?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
um: fix IRQ flag handling naming
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / fs / ext3 / super.c
blob37776800910670e2a675e4fd11378f3d8eb35446
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 "EXT3-fs: %s: aborting transaction: %s in %s\n",
139 caller, errstr, err_fn);
140
141 journal_abort_handle(handle);
142 }
143
144 void ext3_msg(struct super_block *sb, const char *prefix,
145 const char *fmt, ...)
146 {
147 va_list args;
148
149 va_start(args, fmt);
150 printk("%sEXT3-fs (%s): ", prefix, sb->s_id);
151 vprintk(fmt, args);
152 printk("\n");
153 va_end(args);
154 }
155
156 /* Deal with the reporting of failure conditions on a filesystem such as
157 * inconsistencies detected or read IO failures.
158 *
159 * On ext2, we can store the error state of the filesystem in the
160 * superblock. That is not possible on ext3, because we may have other
161 * write ordering constraints on the superblock which prevent us from
162 * writing it out straight away; and given that the journal is about to
163 * be aborted, we can't rely on the current, or future, transactions to
164 * write out the superblock safely.
165 *
166 * We'll just use the journal_abort() error code to record an error in
167 * the journal instead. On recovery, the journal will complain about
168 * that error until we've noted it down and cleared it.
169 */
170
171 static void ext3_handle_error(struct super_block *sb)
172 {
173 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
174
175 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
176 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
177
178 if (sb->s_flags & MS_RDONLY)
179 return;
180
181 if (!test_opt (sb, ERRORS_CONT)) {
182 journal_t *journal = EXT3_SB(sb)->s_journal;
183
184 set_opt(EXT3_SB(sb)->s_mount_opt, ABORT);
185 if (journal)
186 journal_abort(journal, -EIO);
187 }
188 if (test_opt (sb, ERRORS_RO)) {
189 ext3_msg(sb, KERN_CRIT,
190 "error: remounting filesystem read-only");
191 sb->s_flags |= MS_RDONLY;
192 }
193 ext3_commit_super(sb, es, 1);
194 if (test_opt(sb, ERRORS_PANIC))
195 panic("EXT3-fs (%s): panic forced after error\n",
196 sb->s_id);
197 }
198
199 void ext3_error (struct super_block * sb, const char * function,
200 const char * fmt, ...)
201 {
202 va_list args;
203
204 va_start(args, fmt);
205 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
206 vprintk(fmt, args);
207 printk("\n");
208 va_end(args);
209
210 ext3_handle_error(sb);
211 }
212
213 static const char *ext3_decode_error(struct super_block * sb, int errno,
214 char nbuf[16])
215 {
216 char *errstr = NULL;
217
218 switch (errno) {
219 case -EIO:
220 errstr = "IO failure";
221 break;
222 case -ENOMEM:
223 errstr = "Out of memory";
224 break;
225 case -EROFS:
226 if (!sb || EXT3_SB(sb)->s_journal->j_flags & JFS_ABORT)
227 errstr = "Journal has aborted";
228 else
229 errstr = "Readonly filesystem";
230 break;
231 default:
232 /* If the caller passed in an extra buffer for unknown
233 * errors, textualise them now. Else we just return
234 * NULL. */
235 if (nbuf) {
236 /* Check for truncated error codes... */
237 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
238 errstr = nbuf;
239 }
240 break;
241 }
242
243 return errstr;
244 }
245
246 /* __ext3_std_error decodes expected errors from journaling functions
247 * automatically and invokes the appropriate error response. */
248
249 void __ext3_std_error (struct super_block * sb, const char * function,
250 int errno)
251 {
252 char nbuf[16];
253 const char *errstr;
254
255 /* Special case: if the error is EROFS, and we're not already
256 * inside a transaction, then there's really no point in logging
257 * an error. */
258 if (errno == -EROFS && journal_current_handle() == NULL &&
259 (sb->s_flags & MS_RDONLY))
260 return;
261
262 errstr = ext3_decode_error(sb, errno, nbuf);
263 ext3_msg(sb, KERN_CRIT, "error in %s: %s", function, errstr);
264
265 ext3_handle_error(sb);
266 }
267
268 /*
269 * ext3_abort is a much stronger failure handler than ext3_error. The
270 * abort function may be used to deal with unrecoverable failures such
271 * as journal IO errors or ENOMEM at a critical moment in log management.
272 *
273 * We unconditionally force the filesystem into an ABORT|READONLY state,
274 * unless the error response on the fs has been set to panic in which
275 * case we take the easy way out and panic immediately.
276 */
277
278 void ext3_abort (struct super_block * sb, const char * function,
279 const char * fmt, ...)
280 {
281 va_list args;
282
283 va_start(args, fmt);
284 printk(KERN_CRIT "EXT3-fs (%s): error: %s: ", sb->s_id, function);
285 vprintk(fmt, args);
286 printk("\n");
287 va_end(args);
288
289 if (test_opt(sb, ERRORS_PANIC))
290 panic("EXT3-fs: panic from previous error\n");
291
292 if (sb->s_flags & MS_RDONLY)
293 return;
294
295 ext3_msg(sb, KERN_CRIT,
296 "error: remounting filesystem read-only");
297 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
298 sb->s_flags |= MS_RDONLY;
299 set_opt(EXT3_SB(sb)->s_mount_opt, ABORT);
300 if (EXT3_SB(sb)->s_journal)
301 journal_abort(EXT3_SB(sb)->s_journal, -EIO);
302 }
303
304 void ext3_warning (struct super_block * sb, const char * function,
305 const char * fmt, ...)
306 {
307 va_list args;
308
309 va_start(args, fmt);
310 printk(KERN_WARNING "EXT3-fs (%s): warning: %s: ",
311 sb->s_id, function);
312 vprintk(fmt, args);
313 printk("\n");
314 va_end(args);
315 }
316
317 void ext3_update_dynamic_rev(struct super_block *sb)
318 {
319 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
320
321 if (le32_to_cpu(es->s_rev_level) > EXT3_GOOD_OLD_REV)
322 return;
323
324 ext3_msg(sb, KERN_WARNING,
325 "warning: updating to rev %d because of "
326 "new feature flag, running e2fsck is recommended",
327 EXT3_DYNAMIC_REV);
328
329 es->s_first_ino = cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO);
330 es->s_inode_size = cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE);
331 es->s_rev_level = cpu_to_le32(EXT3_DYNAMIC_REV);
332 /* leave es->s_feature_*compat flags alone */
333 /* es->s_uuid will be set by e2fsck if empty */
334
335 /*
336 * The rest of the superblock fields should be zero, and if not it
337 * means they are likely already in use, so leave them alone. We
338 * can leave it up to e2fsck to clean up any inconsistencies there.
339 */
340 }
341
342 /*
343 * Open the external journal device
344 */
345 static struct block_device *ext3_blkdev_get(dev_t dev, struct super_block *sb)
346 {
347 struct block_device *bdev;
348 char b[BDEVNAME_SIZE];
349
350 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
351 if (IS_ERR(bdev))
352 goto fail;
353 return bdev;
354
355 fail:
356 ext3_msg(sb, "error: failed to open journal device %s: %ld",
357 __bdevname(dev, b), PTR_ERR(bdev));
358
359 return NULL;
360 }
361
362 /*
363 * Release the journal device
364 */
365 static int ext3_blkdev_put(struct block_device *bdev)
366 {
367 bd_release(bdev);
368 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
369 }
370
371 static int ext3_blkdev_remove(struct ext3_sb_info *sbi)
372 {
373 struct block_device *bdev;
374 int ret = -ENODEV;
375
376 bdev = sbi->journal_bdev;
377 if (bdev) {
378 ret = ext3_blkdev_put(bdev);
379 sbi->journal_bdev = NULL;
380 }
381 return ret;
382 }
383
384 static inline struct inode *orphan_list_entry(struct list_head *l)
385 {
386 return &list_entry(l, struct ext3_inode_info, i_orphan)->vfs_inode;
387 }
388
389 static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi)
390 {
391 struct list_head *l;
392
393 ext3_msg(sb, KERN_ERR, "error: sb orphan head is %d",
394 le32_to_cpu(sbi->s_es->s_last_orphan));
395
396 ext3_msg(sb, KERN_ERR, "sb_info orphan list:");
397 list_for_each(l, &sbi->s_orphan) {
398 struct inode *inode = orphan_list_entry(l);
399 ext3_msg(sb, KERN_ERR, " "
400 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
401 inode->i_sb->s_id, inode->i_ino, inode,
402 inode->i_mode, inode->i_nlink,
403 NEXT_ORPHAN(inode));
404 }
405 }
406
407 static void ext3_put_super (struct super_block * sb)
408 {
409 struct ext3_sb_info *sbi = EXT3_SB(sb);
410 struct ext3_super_block *es = sbi->s_es;
411 int i, err;
412
413 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
414 ext3_xattr_put_super(sb);
415 err = journal_destroy(sbi->s_journal);
416 sbi->s_journal = NULL;
417 if (err < 0)
418 ext3_abort(sb, __func__, "Couldn't clean up the journal");
419
420 if (!(sb->s_flags & MS_RDONLY)) {
421 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
422 es->s_state = cpu_to_le16(sbi->s_mount_state);
423 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
424 mark_buffer_dirty(sbi->s_sbh);
425 ext3_commit_super(sb, es, 1);
426 }
427
428 for (i = 0; i < sbi->s_gdb_count; i++)
429 brelse(sbi->s_group_desc[i]);
430 kfree(sbi->s_group_desc);
431 percpu_counter_destroy(&sbi->s_freeblocks_counter);
432 percpu_counter_destroy(&sbi->s_freeinodes_counter);
433 percpu_counter_destroy(&sbi->s_dirs_counter);
434 brelse(sbi->s_sbh);
435 #ifdef CONFIG_QUOTA
436 for (i = 0; i < MAXQUOTAS; i++)
437 kfree(sbi->s_qf_names[i]);
438 #endif
439
440 /* Debugging code just in case the in-memory inode orphan list
441 * isn't empty. The on-disk one can be non-empty if we've
442 * detected an error and taken the fs readonly, but the
443 * in-memory list had better be clean by this point. */
444 if (!list_empty(&sbi->s_orphan))
445 dump_orphan_list(sb, sbi);
446 J_ASSERT(list_empty(&sbi->s_orphan));
447
448 invalidate_bdev(sb->s_bdev);
449 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
450 /*
451 * Invalidate the journal device's buffers. We don't want them
452 * floating about in memory - the physical journal device may
453 * hotswapped, and it breaks the `ro-after' testing code.
454 */
455 sync_blockdev(sbi->journal_bdev);
456 invalidate_bdev(sbi->journal_bdev);
457 ext3_blkdev_remove(sbi);
458 }
459 sb->s_fs_info = NULL;
460 kfree(sbi->s_blockgroup_lock);
461 kfree(sbi);
462 }
463
464 static struct kmem_cache *ext3_inode_cachep;
465
466 /*
467 * Called inside transaction, so use GFP_NOFS
468 */
469 static struct inode *ext3_alloc_inode(struct super_block *sb)
470 {
471 struct ext3_inode_info *ei;
472
473 ei = kmem_cache_alloc(ext3_inode_cachep, GFP_NOFS);
474 if (!ei)
475 return NULL;
476 ei->i_block_alloc_info = NULL;
477 ei->vfs_inode.i_version = 1;
478 atomic_set(&ei->i_datasync_tid, 0);
479 atomic_set(&ei->i_sync_tid, 0);
480 return &ei->vfs_inode;
481 }
482
483 static void ext3_destroy_inode(struct inode *inode)
484 {
485 if (!list_empty(&(EXT3_I(inode)->i_orphan))) {
486 printk("EXT3 Inode %p: orphan list check failed!\n",
487 EXT3_I(inode));
488 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
489 EXT3_I(inode), sizeof(struct ext3_inode_info),
490 false);
491 dump_stack();
492 }
493 kmem_cache_free(ext3_inode_cachep, EXT3_I(inode));
494 }
495
496 static void init_once(void *foo)
497 {
498 struct ext3_inode_info *ei = (struct ext3_inode_info *) foo;
499
500 INIT_LIST_HEAD(&ei->i_orphan);
501 #ifdef CONFIG_EXT3_FS_XATTR
502 init_rwsem(&ei->xattr_sem);
503 #endif
504 mutex_init(&ei->truncate_mutex);
505 inode_init_once(&ei->vfs_inode);
506 }
507
508 static int init_inodecache(void)
509 {
510 ext3_inode_cachep = kmem_cache_create("ext3_inode_cache",
511 sizeof(struct ext3_inode_info),
512 0, (SLAB_RECLAIM_ACCOUNT|
513 SLAB_MEM_SPREAD),
514 init_once);
515 if (ext3_inode_cachep == NULL)
516 return -ENOMEM;
517 return 0;
518 }
519
520 static void destroy_inodecache(void)
521 {
522 kmem_cache_destroy(ext3_inode_cachep);
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 char *fmtname = "";
532
533 switch (sbi->s_jquota_fmt) {
534 case QFMT_VFS_OLD:
535 fmtname = "vfsold";
536 break;
537 case QFMT_VFS_V0:
538 fmtname = "vfsv0";
539 break;
540 case QFMT_VFS_V1:
541 fmtname = "vfsv1";
542 break;
543 }
544 seq_printf(seq, ",jqfmt=%s", fmtname);
545 }
546
547 if (sbi->s_qf_names[USRQUOTA])
548 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
549
550 if (sbi->s_qf_names[GRPQUOTA])
551 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
552
553 if (test_opt(sb, USRQUOTA))
554 seq_puts(seq, ",usrquota");
555
556 if (test_opt(sb, GRPQUOTA))
557 seq_puts(seq, ",grpquota");
558 #endif
559 }
560
561 static char *data_mode_string(unsigned long mode)
562 {
563 switch (mode) {
564 case EXT3_MOUNT_JOURNAL_DATA:
565 return "journal";
566 case EXT3_MOUNT_ORDERED_DATA:
567 return "ordered";
568 case EXT3_MOUNT_WRITEBACK_DATA:
569 return "writeback";
570 }
571 return "unknown";
572 }
573
574 /*
575 * Show an option if
576 * - it's set to a non-default value OR
577 * - if the per-sb default is different from the global default
578 */
579 static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs)
580 {
581 struct super_block *sb = vfs->mnt_sb;
582 struct ext3_sb_info *sbi = EXT3_SB(sb);
583 struct ext3_super_block *es = sbi->s_es;
584 unsigned long def_mount_opts;
585
586 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
587
588 if (sbi->s_sb_block != 1)
589 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
590 if (test_opt(sb, MINIX_DF))
591 seq_puts(seq, ",minixdf");
592 if (test_opt(sb, GRPID))
593 seq_puts(seq, ",grpid");
594 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT3_DEFM_BSDGROUPS))
595 seq_puts(seq, ",nogrpid");
596 if (sbi->s_resuid != EXT3_DEF_RESUID ||
597 le16_to_cpu(es->s_def_resuid) != EXT3_DEF_RESUID) {
598 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
599 }
600 if (sbi->s_resgid != EXT3_DEF_RESGID ||
601 le16_to_cpu(es->s_def_resgid) != EXT3_DEF_RESGID) {
602 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
603 }
604 if (test_opt(sb, ERRORS_RO)) {
605 int def_errors = le16_to_cpu(es->s_errors);
606
607 if (def_errors == EXT3_ERRORS_PANIC ||
608 def_errors == EXT3_ERRORS_CONTINUE) {
609 seq_puts(seq, ",errors=remount-ro");
610 }
611 }
612 if (test_opt(sb, ERRORS_CONT))
613 seq_puts(seq, ",errors=continue");
614 if (test_opt(sb, ERRORS_PANIC))
615 seq_puts(seq, ",errors=panic");
616 if (test_opt(sb, NO_UID32))
617 seq_puts(seq, ",nouid32");
618 if (test_opt(sb, DEBUG))
619 seq_puts(seq, ",debug");
620 if (test_opt(sb, OLDALLOC))
621 seq_puts(seq, ",oldalloc");
622 #ifdef CONFIG_EXT3_FS_XATTR
623 if (test_opt(sb, XATTR_USER))
624 seq_puts(seq, ",user_xattr");
625 if (!test_opt(sb, XATTR_USER) &&
626 (def_mount_opts & EXT3_DEFM_XATTR_USER)) {
627 seq_puts(seq, ",nouser_xattr");
628 }
629 #endif
630 #ifdef CONFIG_EXT3_FS_POSIX_ACL
631 if (test_opt(sb, POSIX_ACL))
632 seq_puts(seq, ",acl");
633 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT3_DEFM_ACL))
634 seq_puts(seq, ",noacl");
635 #endif
636 if (!test_opt(sb, RESERVATION))
637 seq_puts(seq, ",noreservation");
638 if (sbi->s_commit_interval) {
639 seq_printf(seq, ",commit=%u",
640 (unsigned) (sbi->s_commit_interval / HZ));
641 }
642
643 /*
644 * Always display barrier state so it's clear what the status is.
645 */
646 seq_puts(seq, ",barrier=");
647 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
648 seq_printf(seq, ",data=%s", data_mode_string(test_opt(sb, DATA_FLAGS)));
649 if (test_opt(sb, DATA_ERR_ABORT))
650 seq_puts(seq, ",data_err=abort");
651
652 if (test_opt(sb, NOLOAD))
653 seq_puts(seq, ",norecovery");
654
655 ext3_show_quota_options(seq, sb);
656
657 return 0;
658 }
659
660
661 static struct inode *ext3_nfs_get_inode(struct super_block *sb,
662 u64 ino, u32 generation)
663 {
664 struct inode *inode;
665
666 if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO)
667 return ERR_PTR(-ESTALE);
668 if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count))
669 return ERR_PTR(-ESTALE);
670
671 /* iget isn't really right if the inode is currently unallocated!!
672 *
673 * ext3_read_inode will return a bad_inode if the inode had been
674 * deleted, so we should be safe.
675 *
676 * Currently we don't know the generation for parent directory, so
677 * a generation of 0 means "accept any"
678 */
679 inode = ext3_iget(sb, ino);
680 if (IS_ERR(inode))
681 return ERR_CAST(inode);
682 if (generation && inode->i_generation != generation) {
683 iput(inode);
684 return ERR_PTR(-ESTALE);
685 }
686
687 return inode;
688 }
689
690 static struct dentry *ext3_fh_to_dentry(struct super_block *sb, struct fid *fid,
691 int fh_len, int fh_type)
692 {
693 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
694 ext3_nfs_get_inode);
695 }
696
697 static struct dentry *ext3_fh_to_parent(struct super_block *sb, struct fid *fid,
698 int fh_len, int fh_type)
699 {
700 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
701 ext3_nfs_get_inode);
702 }
703
704 /*
705 * Try to release metadata pages (indirect blocks, directories) which are
706 * mapped via the block device. Since these pages could have journal heads
707 * which would prevent try_to_free_buffers() from freeing them, we must use
708 * jbd layer's try_to_free_buffers() function to release them.
709 */
710 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
711 gfp_t wait)
712 {
713 journal_t *journal = EXT3_SB(sb)->s_journal;
714
715 WARN_ON(PageChecked(page));
716 if (!page_has_buffers(page))
717 return 0;
718 if (journal)
719 return journal_try_to_free_buffers(journal, page,
720 wait & ~__GFP_WAIT);
721 return try_to_free_buffers(page);
722 }
723
724 #ifdef CONFIG_QUOTA
725 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
726 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
727
728 static int ext3_write_dquot(struct dquot *dquot);
729 static int ext3_acquire_dquot(struct dquot *dquot);
730 static int ext3_release_dquot(struct dquot *dquot);
731 static int ext3_mark_dquot_dirty(struct dquot *dquot);
732 static int ext3_write_info(struct super_block *sb, int type);
733 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
734 char *path);
735 static int ext3_quota_on_mount(struct super_block *sb, int type);
736 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
737 size_t len, loff_t off);
738 static ssize_t ext3_quota_write(struct super_block *sb, int type,
739 const char *data, size_t len, loff_t off);
740
741 static const struct dquot_operations ext3_quota_operations = {
742 .write_dquot = ext3_write_dquot,
743 .acquire_dquot = ext3_acquire_dquot,
744 .release_dquot = ext3_release_dquot,
745 .mark_dirty = ext3_mark_dquot_dirty,
746 .write_info = ext3_write_info,
747 .alloc_dquot = dquot_alloc,
748 .destroy_dquot = dquot_destroy,
749 };
750
751 static const struct quotactl_ops ext3_qctl_operations = {
752 .quota_on = ext3_quota_on,
753 .quota_off = dquot_quota_off,
754 .quota_sync = dquot_quota_sync,
755 .get_info = dquot_get_dqinfo,
756 .set_info = dquot_set_dqinfo,
757 .get_dqblk = dquot_get_dqblk,
758 .set_dqblk = dquot_set_dqblk
759 };
760 #endif
761
762 static const struct super_operations ext3_sops = {
763 .alloc_inode = ext3_alloc_inode,
764 .destroy_inode = ext3_destroy_inode,
765 .write_inode = ext3_write_inode,
766 .dirty_inode = ext3_dirty_inode,
767 .evict_inode = ext3_evict_inode,
768 .put_super = ext3_put_super,
769 .sync_fs = ext3_sync_fs,
770 .freeze_fs = ext3_freeze,
771 .unfreeze_fs = ext3_unfreeze,
772 .statfs = ext3_statfs,
773 .remount_fs = ext3_remount,
774 .show_options = ext3_show_options,
775 #ifdef CONFIG_QUOTA
776 .quota_read = ext3_quota_read,
777 .quota_write = ext3_quota_write,
778 #endif
779 .bdev_try_to_free_page = bdev_try_to_free_page,
780 };
781
782 static const struct export_operations ext3_export_ops = {
783 .fh_to_dentry = ext3_fh_to_dentry,
784 .fh_to_parent = ext3_fh_to_parent,
785 .get_parent = ext3_get_parent,
786 };
787
788 enum {
789 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
790 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
791 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
792 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
793 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
794 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
795 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
796 Opt_data_err_abort, Opt_data_err_ignore,
797 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
798 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
799 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
800 Opt_resize, Opt_usrquota, Opt_grpquota
801 };
802
803 static const match_table_t tokens = {
804 {Opt_bsd_df, "bsddf"},
805 {Opt_minix_df, "minixdf"},
806 {Opt_grpid, "grpid"},
807 {Opt_grpid, "bsdgroups"},
808 {Opt_nogrpid, "nogrpid"},
809 {Opt_nogrpid, "sysvgroups"},
810 {Opt_resgid, "resgid=%u"},
811 {Opt_resuid, "resuid=%u"},
812 {Opt_sb, "sb=%u"},
813 {Opt_err_cont, "errors=continue"},
814 {Opt_err_panic, "errors=panic"},
815 {Opt_err_ro, "errors=remount-ro"},
816 {Opt_nouid32, "nouid32"},
817 {Opt_nocheck, "nocheck"},
818 {Opt_nocheck, "check=none"},
819 {Opt_debug, "debug"},
820 {Opt_oldalloc, "oldalloc"},
821 {Opt_orlov, "orlov"},
822 {Opt_user_xattr, "user_xattr"},
823 {Opt_nouser_xattr, "nouser_xattr"},
824 {Opt_acl, "acl"},
825 {Opt_noacl, "noacl"},
826 {Opt_reservation, "reservation"},
827 {Opt_noreservation, "noreservation"},
828 {Opt_noload, "noload"},
829 {Opt_noload, "norecovery"},
830 {Opt_nobh, "nobh"},
831 {Opt_bh, "bh"},
832 {Opt_commit, "commit=%u"},
833 {Opt_journal_update, "journal=update"},
834 {Opt_journal_inum, "journal=%u"},
835 {Opt_journal_dev, "journal_dev=%u"},
836 {Opt_abort, "abort"},
837 {Opt_data_journal, "data=journal"},
838 {Opt_data_ordered, "data=ordered"},
839 {Opt_data_writeback, "data=writeback"},
840 {Opt_data_err_abort, "data_err=abort"},
841 {Opt_data_err_ignore, "data_err=ignore"},
842 {Opt_offusrjquota, "usrjquota="},
843 {Opt_usrjquota, "usrjquota=%s"},
844 {Opt_offgrpjquota, "grpjquota="},
845 {Opt_grpjquota, "grpjquota=%s"},
846 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
847 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
848 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
849 {Opt_grpquota, "grpquota"},
850 {Opt_noquota, "noquota"},
851 {Opt_quota, "quota"},
852 {Opt_usrquota, "usrquota"},
853 {Opt_barrier, "barrier=%u"},
854 {Opt_barrier, "barrier"},
855 {Opt_nobarrier, "nobarrier"},
856 {Opt_resize, "resize"},
857 {Opt_err, NULL},
858 };
859
860 static ext3_fsblk_t get_sb_block(void **data, struct super_block *sb)
861 {
862 ext3_fsblk_t sb_block;
863 char *options = (char *) *data;
864
865 if (!options || strncmp(options, "sb=", 3) != 0)
866 return 1; /* Default location */
867 options += 3;
868 /*todo: use simple_strtoll with >32bit ext3 */
869 sb_block = simple_strtoul(options, &options, 0);
870 if (*options && *options != ',') {
871 ext3_msg(sb, "error: invalid sb specification: %s",
872 (char *) *data);
873 return 1;
874 }
875 if (*options == ',')
876 options++;
877 *data = (void *) options;
878 return sb_block;
879 }
880
881 #ifdef CONFIG_QUOTA
882 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
883 {
884 struct ext3_sb_info *sbi = EXT3_SB(sb);
885 char *qname;
886
887 if (sb_any_quota_loaded(sb) &&
888 !sbi->s_qf_names[qtype]) {
889 ext3_msg(sb, KERN_ERR,
890 "Cannot change journaled "
891 "quota options when quota turned on");
892 return 0;
893 }
894 qname = match_strdup(args);
895 if (!qname) {
896 ext3_msg(sb, KERN_ERR,
897 "Not enough memory for storing quotafile name");
898 return 0;
899 }
900 if (sbi->s_qf_names[qtype] &&
901 strcmp(sbi->s_qf_names[qtype], qname)) {
902 ext3_msg(sb, KERN_ERR,
903 "%s quota file already specified", QTYPE2NAME(qtype));
904 kfree(qname);
905 return 0;
906 }
907 sbi->s_qf_names[qtype] = qname;
908 if (strchr(sbi->s_qf_names[qtype], '/')) {
909 ext3_msg(sb, KERN_ERR,
910 "quotafile must be on filesystem root");
911 kfree(sbi->s_qf_names[qtype]);
912 sbi->s_qf_names[qtype] = NULL;
913 return 0;
914 }
915 set_opt(sbi->s_mount_opt, QUOTA);
916 return 1;
917 }
918
919 static int clear_qf_name(struct super_block *sb, int qtype) {
920
921 struct ext3_sb_info *sbi = EXT3_SB(sb);
922
923 if (sb_any_quota_loaded(sb) &&
924 sbi->s_qf_names[qtype]) {
925 ext3_msg(sb, KERN_ERR, "Cannot change journaled quota options"
926 " when quota turned on");
927 return 0;
928 }
929 /*
930 * The space will be released later when all options are confirmed
931 * to be correct
932 */
933 sbi->s_qf_names[qtype] = NULL;
934 return 1;
935 }
936 #endif
937
938 static int parse_options (char *options, struct super_block *sb,
939 unsigned int *inum, unsigned long *journal_devnum,
940 ext3_fsblk_t *n_blocks_count, int is_remount)
941 {
942 struct ext3_sb_info *sbi = EXT3_SB(sb);
943 char * p;
944 substring_t args[MAX_OPT_ARGS];
945 int data_opt = 0;
946 int option;
947 #ifdef CONFIG_QUOTA
948 int qfmt;
949 #endif
950
951 if (!options)
952 return 1;
953
954 while ((p = strsep (&options, ",")) != NULL) {
955 int token;
956 if (!*p)
957 continue;
958 /*
959 * Initialize args struct so we know whether arg was
960 * found; some options take optional arguments.
961 */
962 args[0].to = args[0].from = 0;
963 token = match_token(p, tokens, args);
964 switch (token) {
965 case Opt_bsd_df:
966 clear_opt (sbi->s_mount_opt, MINIX_DF);
967 break;
968 case Opt_minix_df:
969 set_opt (sbi->s_mount_opt, MINIX_DF);
970 break;
971 case Opt_grpid:
972 set_opt (sbi->s_mount_opt, GRPID);
973 break;
974 case Opt_nogrpid:
975 clear_opt (sbi->s_mount_opt, GRPID);
976 break;
977 case Opt_resuid:
978 if (match_int(&args[0], &option))
979 return 0;
980 sbi->s_resuid = option;
981 break;
982 case Opt_resgid:
983 if (match_int(&args[0], &option))
984 return 0;
985 sbi->s_resgid = option;
986 break;
987 case Opt_sb:
988 /* handled by get_sb_block() instead of here */
989 /* *sb_block = match_int(&args[0]); */
990 break;
991 case Opt_err_panic:
992 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
993 clear_opt (sbi->s_mount_opt, ERRORS_RO);
994 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
995 break;
996 case Opt_err_ro:
997 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
998 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
999 set_opt (sbi->s_mount_opt, ERRORS_RO);
1000 break;
1001 case Opt_err_cont:
1002 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1003 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1004 set_opt (sbi->s_mount_opt, ERRORS_CONT);
1005 break;
1006 case Opt_nouid32:
1007 set_opt (sbi->s_mount_opt, NO_UID32);
1008 break;
1009 case Opt_nocheck:
1010 clear_opt (sbi->s_mount_opt, CHECK);
1011 break;
1012 case Opt_debug:
1013 set_opt (sbi->s_mount_opt, DEBUG);
1014 break;
1015 case Opt_oldalloc:
1016 set_opt (sbi->s_mount_opt, OLDALLOC);
1017 break;
1018 case Opt_orlov:
1019 clear_opt (sbi->s_mount_opt, OLDALLOC);
1020 break;
1021 #ifdef CONFIG_EXT3_FS_XATTR
1022 case Opt_user_xattr:
1023 set_opt (sbi->s_mount_opt, XATTR_USER);
1024 break;
1025 case Opt_nouser_xattr:
1026 clear_opt (sbi->s_mount_opt, XATTR_USER);
1027 break;
1028 #else
1029 case Opt_user_xattr:
1030 case Opt_nouser_xattr:
1031 ext3_msg(sb, KERN_INFO,
1032 "(no)user_xattr options not supported");
1033 break;
1034 #endif
1035 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1036 case Opt_acl:
1037 set_opt(sbi->s_mount_opt, POSIX_ACL);
1038 break;
1039 case Opt_noacl:
1040 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1041 break;
1042 #else
1043 case Opt_acl:
1044 case Opt_noacl:
1045 ext3_msg(sb, KERN_INFO,
1046 "(no)acl options not supported");
1047 break;
1048 #endif
1049 case Opt_reservation:
1050 set_opt(sbi->s_mount_opt, RESERVATION);
1051 break;
1052 case Opt_noreservation:
1053 clear_opt(sbi->s_mount_opt, RESERVATION);
1054 break;
1055 case Opt_journal_update:
1056 /* @@@ FIXME */
1057 /* Eventually we will want to be able to create
1058 a journal file here. For now, only allow the
1059 user to specify an existing inode to be the
1060 journal file. */
1061 if (is_remount) {
1062 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1063 "journal on remount");
1064 return 0;
1065 }
1066 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1067 break;
1068 case Opt_journal_inum:
1069 if (is_remount) {
1070 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1071 "journal on remount");
1072 return 0;
1073 }
1074 if (match_int(&args[0], &option))
1075 return 0;
1076 *inum = option;
1077 break;
1078 case Opt_journal_dev:
1079 if (is_remount) {
1080 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1081 "journal on remount");
1082 return 0;
1083 }
1084 if (match_int(&args[0], &option))
1085 return 0;
1086 *journal_devnum = option;
1087 break;
1088 case Opt_noload:
1089 set_opt (sbi->s_mount_opt, NOLOAD);
1090 break;
1091 case Opt_commit:
1092 if (match_int(&args[0], &option))
1093 return 0;
1094 if (option < 0)
1095 return 0;
1096 if (option == 0)
1097 option = JBD_DEFAULT_MAX_COMMIT_AGE;
1098 sbi->s_commit_interval = HZ * option;
1099 break;
1100 case Opt_data_journal:
1101 data_opt = EXT3_MOUNT_JOURNAL_DATA;
1102 goto datacheck;
1103 case Opt_data_ordered:
1104 data_opt = EXT3_MOUNT_ORDERED_DATA;
1105 goto datacheck;
1106 case Opt_data_writeback:
1107 data_opt = EXT3_MOUNT_WRITEBACK_DATA;
1108 datacheck:
1109 if (is_remount) {
1110 if (test_opt(sb, DATA_FLAGS) == data_opt)
1111 break;
1112 ext3_msg(sb, KERN_ERR,
1113 "error: cannot change "
1114 "data mode on remount. The filesystem "
1115 "is mounted in data=%s mode and you "
1116 "try to remount it in data=%s mode.",
1117 data_mode_string(test_opt(sb,
1118 DATA_FLAGS)),
1119 data_mode_string(data_opt));
1120 return 0;
1121 } else {
1122 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
1123 sbi->s_mount_opt |= data_opt;
1124 }
1125 break;
1126 case Opt_data_err_abort:
1127 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1128 break;
1129 case Opt_data_err_ignore:
1130 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1131 break;
1132 #ifdef CONFIG_QUOTA
1133 case Opt_usrjquota:
1134 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1135 return 0;
1136 break;
1137 case Opt_grpjquota:
1138 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1139 return 0;
1140 break;
1141 case Opt_offusrjquota:
1142 if (!clear_qf_name(sb, USRQUOTA))
1143 return 0;
1144 break;
1145 case Opt_offgrpjquota:
1146 if (!clear_qf_name(sb, GRPQUOTA))
1147 return 0;
1148 break;
1149 case Opt_jqfmt_vfsold:
1150 qfmt = QFMT_VFS_OLD;
1151 goto set_qf_format;
1152 case Opt_jqfmt_vfsv0:
1153 qfmt = QFMT_VFS_V0;
1154 goto set_qf_format;
1155 case Opt_jqfmt_vfsv1:
1156 qfmt = QFMT_VFS_V1;
1157 set_qf_format:
1158 if (sb_any_quota_loaded(sb) &&
1159 sbi->s_jquota_fmt != qfmt) {
1160 ext3_msg(sb, KERN_ERR, "error: cannot change "
1161 "journaled quota options when "
1162 "quota turned on.");
1163 return 0;
1164 }
1165 sbi->s_jquota_fmt = qfmt;
1166 break;
1167 case Opt_quota:
1168 case Opt_usrquota:
1169 set_opt(sbi->s_mount_opt, QUOTA);
1170 set_opt(sbi->s_mount_opt, USRQUOTA);
1171 break;
1172 case Opt_grpquota:
1173 set_opt(sbi->s_mount_opt, QUOTA);
1174 set_opt(sbi->s_mount_opt, GRPQUOTA);
1175 break;
1176 case Opt_noquota:
1177 if (sb_any_quota_loaded(sb)) {
1178 ext3_msg(sb, KERN_ERR, "error: cannot change "
1179 "quota options when quota turned on.");
1180 return 0;
1181 }
1182 clear_opt(sbi->s_mount_opt, QUOTA);
1183 clear_opt(sbi->s_mount_opt, USRQUOTA);
1184 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1185 break;
1186 #else
1187 case Opt_quota:
1188 case Opt_usrquota:
1189 case Opt_grpquota:
1190 ext3_msg(sb, KERN_ERR,
1191 "error: quota options not supported.");
1192 break;
1193 case Opt_usrjquota:
1194 case Opt_grpjquota:
1195 case Opt_offusrjquota:
1196 case Opt_offgrpjquota:
1197 case Opt_jqfmt_vfsold:
1198 case Opt_jqfmt_vfsv0:
1199 case Opt_jqfmt_vfsv1:
1200 ext3_msg(sb, KERN_ERR,
1201 "error: journaled quota options not "
1202 "supported.");
1203 break;
1204 case Opt_noquota:
1205 break;
1206 #endif
1207 case Opt_abort:
1208 set_opt(sbi->s_mount_opt, ABORT);
1209 break;
1210 case Opt_nobarrier:
1211 clear_opt(sbi->s_mount_opt, BARRIER);
1212 break;
1213 case Opt_barrier:
1214 if (args[0].from) {
1215 if (match_int(&args[0], &option))
1216 return 0;
1217 } else
1218 option = 1; /* No argument, default to 1 */
1219 if (option)
1220 set_opt(sbi->s_mount_opt, BARRIER);
1221 else
1222 clear_opt(sbi->s_mount_opt, BARRIER);
1223 break;
1224 case Opt_ignore:
1225 break;
1226 case Opt_resize:
1227 if (!is_remount) {
1228 ext3_msg(sb, KERN_ERR,
1229 "error: resize option only available "
1230 "for remount");
1231 return 0;
1232 }
1233 if (match_int(&args[0], &option) != 0)
1234 return 0;
1235 *n_blocks_count = option;
1236 break;
1237 case Opt_nobh:
1238 ext3_msg(sb, KERN_WARNING,
1239 "warning: ignoring deprecated nobh option");
1240 break;
1241 case Opt_bh:
1242 ext3_msg(sb, KERN_WARNING,
1243 "warning: ignoring deprecated bh option");
1244 break;
1245 default:
1246 ext3_msg(sb, KERN_ERR,
1247 "error: unrecognized mount option \"%s\" "
1248 "or missing value", p);
1249 return 0;
1250 }
1251 }
1252 #ifdef CONFIG_QUOTA
1253 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1254 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1255 clear_opt(sbi->s_mount_opt, USRQUOTA);
1256 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1257 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1258
1259 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1260 ext3_msg(sb, KERN_ERR, "error: old and new quota "
1261 "format mixing.");
1262 return 0;
1263 }
1264
1265 if (!sbi->s_jquota_fmt) {
1266 ext3_msg(sb, KERN_ERR, "error: journaled quota format "
1267 "not specified.");
1268 return 0;
1269 }
1270 } else {
1271 if (sbi->s_jquota_fmt) {
1272 ext3_msg(sb, KERN_ERR, "error: journaled quota format "
1273 "specified with no journaling "
1274 "enabled.");
1275 return 0;
1276 }
1277 }
1278 #endif
1279 return 1;
1280 }
1281
1282 static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
1283 int read_only)
1284 {
1285 struct ext3_sb_info *sbi = EXT3_SB(sb);
1286 int res = 0;
1287
1288 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
1289 ext3_msg(sb, KERN_ERR,
1290 "error: revision level too high, "
1291 "forcing read-only mode");
1292 res = MS_RDONLY;
1293 }
1294 if (read_only)
1295 return res;
1296 if (!(sbi->s_mount_state & EXT3_VALID_FS))
1297 ext3_msg(sb, KERN_WARNING,
1298 "warning: mounting unchecked fs, "
1299 "running e2fsck is recommended");
1300 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
1301 ext3_msg(sb, KERN_WARNING,
1302 "warning: mounting fs with errors, "
1303 "running e2fsck is recommended");
1304 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1305 le16_to_cpu(es->s_mnt_count) >=
1306 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1307 ext3_msg(sb, KERN_WARNING,
1308 "warning: maximal mount count reached, "
1309 "running e2fsck is recommended");
1310 else if (le32_to_cpu(es->s_checkinterval) &&
1311 (le32_to_cpu(es->s_lastcheck) +
1312 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1313 ext3_msg(sb, KERN_WARNING,
1314 "warning: checktime reached, "
1315 "running e2fsck is recommended");
1316 #if 0
1317 /* @@@ We _will_ want to clear the valid bit if we find
1318 inconsistencies, to force a fsck at reboot. But for
1319 a plain journaled filesystem we can keep it set as
1320 valid forever! :) */
1321 es->s_state &= cpu_to_le16(~EXT3_VALID_FS);
1322 #endif
1323 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1324 es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
1325 le16_add_cpu(&es->s_mnt_count, 1);
1326 es->s_mtime = cpu_to_le32(get_seconds());
1327 ext3_update_dynamic_rev(sb);
1328 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1329
1330 ext3_commit_super(sb, es, 1);
1331 if (test_opt(sb, DEBUG))
1332 ext3_msg(sb, KERN_INFO, "[bs=%lu, gc=%lu, "
1333 "bpg=%lu, ipg=%lu, mo=%04lx]",
1334 sb->s_blocksize,
1335 sbi->s_groups_count,
1336 EXT3_BLOCKS_PER_GROUP(sb),
1337 EXT3_INODES_PER_GROUP(sb),
1338 sbi->s_mount_opt);
1339
1340 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
1341 char b[BDEVNAME_SIZE];
1342 ext3_msg(sb, KERN_INFO, "using external journal on %s",
1343 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1344 } else {
1345 ext3_msg(sb, KERN_INFO, "using internal journal");
1346 }
1347 return res;
1348 }
1349
1350 /* Called at mount-time, super-block is locked */
1351 static int ext3_check_descriptors(struct super_block *sb)
1352 {
1353 struct ext3_sb_info *sbi = EXT3_SB(sb);
1354 int i;
1355
1356 ext3_debug ("Checking group descriptors");
1357
1358 for (i = 0; i < sbi->s_groups_count; i++) {
1359 struct ext3_group_desc *gdp = ext3_get_group_desc(sb, i, NULL);
1360 ext3_fsblk_t first_block = ext3_group_first_block_no(sb, i);
1361 ext3_fsblk_t last_block;
1362
1363 if (i == sbi->s_groups_count - 1)
1364 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
1365 else
1366 last_block = first_block +
1367 (EXT3_BLOCKS_PER_GROUP(sb) - 1);
1368
1369 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
1370 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
1371 {
1372 ext3_error (sb, "ext3_check_descriptors",
1373 "Block bitmap for group %d"
1374 " not in group (block %lu)!",
1375 i, (unsigned long)
1376 le32_to_cpu(gdp->bg_block_bitmap));
1377 return 0;
1378 }
1379 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
1380 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
1381 {
1382 ext3_error (sb, "ext3_check_descriptors",
1383 "Inode bitmap for group %d"
1384 " not in group (block %lu)!",
1385 i, (unsigned long)
1386 le32_to_cpu(gdp->bg_inode_bitmap));
1387 return 0;
1388 }
1389 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
1390 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
1391 last_block)
1392 {
1393 ext3_error (sb, "ext3_check_descriptors",
1394 "Inode table for group %d"
1395 " not in group (block %lu)!",
1396 i, (unsigned long)
1397 le32_to_cpu(gdp->bg_inode_table));
1398 return 0;
1399 }
1400 }
1401
1402 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1403 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1404 return 1;
1405 }
1406
1407
1408 /* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1409 * the superblock) which were deleted from all directories, but held open by
1410 * a process at the time of a crash. We walk the list and try to delete these
1411 * inodes at recovery time (only with a read-write filesystem).
1412 *
1413 * In order to keep the orphan inode chain consistent during traversal (in
1414 * case of crash during recovery), we link each inode into the superblock
1415 * orphan list_head and handle it the same way as an inode deletion during
1416 * normal operation (which journals the operations for us).
1417 *
1418 * We only do an iget() and an iput() on each inode, which is very safe if we
1419 * accidentally point at an in-use or already deleted inode. The worst that
1420 * can happen in this case is that we get a "bit already cleared" message from
1421 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1422 * e2fsck was run on this filesystem, and it must have already done the orphan
1423 * inode cleanup for us, so we can safely abort without any further action.
1424 */
1425 static void ext3_orphan_cleanup (struct super_block * sb,
1426 struct ext3_super_block * es)
1427 {
1428 unsigned int s_flags = sb->s_flags;
1429 int nr_orphans = 0, nr_truncates = 0;
1430 #ifdef CONFIG_QUOTA
1431 int i;
1432 #endif
1433 if (!es->s_last_orphan) {
1434 jbd_debug(4, "no orphan inodes to clean up\n");
1435 return;
1436 }
1437
1438 if (bdev_read_only(sb->s_bdev)) {
1439 ext3_msg(sb, KERN_ERR, "error: write access "
1440 "unavailable, skipping orphan cleanup.");
1441 return;
1442 }
1443
1444 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1445 if (es->s_last_orphan)
1446 jbd_debug(1, "Errors on filesystem, "
1447 "clearing orphan list.\n");
1448 es->s_last_orphan = 0;
1449 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1450 return;
1451 }
1452
1453 if (s_flags & MS_RDONLY) {
1454 ext3_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1455 sb->s_flags &= ~MS_RDONLY;
1456 }
1457 #ifdef CONFIG_QUOTA
1458 /* Needed for iput() to work correctly and not trash data */
1459 sb->s_flags |= MS_ACTIVE;
1460 /* Turn on quotas so that they are updated correctly */
1461 for (i = 0; i < MAXQUOTAS; i++) {
1462 if (EXT3_SB(sb)->s_qf_names[i]) {
1463 int ret = ext3_quota_on_mount(sb, i);
1464 if (ret < 0)
1465 ext3_msg(sb, KERN_ERR,
1466 "error: cannot turn on journaled "
1467 "quota: %d", ret);
1468 }
1469 }
1470 #endif
1471
1472 while (es->s_last_orphan) {
1473 struct inode *inode;
1474
1475 inode = ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1476 if (IS_ERR(inode)) {
1477 es->s_last_orphan = 0;
1478 break;
1479 }
1480
1481 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
1482 dquot_initialize(inode);
1483 if (inode->i_nlink) {
1484 printk(KERN_DEBUG
1485 "%s: truncating inode %lu to %Ld bytes\n",
1486 __func__, inode->i_ino, inode->i_size);
1487 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1488 inode->i_ino, inode->i_size);
1489 ext3_truncate(inode);
1490 nr_truncates++;
1491 } else {
1492 printk(KERN_DEBUG
1493 "%s: deleting unreferenced inode %lu\n",
1494 __func__, inode->i_ino);
1495 jbd_debug(2, "deleting unreferenced inode %lu\n",
1496 inode->i_ino);
1497 nr_orphans++;
1498 }
1499 iput(inode); /* The delete magic happens here! */
1500 }
1501
1502 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1503
1504 if (nr_orphans)
1505 ext3_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1506 PLURAL(nr_orphans));
1507 if (nr_truncates)
1508 ext3_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1509 PLURAL(nr_truncates));
1510 #ifdef CONFIG_QUOTA
1511 /* Turn quotas off */
1512 for (i = 0; i < MAXQUOTAS; i++) {
1513 if (sb_dqopt(sb)->files[i])
1514 dquot_quota_off(sb, i);
1515 }
1516 #endif
1517 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1518 }
1519
1520 /*
1521 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1522 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1523 * We need to be 1 filesystem block less than the 2^32 sector limit.
1524 */
1525 static loff_t ext3_max_size(int bits)
1526 {
1527 loff_t res = EXT3_NDIR_BLOCKS;
1528 int meta_blocks;
1529 loff_t upper_limit;
1530
1531 /* This is calculated to be the largest file size for a
1532 * dense, file such that the total number of
1533 * sectors in the file, including data and all indirect blocks,
1534 * does not exceed 2^32 -1
1535 * __u32 i_blocks representing the total number of
1536 * 512 bytes blocks of the file
1537 */
1538 upper_limit = (1LL << 32) - 1;
1539
1540 /* total blocks in file system block size */
1541 upper_limit >>= (bits - 9);
1542
1543
1544 /* indirect blocks */
1545 meta_blocks = 1;
1546 /* double indirect blocks */
1547 meta_blocks += 1 + (1LL << (bits-2));
1548 /* tripple indirect blocks */
1549 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1550
1551 upper_limit -= meta_blocks;
1552 upper_limit <<= bits;
1553
1554 res += 1LL << (bits-2);
1555 res += 1LL << (2*(bits-2));
1556 res += 1LL << (3*(bits-2));
1557 res <<= bits;
1558 if (res > upper_limit)
1559 res = upper_limit;
1560
1561 if (res > MAX_LFS_FILESIZE)
1562 res = MAX_LFS_FILESIZE;
1563
1564 return res;
1565 }
1566
1567 static ext3_fsblk_t descriptor_loc(struct super_block *sb,
1568 ext3_fsblk_t logic_sb_block,
1569 int nr)
1570 {
1571 struct ext3_sb_info *sbi = EXT3_SB(sb);
1572 unsigned long bg, first_meta_bg;
1573 int has_super = 0;
1574
1575 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1576
1577 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1578 nr < first_meta_bg)
1579 return (logic_sb_block + nr + 1);
1580 bg = sbi->s_desc_per_block * nr;
1581 if (ext3_bg_has_super(sb, bg))
1582 has_super = 1;
1583 return (has_super + ext3_group_first_block_no(sb, bg));
1584 }
1585
1586
1587 static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1588 {
1589 struct buffer_head * bh;
1590 struct ext3_super_block *es = NULL;
1591 struct ext3_sb_info *sbi;
1592 ext3_fsblk_t block;
1593 ext3_fsblk_t sb_block = get_sb_block(&data, sb);
1594 ext3_fsblk_t logic_sb_block;
1595 unsigned long offset = 0;
1596 unsigned int journal_inum = 0;
1597 unsigned long journal_devnum = 0;
1598 unsigned long def_mount_opts;
1599 struct inode *root;
1600 int blocksize;
1601 int hblock;
1602 int db_count;
1603 int i;
1604 int needs_recovery;
1605 int ret = -EINVAL;
1606 __le32 features;
1607 int err;
1608
1609 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1610 if (!sbi)
1611 return -ENOMEM;
1612
1613 sbi->s_blockgroup_lock =
1614 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
1615 if (!sbi->s_blockgroup_lock) {
1616 kfree(sbi);
1617 return -ENOMEM;
1618 }
1619 sb->s_fs_info = sbi;
1620 sbi->s_mount_opt = 0;
1621 sbi->s_resuid = EXT3_DEF_RESUID;
1622 sbi->s_resgid = EXT3_DEF_RESGID;
1623 sbi->s_sb_block = sb_block;
1624
1625 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1626 if (!blocksize) {
1627 ext3_msg(sb, KERN_ERR, "error: unable to set blocksize");
1628 goto out_fail;
1629 }
1630
1631 /*
1632 * The ext3 superblock will not be buffer aligned for other than 1kB
1633 * block sizes. We need to calculate the offset from buffer start.
1634 */
1635 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1636 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1637 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1638 } else {
1639 logic_sb_block = sb_block;
1640 }
1641
1642 if (!(bh = sb_bread(sb, logic_sb_block))) {
1643 ext3_msg(sb, KERN_ERR, "error: unable to read superblock");
1644 goto out_fail;
1645 }
1646 /*
1647 * Note: s_es must be initialized as soon as possible because
1648 * some ext3 macro-instructions depend on its value
1649 */
1650 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1651 sbi->s_es = es;
1652 sb->s_magic = le16_to_cpu(es->s_magic);
1653 if (sb->s_magic != EXT3_SUPER_MAGIC)
1654 goto cantfind_ext3;
1655
1656 /* Set defaults before we parse the mount options */
1657 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1658 if (def_mount_opts & EXT3_DEFM_DEBUG)
1659 set_opt(sbi->s_mount_opt, DEBUG);
1660 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1661 set_opt(sbi->s_mount_opt, GRPID);
1662 if (def_mount_opts & EXT3_DEFM_UID16)
1663 set_opt(sbi->s_mount_opt, NO_UID32);
1664 #ifdef CONFIG_EXT3_FS_XATTR
1665 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1666 set_opt(sbi->s_mount_opt, XATTR_USER);
1667 #endif
1668 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1669 if (def_mount_opts & EXT3_DEFM_ACL)
1670 set_opt(sbi->s_mount_opt, POSIX_ACL);
1671 #endif
1672 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
1673 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1674 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
1675 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1676 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
1677 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
1678
1679 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1680 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1681 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_CONTINUE)
1682 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1683 else
1684 set_opt(sbi->s_mount_opt, ERRORS_RO);
1685
1686 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1687 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1688
1689 set_opt(sbi->s_mount_opt, RESERVATION);
1690
1691 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1692 NULL, 0))
1693 goto failed_mount;
1694
1695 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1696 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
1697
1698 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1699 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1700 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1701 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1702 ext3_msg(sb, KERN_WARNING,
1703 "warning: feature flags set on rev 0 fs, "
1704 "running e2fsck is recommended");
1705 /*
1706 * Check feature flags regardless of the revision level, since we
1707 * previously didn't change the revision level when setting the flags,
1708 * so there is a chance incompat flags are set on a rev 0 filesystem.
1709 */
1710 features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
1711 if (features) {
1712 ext3_msg(sb, KERN_ERR,
1713 "error: couldn't mount because of unsupported "
1714 "optional features (%x)", le32_to_cpu(features));
1715 goto failed_mount;
1716 }
1717 features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
1718 if (!(sb->s_flags & MS_RDONLY) && features) {
1719 ext3_msg(sb, KERN_ERR,
1720 "error: couldn't mount RDWR because of unsupported "
1721 "optional features (%x)", le32_to_cpu(features));
1722 goto failed_mount;
1723 }
1724 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1725
1726 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1727 blocksize > EXT3_MAX_BLOCK_SIZE) {
1728 ext3_msg(sb, KERN_ERR,
1729 "error: couldn't mount because of unsupported "
1730 "filesystem blocksize %d", blocksize);
1731 goto failed_mount;
1732 }
1733
1734 hblock = bdev_logical_block_size(sb->s_bdev);
1735 if (sb->s_blocksize != blocksize) {
1736 /*
1737 * Make sure the blocksize for the filesystem is larger
1738 * than the hardware sectorsize for the machine.
1739 */
1740 if (blocksize < hblock) {
1741 ext3_msg(sb, KERN_ERR,
1742 "error: fsblocksize %d too small for "
1743 "hardware sectorsize %d", blocksize, hblock);
1744 goto failed_mount;
1745 }
1746
1747 brelse (bh);
1748 if (!sb_set_blocksize(sb, blocksize)) {
1749 ext3_msg(sb, KERN_ERR,
1750 "error: bad blocksize %d", blocksize);
1751 goto out_fail;
1752 }
1753 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1754 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1755 bh = sb_bread(sb, logic_sb_block);
1756 if (!bh) {
1757 ext3_msg(sb, KERN_ERR,
1758 "error: can't read superblock on 2nd try");
1759 goto failed_mount;
1760 }
1761 es = (struct ext3_super_block *)(((char *)bh->b_data) + offset);
1762 sbi->s_es = es;
1763 if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
1764 ext3_msg(sb, KERN_ERR,
1765 "error: magic mismatch");
1766 goto failed_mount;
1767 }
1768 }
1769
1770 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1771
1772 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1773 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1774 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1775 } else {
1776 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1777 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1778 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
1779 (!is_power_of_2(sbi->s_inode_size)) ||
1780 (sbi->s_inode_size > blocksize)) {
1781 ext3_msg(sb, KERN_ERR,
1782 "error: unsupported inode size: %d",
1783 sbi->s_inode_size);
1784 goto failed_mount;
1785 }
1786 }
1787 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1788 le32_to_cpu(es->s_log_frag_size);
1789 if (blocksize != sbi->s_frag_size) {
1790 ext3_msg(sb, KERN_ERR,
1791 "error: fragsize %lu != blocksize %u (unsupported)",
1792 sbi->s_frag_size, blocksize);
1793 goto failed_mount;
1794 }
1795 sbi->s_frags_per_block = 1;
1796 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1797 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1798 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1799 if (EXT3_INODE_SIZE(sb) == 0 || EXT3_INODES_PER_GROUP(sb) == 0)
1800 goto cantfind_ext3;
1801 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1802 if (sbi->s_inodes_per_block == 0)
1803 goto cantfind_ext3;
1804 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1805 sbi->s_inodes_per_block;
1806 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1807 sbi->s_sbh = bh;
1808 sbi->s_mount_state = le16_to_cpu(es->s_state);
1809 sbi->s_addr_per_block_bits = ilog2(EXT3_ADDR_PER_BLOCK(sb));
1810 sbi->s_desc_per_block_bits = ilog2(EXT3_DESC_PER_BLOCK(sb));
1811 for (i=0; i < 4; i++)
1812 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1813 sbi->s_def_hash_version = es->s_def_hash_version;
1814 i = le32_to_cpu(es->s_flags);
1815 if (i & EXT2_FLAGS_UNSIGNED_HASH)
1816 sbi->s_hash_unsigned = 3;
1817 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
1818 #ifdef __CHAR_UNSIGNED__
1819 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
1820 sbi->s_hash_unsigned = 3;
1821 #else
1822 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
1823 #endif
1824 }
1825
1826 if (sbi->s_blocks_per_group > blocksize * 8) {
1827 ext3_msg(sb, KERN_ERR,
1828 "#blocks per group too big: %lu",
1829 sbi->s_blocks_per_group);
1830 goto failed_mount;
1831 }
1832 if (sbi->s_frags_per_group > blocksize * 8) {
1833 ext3_msg(sb, KERN_ERR,
1834 "error: #fragments per group too big: %lu",
1835 sbi->s_frags_per_group);
1836 goto failed_mount;
1837 }
1838 if (sbi->s_inodes_per_group > blocksize * 8) {
1839 ext3_msg(sb, KERN_ERR,
1840 "error: #inodes per group too big: %lu",
1841 sbi->s_inodes_per_group);
1842 goto failed_mount;
1843 }
1844
1845 if (generic_check_addressable(sb->s_blocksize_bits,
1846 le32_to_cpu(es->s_blocks_count))) {
1847 ext3_msg(sb, KERN_ERR,
1848 "error: filesystem is too large to mount safely");
1849 if (sizeof(sector_t) < 8)
1850 ext3_msg(sb, KERN_ERR,
1851 "error: CONFIG_LBDAF not enabled");
1852 goto failed_mount;
1853 }
1854
1855 if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1856 goto cantfind_ext3;
1857 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1858 le32_to_cpu(es->s_first_data_block) - 1)
1859 / EXT3_BLOCKS_PER_GROUP(sb)) + 1;
1860 db_count = (sbi->s_groups_count + EXT3_DESC_PER_BLOCK(sb) - 1) /
1861 EXT3_DESC_PER_BLOCK(sb);
1862 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1863 GFP_KERNEL);
1864 if (sbi->s_group_desc == NULL) {
1865 ext3_msg(sb, KERN_ERR,
1866 "error: not enough memory");
1867 goto failed_mount;
1868 }
1869
1870 bgl_lock_init(sbi->s_blockgroup_lock);
1871
1872 for (i = 0; i < db_count; i++) {
1873 block = descriptor_loc(sb, logic_sb_block, i);
1874 sbi->s_group_desc[i] = sb_bread(sb, block);
1875 if (!sbi->s_group_desc[i]) {
1876 ext3_msg(sb, KERN_ERR,
1877 "error: can't read group descriptor %d", i);
1878 db_count = i;
1879 goto failed_mount2;
1880 }
1881 }
1882 if (!ext3_check_descriptors (sb)) {
1883 ext3_msg(sb, KERN_ERR,
1884 "error: group descriptors corrupted");
1885 goto failed_mount2;
1886 }
1887 sbi->s_gdb_count = db_count;
1888 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1889 spin_lock_init(&sbi->s_next_gen_lock);
1890
1891 /* per fileystem reservation list head & lock */
1892 spin_lock_init(&sbi->s_rsv_window_lock);
1893 sbi->s_rsv_window_root = RB_ROOT;
1894 /* Add a single, static dummy reservation to the start of the
1895 * reservation window list --- it gives us a placeholder for
1896 * append-at-start-of-list which makes the allocation logic
1897 * _much_ simpler. */
1898 sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1899 sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1900 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1901 sbi->s_rsv_window_head.rsv_goal_size = 0;
1902 ext3_rsv_window_add(sb, &sbi->s_rsv_window_head);
1903
1904 /*
1905 * set up enough so that it can read an inode
1906 */
1907 sb->s_op = &ext3_sops;
1908 sb->s_export_op = &ext3_export_ops;
1909 sb->s_xattr = ext3_xattr_handlers;
1910 #ifdef CONFIG_QUOTA
1911 sb->s_qcop = &ext3_qctl_operations;
1912 sb->dq_op = &ext3_quota_operations;
1913 #endif
1914 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1915 mutex_init(&sbi->s_orphan_lock);
1916 mutex_init(&sbi->s_resize_lock);
1917
1918 sb->s_root = NULL;
1919
1920 needs_recovery = (es->s_last_orphan != 0 ||
1921 EXT3_HAS_INCOMPAT_FEATURE(sb,
1922 EXT3_FEATURE_INCOMPAT_RECOVER));
1923
1924 /*
1925 * The first inode we look at is the journal inode. Don't try
1926 * root first: it may be modified in the journal!
1927 */
1928 if (!test_opt(sb, NOLOAD) &&
1929 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
1930 if (ext3_load_journal(sb, es, journal_devnum))
1931 goto failed_mount2;
1932 } else if (journal_inum) {
1933 if (ext3_create_journal(sb, es, journal_inum))
1934 goto failed_mount2;
1935 } else {
1936 if (!silent)
1937 ext3_msg(sb, KERN_ERR,
1938 "error: no journal found. "
1939 "mounting ext3 over ext2?");
1940 goto failed_mount2;
1941 }
1942 err = percpu_counter_init(&sbi->s_freeblocks_counter,
1943 ext3_count_free_blocks(sb));
1944 if (!err) {
1945 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1946 ext3_count_free_inodes(sb));
1947 }
1948 if (!err) {
1949 err = percpu_counter_init(&sbi->s_dirs_counter,
1950 ext3_count_dirs(sb));
1951 }
1952 if (err) {
1953 ext3_msg(sb, KERN_ERR, "error: insufficient memory");
1954 goto failed_mount3;
1955 }
1956
1957 /* We have now updated the journal if required, so we can
1958 * validate the data journaling mode. */
1959 switch (test_opt(sb, DATA_FLAGS)) {
1960 case 0:
1961 /* No mode set, assume a default based on the journal
1962 capabilities: ORDERED_DATA if the journal can
1963 cope, else JOURNAL_DATA */
1964 if (journal_check_available_features
1965 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
1966 set_opt(sbi->s_mount_opt, DEFAULT_DATA_MODE);
1967 else
1968 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1969 break;
1970
1971 case EXT3_MOUNT_ORDERED_DATA:
1972 case EXT3_MOUNT_WRITEBACK_DATA:
1973 if (!journal_check_available_features
1974 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
1975 ext3_msg(sb, KERN_ERR,
1976 "error: journal does not support "
1977 "requested data journaling mode");
1978 goto failed_mount3;
1979 }
1980 default:
1981 break;
1982 }
1983
1984 /*
1985 * The journal_load will have done any necessary log recovery,
1986 * so we can safely mount the rest of the filesystem now.
1987 */
1988
1989 root = ext3_iget(sb, EXT3_ROOT_INO);
1990 if (IS_ERR(root)) {
1991 ext3_msg(sb, KERN_ERR, "error: get root inode failed");
1992 ret = PTR_ERR(root);
1993 goto failed_mount3;
1994 }
1995 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1996 iput(root);
1997 ext3_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1998 goto failed_mount3;
1999 }
2000 sb->s_root = d_alloc_root(root);
2001 if (!sb->s_root) {
2002 ext3_msg(sb, KERN_ERR, "error: get root dentry failed");
2003 iput(root);
2004 ret = -ENOMEM;
2005 goto failed_mount3;
2006 }
2007
2008 ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY);
2009
2010 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
2011 ext3_orphan_cleanup(sb, es);
2012 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
2013 if (needs_recovery)
2014 ext3_msg(sb, KERN_INFO, "recovery complete");
2015 ext3_mark_recovery_complete(sb, es);
2016 ext3_msg(sb, KERN_INFO, "mounted filesystem with %s data mode",
2017 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
2018 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
2019 "writeback");
2020
2021 return 0;
2022
2023 cantfind_ext3:
2024 if (!silent)
2025 ext3_msg(sb, KERN_INFO,
2026 "error: can't find ext3 filesystem on dev %s.",
2027 sb->s_id);
2028 goto failed_mount;
2029
2030 failed_mount3:
2031 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2032 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2033 percpu_counter_destroy(&sbi->s_dirs_counter);
2034 journal_destroy(sbi->s_journal);
2035 failed_mount2:
2036 for (i = 0; i < db_count; i++)
2037 brelse(sbi->s_group_desc[i]);
2038 kfree(sbi->s_group_desc);
2039 failed_mount:
2040 #ifdef CONFIG_QUOTA
2041 for (i = 0; i < MAXQUOTAS; i++)
2042 kfree(sbi->s_qf_names[i]);
2043 #endif
2044 ext3_blkdev_remove(sbi);
2045 brelse(bh);
2046 out_fail:
2047 sb->s_fs_info = NULL;
2048 kfree(sbi->s_blockgroup_lock);
2049 kfree(sbi);
2050 return ret;
2051 }
2052
2053 /*
2054 * Setup any per-fs journal parameters now. We'll do this both on
2055 * initial mount, once the journal has been initialised but before we've
2056 * done any recovery; and again on any subsequent remount.
2057 */
2058 static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
2059 {
2060 struct ext3_sb_info *sbi = EXT3_SB(sb);
2061
2062 if (sbi->s_commit_interval)
2063 journal->j_commit_interval = sbi->s_commit_interval;
2064 /* We could also set up an ext3-specific default for the commit
2065 * interval here, but for now we'll just fall back to the jbd
2066 * default. */
2067
2068 spin_lock(&journal->j_state_lock);
2069 if (test_opt(sb, BARRIER))
2070 journal->j_flags |= JFS_BARRIER;
2071 else
2072 journal->j_flags &= ~JFS_BARRIER;
2073 if (test_opt(sb, DATA_ERR_ABORT))
2074 journal->j_flags |= JFS_ABORT_ON_SYNCDATA_ERR;
2075 else
2076 journal->j_flags &= ~JFS_ABORT_ON_SYNCDATA_ERR;
2077 spin_unlock(&journal->j_state_lock);
2078 }
2079
2080 static journal_t *ext3_get_journal(struct super_block *sb,
2081 unsigned int journal_inum)
2082 {
2083 struct inode *journal_inode;
2084 journal_t *journal;
2085
2086 /* First, test for the existence of a valid inode on disk. Bad
2087 * things happen if we iget() an unused inode, as the subsequent
2088 * iput() will try to delete it. */
2089
2090 journal_inode = ext3_iget(sb, journal_inum);
2091 if (IS_ERR(journal_inode)) {
2092 ext3_msg(sb, KERN_ERR, "error: no journal found");
2093 return NULL;
2094 }
2095 if (!journal_inode->i_nlink) {
2096 make_bad_inode(journal_inode);
2097 iput(journal_inode);
2098 ext3_msg(sb, KERN_ERR, "error: journal inode is deleted");
2099 return NULL;
2100 }
2101
2102 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2103 journal_inode, journal_inode->i_size);
2104 if (!S_ISREG(journal_inode->i_mode)) {
2105 ext3_msg(sb, KERN_ERR, "error: invalid journal inode");
2106 iput(journal_inode);
2107 return NULL;
2108 }
2109
2110 journal = journal_init_inode(journal_inode);
2111 if (!journal) {
2112 ext3_msg(sb, KERN_ERR, "error: could not load journal inode");
2113 iput(journal_inode);
2114 return NULL;
2115 }
2116 journal->j_private = sb;
2117 ext3_init_journal_params(sb, journal);
2118 return journal;
2119 }
2120
2121 static journal_t *ext3_get_dev_journal(struct super_block *sb,
2122 dev_t j_dev)
2123 {
2124 struct buffer_head * bh;
2125 journal_t *journal;
2126 ext3_fsblk_t start;
2127 ext3_fsblk_t len;
2128 int hblock, blocksize;
2129 ext3_fsblk_t sb_block;
2130 unsigned long offset;
2131 struct ext3_super_block * es;
2132 struct block_device *bdev;
2133
2134 bdev = ext3_blkdev_get(j_dev, sb);
2135 if (bdev == NULL)
2136 return NULL;
2137
2138 if (bd_claim(bdev, sb)) {
2139 ext3_msg(sb, KERN_ERR,
2140 "error: failed to claim external journal device");
2141 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
2142 return NULL;
2143 }
2144
2145 blocksize = sb->s_blocksize;
2146 hblock = bdev_logical_block_size(bdev);
2147 if (blocksize < hblock) {
2148 ext3_msg(sb, KERN_ERR,
2149 "error: blocksize too small for journal device");
2150 goto out_bdev;
2151 }
2152
2153 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
2154 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
2155 set_blocksize(bdev, blocksize);
2156 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2157 ext3_msg(sb, KERN_ERR, "error: couldn't read superblock of "
2158 "external journal");
2159 goto out_bdev;
2160 }
2161
2162 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
2163 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
2164 !(le32_to_cpu(es->s_feature_incompat) &
2165 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2166 ext3_msg(sb, KERN_ERR, "error: external journal has "
2167 "bad superblock");
2168 brelse(bh);
2169 goto out_bdev;
2170 }
2171
2172 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2173 ext3_msg(sb, KERN_ERR, "error: journal UUID does not match");
2174 brelse(bh);
2175 goto out_bdev;
2176 }
2177
2178 len = le32_to_cpu(es->s_blocks_count);
2179 start = sb_block + 1;
2180 brelse(bh); /* we're done with the superblock */
2181
2182 journal = journal_init_dev(bdev, sb->s_bdev,
2183 start, len, blocksize);
2184 if (!journal) {
2185 ext3_msg(sb, KERN_ERR,
2186 "error: failed to create device journal");
2187 goto out_bdev;
2188 }
2189 journal->j_private = sb;
2190 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2191 wait_on_buffer(journal->j_sb_buffer);
2192 if (!buffer_uptodate(journal->j_sb_buffer)) {
2193 ext3_msg(sb, KERN_ERR, "I/O error on journal device");
2194 goto out_journal;
2195 }
2196 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2197 ext3_msg(sb, KERN_ERR,
2198 "error: external journal has more than one "
2199 "user (unsupported) - %d",
2200 be32_to_cpu(journal->j_superblock->s_nr_users));
2201 goto out_journal;
2202 }
2203 EXT3_SB(sb)->journal_bdev = bdev;
2204 ext3_init_journal_params(sb, journal);
2205 return journal;
2206 out_journal:
2207 journal_destroy(journal);
2208 out_bdev:
2209 ext3_blkdev_put(bdev);
2210 return NULL;
2211 }
2212
2213 static int ext3_load_journal(struct super_block *sb,
2214 struct ext3_super_block *es,
2215 unsigned long journal_devnum)
2216 {
2217 journal_t *journal;
2218 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2219 dev_t journal_dev;
2220 int err = 0;
2221 int really_read_only;
2222
2223 if (journal_devnum &&
2224 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2225 ext3_msg(sb, KERN_INFO, "external journal device major/minor "
2226 "numbers have changed");
2227 journal_dev = new_decode_dev(journal_devnum);
2228 } else
2229 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2230
2231 really_read_only = bdev_read_only(sb->s_bdev);
2232
2233 /*
2234 * Are we loading a blank journal or performing recovery after a
2235 * crash? For recovery, we need to check in advance whether we
2236 * can get read-write access to the device.
2237 */
2238
2239 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
2240 if (sb->s_flags & MS_RDONLY) {
2241 ext3_msg(sb, KERN_INFO,
2242 "recovery required on readonly filesystem");
2243 if (really_read_only) {
2244 ext3_msg(sb, KERN_ERR, "error: write access "
2245 "unavailable, cannot proceed");
2246 return -EROFS;
2247 }
2248 ext3_msg(sb, KERN_INFO,
2249 "write access will be enabled during recovery");
2250 }
2251 }
2252
2253 if (journal_inum && journal_dev) {
2254 ext3_msg(sb, KERN_ERR, "error: filesystem has both journal "
2255 "and inode journals");
2256 return -EINVAL;
2257 }
2258
2259 if (journal_inum) {
2260 if (!(journal = ext3_get_journal(sb, journal_inum)))
2261 return -EINVAL;
2262 } else {
2263 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
2264 return -EINVAL;
2265 }
2266
2267 if (!(journal->j_flags & JFS_BARRIER))
2268 printk(KERN_INFO "EXT3-fs: barriers not enabled\n");
2269
2270 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2271 err = journal_update_format(journal);
2272 if (err) {
2273 ext3_msg(sb, KERN_ERR, "error updating journal");
2274 journal_destroy(journal);
2275 return err;
2276 }
2277 }
2278
2279 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
2280 err = journal_wipe(journal, !really_read_only);
2281 if (!err)
2282 err = journal_load(journal);
2283
2284 if (err) {
2285 ext3_msg(sb, KERN_ERR, "error loading journal");
2286 journal_destroy(journal);
2287 return err;
2288 }
2289
2290 EXT3_SB(sb)->s_journal = journal;
2291 ext3_clear_journal_err(sb, es);
2292
2293 if (journal_devnum &&
2294 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2295 es->s_journal_dev = cpu_to_le32(journal_devnum);
2296
2297 /* Make sure we flush the recovery flag to disk. */
2298 ext3_commit_super(sb, es, 1);
2299 }
2300
2301 return 0;
2302 }
2303
2304 static int ext3_create_journal(struct super_block *sb,
2305 struct ext3_super_block *es,
2306 unsigned int journal_inum)
2307 {
2308 journal_t *journal;
2309 int err;
2310
2311 if (sb->s_flags & MS_RDONLY) {
2312 ext3_msg(sb, KERN_ERR,
2313 "error: readonly filesystem when trying to "
2314 "create journal");
2315 return -EROFS;
2316 }
2317
2318 journal = ext3_get_journal(sb, journal_inum);
2319 if (!journal)
2320 return -EINVAL;
2321
2322 ext3_msg(sb, KERN_INFO, "creating new journal on inode %u",
2323 journal_inum);
2324
2325 err = journal_create(journal);
2326 if (err) {
2327 ext3_msg(sb, KERN_ERR, "error creating journal");
2328 journal_destroy(journal);
2329 return -EIO;
2330 }
2331
2332 EXT3_SB(sb)->s_journal = journal;
2333
2334 ext3_update_dynamic_rev(sb);
2335 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2336 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
2337
2338 es->s_journal_inum = cpu_to_le32(journal_inum);
2339
2340 /* Make sure we flush the recovery flag to disk. */
2341 ext3_commit_super(sb, es, 1);
2342
2343 return 0;
2344 }
2345
2346 static int ext3_commit_super(struct super_block *sb,
2347 struct ext3_super_block *es,
2348 int sync)
2349 {
2350 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
2351 int error = 0;
2352
2353 if (!sbh)
2354 return error;
2355 /*
2356 * If the file system is mounted read-only, don't update the
2357 * superblock write time. This avoids updating the superblock
2358 * write time when we are mounting the root file system
2359 * read/only but we need to replay the journal; at that point,
2360 * for people who are east of GMT and who make their clock
2361 * tick in localtime for Windows bug-for-bug compatibility,
2362 * the clock is set in the future, and this will cause e2fsck
2363 * to complain and force a full file system check.
2364 */
2365 if (!(sb->s_flags & MS_RDONLY))
2366 es->s_wtime = cpu_to_le32(get_seconds());
2367 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
2368 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
2369 BUFFER_TRACE(sbh, "marking dirty");
2370 mark_buffer_dirty(sbh);
2371 if (sync)
2372 error = sync_dirty_buffer(sbh);
2373 return error;
2374 }
2375
2376
2377 /*
2378 * Have we just finished recovery? If so, and if we are mounting (or
2379 * remounting) the filesystem readonly, then we will end up with a
2380 * consistent fs on disk. Record that fact.
2381 */
2382 static void ext3_mark_recovery_complete(struct super_block * sb,
2383 struct ext3_super_block * es)
2384 {
2385 journal_t *journal = EXT3_SB(sb)->s_journal;
2386
2387 journal_lock_updates(journal);
2388 if (journal_flush(journal) < 0)
2389 goto out;
2390
2391 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
2392 sb->s_flags & MS_RDONLY) {
2393 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2394 ext3_commit_super(sb, es, 1);
2395 }
2396
2397 out:
2398 journal_unlock_updates(journal);
2399 }
2400
2401 /*
2402 * If we are mounting (or read-write remounting) a filesystem whose journal
2403 * has recorded an error from a previous lifetime, move that error to the
2404 * main filesystem now.
2405 */
2406 static void ext3_clear_journal_err(struct super_block *sb,
2407 struct ext3_super_block *es)
2408 {
2409 journal_t *journal;
2410 int j_errno;
2411 const char *errstr;
2412
2413 journal = EXT3_SB(sb)->s_journal;
2414
2415 /*
2416 * Now check for any error status which may have been recorded in the
2417 * journal by a prior ext3_error() or ext3_abort()
2418 */
2419
2420 j_errno = journal_errno(journal);
2421 if (j_errno) {
2422 char nbuf[16];
2423
2424 errstr = ext3_decode_error(sb, j_errno, nbuf);
2425 ext3_warning(sb, __func__, "Filesystem error recorded "
2426 "from previous mount: %s", errstr);
2427 ext3_warning(sb, __func__, "Marking fs in need of "
2428 "filesystem check.");
2429
2430 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
2431 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
2432 ext3_commit_super (sb, es, 1);
2433
2434 journal_clear_err(journal);
2435 }
2436 }
2437
2438 /*
2439 * Force the running and committing transactions to commit,
2440 * and wait on the commit.
2441 */
2442 int ext3_force_commit(struct super_block *sb)
2443 {
2444 journal_t *journal;
2445 int ret;
2446
2447 if (sb->s_flags & MS_RDONLY)
2448 return 0;
2449
2450 journal = EXT3_SB(sb)->s_journal;
2451 ret = ext3_journal_force_commit(journal);
2452 return ret;
2453 }
2454
2455 static int ext3_sync_fs(struct super_block *sb, int wait)
2456 {
2457 tid_t target;
2458
2459 if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
2460 if (wait)
2461 log_wait_commit(EXT3_SB(sb)->s_journal, target);
2462 }
2463 return 0;
2464 }
2465
2466 /*
2467 * LVM calls this function before a (read-only) snapshot is created. This
2468 * gives us a chance to flush the journal completely and mark the fs clean.
2469 */
2470 static int ext3_freeze(struct super_block *sb)
2471 {
2472 int error = 0;
2473 journal_t *journal;
2474
2475 if (!(sb->s_flags & MS_RDONLY)) {
2476 journal = EXT3_SB(sb)->s_journal;
2477
2478 /* Now we set up the journal barrier. */
2479 journal_lock_updates(journal);
2480
2481 /*
2482 * We don't want to clear needs_recovery flag when we failed
2483 * to flush the journal.
2484 */
2485 error = journal_flush(journal);
2486 if (error < 0)
2487 goto out;
2488
2489 /* Journal blocked and flushed, clear needs_recovery flag. */
2490 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2491 error = ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2492 if (error)
2493 goto out;
2494 }
2495 return 0;
2496
2497 out:
2498 journal_unlock_updates(journal);
2499 return error;
2500 }
2501
2502 /*
2503 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2504 * flag here, even though the filesystem is not technically dirty yet.
2505 */
2506 static int ext3_unfreeze(struct super_block *sb)
2507 {
2508 if (!(sb->s_flags & MS_RDONLY)) {
2509 lock_super(sb);
2510 /* Reser the needs_recovery flag before the fs is unlocked. */
2511 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2512 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2513 unlock_super(sb);
2514 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2515 }
2516 return 0;
2517 }
2518
2519 static int ext3_remount (struct super_block * sb, int * flags, char * data)
2520 {
2521 struct ext3_super_block * es;
2522 struct ext3_sb_info *sbi = EXT3_SB(sb);
2523 ext3_fsblk_t n_blocks_count = 0;
2524 unsigned long old_sb_flags;
2525 struct ext3_mount_options old_opts;
2526 int enable_quota = 0;
2527 int err;
2528 #ifdef CONFIG_QUOTA
2529 int i;
2530 #endif
2531
2532 /* Store the original options */
2533 lock_super(sb);
2534 old_sb_flags = sb->s_flags;
2535 old_opts.s_mount_opt = sbi->s_mount_opt;
2536 old_opts.s_resuid = sbi->s_resuid;
2537 old_opts.s_resgid = sbi->s_resgid;
2538 old_opts.s_commit_interval = sbi->s_commit_interval;
2539 #ifdef CONFIG_QUOTA
2540 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2541 for (i = 0; i < MAXQUOTAS; i++)
2542 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2543 #endif
2544
2545 /*
2546 * Allow the "check" option to be passed as a remount option.
2547 */
2548 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2549 err = -EINVAL;
2550 goto restore_opts;
2551 }
2552
2553 if (test_opt(sb, ABORT))
2554 ext3_abort(sb, __func__, "Abort forced by user");
2555
2556 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2557 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
2558
2559 es = sbi->s_es;
2560
2561 ext3_init_journal_params(sb, sbi->s_journal);
2562
2563 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2564 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
2565 if (test_opt(sb, ABORT)) {
2566 err = -EROFS;
2567 goto restore_opts;
2568 }
2569
2570 if (*flags & MS_RDONLY) {
2571 err = dquot_suspend(sb, -1);
2572 if (err < 0)
2573 goto restore_opts;
2574
2575 /*
2576 * First of all, the unconditional stuff we have to do
2577 * to disable replay of the journal when we next remount
2578 */
2579 sb->s_flags |= MS_RDONLY;
2580
2581 /*
2582 * OK, test if we are remounting a valid rw partition
2583 * readonly, and if so set the rdonly flag and then
2584 * mark the partition as valid again.
2585 */
2586 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2587 (sbi->s_mount_state & EXT3_VALID_FS))
2588 es->s_state = cpu_to_le16(sbi->s_mount_state);
2589
2590 ext3_mark_recovery_complete(sb, es);
2591 } else {
2592 __le32 ret;
2593 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2594 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
2595 ext3_msg(sb, KERN_WARNING,
2596 "warning: couldn't remount RDWR "
2597 "because of unsupported optional "
2598 "features (%x)", le32_to_cpu(ret));
2599 err = -EROFS;
2600 goto restore_opts;
2601 }
2602
2603 /*
2604 * If we have an unprocessed orphan list hanging
2605 * around from a previously readonly bdev mount,
2606 * require a full umount/remount for now.
2607 */
2608 if (es->s_last_orphan) {
2609 ext3_msg(sb, KERN_WARNING, "warning: couldn't "
2610 "remount RDWR because of unprocessed "
2611 "orphan inode list. Please "
2612 "umount/remount instead.");
2613 err = -EINVAL;
2614 goto restore_opts;
2615 }
2616
2617 /*
2618 * Mounting a RDONLY partition read-write, so reread
2619 * and store the current valid flag. (It may have
2620 * been changed by e2fsck since we originally mounted
2621 * the partition.)
2622 */
2623 ext3_clear_journal_err(sb, es);
2624 sbi->s_mount_state = le16_to_cpu(es->s_state);
2625 if ((err = ext3_group_extend(sb, es, n_blocks_count)))
2626 goto restore_opts;
2627 if (!ext3_setup_super (sb, es, 0))
2628 sb->s_flags &= ~MS_RDONLY;
2629 enable_quota = 1;
2630 }
2631 }
2632 #ifdef CONFIG_QUOTA
2633 /* Release old quota file names */
2634 for (i = 0; i < MAXQUOTAS; i++)
2635 if (old_opts.s_qf_names[i] &&
2636 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2637 kfree(old_opts.s_qf_names[i]);
2638 #endif
2639 unlock_super(sb);
2640
2641 if (enable_quota)
2642 dquot_resume(sb, -1);
2643 return 0;
2644 restore_opts:
2645 sb->s_flags = old_sb_flags;
2646 sbi->s_mount_opt = old_opts.s_mount_opt;
2647 sbi->s_resuid = old_opts.s_resuid;
2648 sbi->s_resgid = old_opts.s_resgid;
2649 sbi->s_commit_interval = old_opts.s_commit_interval;
2650 #ifdef CONFIG_QUOTA
2651 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2652 for (i = 0; i < MAXQUOTAS; i++) {
2653 if (sbi->s_qf_names[i] &&
2654 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2655 kfree(sbi->s_qf_names[i]);
2656 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2657 }
2658 #endif
2659 unlock_super(sb);
2660 return err;
2661 }
2662
2663 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf)
2664 {
2665 struct super_block *sb = dentry->d_sb;
2666 struct ext3_sb_info *sbi = EXT3_SB(sb);
2667 struct ext3_super_block *es = sbi->s_es;
2668 u64 fsid;
2669
2670 if (test_opt(sb, MINIX_DF)) {
2671 sbi->s_overhead_last = 0;
2672 } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
2673 unsigned long ngroups = sbi->s_groups_count, i;
2674 ext3_fsblk_t overhead = 0;
2675 smp_rmb();
2676
2677 /*
2678 * Compute the overhead (FS structures). This is constant
2679 * for a given filesystem unless the number of block groups
2680 * changes so we cache the previous value until it does.
2681 */
2682
2683 /*
2684 * All of the blocks before first_data_block are
2685 * overhead
2686 */
2687 overhead = le32_to_cpu(es->s_first_data_block);
2688
2689 /*
2690 * Add the overhead attributed to the superblock and
2691 * block group descriptors. If the sparse superblocks
2692 * feature is turned on, then not all groups have this.
2693 */
2694 for (i = 0; i < ngroups; i++) {
2695 overhead += ext3_bg_has_super(sb, i) +
2696 ext3_bg_num_gdb(sb, i);
2697 cond_resched();
2698 }
2699
2700 /*
2701 * Every block group has an inode bitmap, a block
2702 * bitmap, and an inode table.
2703 */
2704 overhead += ngroups * (2 + sbi->s_itb_per_group);
2705 sbi->s_overhead_last = overhead;
2706 smp_wmb();
2707 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
2708 }
2709
2710 buf->f_type = EXT3_SUPER_MAGIC;
2711 buf->f_bsize = sb->s_blocksize;
2712 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
2713 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
2714 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2715 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2716 buf->f_bavail = 0;
2717 buf->f_files = le32_to_cpu(es->s_inodes_count);
2718 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
2719 buf->f_namelen = EXT3_NAME_LEN;
2720 fsid = le64_to_cpup((void *)es->s_uuid) ^
2721 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2722 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2723 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
2724 return 0;
2725 }
2726
2727 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2728 * is locked for write. Otherwise the are possible deadlocks:
2729 * Process 1 Process 2
2730 * ext3_create() quota_sync()
2731 * journal_start() write_dquot()
2732 * dquot_initialize() down(dqio_mutex)
2733 * down(dqio_mutex) journal_start()
2734 *
2735 */
2736
2737 #ifdef CONFIG_QUOTA
2738
2739 static inline struct inode *dquot_to_inode(struct dquot *dquot)
2740 {
2741 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2742 }
2743
2744 static int ext3_write_dquot(struct dquot *dquot)
2745 {
2746 int ret, err;
2747 handle_t *handle;
2748 struct inode *inode;
2749
2750 inode = dquot_to_inode(dquot);
2751 handle = ext3_journal_start(inode,
2752 EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
2753 if (IS_ERR(handle))
2754 return PTR_ERR(handle);
2755 ret = dquot_commit(dquot);
2756 err = ext3_journal_stop(handle);
2757 if (!ret)
2758 ret = err;
2759 return ret;
2760 }
2761
2762 static int ext3_acquire_dquot(struct dquot *dquot)
2763 {
2764 int ret, err;
2765 handle_t *handle;
2766
2767 handle = ext3_journal_start(dquot_to_inode(dquot),
2768 EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
2769 if (IS_ERR(handle))
2770 return PTR_ERR(handle);
2771 ret = dquot_acquire(dquot);
2772 err = ext3_journal_stop(handle);
2773 if (!ret)
2774 ret = err;
2775 return ret;
2776 }
2777
2778 static int ext3_release_dquot(struct dquot *dquot)
2779 {
2780 int ret, err;
2781 handle_t *handle;
2782
2783 handle = ext3_journal_start(dquot_to_inode(dquot),
2784 EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
2785 if (IS_ERR(handle)) {
2786 /* Release dquot anyway to avoid endless cycle in dqput() */
2787 dquot_release(dquot);
2788 return PTR_ERR(handle);
2789 }
2790 ret = dquot_release(dquot);
2791 err = ext3_journal_stop(handle);
2792 if (!ret)
2793 ret = err;
2794 return ret;
2795 }
2796
2797 static int ext3_mark_dquot_dirty(struct dquot *dquot)
2798 {
2799 /* Are we journaling quotas? */
2800 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2801 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2802 dquot_mark_dquot_dirty(dquot);
2803 return ext3_write_dquot(dquot);
2804 } else {
2805 return dquot_mark_dquot_dirty(dquot);
2806 }
2807 }
2808
2809 static int ext3_write_info(struct super_block *sb, int type)
2810 {
2811 int ret, err;
2812 handle_t *handle;
2813
2814 /* Data block + inode block */
2815 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2816 if (IS_ERR(handle))
2817 return PTR_ERR(handle);
2818 ret = dquot_commit_info(sb, type);
2819 err = ext3_journal_stop(handle);
2820 if (!ret)
2821 ret = err;
2822 return ret;
2823 }
2824
2825 /*
2826 * Turn on quotas during mount time - we need to find
2827 * the quota file and such...
2828 */
2829 static int ext3_quota_on_mount(struct super_block *sb, int type)
2830 {
2831 return dquot_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2832 EXT3_SB(sb)->s_jquota_fmt, type);
2833 }
2834
2835 /*
2836 * Standard function to be called on quota_on
2837 */
2838 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
2839 char *name)
2840 {
2841 int err;
2842 struct path path;
2843
2844 if (!test_opt(sb, QUOTA))
2845 return -EINVAL;
2846
2847 err = kern_path(name, LOOKUP_FOLLOW, &path);
2848 if (err)
2849 return err;
2850
2851 /* Quotafile not on the same filesystem? */
2852 if (path.mnt->mnt_sb != sb) {
2853 path_put(&path);
2854 return -EXDEV;
2855 }
2856 /* Journaling quota? */
2857 if (EXT3_SB(sb)->s_qf_names[type]) {
2858 /* Quotafile not of fs root? */
2859 if (path.dentry->d_parent != sb->s_root)
2860 ext3_msg(sb, KERN_WARNING,
2861 "warning: Quota file not on filesystem root. "
2862 "Journaled quota will not work.");
2863 }
2864
2865 /*
2866 * When we journal data on quota file, we have to flush journal to see
2867 * all updates to the file when we bypass pagecache...
2868 */
2869 if (ext3_should_journal_data(path.dentry->d_inode)) {
2870 /*
2871 * We don't need to lock updates but journal_flush() could
2872 * otherwise be livelocked...
2873 */
2874 journal_lock_updates(EXT3_SB(sb)->s_journal);
2875 err = journal_flush(EXT3_SB(sb)->s_journal);
2876 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2877 if (err) {
2878 path_put(&path);
2879 return err;
2880 }
2881 }
2882
2883 err = dquot_quota_on_path(sb, type, format_id, &path);
2884 path_put(&path);
2885 return err;
2886 }
2887
2888 /* Read data from quotafile - avoid pagecache and such because we cannot afford
2889 * acquiring the locks... As quota files are never truncated and quota code
2890 * itself serializes the operations (and noone else should touch the files)
2891 * we don't have to be afraid of races */
2892 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
2893 size_t len, loff_t off)
2894 {
2895 struct inode *inode = sb_dqopt(sb)->files[type];
2896 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2897 int err = 0;
2898 int offset = off & (sb->s_blocksize - 1);
2899 int tocopy;
2900 size_t toread;
2901 struct buffer_head *bh;
2902 loff_t i_size = i_size_read(inode);
2903
2904 if (off > i_size)
2905 return 0;
2906 if (off+len > i_size)
2907 len = i_size-off;
2908 toread = len;
2909 while (toread > 0) {
2910 tocopy = sb->s_blocksize - offset < toread ?
2911 sb->s_blocksize - offset : toread;
2912 bh = ext3_bread(NULL, inode, blk, 0, &err);
2913 if (err)
2914 return err;
2915 if (!bh) /* A hole? */
2916 memset(data, 0, tocopy);
2917 else
2918 memcpy(data, bh->b_data+offset, tocopy);
2919 brelse(bh);
2920 offset = 0;
2921 toread -= tocopy;
2922 data += tocopy;
2923 blk++;
2924 }
2925 return len;
2926 }
2927
2928 /* Write to quotafile (we know the transaction is already started and has
2929 * enough credits) */
2930 static ssize_t ext3_quota_write(struct super_block *sb, int type,
2931 const char *data, size_t len, loff_t off)
2932 {
2933 struct inode *inode = sb_dqopt(sb)->files[type];
2934 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2935 int err = 0;
2936 int offset = off & (sb->s_blocksize - 1);
2937 int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
2938 struct buffer_head *bh;
2939 handle_t *handle = journal_current_handle();
2940
2941 if (!handle) {
2942 ext3_msg(sb, KERN_WARNING,
2943 "warning: quota write (off=%llu, len=%llu)"
2944 " cancelled because transaction is not started.",
2945 (unsigned long long)off, (unsigned long long)len);
2946 return -EIO;
2947 }
2948
2949 /*
2950 * Since we account only one data block in transaction credits,
2951 * then it is impossible to cross a block boundary.
2952 */
2953 if (sb->s_blocksize - offset < len) {
2954 ext3_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
2955 " cancelled because not block aligned",
2956 (unsigned long long)off, (unsigned long long)len);
2957 return -EIO;
2958 }
2959 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
2960 bh = ext3_bread(handle, inode, blk, 1, &err);
2961 if (!bh)
2962 goto out;
2963 if (journal_quota) {
2964 err = ext3_journal_get_write_access(handle, bh);
2965 if (err) {
2966 brelse(bh);
2967 goto out;
2968 }
2969 }
2970 lock_buffer(bh);
2971 memcpy(bh->b_data+offset, data, len);
2972 flush_dcache_page(bh->b_page);
2973 unlock_buffer(bh);
2974 if (journal_quota)
2975 err = ext3_journal_dirty_metadata(handle, bh);
2976 else {
2977 /* Always do at least ordered writes for quotas */
2978 err = ext3_journal_dirty_data(handle, bh);
2979 mark_buffer_dirty(bh);
2980 }
2981 brelse(bh);
2982 out:
2983 if (err) {
2984 mutex_unlock(&inode->i_mutex);
2985 return err;
2986 }
2987 if (inode->i_size < off + len) {
2988 i_size_write(inode, off + len);
2989 EXT3_I(inode)->i_disksize = inode->i_size;
2990 }
2991 inode->i_version++;
2992 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2993 ext3_mark_inode_dirty(handle, inode);
2994 mutex_unlock(&inode->i_mutex);
2995 return len;
2996 }
2997
2998 #endif
2999
3000 static int ext3_get_sb(struct file_system_type *fs_type,
3001 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3002 {
3003 return get_sb_bdev(fs_type, flags, dev_name, data, ext3_fill_super, mnt);
3004 }
3005
3006 static struct file_system_type ext3_fs_type = {
3007 .owner = THIS_MODULE,
3008 .name = "ext3",
3009 .get_sb = ext3_get_sb,
3010 .kill_sb = kill_block_super,
3011 .fs_flags = FS_REQUIRES_DEV,
3012 };
3013
3014 static int __init init_ext3_fs(void)
3015 {
3016 int err = init_ext3_xattr();
3017 if (err)
3018 return err;
3019 err = init_inodecache();
3020 if (err)
3021 goto out1;
3022 err = register_filesystem(&ext3_fs_type);
3023 if (err)
3024 goto out;
3025 return 0;
3026 out:
3027 destroy_inodecache();
3028 out1:
3029 exit_ext3_xattr();
3030 return err;
3031 }
3032
3033 static void __exit exit_ext3_fs(void)
3034 {
3035 unregister_filesystem(&ext3_fs_type);
3036 destroy_inodecache();
3037 exit_ext3_xattr();
3038 }
3039
3040 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3041 MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
3042 MODULE_LICENSE("GPL");
3043 module_init(init_ext3_fs)
3044 module_exit(exit_ext3_fs)
Linux 2.6 ibm-acpi/thinkpad-acpi driver git tree