return "unknown";
559 }
560
561 /*
562 * Show an option if
563 * - it's set to a non-default value OR
564 * - if the per-sb default is different from the global default
565 */
566 static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs)
567 {
568 struct super_block *sb = vfs->mnt_sb;
569 struct ext3_sb_info *sbi = EXT3_SB(sb);
570 struct ext3_super_block *es = sbi->s_es;
571 unsigned long def_mount_opts;
572
573 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
574
575 if (sbi->s_sb_block != 1)
576 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
577 if (test_opt(sb, MINIX_DF))
578 seq_puts(seq, ",minixdf");
579 if (test_opt(sb, GRPID))
580 seq_puts(seq, ",grpid");
581 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT3_DEFM_BSDGROUPS))
582 seq_puts(seq, ",nogrpid");
583 if (sbi->s_resuid != EXT3_DEF_RESUID ||
584 le16_to_cpu(es->s_def_resuid) != EXT3_DEF_RESUID) {
585 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
586 }
587 if (sbi->s_resgid != EXT3_DEF_RESGID ||
588 le16_to_cpu(es->s_def_resgid) != EXT3_DEF_RESGID) {
589 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
590 }
591 if (test_opt(sb, ERRORS_RO)) {
592 int def_errors = le16_to_cpu(es->s_errors);
593
594 if (def_errors == EXT3_ERRORS_PANIC ||
595 def_errors == EXT3_ERRORS_CONTINUE) {
596 seq_puts(seq, ",errors=remount-ro");
597 }
598 }
599 if (test_opt(sb, ERRORS_CONT))
600 seq_puts(seq, ",errors=continue");
601 if (test_opt(sb, ERRORS_PANIC))
602 seq_puts(seq, ",errors=panic");
603 if (test_opt(sb, NO_UID32))
604 seq_puts(seq, ",nouid32");
605 if (test_opt(sb, DEBUG))
606 seq_puts(seq, ",debug");
607 if (test_opt(sb, OLDALLOC))
608 seq_puts(seq, ",oldalloc");
609 #ifdef CONFIG_EXT3_FS_XATTR
610 if (test_opt(sb, XATTR_USER))
611 seq_puts(seq, ",user_xattr");
612 if (!test_opt(sb, XATTR_USER) &&
613 (def_mount_opts & EXT3_DEFM_XATTR_USER)) {
614 seq_puts(seq, ",nouser_xattr");
615 }
616 #endif
617 #ifdef CONFIG_EXT3_FS_POSIX_ACL
618 if (test_opt(sb, POSIX_ACL))
619 seq_puts(seq, ",acl");
620 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT3_DEFM_ACL))
621 seq_puts(seq, ",noacl");
622 #endif
623 if (!test_opt(sb, RESERVATION))
624 seq_puts(seq, ",noreservation");
625 if (sbi->s_commit_interval) {
626 seq_printf(seq, ",commit=%u",
627 (unsigned) (sbi->s_commit_interval / HZ));
628 }
629 if (test_opt(sb, BARRIER))
630 seq_puts(seq, ",barrier=1");
631 if (test_opt(sb, NOBH))
632 seq_puts(seq, ",nobh");
633
634 seq_printf(seq, ",data=%s", data_mode_string(sbi->s_mount_opt &
635 EXT3_MOUNT_DATA_FLAGS));
636 if (test_opt(sb, DATA_ERR_ABORT))
637 seq_puts(seq, ",data_err=abort");
638
639 ext3_show_quota_options(seq, sb);
640
641 return 0;
642 }
643
644
645 static struct inode *ext3_nfs_get_inode(struct super_block *sb,
646 u64 ino, u32 generation)
647 {
648 struct inode *inode;
649
650 if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO)
651 return ERR_PTR(-ESTALE);
652 if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count))
653 return ERR_PTR(-ESTALE);
654
655 /* iget isn't really right if the inode is currently unallocated!!
656 *
657 * ext3_read_inode will return a bad_inode if the inode had been
658 * deleted, so we should be safe.
659 *
660 * Currently we don't know the generation for parent directory, so
661 * a generation of 0 means "accept any"
662 */
663 inode = ext3_iget(sb, ino);
664 if (IS_ERR(inode))
665 return ERR_CAST(inode);
666 if (generation && inode->i_generation != generation) {
667 iput(inode);
668 return ERR_PTR(-ESTALE);
669 }
670
671 return inode;
672 }
673
674 static struct dentry *ext3_fh_to_dentry(struct super_block *sb, struct fid *fid,
675 int fh_len, int fh_type)
676 {
677 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
678 ext3_nfs_get_inode);
679 }
680
681 static struct dentry *ext3_fh_to_parent(struct super_block *sb, struct fid *fid,
682 int fh_len, int fh_type)
683 {
684 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
685 ext3_nfs_get_inode);
686 }
687
688 /*
689 * Try to release metadata pages (indirect blocks, directories) which are
690 * mapped via the block device. Since these pages could have journal heads
691 * which would prevent try_to_free_buffers() from freeing them, we must use
692 * jbd layer's try_to_free_buffers() function to release them.
693 */
694 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
695 gfp_t wait)
696 {
697 journal_t *journal = EXT3_SB(sb)->s_journal;
698
699 WARN_ON(PageChecked(page));
700 if (!page_has_buffers(page))
701 return 0;
702 if (journal)
703 return journal_try_to_free_buffers(journal, page,
704 wait & ~__GFP_WAIT);
705 return try_to_free_buffers(page);
706 }
707
708 #ifdef CONFIG_QUOTA
709 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
710 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
711
712 static int ext3_write_dquot(struct dquot *dquot);
713 static int ext3_acquire_dquot(struct dquot *dquot);
714 static int ext3_release_dquot(struct dquot *dquot);
715 static int ext3_mark_dquot_dirty(struct dquot *dquot);
716 static int ext3_write_info(struct super_block *sb, int type);
717 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
718 char *path, int remount);
719 static int ext3_quota_on_mount(struct super_block *sb, int type);
720 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
721 size_t len, loff_t off);
722 static ssize_t ext3_quota_write(struct super_block *sb, int type,
723 const char *data, size_t len, loff_t off);
724
725 static const struct dquot_operations ext3_quota_operations = {
726 .initialize = dquot_initialize,
727 .drop = dquot_drop,
728 .alloc_space = dquot_alloc_space,
729 .alloc_inode = dquot_alloc_inode,
730 .free_space = dquot_free_space,
731 .free_inode = dquot_free_inode,
732 .transfer = dquot_transfer,
733 .write_dquot = ext3_write_dquot,
734 .acquire_dquot = ext3_acquire_dquot,
735 .release_dquot = ext3_release_dquot,
736 .mark_dirty = ext3_mark_dquot_dirty,
737 .write_info = ext3_write_info,
738 .alloc_dquot = dquot_alloc,
739 .destroy_dquot = dquot_destroy,
740 };
741
742 static const struct quotactl_ops ext3_qctl_operations = {
743 .quota_on = ext3_quota_on,
744 .quota_off = vfs_quota_off,
745 .quota_sync = vfs_quota_sync,
746 .get_info = vfs_get_dqinfo,
747 .set_info = vfs_set_dqinfo,
748 .get_dqblk = vfs_get_dqblk,
749 .set_dqblk = vfs_set_dqblk
750 };
751 #endif
752
753 static const struct super_operations ext3_sops = {
754 .alloc_inode = ext3_alloc_inode,
755 .destroy_inode = ext3_destroy_inode,
756 .write_inode = ext3_write_inode,
757 .dirty_inode = ext3_dirty_inode,
758 .delete_inode = ext3_delete_inode,
759 .put_super = ext3_put_super,
760 .sync_fs = ext3_sync_fs,
761 .freeze_fs = ext3_freeze,
762 .unfreeze_fs = ext3_unfreeze,
763 .statfs = ext3_statfs,
764 .remount_fs = ext3_remount,
765 .clear_inode = ext3_clear_inode,
766 .show_options = ext3_show_options,
767 #ifdef CONFIG_QUOTA
768 .quota_read = ext3_quota_read,
769 .quota_write = ext3_quota_write,
770 #endif
771 .bdev_try_to_free_page = bdev_try_to_free_page,
772 };
773
774 static const struct export_operations ext3_export_ops = {
775 .fh_to_dentry = ext3_fh_to_dentry,
776 .fh_to_parent = ext3_fh_to_parent,
777 .get_parent = ext3_get_parent,
778 };
779
780 enum {
781 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
782 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
783 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
784 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
785 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
786 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
787 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
788 Opt_data_err_abort, Opt_data_err_ignore,
789 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
790 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
791 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
792 Opt_grpquota
793 };
794
795 static const match_table_t tokens = {
796 {Opt_bsd_df, "bsddf"},
797 {Opt_minix_df, "minixdf"},
798 {Opt_grpid, "grpid"},
799 {Opt_grpid, "bsdgroups"},
800 {Opt_nogrpid, "nogrpid"},
801 {Opt_nogrpid, "sysvgroups"},
802 {Opt_resgid, "resgid=%u"},
803 {Opt_resuid, "resuid=%u"},
804 {Opt_sb, "sb=%u"},
805 {Opt_err_cont, "errors=continue"},
806 {Opt_err_panic, "errors=panic"},
807 {Opt_err_ro, "errors=remount-ro"},
808 {Opt_nouid32, "nouid32"},
809 {Opt_nocheck, "nocheck"},
810 {Opt_nocheck, "check=none"},
811 {Opt_debug, "debug"},
812 {Opt_oldalloc, "oldalloc"},
813 {Opt_orlov, "orlov"},
814 {Opt_user_xattr, "user_xattr"},
815 {Opt_nouser_xattr, "nouser_xattr"},
816 {Opt_acl, "acl"},
817 {Opt_noacl, "noacl"},
818 {Opt_reservation, "reservation"},
819 {Opt_noreservation, "noreservation"},
820 {Opt_noload, "noload"},
821 {Opt_nobh, "nobh"},
822 {Opt_bh, "bh"},
823 {Opt_commit, "commit=%u"},
824 {Opt_journal_update, "journal=update"},
825 {Opt_journal_inum, "journal=%u"},
826 {Opt_journal_dev, "journal_dev=%u"},
827 {Opt_abort, "abort"},
828 {Opt_data_journal, "data=journal"},
829 {Opt_data_ordered, "data=ordered"},
830 {Opt_data_writeback, "data=writeback"},
831 {Opt_data_err_abort, "data_err=abort"},
832 {Opt_data_err_ignore, "data_err=ignore"},
833 {Opt_offusrjquota, "usrjquota="},
834 {Opt_usrjquota, "usrjquota=%s"},
835 {Opt_offgrpjquota, "grpjquota="},
836 {Opt_grpjquota, "grpjquota=%s"},
837 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
838 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
839 {Opt_grpquota, "grpquota"},
840 {Opt_noquota, "noquota"},
841 {Opt_quota, "quota"},
842 {Opt_usrquota, "usrquota"},
843 {Opt_barrier, "barrier=%u"},
844 {Opt_resize, "resize"},
845 {Opt_err, NULL},
846 };
847
848 static ext3_fsblk_t get_sb_block(void **data)
849 {
850 ext3_fsblk_t sb_block;
851 char *options = (char *) *data;
852
853 if (!options || strncmp(options, "sb=", 3) != 0)
854 return 1; /* Default location */
855 options += 3;
856 /*todo: use simple_strtoll with >32bit ext3 */
857 sb_block = simple_strtoul(options, &options, 0);
858 if (*options && *options != ',') {
859 printk("EXT3-fs: Invalid sb specification: %s\n",
860 (char *) *data);
861 return 1;
862 }
863 if (*options == ',')
864 options++;
865 *data = (void *) options;
866 return sb_block;
867 }
868
869 static int parse_options (char *options, struct super_block *sb,
870 unsigned int *inum, unsigned long *journal_devnum,
871 ext3_fsblk_t *n_blocks_count, int is_remount)
872 {
873 struct ext3_sb_info *sbi = EXT3_SB(sb);
874 char * p;
875 substring_t args[MAX_OPT_ARGS];
876 int data_opt = 0;
877 int option;
878 #ifdef CONFIG_QUOTA
879 int qtype, qfmt;
880 char *qname;
881 #endif
882
883 if (!options)
884 return 1;
885
886 while ((p = strsep (&options, ",")) != NULL) {
887 int token;
888 if (!*p)
889 continue;
890
891 token = match_token(p, tokens, args);
892 switch (token) {
893 case Opt_bsd_df:
894 clear_opt (sbi->s_mount_opt, MINIX_DF);
895 break;
896 case Opt_minix_df:
897 set_opt (sbi->s_mount_opt, MINIX_DF);
898 break;
899 case Opt_grpid:
900 set_opt (sbi->s_mount_opt, GRPID);
901 break;
902 case Opt_nogrpid:
903 clear_opt (sbi->s_mount_opt, GRPID);
904 break;
905 case Opt_resuid:
906 if (match_int(&args[0], &option))
907 return 0;
908 sbi->s_resuid = option;
909 break;
910 case Opt_resgid:
911 if (match_int(&args[0], &option))
912 return 0;
913 sbi->s_resgid = option;
914 break;
915 case Opt_sb:
916 /* handled by get_sb_block() instead of here */
917 /* *sb_block = match_int(&args[0]); */
918 break;
919 case Opt_err_panic:
920 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
921 clear_opt (sbi->s_mount_opt, ERRORS_RO);
922 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
923 break;
924 case Opt_err_ro:
925 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
926 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
927 set_opt (sbi->s_mount_opt, ERRORS_RO);
928 break;
929 case Opt_err_cont:
930 clear_opt (sbi->s_mount_opt, ERRORS_RO);
931 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
932 set_opt (sbi->s_mount_opt, ERRORS_CONT);
933 break;
934 case Opt_nouid32:
935 set_opt (sbi->s_mount_opt, NO_UID32);
936 break;
937 case Opt_nocheck:
938 clear_opt (sbi->s_mount_opt, CHECK);
939 break;
940 case Opt_debug:
941 set_opt (sbi->s_mount_opt, DEBUG);
942 break;
943 case Opt_oldalloc:
944 set_opt (sbi->s_mount_opt, OLDALLOC);
945 break;
946 case Opt_orlov:
947 clear_opt (sbi->s_mount_opt, OLDALLOC);
948 break;
949 #ifdef CONFIG_EXT3_FS_XATTR
950 case Opt_user_xattr:
951 set_opt (sbi->s_mount_opt, XATTR_USER);
952 break;
953 case Opt_nouser_xattr:
954 clear_opt (sbi->s_mount_opt, XATTR_USER);
955 break;
956 #else
957 case Opt_user_xattr:
958 case Opt_nouser_xattr:
959 printk("EXT3 (no)user_xattr options not supported\n");
960 break;
961 #endif
962 #ifdef CONFIG_EXT3_FS_POSIX_ACL
963 case Opt_acl:
964 set_opt(sbi->s_mount_opt, POSIX_ACL);
965 break;
966 case Opt_noacl:
967 clear_opt(sbi->s_mount_opt, POSIX_ACL);
968 break;
969 #else
970 case Opt_acl:
971 case Opt_noacl:
972 printk("EXT3 (no)acl options not supported\n");
973 break;
974 #endif
975 case Opt_reservation:
976 set_opt(sbi->s_mount_opt, RESERVATION);
977 break;
978 case Opt_noreservation:
979 clear_opt(sbi->s_mount_opt, RESERVATION);
980 break;
981 case Opt_journal_update:
982 /* @@@ FIXME */
983 /* Eventually we will want to be able to create
984 a journal file here. For now, only allow the
985 user to specify an existing inode to be the
986 journal file. */
987 if (is_remount) {
988 printk(KERN_ERR "EXT3-fs: cannot specify "
989 "journal on remount\n");
990 return 0;
991 }
992 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
993 break;
994 case Opt_journal_inum:
995 if (is_remount) {
996 printk(KERN_ERR "EXT3-fs: cannot specify "
997 "journal on remount\n");
998 return 0;
999 }
1000 if (match_int(&args[0], &option))
1001 return 0;
1002 *inum = option;
1003 break;
1004 case Opt_journal_dev:
1005 if (is_remount) {
1006 printk(KERN_ERR "EXT3-fs: cannot specify "
1007 "journal on remount\n");
1008 return 0;
1009 }
1010 if (match_int(&args[0], &option))
1011 return 0;
1012 *journal_devnum = option;
1013 break;
1014 case Opt_noload:
1015 set_opt (sbi->s_mount_opt, NOLOAD);
1016 break;
1017 case Opt_commit:
1018 if (match_int(&args[0], &option))
1019 return 0;
1020 if (option < 0)
1021 return 0;
1022 if (option == 0)
1023 option = JBD_DEFAULT_MAX_COMMIT_AGE;
1024 sbi->s_commit_interval = HZ * option;
1025 break;
1026 case Opt_data_journal:
1027 data_opt = EXT3_MOUNT_JOURNAL_DATA;
1028 goto datacheck;
1029 case Opt_data_ordered:
1030 data_opt = EXT3_MOUNT_ORDERED_DATA;
1031 goto datacheck;
1032 case Opt_data_writeback:
1033 data_opt = EXT3_MOUNT_WRITEBACK_DATA;
1034 datacheck:
1035 if (is_remount) {
1036 if ((sbi->s_mount_opt & EXT3_MOUNT_DATA_FLAGS)
1037 == data_opt)
1038 break;
1039 printk(KERN_ERR
1040 "EXT3-fs (device %s): Cannot change "
1041 "data mode on remount. The filesystem "
1042 "is mounted in data=%s mode and you "
1043 "try to remount it in data=%s mode.\n",
1044 sb->s_id,
1045 data_mode_string(sbi->s_mount_opt &
1046 EXT3_MOUNT_DATA_FLAGS),
1047 data_mode_string(data_opt));
1048 return 0;
1049 } else {
1050 sbi->s_mount_opt &= ~EXT3_MOUNT_DATA_FLAGS;
1051 sbi->s_mount_opt |= data_opt;
1052 }
1053 break;
1054 case Opt_data_err_abort:
1055 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1056 break;
1057 case Opt_data_err_ignore:
1058 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1059 break;
1060 #ifdef CONFIG_QUOTA
1061 case Opt_usrjquota:
1062 qtype = USRQUOTA;
1063 goto set_qf_name;
1064 case Opt_grpjquota:
1065 qtype = GRPQUOTA;
1066 set_qf_name:
1067 if (sb_any_quota_loaded(sb) &&
1068 !sbi->s_qf_names[qtype]) {
1069 printk(KERN_ERR
1070 "EXT3-fs: Cannot change journaled "
1071 "quota options when quota turned on.\n");
1072 return 0;
1073 }
1074 qname = match_strdup(&args[0]);
1075 if (!qname) {
1076 printk(KERN_ERR
1077 "EXT3-fs: not enough memory for "
1078 "storing quotafile name.\n");
1079 return 0;
1080 }
1081 if (sbi->s_qf_names[qtype] &&
1082 strcmp(sbi->s_qf_names[qtype], qname)) {
1083 printk(KERN_ERR
1084 "EXT3-fs: %s quota file already "
1085 "specified.\n", QTYPE2NAME(qtype));
1086 kfree(qname);
1087 return 0;
1088 }
1089 sbi->s_qf_names[qtype] = qname;
1090 if (strchr(sbi->s_qf_names[qtype], '/')) {
1091 printk(KERN_ERR
1092 "EXT3-fs: quotafile must be on "
1093 "filesystem root.\n");
1094 kfree(sbi->s_qf_names[qtype]);
1095 sbi->s_qf_names[qtype] = NULL;
1096 return 0;
1097 }
1098 set_opt(sbi->s_mount_opt, QUOTA);
1099 break;
1100 case Opt_offusrjquota:
1101 qtype = USRQUOTA;
1102 goto clear_qf_name;
1103 case Opt_offgrpjquota:
1104 qtype = GRPQUOTA;
1105 clear_qf_name:
1106 if (sb_any_quota_loaded(sb) &&
1107 sbi->s_qf_names[qtype]) {
1108 printk(KERN_ERR "EXT3-fs: Cannot change "
1109 "journaled quota options when "
1110 "quota turned on.\n");
1111 return 0;
1112 }
1113 /*
1114 * The space will be released later when all options
1115 * are confirmed to be correct
1116 */
1117 sbi->s_qf_names[qtype] = NULL;
1118 break;
1119 case Opt_jqfmt_vfsold:
1120 qfmt = QFMT_VFS_OLD;
1121 goto set_qf_format;
1122 case Opt_jqfmt_vfsv0:
1123 qfmt = QFMT_VFS_V0;
1124 set_qf_format:
1125 if (sb_any_quota_loaded(sb) &&
1126 sbi->s_jquota_fmt != qfmt) {
1127 printk(KERN_ERR "EXT3-fs: Cannot change "
1128 "journaled quota options when "
1129 "quota turned on.\n");
1130 return 0;
1131 }
1132 sbi->s_jquota_fmt = qfmt;
1133 break;
1134 case Opt_quota:
1135 case Opt_usrquota:
1136 set_opt(sbi->s_mount_opt, QUOTA);
1137 set_opt(sbi->s_mount_opt, USRQUOTA);
1138 break;
1139 case Opt_grpquota:
1140 set_opt(sbi->s_mount_opt, QUOTA);
1141 set_opt(sbi->s_mount_opt, GRPQUOTA);
1142 break;
1143 case Opt_noquota:
1144 if (sb_any_quota_loaded(sb)) {
1145 printk(KERN_ERR "EXT3-fs: Cannot change quota "
1146 "options when quota turned on.\n");
1147 return 0;
1148 }
1149 clear_opt(sbi->s_mount_opt, QUOTA);
1150 clear_opt(sbi->s_mount_opt, USRQUOTA);
1151 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1152 break;
1153 #else
1154 case Opt_quota:
1155 case Opt_usrquota:
1156 case Opt_grpquota:
1157 printk(KERN_ERR
1158 "EXT3-fs: quota options not supported.\n");
1159 break;
1160 case Opt_usrjquota:
1161 case Opt_grpjquota:
1162 case Opt_offusrjquota:
1163 case Opt_offgrpjquota:
1164 case Opt_jqfmt_vfsold:
1165 case Opt_jqfmt_vfsv0:
1166 printk(KERN_ERR
1167 "EXT3-fs: journaled quota options not "
1168 "supported.\n");
1169 break;
1170 case Opt_noquota:
1171 break;
1172 #endif
1173 case Opt_abort:
1174 set_opt(sbi->s_mount_opt, ABORT);
1175 break;
1176 case Opt_barrier:
1177 if (match_int(&args[0], &option))
1178 return 0;
1179 if (option)
1180 set_opt(sbi->s_mount_opt, BARRIER);
1181 else
1182 clear_opt(sbi->s_mount_opt, BARRIER);
1183 break;
1184 case Opt_ignore:
1185 break;
1186 case Opt_resize:
1187 if (!is_remount) {
1188 printk("EXT3-fs: resize option only available "
1189 "for remount\n");
1190 return 0;
1191 }
1192 if (match_int(&args[0], &option) != 0)
1193 return 0;
1194 *n_blocks_count = option;
1195 break;
1196 case Opt_nobh:
1197 set_opt(sbi->s_mount_opt, NOBH);
1198 break;
1199 case Opt_bh:
1200 clear_opt(sbi->s_mount_opt, NOBH);
1201 break;
1202 default:
1203 printk (KERN_ERR
1204 "EXT3-fs: Unrecognized mount option \"%s\" "
1205 "or missing value\n", p);
1206 return 0;
1207 }
1208 }
1209 #ifdef CONFIG_QUOTA
1210 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1211 if ((sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA) &&
1212 sbi->s_qf_names[USRQUOTA])
1213 clear_opt(sbi->s_mount_opt, USRQUOTA);
1214
1215 if ((sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA) &&
1216 sbi->s_qf_names[GRPQUOTA])
1217 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1218
1219 if ((sbi->s_qf_names[USRQUOTA] &&
1220 (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)) ||
1221 (sbi->s_qf_names[GRPQUOTA] &&
1222 (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA))) {
1223 printk(KERN_ERR "EXT3-fs: old and new quota "
1224 "format mixing.\n");
1225 return 0;
1226 }
1227
1228 if (!sbi->s_jquota_fmt) {
1229 printk(KERN_ERR "EXT3-fs: journaled quota format "
1230 "not specified.\n");
1231 return 0;
1232 }
1233 } else {
1234 if (sbi->s_jquota_fmt) {
1235 printk(KERN_ERR "EXT3-fs: journaled quota format "
1236 "specified with no journaling "
1237 "enabled.\n");
1238 return 0;
1239 }
1240 }
1241 #endif
1242 return 1;
1243 }
1244
1245 static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
1246 int read_only)
1247 {
1248 struct ext3_sb_info *sbi = EXT3_SB(sb);
1249 int res = 0;
1250
1251 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
1252 printk (KERN_ERR "EXT3-fs warning: revision level too high, "
1253 "forcing read-only mode\n");
1254 res = MS_RDONLY;
1255 }
1256 if (read_only)
1257 return res;
1258 if (!(sbi->s_mount_state & EXT3_VALID_FS))
1259 printk (KERN_WARNING "EXT3-fs warning: mounting unchecked fs, "
1260 "running e2fsck is recommended\n");
1261 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
1262 printk (KERN_WARNING
1263 "EXT3-fs warning: mounting fs with errors, "
1264 "running e2fsck is recommended\n");
1265 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1266 le16_to_cpu(es->s_mnt_count) >=
1267 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1268 printk (KERN_WARNING
1269 "EXT3-fs warning: maximal mount count reached, "
1270 "running e2fsck is recommended\n");
1271 else if (le32_to_cpu(es->s_checkinterval) &&
1272 (le32_to_cpu(es->s_lastcheck) +
1273 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1274 printk (KERN_WARNING
1275 "EXT3-fs warning: checktime reached, "
1276 "running e2fsck is recommended\n");
1277 #if 0
1278 /* @@@ We _will_ want to clear the valid bit if we find
1279 inconsistencies, to force a fsck at reboot. But for
1280 a plain journaled filesystem we can keep it set as
1281 valid forever! :) */
1282 es->s_state &= cpu_to_le16(~EXT3_VALID_FS);
1283 #endif
1284 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1285 es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
1286 le16_add_cpu(&es->s_mnt_count, 1);
1287 es->s_mtime = cpu_to_le32(get_seconds());
1288 ext3_update_dynamic_rev(sb);
1289 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1290
1291 ext3_commit_super(sb, es, 1);
1292 if (test_opt(sb, DEBUG))
1293 printk(KERN_INFO "[EXT3 FS bs=%lu, gc=%lu, "
1294 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1295 sb->s_blocksize,
1296 sbi->s_groups_count,
1297 EXT3_BLOCKS_PER_GROUP(sb),
1298 EXT3_INODES_PER_GROUP(sb),
1299 sbi->s_mount_opt);
1300
1301 printk(KERN_INFO "EXT3 FS on %s, ", sb->s_id);
1302 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
1303 char b[BDEVNAME_SIZE];
1304
1305 printk("external journal on %s\n",
1306 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1307 } else {
1308 printk("internal journal\n");
1309 }
1310 return res;
1311 }
1312
1313 /* Called at mount-time, super-block is locked */
1314 static int ext3_check_descriptors(struct super_block *sb)
1315 {
1316 struct ext3_sb_info *sbi = EXT3_SB(sb);
1317 int i;
1318
1319 ext3_debug ("Checking group descriptors");
1320
1321 for (i = 0; i < sbi->s_groups_count; i++) {
1322 struct ext3_group_desc *gdp = ext3_get_group_desc(sb, i, NULL);
1323 ext3_fsblk_t first_block = ext3_group_first_block_no(sb, i);
1324 ext3_fsblk_t last_block;
1325
1326 if (i == sbi->s_groups_count - 1)
1327 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
1328 else
1329 last_block = first_block +
1330 (EXT3_BLOCKS_PER_GROUP(sb) - 1);
1331
1332 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
1333 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
1334 {
1335 ext3_error (sb, "ext3_check_descriptors",
1336 "Block bitmap for group %d"
1337 " not in group (block %lu)!",
1338 i, (unsigned long)
1339 le32_to_cpu(gdp->bg_block_bitmap));
1340 return 0;
1341 }
1342 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
1343 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
1344 {
1345 ext3_error (sb, "ext3_check_descriptors",
1346 "Inode bitmap for group %d"
1347 " not in group (block %lu)!",
1348 i, (unsigned long)
1349 le32_to_cpu(gdp->bg_inode_bitmap));
1350 return 0;
1351 }
1352 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
1353 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
1354 last_block)
1355 {
1356 ext3_error (sb, "ext3_check_descriptors",
1357 "Inode table for group %d"
1358 " not in group (block %lu)!",
1359 i, (unsigned long)
1360 le32_to_cpu(gdp->bg_inode_table));
1361 return 0;
1362 }
1363 }
1364
1365 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1366 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1367 return 1;
1368 }
1369
1370
1371 /* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1372 * the superblock) which were deleted from all directories, but held open by
1373 * a process at the time of a crash. We walk the list and try to delete these
1374 * inodes at recovery time (only with a read-write filesystem).
1375 *
1376 * In order to keep the orphan inode chain consistent during traversal (in
1377 * case of crash during recovery), we link each inode into the superblock
1378 * orphan list_head and handle it the same way as an inode deletion during
1379 * normal operation (which journals the operations for us).
1380 *
1381 * We only do an iget() and an iput() on each inode, which is very safe if we
1382 * accidentally point at an in-use or already deleted inode. The worst that
1383 * can happen in this case is that we get a "bit already cleared" message from
1384 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1385 * e2fsck was run on this filesystem, and it must have already done the orphan
1386 * inode cleanup for us, so we can safely abort without any further action.
1387 */
1388 static void ext3_orphan_cleanup (struct super_block * sb,
1389 struct ext3_super_block * es)
1390 {
1391 unsigned int s_flags = sb->s_flags;
1392 int nr_orphans = 0, nr_truncates = 0;
1393 #ifdef CONFIG_QUOTA
1394 int i;
1395 #endif
1396 if (!es->s_last_orphan) {
1397 jbd_debug(4, "no orphan inodes to clean up\n");
1398 return;
1399 }
1400
1401 if (bdev_read_only(sb->s_bdev)) {
1402 printk(KERN_ERR "EXT3-fs: write access "
1403 "unavailable, skipping orphan cleanup.\n");
1404 return;
1405 }
1406
1407 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1408 if (es->s_last_orphan)
1409 jbd_debug(1, "Errors on filesystem, "
1410 "clearing orphan list.\n");
1411 es->s_last_orphan = 0;
1412 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1413 return;
1414 }
1415
1416 if (s_flags & MS_RDONLY) {
1417 printk(KERN_INFO "EXT3-fs: %s: orphan cleanup on readonly fs\n",
1418 sb->s_id);
1419 sb->s_flags &= ~MS_RDONLY;
1420 }
1421 #ifdef CONFIG_QUOTA
1422 /* Needed for iput() to work correctly and not trash data */
1423 sb->s_flags |= MS_ACTIVE;
1424 /* Turn on quotas so that they are updated correctly */
1425 for (i = 0; i < MAXQUOTAS; i++) {
1426 if (EXT3_SB(sb)->s_qf_names[i]) {
1427 int ret = ext3_quota_on_mount(sb, i);
1428 if (ret < 0)
1429 printk(KERN_ERR
1430 "EXT3-fs: Cannot turn on journaled "
1431 "quota: error %d\n", ret);
1432 }
1433 }
1434 #endif
1435
1436 while (es->s_last_orphan) {
1437 struct inode *inode;
1438
1439 inode = ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1440 if (IS_ERR(inode)) {
1441 es->s_last_orphan = 0;
1442 break;
1443 }
1444
1445 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
1446 vfs_dq_init(inode);
1447 if (inode->i_nlink) {
1448 printk(KERN_DEBUG
1449 "%s: truncating inode %lu to %Ld bytes\n",
1450 __func__, inode->i_ino, inode->i_size);
1451 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1452 inode->i_ino, inode->i_size);
1453 ext3_truncate(inode);
1454 nr_truncates++;
1455 } else {
1456 printk(KERN_DEBUG
1457 "%s: deleting unreferenced inode %lu\n",
1458 __func__, inode->i_ino);
1459 jbd_debug(2, "deleting unreferenced inode %lu\n",
1460 inode->i_ino);
1461 nr_orphans++;
1462 }
1463 iput(inode); /* The delete magic happens here! */
1464 }
1465
1466 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1467
1468 if (nr_orphans)
1469 printk(KERN_INFO "EXT3-fs: %s: %d orphan inode%s deleted\n",
1470 sb->s_id, PLURAL(nr_orphans));
1471 if (nr_truncates)
1472 printk(KERN_INFO "EXT3-fs: %s: %d truncate%s cleaned up\n",
1473 sb->s_id, PLURAL(nr_truncates));
1474 #ifdef CONFIG_QUOTA
1475 /* Turn quotas off */
1476 for (i = 0; i < MAXQUOTAS; i++) {
1477 if (sb_dqopt(sb)->files[i])
1478 vfs_quota_off(sb, i, 0);
1479 }
1480 #endif
1481 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1482 }
1483
1484 /*
1485 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1486 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1487 * We need to be 1 filesystem block less than the 2^32 sector limit.
1488 */
1489 static loff_t ext3_max_size(int bits)
1490 {
1491 loff_t res = EXT3_NDIR_BLOCKS;
1492 int meta_blocks;
1493 loff_t upper_limit;
1494
1495 /* This is calculated to be the largest file size for a
1496 * dense, file such that the total number of
1497 * sectors in the file, including data and all indirect blocks,
1498 * does not exceed 2^32 -1
1499 * __u32 i_blocks representing the total number of
1500 * 512 bytes blocks of the file
1501 */
1502 upper_limit = (1LL << 32) - 1;
1503
1504 /* total blocks in file system block size */
1505 upper_limit >>= (bits - 9);
1506
1507
1508 /* indirect blocks */
1509 meta_blocks = 1;
1510 /* double indirect blocks */
1511 meta_blocks += 1 + (1LL << (bits-2));
1512 /* tripple indirect blocks */
1513 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1514
1515 upper_limit -= meta_blocks;
1516 upper_limit <<= bits;
1517
1518 res += 1LL << (bits-2);
1519 res += 1LL << (2*(bits-2));
1520 res += 1LL << (3*(bits-2));
1521 res <<= bits;
1522 if (res > upper_limit)
1523 res = upper_limit;
1524
1525 if (res > MAX_LFS_FILESIZE)
1526 res = MAX_LFS_FILESIZE;
1527
1528 return res;
1529 }
1530
1531 static ext3_fsblk_t descriptor_loc(struct super_block *sb,
1532 ext3_fsblk_t logic_sb_block,
1533 int nr)
1534 {
1535 struct ext3_sb_info *sbi = EXT3_SB(sb);
1536 unsigned long bg, first_meta_bg;
1537 int has_super = 0;
1538
1539 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1540
1541 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1542 nr < first_meta_bg)
1543 return (logic_sb_block + nr + 1);
1544 bg = sbi->s_desc_per_block * nr;
1545 if (ext3_bg_has_super(sb, bg))
1546 has_super = 1;
1547 return (has_super + ext3_group_first_block_no(sb, bg));
1548 }
1549
1550
1551 static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1552 {
1553 struct buffer_head * bh;
1554 struct ext3_super_block *es = NULL;
1555 struct ext3_sb_info *sbi;
1556 ext3_fsblk_t block;
1557 ext3_fsblk_t sb_block = get_sb_block(&data);
1558 ext3_fsblk_t logic_sb_block;
1559 unsigned long offset = 0;
1560 unsigned int journal_inum = 0;
1561 unsigned long journal_devnum = 0;
1562 unsigned long def_mount_opts;
1563 struct inode *root;
1564 int blocksize;
1565 int hblock;
1566 int db_count;
1567 int i;
1568 int needs_recovery;
1569 int ret = -EINVAL;
1570 __le32 features;
1571 int err;
1572
1573 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1574 if (!sbi)
1575 return -ENOMEM;
1576
1577 sbi->s_blockgroup_lock =
1578 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
1579 if (!sbi->s_blockgroup_lock) {
1580 kfree(sbi);
1581 return -ENOMEM;
1582 }
1583 sb->s_fs_info = sbi;
1584 sbi->s_mount_opt = 0;
1585 sbi->s_resuid = EXT3_DEF_RESUID;
1586 sbi->s_resgid = EXT3_DEF_RESGID;
1587 sbi->s_sb_block = sb_block;
1588
1589 unlock_kernel();
1590
1591 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1592 if (!blocksize) {
1593 printk(KERN_ERR "EXT3-fs: unable to set blocksize\n");
1594 goto out_fail;
1595 }
1596
1597 /*
1598 * The ext3 superblock will not be buffer aligned for other than 1kB
1599 * block sizes. We need to calculate the offset from buffer start.
1600 */
1601 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1602 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1603 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1604 } else {
1605 logic_sb_block = sb_block;
1606 }
1607
1608 if (!(bh = sb_bread(sb, logic_sb_block))) {
1609 printk (KERN_ERR "EXT3-fs: unable to read superblock\n");
1610 goto out_fail;
1611 }
1612 /*
1613 * Note: s_es must be initialized as soon as possible because
1614 * some ext3 macro-instructions depend on its value
1615 */
1616 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1617 sbi->s_es = es;
1618 sb->s_magic = le16_to_cpu(es->s_magic);
1619 if (sb->s_magic != EXT3_SUPER_MAGIC)
1620 goto cantfind_ext3;
1621
1622 /* Set defaults before we parse the mount options */
1623 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1624 if (def_mount_opts & EXT3_DEFM_DEBUG)
1625 set_opt(sbi->s_mount_opt, DEBUG);
1626 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1627 set_opt(sbi->s_mount_opt, GRPID);
1628 if (def_mount_opts & EXT3_DEFM_UID16)
1629 set_opt(sbi->s_mount_opt, NO_UID32);
1630 #ifdef CONFIG_EXT3_FS_XATTR
1631 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1632 set_opt(sbi->s_mount_opt, XATTR_USER);
1633 #endif
1634 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1635 if (def_mount_opts & EXT3_DEFM_ACL)
1636 set_opt(sbi->s_mount_opt, POSIX_ACL);
1637 #endif
1638 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
1639 sbi->s_mount_opt |= EXT3_MOUNT_JOURNAL_DATA;
1640 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
1641 sbi->s_mount_opt |= EXT3_MOUNT_ORDERED_DATA;
1642 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
1643 sbi->s_mount_opt |= EXT3_MOUNT_WRITEBACK_DATA;
1644
1645 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1646 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1647 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_CONTINUE)
1648 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1649 else
1650 set_opt(sbi->s_mount_opt, ERRORS_RO);
1651
1652 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1653 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1654
1655 set_opt(sbi->s_mount_opt, RESERVATION);
1656
1657 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1658 NULL, 0))
1659 goto failed_mount;
1660
1661 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1662 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1663
1664 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1665 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1666 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1667 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1668 printk(KERN_WARNING
1669 "EXT3-fs warning: feature flags set on rev 0 fs, "
1670 "running e2fsck is recommended\n");
1671 /*
1672 * Check feature flags regardless of the revision level, since we
1673 * previously didn't change the revision level when setting the flags,
1674 * so there is a chance incompat flags are set on a rev 0 filesystem.
1675 */
1676 features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
1677 if (features) {
1678 printk(KERN_ERR "EXT3-fs: %s: couldn't mount because of "
1679 "unsupported optional features (%x).\n",
1680 sb->s_id, le32_to_cpu(features));
1681 goto failed_mount;
1682 }
1683 features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
1684 if (!(sb->s_flags & MS_RDONLY) && features) {
1685 printk(KERN_ERR "EXT3-fs: %s: couldn't mount RDWR because of "
1686 "unsupported optional features (%x).\n",
1687 sb->s_id, le32_to_cpu(features));
1688 goto failed_mount;
1689 }
1690 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1691
1692 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1693 blocksize > EXT3_MAX_BLOCK_SIZE) {
1694 printk(KERN_ERR
1695 "EXT3-fs: Unsupported filesystem blocksize %d on %s.\n",
1696 blocksize, sb->s_id);
1697 goto failed_mount;
1698 }
1699
1700 hblock = bdev_logical_block_size(sb->s_bdev);
1701 if (sb->s_blocksize != blocksize) {
1702 /*
1703 * Make sure the blocksize for the filesystem is larger
1704 * than the hardware sectorsize for the machine.
1705 */
1706 if (blocksize < hblock) {
1707 printk(KERN_ERR "EXT3-fs: blocksize %d too small for "
1708 "device blocksize %d.\n", blocksize, hblock);
1709 goto failed_mount;
1710 }
1711
1712 brelse (bh);
1713 if (!sb_set_blocksize(sb, blocksize)) {
1714 printk(KERN_ERR "EXT3-fs: bad blocksize %d.\n",
1715 blocksize);
1716 goto out_fail;
1717 }
1718 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1719 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1720 bh = sb_bread(sb, logic_sb_block);
1721 if (!bh) {
1722 printk(KERN_ERR
1723 "EXT3-fs: Can't read superblock on 2nd try.\n");
1724 goto failed_mount;
1725 }
1726 es = (struct ext3_super_block *)(((char *)bh->b_data) + offset);
1727 sbi->s_es = es;
1728 if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
1729 printk (KERN_ERR
1730 "EXT3-fs: Magic mismatch, very weird !\n");
1731 goto failed_mount;
1732 }
1733 }
1734
1735 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1736
1737 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1738 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1739 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1740 } else {
1741 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1742 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1743 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
1744 (!is_power_of_2(sbi->s_inode_size)) ||
1745 (sbi->s_inode_size > blocksize)) {
1746 printk (KERN_ERR
1747 "EXT3-fs: unsupported inode size: %d\n",
1748 sbi->s_inode_size);
1749 goto failed_mount;
1750 }
1751 }
1752 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1753 le32_to_cpu(es->s_log_frag_size);
1754 if (blocksize != sbi->s_frag_size) {
1755 printk(KERN_ERR
1756 "EXT3-fs: fragsize %lu != blocksize %u (unsupported)\n",
1757 sbi->s_frag_size, blocksize);
1758 goto failed_mount;
1759 }
1760 sbi->s_frags_per_block = 1;
1761 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1762 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1763 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1764 if (EXT3_INODE_SIZE(sb) == 0 || EXT3_INODES_PER_GROUP(sb) == 0)
1765 goto cantfind_ext3;
1766 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1767 if (sbi->s_inodes_per_block == 0)
1768 goto cantfind_ext3;
1769 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1770 sbi->s_inodes_per_block;
1771 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1772 sbi->s_sbh = bh;
1773 sbi->s_mount_state = le16_to_cpu(es->s_state);
1774 sbi->s_addr_per_block_bits = ilog2(EXT3_ADDR_PER_BLOCK(sb));
1775 sbi->s_desc_per_block_bits = ilog2(EXT3_DESC_PER_BLOCK(sb));
1776 for (i=0; i < 4; i++)
1777 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1778 sbi->s_def_hash_version = es->s_def_hash_version;
1779 i = le32_to_cpu(es->s_flags);
1780 if (i & EXT2_FLAGS_UNSIGNED_HASH)
1781 sbi->s_hash_unsigned = 3;
1782 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
1783 #ifdef __CHAR_UNSIGNED__
1784 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
1785 sbi->s_hash_unsigned = 3;
1786 #else
1787 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
1788 #endif
1789 }
1790
1791 if (sbi->s_blocks_per_group > blocksize * 8) {
1792 printk (KERN_ERR
1793 "EXT3-fs: #blocks per group too big: %lu\n",
1794 sbi->s_blocks_per_group);
1795 goto failed_mount;
1796 }
1797 if (sbi->s_frags_per_group > blocksize * 8) {
1798 printk (KERN_ERR
1799 "EXT3-fs: #fragments per group too big: %lu\n",
1800 sbi->s_frags_per_group);
1801 goto failed_mount;
1802 }
1803 if (sbi->s_inodes_per_group > blocksize * 8) {
1804 printk (KERN_ERR
1805 "EXT3-fs: #inodes per group too big: %lu\n",
1806 sbi->s_inodes_per_group);
1807 goto failed_mount;
1808 }
1809
1810 if (le32_to_cpu(es->s_blocks_count) >
1811 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
1812 printk(KERN_ERR "EXT3-fs: filesystem on %s:"
1813 " too large to mount safely\n", sb->s_id);
1814 if (sizeof(sector_t) < 8)
1815 printk(KERN_WARNING "EXT3-fs: CONFIG_LBDAF not "
1816 "enabled\n");
1817 goto failed_mount;
1818 }
1819
1820 if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1821 goto cantfind_ext3;
1822 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1823 le32_to_cpu(es->s_first_data_block) - 1)
1824 / EXT3_BLOCKS_PER_GROUP(sb)) + 1;
1825 db_count = (sbi->s_groups_count + EXT3_DESC_PER_BLOCK(sb) - 1) /
1826 EXT3_DESC_PER_BLOCK(sb);
1827 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1828 GFP_KERNEL);
1829 if (sbi->s_group_desc == NULL) {
1830 printk (KERN_ERR "EXT3-fs: not enough memory\n");
1831 goto failed_mount;
1832 }
1833
1834 bgl_lock_init(sbi->s_blockgroup_lock);
1835
1836 for (i = 0; i < db_count; i++) {
1837 block = descriptor_loc(sb, logic_sb_block, i);
1838 sbi->s_group_desc[i] = sb_bread(sb, block);
1839 if (!sbi->s_group_desc[i]) {
1840 printk (KERN_ERR "EXT3-fs: "
1841 "can't read group descriptor %d\n", i);
1842 db_count = i;
1843 goto failed_mount2;
1844 }
1845 }
1846 if (!ext3_check_descriptors (sb)) {
1847 printk(KERN_ERR "EXT3-fs: group descriptors corrupted!\n");
1848 goto failed_mount2;
1849 }
1850 sbi->s_gdb_count = db_count;
1851 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1852 spin_lock_init(&sbi->s_next_gen_lock);
1853
1854 err = percpu_counter_init(&sbi->s_freeblocks_counter,
1855 ext3_count_free_blocks(sb));
1856 if (!err) {
1857 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1858 ext3_count_free_inodes(sb));
1859 }
1860 if (!err) {
1861 err = percpu_counter_init(&sbi->s_dirs_counter,
1862 ext3_count_dirs(sb));
1863 }
1864 if (err) {
1865 printk(KERN_ERR "EXT3-fs: insufficient memory\n");
1866 goto failed_mount3;
1867 }
1868
1869 /* per fileystem reservation list head & lock */
1870 spin_lock_init(&sbi->s_rsv_window_lock);
1871 sbi->s_rsv_window_root = RB_ROOT;
1872 /* Add a single, static dummy reservation to the start of the
1873 * reservation window list --- it gives us a placeholder for
1874 * append-at-start-of-list which makes the allocation logic
1875 * _much_ simpler. */
1876 sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1877 sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1878 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1879 sbi->s_rsv_window_head.rsv_goal_size = 0;
1880 ext3_rsv_window_add(sb, &sbi->s_rsv_window_head);
1881
1882 /*
1883 * set up enough so that it can read an inode
1884 */
1885 sb->s_op = &ext3_sops;
1886 sb->s_export_op = &ext3_export_ops;
1887 sb->s_xattr = ext3_xattr_handlers;
1888 #ifdef CONFIG_QUOTA
1889 sb->s_qcop = &ext3_qctl_operations;
1890 sb->dq_op = &ext3_quota_operations;
1891 #endif
1892 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1893
1894 sb->s_root = NULL;
1895
1896 needs_recovery = (es->s_last_orphan != 0 ||
1897 EXT3_HAS_INCOMPAT_FEATURE(sb,
1898 EXT3_FEATURE_INCOMPAT_RECOVER));
1899
1900 /*
1901 * The first inode we look at is the journal inode. Don't try
1902 * root first: it may be modified in the journal!
1903 */
1904 if (!test_opt(sb, NOLOAD) &&
1905 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
1906 if (ext3_load_journal(sb, es, journal_devnum))
1907 goto failed_mount3;
1908 } else if (journal_inum) {
1909 if (ext3_create_journal(sb, es, journal_inum))
1910 goto failed_mount3;
1911 } else {
1912 if (!silent)
1913 printk (KERN_ERR
1914 "ext3: No journal on filesystem on %s\n",
1915 sb->s_id);
1916 goto failed_mount3;
1917 }
1918
1919 /* We have now updated the journal if required, so we can
1920 * validate the data journaling mode. */
1921 switch (test_opt(sb, DATA_FLAGS)) {
1922 case 0:
1923 /* No mode set, assume a default based on the journal
1924 capabilities: ORDERED_DATA if the journal can
1925 cope, else JOURNAL_DATA */
1926 if (journal_check_available_features
1927 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
1928 set_opt(sbi->s_mount_opt, DEFAULT_DATA_MODE);
1929 else
1930 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1931 break;
1932
1933 case EXT3_MOUNT_ORDERED_DATA:
1934 case EXT3_MOUNT_WRITEBACK_DATA:
1935 if (!journal_check_available_features
1936 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
1937 printk(KERN_ERR "EXT3-fs: Journal does not support "
1938 "requested data journaling mode\n");
1939 goto failed_mount4;
1940 }
1941 default:
1942 break;
1943 }
1944
1945 if (test_opt(sb, NOBH)) {
1946 if (!(test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)) {
1947 printk(KERN_WARNING "EXT3-fs: Ignoring nobh option - "
1948 "its supported only with writeback mode\n");
1949 clear_opt(sbi->s_mount_opt, NOBH);
1950 }
1951 }
1952 /*
1953 * The journal_load will have done any necessary log recovery,
1954 * so we can safely mount the rest of the filesystem now.
1955 */
1956
1957 root = ext3_iget(sb, EXT3_ROOT_INO);
1958 if (IS_ERR(root)) {
1959 printk(KERN_ERR "EXT3-fs: get root inode failed\n");
1960 ret = PTR_ERR(root);
1961 goto failed_mount4;
1962 }
1963 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1964 iput(root);
1965 printk(KERN_ERR "EXT3-fs: corrupt root inode, run e2fsck\n");
1966 goto failed_mount4;
1967 }
1968 sb->s_root = d_alloc_root(root);
1969 if (!sb->s_root) {
1970 printk(KERN_ERR "EXT3-fs: get root dentry failed\n");
1971 iput(root);
1972 ret = -ENOMEM;
1973 goto failed_mount4;
1974 }
1975
1976 ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1977 /*
1978 * akpm: core read_super() calls in here with the superblock locked.
1979 * That deadlocks, because orphan cleanup needs to lock the superblock
1980 * in numerous places. Here we just pop the lock - it's relatively
1981 * harmless, because we are now ready to accept write_super() requests,
1982 * and aviro says that's the only reason for hanging onto the
1983 * superblock lock.
1984 */
1985 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
1986 ext3_orphan_cleanup(sb, es);
1987 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
1988 if (needs_recovery)
1989 printk (KERN_INFO "EXT3-fs: recovery complete.\n");
1990 ext3_mark_recovery_complete(sb, es);
1991 printk (KERN_INFO "EXT3-fs: mounted filesystem with %s data mode.\n",
1992 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
1993 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
1994 "writeback");
1995
1996 lock_kernel();
1997 return 0;
1998
1999 cantfind_ext3:
2000 if (!silent)
2001 printk(KERN_ERR "VFS: Can't find ext3 filesystem on dev %s.\n",
2002 sb->s_id);
2003 goto failed_mount;
2004
2005 failed_mount4:
2006 journal_destroy(sbi->s_journal);
2007 failed_mount3:
2008 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2009 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2010 percpu_counter_destroy(&sbi->s_dirs_counter);
2011 failed_mount2:
2012 for (i = 0; i < db_count; i++)
2013 brelse(sbi->s_group_desc[i]);
2014 kfree(sbi->s_group_desc);
2015 failed_mount:
2016 #ifdef CONFIG_QUOTA
2017 for (i = 0; i < MAXQUOTAS; i++)
2018 kfree(sbi->s_qf_names[i]);
2019 #endif
2020 ext3_blkdev_remove(sbi);
2021 brelse(bh);
2022 out_fail:
2023 sb->s_fs_info = NULL;
2024 kfree(sbi->s_blockgroup_lock);
2025 kfree(sbi);
2026 lock_kernel();
2027 return ret;
2028 }
2029
2030 /*
2031 * Setup any per-fs journal parameters now. We'll do this both on
2032 * initial mount, once the journal has been initialised but before we've
2033 * done any recovery; and again on any subsequent remount.
2034 */
2035 static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
2036 {
2037 struct ext3_sb_info *sbi = EXT3_SB(sb);
2038
2039 if (sbi->s_commit_interval)
2040 journal->j_commit_interval = sbi->s_commit_interval;
2041 /* We could also set up an ext3-specific default for the commit
2042 * interval here, but for now we'll just fall back to the jbd
2043 * default. */
2044
2045 spin_lock(&journal->j_state_lock);
2046 if (test_opt(sb, BARRIER))
2047 journal->j_flags |= JFS_BARRIER;
2048 else
2049 journal->j_flags &= ~JFS_BARRIER;
2050 if (test_opt(sb, DATA_ERR_ABORT))
2051 journal->j_flags |= JFS_ABORT_ON_SYNCDATA_ERR;
2052 else
2053 journal->j_flags &= ~JFS_ABORT_ON_SYNCDATA_ERR;
2054 spin_unlock(&journal->j_state_lock);
2055 }
2056
2057 static journal_t *ext3_get_journal(struct super_block *sb,
2058 unsigned int journal_inum)
2059 {
2060 struct inode *journal_inode;
2061 journal_t *journal;
2062
2063 /* First, test for the existence of a valid inode on disk. Bad
2064 * things happen if we iget() an unused inode, as the subsequent
2065 * iput() will try to delete it. */
2066
2067 journal_inode = ext3_iget(sb, journal_inum);
2068 if (IS_ERR(journal_inode)) {
2069 printk(KERN_ERR "EXT3-fs: no journal found.\n");
2070 return NULL;
2071 }
2072 if (!journal_inode->i_nlink) {
2073 make_bad_inode(journal_inode);
2074 iput(journal_inode);
2075 printk(KERN_ERR "EXT3-fs: journal inode is deleted.\n");
2076 return NULL;
2077 }
2078
2079 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2080 journal_inode, journal_inode->i_size);
2081 if (!S_ISREG(journal_inode->i_mode)) {
2082 printk(KERN_ERR "EXT3-fs: invalid journal inode.\n");
2083 iput(journal_inode);
2084 return NULL;
2085 }
2086
2087 journal = journal_init_inode(journal_inode);
2088 if (!journal) {
2089 printk(KERN_ERR "EXT3-fs: Could not load journal inode\n");
2090 iput(journal_inode);
2091 return NULL;
2092 }
2093 journal->j_private = sb;
2094 ext3_init_journal_params(sb, journal);
2095 return journal;
2096 }
2097
2098 static journal_t *ext3_get_dev_journal(struct super_block *sb,
2099 dev_t j_dev)
2100 {
2101 struct buffer_head * bh;
2102 journal_t *journal;
2103 ext3_fsblk_t start;
2104 ext3_fsblk_t len;
2105 int hblock, blocksize;
2106 ext3_fsblk_t sb_block;
2107 unsigned long offset;
2108 struct ext3_super_block * es;
2109 struct block_device *bdev;
2110
2111 bdev = ext3_blkdev_get(j_dev);
2112 if (bdev == NULL)
2113 return NULL;
2114
2115 if (bd_claim(bdev, sb)) {
2116 printk(KERN_ERR
2117 "EXT3: failed to claim external journal device.\n");
2118 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
2119 return NULL;
2120 }
2121
2122 blocksize = sb->s_blocksize;
2123 hblock = bdev_logical_block_size(bdev);
2124 if (blocksize < hblock) {
2125 printk(KERN_ERR
2126 "EXT3-fs: blocksize too small for journal device.\n");
2127 goto out_bdev;
2128 }
2129
2130 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
2131 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
2132 set_blocksize(bdev, blocksize);
2133 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2134 printk(KERN_ERR "EXT3-fs: couldn't read superblock of "
2135 "external journal\n");
2136 goto out_bdev;
2137 }
2138
2139 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
2140 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
2141 !(le32_to_cpu(es->s_feature_incompat) &
2142 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2143 printk(KERN_ERR "EXT3-fs: external journal has "
2144 "bad superblock\n");
2145 brelse(bh);
2146 goto out_bdev;
2147 }
2148
2149 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2150 printk(KERN_ERR "EXT3-fs: journal UUID does not match\n");
2151 brelse(bh);
2152 goto out_bdev;
2153 }
2154
2155 len = le32_to_cpu(es->s_blocks_count);
2156 start = sb_block + 1;
2157 brelse(bh); /* we're done with the superblock */
2158
2159 journal = journal_init_dev(bdev, sb->s_bdev,
2160 start, len, blocksize);
2161 if (!journal) {
2162 printk(KERN_ERR "EXT3-fs: failed to create device journal\n");
2163 goto out_bdev;
2164 }
2165 journal->j_private = sb;
2166 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2167 wait_on_buffer(journal->j_sb_buffer);
2168 if (!buffer_uptodate(journal->j_sb_buffer)) {
2169 printk(KERN_ERR "EXT3-fs: I/O error on journal device\n");
2170 goto out_journal;
2171 }
2172 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2173 printk(KERN_ERR "EXT3-fs: External journal has more than one "
2174 "user (unsupported) - %d\n",
2175 be32_to_cpu(journal->j_superblock->s_nr_users));
2176 goto out_journal;
2177 }
2178 EXT3_SB(sb)->journal_bdev = bdev;
2179 ext3_init_journal_params(sb, journal);
2180 return journal;
2181 out_journal:
2182 journal_destroy(journal);
2183 out_bdev:
2184 ext3_blkdev_put(bdev);
2185 return NULL;
2186 }
2187
2188 static int ext3_load_journal(struct super_block *sb,
2189 struct ext3_super_block *es,
2190 unsigned long journal_devnum)
2191 {
2192 journal_t *journal;
2193 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2194 dev_t journal_dev;
2195 int err = 0;
2196 int really_read_only;
2197
2198 if (journal_devnum &&
2199 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2200 printk(KERN_INFO "EXT3-fs: external journal device major/minor "
2201 "numbers have changed\n");
2202 journal_dev = new_decode_dev(journal_devnum);
2203 } else
2204 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2205
2206 really_read_only = bdev_read_only(sb->s_bdev);
2207
2208 /*
2209 * Are we loading a blank journal or performing recovery after a
2210 * crash? For recovery, we need to check in advance whether we
2211 * can get read-write access to the device.
2212 */
2213
2214 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
2215 if (sb->s_flags & MS_RDONLY) {
2216 printk(KERN_INFO "EXT3-fs: INFO: recovery "
2217 "required on readonly filesystem.\n");
2218 if (really_read_only) {
2219 printk(KERN_ERR "EXT3-fs: write access "
2220 "unavailable, cannot proceed.\n");
2221 return -EROFS;
2222 }
2223 printk (KERN_INFO "EXT3-fs: write access will "
2224 "be enabled during recovery.\n");
2225 }
2226 }
2227
2228 if (journal_inum && journal_dev) {
2229 printk(KERN_ERR "EXT3-fs: filesystem has both journal "
2230 "and inode journals!\n");
2231 return -EINVAL;
2232 }
2233
2234 if (journal_inum) {
2235 if (!(journal = ext3_get_journal(sb, journal_inum)))
2236 return -EINVAL;
2237 } else {
2238 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
2239 return -EINVAL;
2240 }
2241
2242 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2243 err = journal_update_format(journal);
2244 if (err) {
2245 printk(KERN_ERR "EXT3-fs: error updating journal.\n");
2246 journal_destroy(journal);
2247 return err;
2248 }
2249 }
2250
2251 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
2252 err = journal_wipe(journal, !really_read_only);
2253 if (!err)
2254 err = journal_load(journal);
2255
2256 if (err) {
2257 printk(KERN_ERR "EXT3-fs: error loading journal.\n");
2258 journal_destroy(journal);
2259 return err;
2260 }
2261
2262 EXT3_SB(sb)->s_journal = journal;
2263 ext3_clear_journal_err(sb, es);
2264
2265 if (journal_devnum &&
2266 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2267 es->s_journal_dev = cpu_to_le32(journal_devnum);
2268
2269 /* Make sure we flush the recovery flag to disk. */
2270 ext3_commit_super(sb, es, 1);
2271 }
2272
2273 return 0;
2274 }
2275
2276 static int ext3_create_journal(struct super_block * sb,
2277 struct ext3_super_block * es,
2278 unsigned int journal_inum)
2279 {
2280 journal_t *journal;
2281 int err;
2282
2283 if (sb->s_flags & MS_RDONLY) {
2284 printk(KERN_ERR "EXT3-fs: readonly filesystem when trying to "
2285 "create journal.\n");
2286 return -EROFS;
2287 }
2288
2289 journal = ext3_get_journal(sb, journal_inum);
2290 if (!journal)
2291 return -EINVAL;
2292
2293 printk(KERN_INFO "EXT3-fs: creating new journal on inode %u\n",
2294 journal_inum);
2295
2296 err = journal_create(journal);
2297 if (err) {
2298 printk(KERN_ERR "EXT3-fs: error creating journal.\n");
2299 journal_destroy(journal);
2300 return -EIO;
2301 }
2302
2303 EXT3_SB(sb)->s_journal = journal;
2304
2305 ext3_update_dynamic_rev(sb);
2306 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2307 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
2308
2309 es->s_journal_inum = cpu_to_le32(journal_inum);
2310
2311 /* Make sure we flush the recovery flag to disk. */
2312 ext3_commit_super(sb, es, 1);
2313
2314 return 0;
2315 }
2316
2317 static int ext3_commit_super(struct super_block *sb,
2318 struct ext3_super_block *es,
2319 int sync)
2320 {
2321 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
2322 int error = 0;
2323
2324 if (!sbh)
2325 return error;
2326 /*
2327 * If the file system is mounted read-only, don't update the
2328 * superblock write time. This avoids updating the superblock
2329 * write time when we are mounting the root file system
2330 * read/only but we need to replay the journal; at that point,
2331 * for people who are east of GMT and who make their clock
2332 * tick in localtime for Windows bug-for-bug compatibility,
2333 * the clock is set in the future, and this will cause e2fsck
2334 * to complain and force a full file system check.
2335 */
2336 if (!(sb->s_flags & MS_RDONLY))
2337 es->s_wtime = cpu_to_le32(get_seconds());
2338 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
2339 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
2340 BUFFER_TRACE(sbh, "marking dirty");
2341 mark_buffer_dirty(sbh);
2342 if (sync)
2343 error = sync_dirty_buffer(sbh);
2344 return error;
2345 }
2346
2347
2348 /*
2349 * Have we just finished recovery? If so, and if we are mounting (or
2350 * remounting) the filesystem readonly, then we will end up with a
2351 * consistent fs on disk. Record that fact.
2352 */
2353 static void ext3_mark_recovery_complete(struct super_block * sb,
2354 struct ext3_super_block * es)
2355 {
2356 journal_t *journal = EXT3_SB(sb)->s_journal;
2357
2358 journal_lock_updates(journal);
2359 if (journal_flush(journal) < 0)
2360 goto out;
2361
2362 lock_super(sb);
2363 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
2364 sb->s_flags & MS_RDONLY) {
2365 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2366 ext3_commit_super(sb, es, 1);
2367 }
2368 unlock_super(sb);
2369
2370 out:
2371 journal_unlock_updates(journal);
2372 }
2373
2374 /*
2375 * If we are mounting (or read-write remounting) a filesystem whose journal
2376 * has recorded an error from a previous lifetime, move that error to the
2377 * main filesystem now.
2378 */
2379 static void ext3_clear_journal_err(struct super_block * sb,
2380 struct ext3_super_block * es)
2381 {
2382 journal_t *journal;
2383 int j_errno;
2384 const char *errstr;
2385
2386 journal = EXT3_SB(sb)->s_journal;
2387
2388 /*
2389 * Now check for any error status which may have been recorded in the
2390 * journal by a prior ext3_error() or ext3_abort()
2391 */
2392
2393 j_errno = journal_errno(journal);
2394 if (j_errno) {
2395 char nbuf[16];
2396
2397 errstr = ext3_decode_error(sb, j_errno, nbuf);
2398 ext3_warning(sb, __func__, "Filesystem error recorded "
2399 "from previous mount: %s", errstr);
2400 ext3_warning(sb, __func__, "Marking fs in need of "
2401 "filesystem check.");
2402
2403 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
2404 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
2405 ext3_commit_super (sb, es, 1);
2406
2407 journal_clear_err(journal);
2408 }
2409 }
2410
2411 /*
2412 * Force the running and committing transactions to commit,
2413 * and wait on the commit.
2414 */
2415 int ext3_force_commit(struct super_block *sb)
2416 {
2417 journal_t *journal;
2418 int ret;
2419
2420 if (sb->s_flags & MS_RDONLY)
2421 return 0;
2422
2423 journal = EXT3_SB(sb)->s_journal;
2424 ret = ext3_journal_force_commit(journal);
2425 return ret;
2426 }
2427
2428 static int ext3_sync_fs(struct super_block *sb, int wait)
2429 {
2430 tid_t target;
2431
2432 if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
2433 if (wait)
2434 log_wait_commit(EXT3_SB(sb)->s_journal, target);
2435 }
2436 return 0;
2437 }
2438
2439 /*
2440 * LVM calls this function before a (read-only) snapshot is created. This
2441 * gives us a chance to flush the journal completely and mark the fs clean.
2442 */
2443 static int ext3_freeze(struct super_block *sb)
2444 {
2445 int error = 0;
2446 journal_t *journal;
2447
2448 if (!(sb->s_flags & MS_RDONLY)) {
2449 journal = EXT3_SB(sb)->s_journal;
2450
2451 /* Now we set up the journal barrier. */
2452 journal_lock_updates(journal);
2453
2454 /*
2455 * We don't want to clear needs_recovery flag when we failed
2456 * to flush the journal.
2457 */
2458 error = journal_flush(journal);
2459 if (error < 0)
2460 goto out;
2461
2462 /* Journal blocked and flushed, clear needs_recovery flag. */
2463 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2464 error = ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2465 if (error)
2466 goto out;
2467 }
2468 return 0;
2469
2470 out:
2471 journal_unlock_updates(journal);
2472 return error;
2473 }
2474
2475 /*
2476 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2477 * flag here, even though the filesystem is not technically dirty yet.
2478 */
2479 static int ext3_unfreeze(struct super_block *sb)
2480 {
2481 if (!(sb->s_flags & MS_RDONLY)) {
2482 lock_super(sb);
2483 /* Reser the needs_recovery flag before the fs is unlocked. */
2484 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2485 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2486 unlock_super(sb);
2487 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2488 }
2489 return 0;
2490 }
2491
2492 static int ext3_remount (struct super_block * sb, int * flags, char * data)
2493 {
2494 struct ext3_super_block * es;
2495 struct ext3_sb_info *sbi = EXT3_SB(sb);
2496 ext3_fsblk_t n_blocks_count = 0;
2497 unsigned long old_sb_flags;
2498 struct ext3_mount_options old_opts;
2499 int err;
2500 #ifdef CONFIG_QUOTA
2501 int i;
2502 #endif
2503
2504 lock_kernel();
2505
2506 /* Store the original options */
2507 lock_super(sb);
2508 old_sb_flags = sb->s_flags;
2509 old_opts.s_mount_opt = sbi->s_mount_opt;
2510 old_opts.s_resuid = sbi->s_resuid;
2511 old_opts.s_resgid = sbi->s_resgid;
2512 old_opts.s_commit_interval = sbi->s_commit_interval;
2513 #ifdef CONFIG_QUOTA
2514 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2515 for (i = 0; i < MAXQUOTAS; i++)
2516 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2517 #endif
2518
2519 /*
2520 * Allow the "check" option to be passed as a remount option.
2521 */
2522 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2523 err = -EINVAL;
2524 goto restore_opts;
2525 }
2526
2527 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT)
2528 ext3_abort(sb, __func__, "Abort forced by user");
2529
2530 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2531 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2532
2533 es = sbi->s_es;
2534
2535 ext3_init_journal_params(sb, sbi->s_journal);
2536
2537 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2538 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
2539 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT) {
2540 err = -EROFS;
2541 goto restore_opts;
2542 }
2543
2544 if (*flags & MS_RDONLY) {
2545 /*
2546 * First of all, the unconditional stuff we have to do
2547 * to disable replay of the journal when we next remount
2548 */
2549 sb->s_flags |= MS_RDONLY;
2550
2551 /*
2552 * OK, test if we are remounting a valid rw partition
2553 * readonly, and if so set the rdonly flag and then
2554 * mark the partition as valid again.
2555 */
2556 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2557 (sbi->s_mount_state & EXT3_VALID_FS))
2558 es->s_state = cpu_to_le16(sbi->s_mount_state);
2559
2560 /*
2561 * We have to unlock super so that we can wait for
2562 * transactions.
2563 */
2564 unlock_super(sb);
2565 ext3_mark_recovery_complete(sb, es);
2566 lock_super(sb);
2567 } else {
2568 __le32 ret;
2569 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2570 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
2571 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2572 "remount RDWR because of unsupported "
2573 "optional features (%x).\n",
2574 sb->s_id, le32_to_cpu(ret));
2575 err = -EROFS;
2576 goto restore_opts;
2577 }
2578
2579 /*
2580 * If we have an unprocessed orphan list hanging
2581 * around from a previously readonly bdev mount,
2582 * require a full umount/remount for now.
2583 */
2584 if (es->s_last_orphan) {
2585 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2586 "remount RDWR because of unprocessed "
2587 "orphan inode list. Please "
2588 "umount/remount instead.\n",
2589 sb->s_id);
2590 err = -EINVAL;
2591 goto restore_opts;
2592 }
2593
2594 /*
2595 * Mounting a RDONLY partition read-write, so reread
2596 * and store the current valid flag. (It may have
2597 * been changed by e2fsck since we originally mounted
2598 * the partition.)
2599 */
2600 ext3_clear_journal_err(sb, es);
2601 sbi->s_mount_state = le16_to_cpu(es->s_state);
2602 if ((err = ext3_group_extend(sb, es, n_blocks_count)))
2603 goto restore_opts;
2604 if (!ext3_setup_super (sb, es, 0))
2605 sb->s_flags &= ~MS_RDONLY;
2606 }
2607 }
2608 #ifdef CONFIG_QUOTA
2609 /* Release old quota file names */
2610 for (i = 0; i < MAXQUOTAS; i++)
2611 if (old_opts.s_qf_names[i] &&
2612 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2613 kfree(old_opts.s_qf_names[i]);
2614 #endif
2615 unlock_super(sb);
2616 unlock_kernel();
2617 return 0;
2618 restore_opts:
2619 sb->s_flags = old_sb_flags;
2620 sbi->s_mount_opt = old_opts.s_mount_opt;
2621 sbi->s_resuid = old_opts.s_resuid;
2622 sbi->s_resgid = old_opts.s_resgid;
2623 sbi->s_commit_interval = old_opts.s_commit_interval;
2624 #ifdef CONFIG_QUOTA
2625 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2626 for (i = 0; i < MAXQUOTAS; i++) {
2627 if (sbi->s_qf_names[i] &&
2628 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2629 kfree(sbi->s_qf_names[i]);
2630 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2631 }
2632 #endif
2633 unlock_super(sb);
2634 unlock_kernel();
2635 return err;
2636 }
2637
2638 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf)
2639 {
2640 struct super_block *sb = dentry->d_sb;
2641 struct ext3_sb_info *sbi = EXT3_SB(sb);
2642 struct ext3_super_block *es = sbi->s_es;
2643 u64 fsid;
2644
2645 if (test_opt(sb, MINIX_DF)) {
2646 sbi->s_overhead_last = 0;
2647 } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
2648 unsigned long ngroups = sbi->s_groups_count, i;
2649 ext3_fsblk_t overhead = 0;
2650 smp_rmb();
2651
2652 /*
2653 * Compute the overhead (FS structures). This is constant
2654 * for a given filesystem unless the number of block groups
2655 * changes so we cache the previous value until it does.
2656 */
2657
2658 /*
2659 * All of the blocks before first_data_block are
2660 * overhead
2661 */
2662 overhead = le32_to_cpu(es->s_first_data_block);
2663
2664 /*
2665 * Add the overhead attributed to the superblock and
2666 * block group descriptors. If the sparse superblocks
2667 * feature is turned on, then not all groups have this.
2668 */
2669 for (i = 0; i < ngroups; i++) {
2670 overhead += ext3_bg_has_super(sb, i) +
2671 ext3_bg_num_gdb(sb, i);
2672 cond_resched();
2673 }
2674
2675 /*
2676 * Every block group has an inode bitmap, a block
2677 * bitmap, and an inode table.
2678 */
2679 overhead += ngroups * (2 + sbi->s_itb_per_group);
2680 sbi->s_overhead_last = overhead;
2681 smp_wmb();
2682 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
2683 }
2684
2685 buf->f_type = EXT3_SUPER_MAGIC;
2686 buf->f_bsize = sb->s_blocksize;
2687 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
2688 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
2689 es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
2690 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2691 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2692 buf->f_bavail = 0;
2693 buf->f_files = le32_to_cpu(es->s_inodes_count);
2694 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
2695 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
2696 buf->f_namelen = EXT3_NAME_LEN;
2697 fsid = le64_to_cpup((void *)es->s_uuid) ^
2698 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2699 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2700 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
2701 return 0;
2702 }
2703
2704 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2705 * is locked for write. Otherwise the are possible deadlocks:
2706 * Process 1 Process 2
2707 * ext3_create() quota_sync()
2708 * journal_start() write_dquot()
2709 * vfs_dq_init() down(dqio_mutex)
2710 * down(dqio_mutex) journal_start()
2711 *
2712 */
2713
2714 #ifdef CONFIG_QUOTA
2715
2716 static inline struct inode *dquot_to_inode(struct dquot *dquot)
2717 {
2718 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2719 }
2720
2721 static int ext3_write_dquot(struct dquot *dquot)
2722 {
2723 int ret, err;
2724 handle_t *handle;
2725 struct inode *inode;
2726
2727 inode = dquot_to_inode(dquot);
2728 handle = ext3_journal_start(inode,
2729 EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
2730 if (IS_ERR(handle))
2731 return PTR_ERR(handle);
2732 ret = dquot_commit(dquot);
2733 err = ext3_journal_stop(handle);
2734 if (!ret)
2735 ret = err;
2736 return ret;
2737 }
2738
2739 static int ext3_acquire_dquot(struct dquot *dquot)
2740 {
2741 int ret, err;
2742 handle_t *handle;
2743
2744 handle = ext3_journal_start(dquot_to_inode(dquot),
2745 EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
2746 if (IS_ERR(handle))
2747 return PTR_ERR(handle);
2748 ret = dquot_acquire(dquot);
2749 err = ext3_journal_stop(handle);
2750 if (!ret)
2751 ret = err;
2752 return ret;
2753 }
2754
2755 static int ext3_release_dquot(struct dquot *dquot)
2756 {
2757 int ret, err;
2758 handle_t *handle;
2759
2760 handle = ext3_journal_start(dquot_to_inode(dquot),
2761 EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
2762 if (IS_ERR(handle)) {
2763 /* Release dquot anyway to avoid endless cycle in dqput() */
2764 dquot_release(dquot);
2765 return PTR_ERR(handle);
2766 }
2767 ret = dquot_release(dquot);
2768 err = ext3_journal_stop(handle);
2769 if (!ret)
2770 ret = err;
2771 return ret;
2772 }
2773
2774 static int ext3_mark_dquot_dirty(struct dquot *dquot)
2775 {
2776 /* Are we journaling quotas? */
2777 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2778 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2779 dquot_mark_dquot_dirty(dquot);
2780 return ext3_write_dquot(dquot);
2781 } else {
2782 return dquot_mark_dquot_dirty(dquot);
2783 }
2784 }
2785
2786 static int ext3_write_info(struct super_block *sb, int type)
2787 {
2788 int ret, err;
2789 handle_t *handle;
2790
2791 /* Data block + inode block */
2792 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2793 if (IS_ERR(handle))
2794 return PTR_ERR(handle);
2795 ret = dquot_commit_info(sb, type);
2796 err = ext3_journal_stop(handle);
2797 if (!ret)
2798 ret = err;
2799 return ret;
2800 }
2801
2802 /*
2803 * Turn on quotas during mount time - we need to find
2804 * the quota file and such...
2805 */
2806 static int ext3_quota_on_mount(struct super_block *sb, int type)
2807 {
2808 return vfs_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2809 EXT3_SB(sb)->s_jquota_fmt, type);
2810 }
2811
2812 /*
2813 * Standard function to be called on quota_on
2814 */
2815 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
2816 char *name, int remount)
2817 {
2818 int err;
2819 struct path path;
2820
2821 if (!test_opt(sb, QUOTA))
2822 return -EINVAL;
2823 /* When remounting, no checks are needed and in fact, name is NULL */
2824 if (remount)
2825 return vfs_quota_on(sb, type, format_id, name, remount);
2826
2827 err = kern_path(name, LOOKUP_FOLLOW, &path);
2828 if (err)
2829 return err;
2830
2831 /* Quotafile not on the same filesystem? */
2832 if (path.mnt->mnt_sb != sb) {
2833 path_put(&path);
2834 return -EXDEV;
2835 }
2836 /* Journaling quota? */
2837 if (EXT3_SB(sb)->s_qf_names[type]) {
2838 /* Quotafile not of fs root? */
2839 if (path.dentry->d_parent != sb->s_root)
2840 printk(KERN_WARNING
2841 "EXT3-fs: Quota file not on filesystem root. "
2842 "Journaled quota will not work.\n");
2843 }
2844
2845 /*
2846 * When we journal data on quota file, we have to flush journal to see
2847 * all updates to the file when we bypass pagecache...
2848 */
2849 if (ext3_should_journal_data(path.dentry->d_inode)) {
2850 /*
2851 * We don't need to lock updates but journal_flush() could
2852 * otherwise be livelocked...
2853 */
2854 journal_lock_updates(EXT3_SB(sb)->s_journal);
2855 err = journal_flush(EXT3_SB(sb)->s_journal);
2856 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2857 if (err) {
2858 path_put(&path);
2859 return err;
2860 }
2861 }
2862
2863 err = vfs_quota_on_path(sb, type, format_id, &path);
2864 path_put(&path);
2865 return err;
2866 }
2867
2868 /* Read data from quotafile - avoid pagecache and such because we cannot afford
2869 * acquiring the locks... As quota files are never truncated and quota code
2870 * itself serializes the operations (and noone else should touch the files)
2871 * we don't have to be afraid of races */
2872 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
2873 size_t len, loff_t off)
2874 {
2875 struct inode *inode = sb_dqopt(sb)->files[type];
2876 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2877 int err = 0;
2878 int offset = off & (sb->s_blocksize - 1);
2879 int tocopy;
2880 size_t toread;
2881 struct buffer_head *bh;
2882 loff_t i_size = i_size_read(inode);
2883
2884 if (off > i_size)
2885 return 0;
2886 if (off+len > i_size)
2887 len = i_size-off;
2888 toread = len;
2889 while (toread > 0) {
2890 tocopy = sb->s_blocksize - offset < toread ?
2891 sb->s_blocksize - offset : toread;
2892 bh = ext3_bread(NULL, inode, blk, 0, &err);
2893 if (err)
2894 return err;
2895 if (!bh) /* A hole? */
2896 memset(data, 0, tocopy);
2897 else
2898 memcpy(data, bh->b_data+offset, tocopy);
2899 brelse(bh);
2900 offset = 0;
2901 toread -= tocopy;
2902 data += tocopy;
2903 blk++;
2904 }
2905 return len;
2906 }
2907
2908 /* Write to quotafile (we know the transaction is already started and has
2909 * enough credits) */
2910 static ssize_t ext3_quota_write(struct super_block *sb, int type,
2911 const char *data, size_t len, loff_t off)
2912 {
2913 struct inode *inode = sb_dqopt(sb)->files[type];
2914 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2915 int err = 0;
2916 int offset = off & (sb->s_blocksize - 1);
2917 int tocopy;
2918 int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
2919 size_t towrite = len;
2920 struct buffer_head *bh;
2921 handle_t *handle = journal_current_handle();
2922
2923 if (!handle) {
2924 printk(KERN_WARNING "EXT3-fs: Quota write (off=%Lu, len=%Lu)"
2925 " cancelled because transaction is not started.\n",
2926 (unsigned long long)off, (unsigned long long)len);
2927 return -EIO;
2928 }
2929 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
2930 while (towrite > 0) {
2931 tocopy = sb->s_blocksize - offset < towrite ?
2932 sb->s_blocksize - offset : towrite;
2933 bh = ext3_bread(handle, inode, blk, 1, &err);
2934 if (!bh)
2935 goto out;
2936 if (journal_quota) {
2937 err = ext3_journal_get_write_access(handle, bh);
2938 if (err) {
2939 brelse(bh);
2940 goto out;
2941 }
2942 }
2943 lock_buffer(bh);
2944 memcpy(bh->b_data+offset, data, tocopy);
2945 flush_dcache_page(bh->b_page);
2946 unlock_buffer(bh);
2947 if (journal_quota)
2948 err = ext3_journal_dirty_metadata(handle, bh);
2949 else {
2950 /* Always do at least ordered writes for quotas */
2951 err = ext3_journal_dirty_data(handle, bh);
2952 mark_buffer_dirty(bh);
2953 }
2954 brelse(bh);
2955 if (err)
2956 goto out;
2957 offset = 0;
2958 towrite -= tocopy;
2959 data += tocopy;
2960 blk++;
2961 }
2962 out:
2963 if (len == towrite) {
2964 mutex_unlock(&inode->i_mutex);
2965 return err;
2966 }
2967 if (inode->i_size < off+len-towrite) {
2968 i_size_write(inode, off+len-towrite);
2969 EXT3_I(inode)->i_disksize = inode->i_size;
2970 }
2971 inode->i_version++;
2972 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2973 ext3_mark_inode_dirty(handle, inode);
2974 mutex_unlock(&inode->i_mutex);
2975 return len - towrite;
2976 }
2977
2978 #endif
2979
2980 static int ext3_get_sb(struct file_system_type *fs_type,
2981 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
2982 {
2983 return get_sb_bdev(fs_type, flags, dev_name, data, ext3_fill_super, mnt);
2984 }
2985
2986 static struct file_system_type ext3_fs_type = {
2987 .owner = THIS_MODULE,
2988 .name = "ext3",
2989 .get_sb = ext3_get_sb,
2990 .kill_sb = kill_block_super,
2991 .fs_flags = FS_REQUIRES_DEV,
2992 };
2993
2994 static int __init init_ext3_fs(void)
2995 {
2996 int err = init_ext3_xattr();
2997 if (err)
2998 return err;
2999 err = init_inodecache();
3000 if (err)
3001 goto out1;
3002 err = register_filesystem(&ext3_fs_type);
3003 if (err)
3004 goto out;
3005 return 0;
3006 out:
3007 destroy_inodecache();
3008 out1:
3009 exit_ext3_xattr();
3010 return err;
3011 }
3012
3013 static void __exit exit_ext3_fs(void)
3014 {
3015 unregister_filesystem(&ext3_fs_type);
3016 destroy_inodecache();
3017 exit_ext3_xattr();
3018 }
3019
3020 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3021 MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
3022 MODULE_LICENSE("GPL");
3023 module_init(init_ext3_fs)
3024 module_exit(exit_ext3_fs)
Linux with added FPC-III support