?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Linux 2.6.30-rc7
[linux-2.6/mini2440.git] / fs / ext4 / super.c
blob2958f4e6f222a8956f595145b1fe31e16e4eec9d
1 /*
2 * linux/fs/ext4/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/jbd2.h>
24 #include <linux/slab.h>
25 #include <linux/init.h>
26 #include <linux/blkdev.h>
27 #include <linux/parser.h>
28 #include <linux/smp_lock.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/proc_fs.h>
38 #include <linux/ctype.h>
39 #include <linux/marker.h>
40 #include <linux/log2.h>
41 #include <linux/crc16.h>
42 #include <asm/uaccess.h>
43
44 #include "ext4.h"
45 #include "ext4_jbd2.h"
46 #include "xattr.h"
47 #include "acl.h"
48 #include "namei.h"
49 #include "group.h"
50
51 struct proc_dir_entry *ext4_proc_root;
52 static struct kset *ext4_kset;
53
54 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
55 unsigned long journal_devnum);
56 static int ext4_commit_super(struct super_block *sb,
57 struct ext4_super_block *es, int sync);
58 static void ext4_mark_recovery_complete(struct super_block *sb,
59 struct ext4_super_block *es);
60 static void ext4_clear_journal_err(struct super_block *sb,
61 struct ext4_super_block *es);
62 static int ext4_sync_fs(struct super_block *sb, int wait);
63 static const char *ext4_decode_error(struct super_block *sb, int errno,
64 char nbuf[16]);
65 static int ext4_remount(struct super_block *sb, int *flags, char *data);
66 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
67 static int ext4_unfreeze(struct super_block *sb);
68 static void ext4_write_super(struct super_block *sb);
69 static int ext4_freeze(struct super_block *sb);
70
71
72 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
73 struct ext4_group_desc *bg)
74 {
75 return le32_to_cpu(bg->bg_block_bitmap_lo) |
76 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
77 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
78 }
79
80 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
81 struct ext4_group_desc *bg)
82 {
83 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
84 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
85 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
86 }
87
88 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
89 struct ext4_group_desc *bg)
90 {
91 return le32_to_cpu(bg->bg_inode_table_lo) |
92 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
93 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
94 }
95
96 __u32 ext4_free_blks_count(struct super_block *sb,
97 struct ext4_group_desc *bg)
98 {
99 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
100 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
101 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
102 }
103
104 __u32 ext4_free_inodes_count(struct super_block *sb,
105 struct ext4_group_desc *bg)
106 {
107 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
108 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
109 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
110 }
111
112 __u32 ext4_used_dirs_count(struct super_block *sb,
113 struct ext4_group_desc *bg)
114 {
115 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
116 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
117 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
118 }
119
120 __u32 ext4_itable_unused_count(struct super_block *sb,
121 struct ext4_group_desc *bg)
122 {
123 return le16_to_cpu(bg->bg_itable_unused_lo) |
124 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
125 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
126 }
127
128 void ext4_block_bitmap_set(struct super_block *sb,
129 struct ext4_group_desc *bg, ext4_fsblk_t blk)
130 {
131 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
132 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
133 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
134 }
135
136 void ext4_inode_bitmap_set(struct super_block *sb,
137 struct ext4_group_desc *bg, ext4_fsblk_t blk)
138 {
139 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
140 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
141 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
142 }
143
144 void ext4_inode_table_set(struct super_block *sb,
145 struct ext4_group_desc *bg, ext4_fsblk_t blk)
146 {
147 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
148 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
149 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
150 }
151
152 void ext4_free_blks_set(struct super_block *sb,
153 struct ext4_group_desc *bg, __u32 count)
154 {
155 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
156 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
157 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
158 }
159
160 void ext4_free_inodes_set(struct super_block *sb,
161 struct ext4_group_desc *bg, __u32 count)
162 {
163 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
164 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
165 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
166 }
167
168 void ext4_used_dirs_set(struct super_block *sb,
169 struct ext4_group_desc *bg, __u32 count)
170 {
171 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
172 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
173 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
174 }
175
176 void ext4_itable_unused_set(struct super_block *sb,
177 struct ext4_group_desc *bg, __u32 count)
178 {
179 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
180 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
181 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
182 }
183
184 /*
185 * Wrappers for jbd2_journal_start/end.
186 *
187 * The only special thing we need to do here is to make sure that all
188 * journal_end calls result in the superblock being marked dirty, so
189 * that sync() will call the filesystem's write_super callback if
190 * appropriate.
191 */
192 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
193 {
194 journal_t *journal;
195
196 if (sb->s_flags & MS_RDONLY)
197 return ERR_PTR(-EROFS);
198
199 /* Special case here: if the journal has aborted behind our
200 * backs (eg. EIO in the commit thread), then we still need to
201 * take the FS itself readonly cleanly. */
202 journal = EXT4_SB(sb)->s_journal;
203 if (journal) {
204 if (is_journal_aborted(journal)) {
205 ext4_abort(sb, __func__,
206 "Detected aborted journal");
207 return ERR_PTR(-EROFS);
208 }
209 return jbd2_journal_start(journal, nblocks);
210 }
211 /*
212 * We're not journaling, return the appropriate indication.
213 */
214 current->journal_info = EXT4_NOJOURNAL_HANDLE;
215 return current->journal_info;
216 }
217
218 /*
219 * The only special thing we need to do here is to make sure that all
220 * jbd2_journal_stop calls result in the superblock being marked dirty, so
221 * that sync() will call the filesystem's write_super callback if
222 * appropriate.
223 */
224 int __ext4_journal_stop(const char *where, handle_t *handle)
225 {
226 struct super_block *sb;
227 int err;
228 int rc;
229
230 if (!ext4_handle_valid(handle)) {
231 /*
232 * Do this here since we don't call jbd2_journal_stop() in
233 * no-journal mode.
234 */
235 current->journal_info = NULL;
236 return 0;
237 }
238 sb = handle->h_transaction->t_journal->j_private;
239 err = handle->h_err;
240 rc = jbd2_journal_stop(handle);
241
242 if (!err)
243 err = rc;
244 if (err)
245 __ext4_std_error(sb, where, err);
246 return err;
247 }
248
249 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
250 struct buffer_head *bh, handle_t *handle, int err)
251 {
252 char nbuf[16];
253 const char *errstr = ext4_decode_error(NULL, err, nbuf);
254
255 BUG_ON(!ext4_handle_valid(handle));
256
257 if (bh)
258 BUFFER_TRACE(bh, "abort");
259
260 if (!handle->h_err)
261 handle->h_err = err;
262
263 if (is_handle_aborted(handle))
264 return;
265
266 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
267 caller, errstr, err_fn);
268
269 jbd2_journal_abort_handle(handle);
270 }
271
272 /* Deal with the reporting of failure conditions on a filesystem such as
273 * inconsistencies detected or read IO failures.
274 *
275 * On ext2, we can store the error state of the filesystem in the
276 * superblock. That is not possible on ext4, because we may have other
277 * write ordering constraints on the superblock which prevent us from
278 * writing it out straight away; and given that the journal is about to
279 * be aborted, we can't rely on the current, or future, transactions to
280 * write out the superblock safely.
281 *
282 * We'll just use the jbd2_journal_abort() error code to record an error in
283 * the journal instead. On recovery, the journal will compain about
284 * that error until we've noted it down and cleared it.
285 */
286
287 static void ext4_handle_error(struct super_block *sb)
288 {
289 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
290
291 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
292 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
293
294 if (sb->s_flags & MS_RDONLY)
295 return;
296
297 if (!test_opt(sb, ERRORS_CONT)) {
298 journal_t *journal = EXT4_SB(sb)->s_journal;
299
300 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
301 if (journal)
302 jbd2_journal_abort(journal, -EIO);
303 }
304 if (test_opt(sb, ERRORS_RO)) {
305 printk(KERN_CRIT "Remounting filesystem read-only\n");
306 sb->s_flags |= MS_RDONLY;
307 }
308 ext4_commit_super(sb, es, 1);
309 if (test_opt(sb, ERRORS_PANIC))
310 panic("EXT4-fs (device %s): panic forced after error\n",
311 sb->s_id);
312 }
313
314 void ext4_error(struct super_block *sb, const char *function,
315 const char *fmt, ...)
316 {
317 va_list args;
318
319 va_start(args, fmt);
320 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
321 vprintk(fmt, args);
322 printk("\n");
323 va_end(args);
324
325 ext4_handle_error(sb);
326 }
327
328 static const char *ext4_decode_error(struct super_block *sb, int errno,
329 char nbuf[16])
330 {
331 char *errstr = NULL;
332
333 switch (errno) {
334 case -EIO:
335 errstr = "IO failure";
336 break;
337 case -ENOMEM:
338 errstr = "Out of memory";
339 break;
340 case -EROFS:
341 if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
342 errstr = "Journal has aborted";
343 else
344 errstr = "Readonly filesystem";
345 break;
346 default:
347 /* If the caller passed in an extra buffer for unknown
348 * errors, textualise them now. Else we just return
349 * NULL. */
350 if (nbuf) {
351 /* Check for truncated error codes... */
352 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
353 errstr = nbuf;
354 }
355 break;
356 }
357
358 return errstr;
359 }
360
361 /* __ext4_std_error decodes expected errors from journaling functions
362 * automatically and invokes the appropriate error response. */
363
364 void __ext4_std_error(struct super_block *sb, const char *function, int errno)
365 {
366 char nbuf[16];
367 const char *errstr;
368
369 /* Special case: if the error is EROFS, and we're not already
370 * inside a transaction, then there's really no point in logging
371 * an error. */
372 if (errno == -EROFS && journal_current_handle() == NULL &&
373 (sb->s_flags & MS_RDONLY))
374 return;
375
376 errstr = ext4_decode_error(sb, errno, nbuf);
377 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
378 sb->s_id, function, errstr);
379
380 ext4_handle_error(sb);
381 }
382
383 /*
384 * ext4_abort is a much stronger failure handler than ext4_error. The
385 * abort function may be used to deal with unrecoverable failures such
386 * as journal IO errors or ENOMEM at a critical moment in log management.
387 *
388 * We unconditionally force the filesystem into an ABORT|READONLY state,
389 * unless the error response on the fs has been set to panic in which
390 * case we take the easy way out and panic immediately.
391 */
392
393 void ext4_abort(struct super_block *sb, const char *function,
394 const char *fmt, ...)
395 {
396 va_list args;
397
398 printk(KERN_CRIT "ext4_abort called.\n");
399
400 va_start(args, fmt);
401 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
402 vprintk(fmt, args);
403 printk("\n");
404 va_end(args);
405
406 if (test_opt(sb, ERRORS_PANIC))
407 panic("EXT4-fs panic from previous error\n");
408
409 if (sb->s_flags & MS_RDONLY)
410 return;
411
412 printk(KERN_CRIT "Remounting filesystem read-only\n");
413 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
414 sb->s_flags |= MS_RDONLY;
415 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
416 if (EXT4_SB(sb)->s_journal)
417 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
418 }
419
420 void ext4_warning(struct super_block *sb, const char *function,
421 const char *fmt, ...)
422 {
423 va_list args;
424
425 va_start(args, fmt);
426 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
427 sb->s_id, function);
428 vprintk(fmt, args);
429 printk("\n");
430 va_end(args);
431 }
432
433 void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp,
434 const char *function, const char *fmt, ...)
435 __releases(bitlock)
436 __acquires(bitlock)
437 {
438 va_list args;
439 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
440
441 va_start(args, fmt);
442 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
443 vprintk(fmt, args);
444 printk("\n");
445 va_end(args);
446
447 if (test_opt(sb, ERRORS_CONT)) {
448 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
449 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
450 ext4_commit_super(sb, es, 0);
451 return;
452 }
453 ext4_unlock_group(sb, grp);
454 ext4_handle_error(sb);
455 /*
456 * We only get here in the ERRORS_RO case; relocking the group
457 * may be dangerous, but nothing bad will happen since the
458 * filesystem will have already been marked read/only and the
459 * journal has been aborted. We return 1 as a hint to callers
460 * who might what to use the return value from
461 * ext4_grp_locked_error() to distinguish beween the
462 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
463 * aggressively from the ext4 function in question, with a
464 * more appropriate error code.
465 */
466 ext4_lock_group(sb, grp);
467 return;
468 }
469
470
471 void ext4_update_dynamic_rev(struct super_block *sb)
472 {
473 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
474
475 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
476 return;
477
478 ext4_warning(sb, __func__,
479 "updating to rev %d because of new feature flag, "
480 "running e2fsck is recommended",
481 EXT4_DYNAMIC_REV);
482
483 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
484 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
485 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
486 /* leave es->s_feature_*compat flags alone */
487 /* es->s_uuid will be set by e2fsck if empty */
488
489 /*
490 * The rest of the superblock fields should be zero, and if not it
491 * means they are likely already in use, so leave them alone. We
492 * can leave it up to e2fsck to clean up any inconsistencies there.
493 */
494 }
495
496 /*
497 * Open the external journal device
498 */
499 static struct block_device *ext4_blkdev_get(dev_t dev)
500 {
501 struct block_device *bdev;
502 char b[BDEVNAME_SIZE];
503
504 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
505 if (IS_ERR(bdev))
506 goto fail;
507 return bdev;
508
509 fail:
510 printk(KERN_ERR "EXT4-fs: failed to open journal device %s: %ld\n",
511 __bdevname(dev, b), PTR_ERR(bdev));
512 return NULL;
513 }
514
515 /*
516 * Release the journal device
517 */
518 static int ext4_blkdev_put(struct block_device *bdev)
519 {
520 bd_release(bdev);
521 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
522 }
523
524 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
525 {
526 struct block_device *bdev;
527 int ret = -ENODEV;
528
529 bdev = sbi->journal_bdev;
530 if (bdev) {
531 ret = ext4_blkdev_put(bdev);
532 sbi->journal_bdev = NULL;
533 }
534 return ret;
535 }
536
537 static inline struct inode *orphan_list_entry(struct list_head *l)
538 {
539 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
540 }
541
542 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
543 {
544 struct list_head *l;
545
546 printk(KERN_ERR "sb orphan head is %d\n",
547 le32_to_cpu(sbi->s_es->s_last_orphan));
548
549 printk(KERN_ERR "sb_info orphan list:\n");
550 list_for_each(l, &sbi->s_orphan) {
551 struct inode *inode = orphan_list_entry(l);
552 printk(KERN_ERR " "
553 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
554 inode->i_sb->s_id, inode->i_ino, inode,
555 inode->i_mode, inode->i_nlink,
556 NEXT_ORPHAN(inode));
557 }
558 }
559
560 static void ext4_put_super(struct super_block *sb)
561 {
562 struct ext4_sb_info *sbi = EXT4_SB(sb);
563 struct ext4_super_block *es = sbi->s_es;
564 int i, err;
565
566 ext4_mb_release(sb);
567 ext4_ext_release(sb);
568 ext4_xattr_put_super(sb);
569 if (sbi->s_journal) {
570 err = jbd2_journal_destroy(sbi->s_journal);
571 sbi->s_journal = NULL;
572 if (err < 0)
573 ext4_abort(sb, __func__,
574 "Couldn't clean up the journal");
575 }
576 if (!(sb->s_flags & MS_RDONLY)) {
577 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
578 es->s_state = cpu_to_le16(sbi->s_mount_state);
579 ext4_commit_super(sb, es, 1);
580 }
581 if (sbi->s_proc) {
582 remove_proc_entry(sb->s_id, ext4_proc_root);
583 }
584 kobject_del(&sbi->s_kobj);
585
586 for (i = 0; i < sbi->s_gdb_count; i++)
587 brelse(sbi->s_group_desc[i]);
588 kfree(sbi->s_group_desc);
589 kfree(sbi->s_flex_groups);
590 percpu_counter_destroy(&sbi->s_freeblocks_counter);
591 percpu_counter_destroy(&sbi->s_freeinodes_counter);
592 percpu_counter_destroy(&sbi->s_dirs_counter);
593 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
594 brelse(sbi->s_sbh);
595 #ifdef CONFIG_QUOTA
596 for (i = 0; i < MAXQUOTAS; i++)
597 kfree(sbi->s_qf_names[i]);
598 #endif
599
600 /* Debugging code just in case the in-memory inode orphan list
601 * isn't empty. The on-disk one can be non-empty if we've
602 * detected an error and taken the fs readonly, but the
603 * in-memory list had better be clean by this point. */
604 if (!list_empty(&sbi->s_orphan))
605 dump_orphan_list(sb, sbi);
606 J_ASSERT(list_empty(&sbi->s_orphan));
607
608 invalidate_bdev(sb->s_bdev);
609 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
610 /*
611 * Invalidate the journal device's buffers. We don't want them
612 * floating about in memory - the physical journal device may
613 * hotswapped, and it breaks the `ro-after' testing code.
614 */
615 sync_blockdev(sbi->journal_bdev);
616 invalidate_bdev(sbi->journal_bdev);
617 ext4_blkdev_remove(sbi);
618 }
619 sb->s_fs_info = NULL;
620 /*
621 * Now that we are completely done shutting down the
622 * superblock, we need to actually destroy the kobject.
623 */
624 unlock_kernel();
625 unlock_super(sb);
626 kobject_put(&sbi->s_kobj);
627 wait_for_completion(&sbi->s_kobj_unregister);
628 lock_super(sb);
629 lock_kernel();
630 kfree(sbi->s_blockgroup_lock);
631 kfree(sbi);
632 return;
633 }
634
635 static struct kmem_cache *ext4_inode_cachep;
636
637 /*
638 * Called inside transaction, so use GFP_NOFS
639 */
640 static struct inode *ext4_alloc_inode(struct super_block *sb)
641 {
642 struct ext4_inode_info *ei;
643
644 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
645 if (!ei)
646 return NULL;
647 #ifdef CONFIG_EXT4_FS_POSIX_ACL
648 ei->i_acl = EXT4_ACL_NOT_CACHED;
649 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
650 #endif
651 ei->vfs_inode.i_version = 1;
652 ei->vfs_inode.i_data.writeback_index = 0;
653 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
654 INIT_LIST_HEAD(&ei->i_prealloc_list);
655 spin_lock_init(&ei->i_prealloc_lock);
656 /*
657 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
658 * therefore it can be null here. Don't check it, just initialize
659 * jinode.
660 */
661 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
662 ei->i_reserved_data_blocks = 0;
663 ei->i_reserved_meta_blocks = 0;
664 ei->i_allocated_meta_blocks = 0;
665 ei->i_delalloc_reserved_flag = 0;
666 spin_lock_init(&(ei->i_block_reservation_lock));
667 return &ei->vfs_inode;
668 }
669
670 static void ext4_destroy_inode(struct inode *inode)
671 {
672 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
673 printk("EXT4 Inode %p: orphan list check failed!\n",
674 EXT4_I(inode));
675 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
676 EXT4_I(inode), sizeof(struct ext4_inode_info),
677 true);
678 dump_stack();
679 }
680 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
681 }
682
683 static void init_once(void *foo)
684 {
685 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
686
687 INIT_LIST_HEAD(&ei->i_orphan);
688 #ifdef CONFIG_EXT4_FS_XATTR
689 init_rwsem(&ei->xattr_sem);
690 #endif
691 init_rwsem(&ei->i_data_sem);
692 inode_init_once(&ei->vfs_inode);
693 }
694
695 static int init_inodecache(void)
696 {
697 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
698 sizeof(struct ext4_inode_info),
699 0, (SLAB_RECLAIM_ACCOUNT|
700 SLAB_MEM_SPREAD),
701 init_once);
702 if (ext4_inode_cachep == NULL)
703 return -ENOMEM;
704 return 0;
705 }
706
707 static void destroy_inodecache(void)
708 {
709 kmem_cache_destroy(ext4_inode_cachep);
710 }
711
712 static void ext4_clear_inode(struct inode *inode)
713 {
714 #ifdef CONFIG_EXT4_FS_POSIX_ACL
715 if (EXT4_I(inode)->i_acl &&
716 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
717 posix_acl_release(EXT4_I(inode)->i_acl);
718 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
719 }
720 if (EXT4_I(inode)->i_default_acl &&
721 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
722 posix_acl_release(EXT4_I(inode)->i_default_acl);
723 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
724 }
725 #endif
726 ext4_discard_preallocations(inode);
727 if (EXT4_JOURNAL(inode))
728 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
729 &EXT4_I(inode)->jinode);
730 }
731
732 static inline void ext4_show_quota_options(struct seq_file *seq,
733 struct super_block *sb)
734 {
735 #if defined(CONFIG_QUOTA)
736 struct ext4_sb_info *sbi = EXT4_SB(sb);
737
738 if (sbi->s_jquota_fmt)
739 seq_printf(seq, ",jqfmt=%s",
740 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
741
742 if (sbi->s_qf_names[USRQUOTA])
743 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
744
745 if (sbi->s_qf_names[GRPQUOTA])
746 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
747
748 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
749 seq_puts(seq, ",usrquota");
750
751 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
752 seq_puts(seq, ",grpquota");
753 #endif
754 }
755
756 /*
757 * Show an option if
758 * - it's set to a non-default value OR
759 * - if the per-sb default is different from the global default
760 */
761 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
762 {
763 int def_errors;
764 unsigned long def_mount_opts;
765 struct super_block *sb = vfs->mnt_sb;
766 struct ext4_sb_info *sbi = EXT4_SB(sb);
767 struct ext4_super_block *es = sbi->s_es;
768
769 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
770 def_errors = le16_to_cpu(es->s_errors);
771
772 if (sbi->s_sb_block != 1)
773 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
774 if (test_opt(sb, MINIX_DF))
775 seq_puts(seq, ",minixdf");
776 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
777 seq_puts(seq, ",grpid");
778 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
779 seq_puts(seq, ",nogrpid");
780 if (sbi->s_resuid != EXT4_DEF_RESUID ||
781 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
782 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
783 }
784 if (sbi->s_resgid != EXT4_DEF_RESGID ||
785 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
786 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
787 }
788 if (test_opt(sb, ERRORS_RO)) {
789 if (def_errors == EXT4_ERRORS_PANIC ||
790 def_errors == EXT4_ERRORS_CONTINUE) {
791 seq_puts(seq, ",errors=remount-ro");
792 }
793 }
794 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
795 seq_puts(seq, ",errors=continue");
796 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
797 seq_puts(seq, ",errors=panic");
798 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
799 seq_puts(seq, ",nouid32");
800 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
801 seq_puts(seq, ",debug");
802 if (test_opt(sb, OLDALLOC))
803 seq_puts(seq, ",oldalloc");
804 #ifdef CONFIG_EXT4_FS_XATTR
805 if (test_opt(sb, XATTR_USER) &&
806 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
807 seq_puts(seq, ",user_xattr");
808 if (!test_opt(sb, XATTR_USER) &&
809 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
810 seq_puts(seq, ",nouser_xattr");
811 }
812 #endif
813 #ifdef CONFIG_EXT4_FS_POSIX_ACL
814 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
815 seq_puts(seq, ",acl");
816 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
817 seq_puts(seq, ",noacl");
818 #endif
819 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
820 seq_printf(seq, ",commit=%u",
821 (unsigned) (sbi->s_commit_interval / HZ));
822 }
823 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
824 seq_printf(seq, ",min_batch_time=%u",
825 (unsigned) sbi->s_min_batch_time);
826 }
827 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
828 seq_printf(seq, ",max_batch_time=%u",
829 (unsigned) sbi->s_min_batch_time);
830 }
831
832 /*
833 * We're changing the default of barrier mount option, so
834 * let's always display its mount state so it's clear what its
835 * status is.
836 */
837 seq_puts(seq, ",barrier=");
838 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
839 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
840 seq_puts(seq, ",journal_async_commit");
841 if (test_opt(sb, NOBH))
842 seq_puts(seq, ",nobh");
843 if (test_opt(sb, I_VERSION))
844 seq_puts(seq, ",i_version");
845 if (!test_opt(sb, DELALLOC))
846 seq_puts(seq, ",nodelalloc");
847
848
849 if (sbi->s_stripe)
850 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
851 /*
852 * journal mode get enabled in different ways
853 * So just print the value even if we didn't specify it
854 */
855 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
856 seq_puts(seq, ",data=journal");
857 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
858 seq_puts(seq, ",data=ordered");
859 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
860 seq_puts(seq, ",data=writeback");
861
862 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
863 seq_printf(seq, ",inode_readahead_blks=%u",
864 sbi->s_inode_readahead_blks);
865
866 if (test_opt(sb, DATA_ERR_ABORT))
867 seq_puts(seq, ",data_err=abort");
868
869 if (test_opt(sb, NO_AUTO_DA_ALLOC))
870 seq_puts(seq, ",noauto_da_alloc");
871
872 ext4_show_quota_options(seq, sb);
873 return 0;
874 }
875
876
877 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
878 u64 ino, u32 generation)
879 {
880 struct inode *inode;
881
882 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
883 return ERR_PTR(-ESTALE);
884 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
885 return ERR_PTR(-ESTALE);
886
887 /* iget isn't really right if the inode is currently unallocated!!
888 *
889 * ext4_read_inode will return a bad_inode if the inode had been
890 * deleted, so we should be safe.
891 *
892 * Currently we don't know the generation for parent directory, so
893 * a generation of 0 means "accept any"
894 */
895 inode = ext4_iget(sb, ino);
896 if (IS_ERR(inode))
897 return ERR_CAST(inode);
898 if (generation && inode->i_generation != generation) {
899 iput(inode);
900 return ERR_PTR(-ESTALE);
901 }
902
903 return inode;
904 }
905
906 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
907 int fh_len, int fh_type)
908 {
909 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
910 ext4_nfs_get_inode);
911 }
912
913 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
914 int fh_len, int fh_type)
915 {
916 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
917 ext4_nfs_get_inode);
918 }
919
920 /*
921 * Try to release metadata pages (indirect blocks, directories) which are
922 * mapped via the block device. Since these pages could have journal heads
923 * which would prevent try_to_free_buffers() from freeing them, we must use
924 * jbd2 layer's try_to_free_buffers() function to release them.
925 */
926 static int bdev_try_to_free_page(struct super_block *sb, struct page *page, gfp_t wait)
927 {
928 journal_t *journal = EXT4_SB(sb)->s_journal;
929
930 WARN_ON(PageChecked(page));
931 if (!page_has_buffers(page))
932 return 0;
933 if (journal)
934 return jbd2_journal_try_to_free_buffers(journal, page,
935 wait & ~__GFP_WAIT);
936 return try_to_free_buffers(page);
937 }
938
939 #ifdef CONFIG_QUOTA
940 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
941 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
942
943 static int ext4_write_dquot(struct dquot *dquot);
944 static int ext4_acquire_dquot(struct dquot *dquot);
945 static int ext4_release_dquot(struct dquot *dquot);
946 static int ext4_mark_dquot_dirty(struct dquot *dquot);
947 static int ext4_write_info(struct super_block *sb, int type);
948 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
949 char *path, int remount);
950 static int ext4_quota_on_mount(struct super_block *sb, int type);
951 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
952 size_t len, loff_t off);
953 static ssize_t ext4_quota_write(struct super_block *sb, int type,
954 const char *data, size_t len, loff_t off);
955
956 static struct dquot_operations ext4_quota_operations = {
957 .initialize = dquot_initialize,
958 .drop = dquot_drop,
959 .alloc_space = dquot_alloc_space,
960 .reserve_space = dquot_reserve_space,
961 .claim_space = dquot_claim_space,
962 .release_rsv = dquot_release_reserved_space,
963 .get_reserved_space = ext4_get_reserved_space,
964 .alloc_inode = dquot_alloc_inode,
965 .free_space = dquot_free_space,
966 .free_inode = dquot_free_inode,
967 .transfer = dquot_transfer,
968 .write_dquot = ext4_write_dquot,
969 .acquire_dquot = ext4_acquire_dquot,
970 .release_dquot = ext4_release_dquot,
971 .mark_dirty = ext4_mark_dquot_dirty,
972 .write_info = ext4_write_info,
973 .alloc_dquot = dquot_alloc,
974 .destroy_dquot = dquot_destroy,
975 };
976
977 static struct quotactl_ops ext4_qctl_operations = {
978 .quota_on = ext4_quota_on,
979 .quota_off = vfs_quota_off,
980 .quota_sync = vfs_quota_sync,
981 .get_info = vfs_get_dqinfo,
982 .set_info = vfs_set_dqinfo,
983 .get_dqblk = vfs_get_dqblk,
984 .set_dqblk = vfs_set_dqblk
985 };
986 #endif
987
988 static const struct super_operations ext4_sops = {
989 .alloc_inode = ext4_alloc_inode,
990 .destroy_inode = ext4_destroy_inode,
991 .write_inode = ext4_write_inode,
992 .dirty_inode = ext4_dirty_inode,
993 .delete_inode = ext4_delete_inode,
994 .put_super = ext4_put_super,
995 .write_super = ext4_write_super,
996 .sync_fs = ext4_sync_fs,
997 .freeze_fs = ext4_freeze,
998 .unfreeze_fs = ext4_unfreeze,
999 .statfs = ext4_statfs,
1000 .remount_fs = ext4_remount,
1001 .clear_inode = ext4_clear_inode,
1002 .show_options = ext4_show_options,
1003 #ifdef CONFIG_QUOTA
1004 .quota_read = ext4_quota_read,
1005 .quota_write = ext4_quota_write,
1006 #endif
1007 .bdev_try_to_free_page = bdev_try_to_free_page,
1008 };
1009
1010 static const struct export_operations ext4_export_ops = {
1011 .fh_to_dentry = ext4_fh_to_dentry,
1012 .fh_to_parent = ext4_fh_to_parent,
1013 .get_parent = ext4_get_parent,
1014 };
1015
1016 enum {
1017 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1018 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1019 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1020 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1021 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1022 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1023 Opt_journal_update, Opt_journal_dev,
1024 Opt_journal_checksum, Opt_journal_async_commit,
1025 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1026 Opt_data_err_abort, Opt_data_err_ignore,
1027 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1028 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
1029 Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err, Opt_resize,
1030 Opt_usrquota, Opt_grpquota, Opt_i_version,
1031 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1032 Opt_inode_readahead_blks, Opt_journal_ioprio
1033 };
1034
1035 static const match_table_t tokens = {
1036 {Opt_bsd_df, "bsddf"},
1037 {Opt_minix_df, "minixdf"},
1038 {Opt_grpid, "grpid"},
1039 {Opt_grpid, "bsdgroups"},
1040 {Opt_nogrpid, "nogrpid"},
1041 {Opt_nogrpid, "sysvgroups"},
1042 {Opt_resgid, "resgid=%u"},
1043 {Opt_resuid, "resuid=%u"},
1044 {Opt_sb, "sb=%u"},
1045 {Opt_err_cont, "errors=continue"},
1046 {Opt_err_panic, "errors=panic"},
1047 {Opt_err_ro, "errors=remount-ro"},
1048 {Opt_nouid32, "nouid32"},
1049 {Opt_debug, "debug"},
1050 {Opt_oldalloc, "oldalloc"},
1051 {Opt_orlov, "orlov"},
1052 {Opt_user_xattr, "user_xattr"},
1053 {Opt_nouser_xattr, "nouser_xattr"},
1054 {Opt_acl, "acl"},
1055 {Opt_noacl, "noacl"},
1056 {Opt_noload, "noload"},
1057 {Opt_nobh, "nobh"},
1058 {Opt_bh, "bh"},
1059 {Opt_commit, "commit=%u"},
1060 {Opt_min_batch_time, "min_batch_time=%u"},
1061 {Opt_max_batch_time, "max_batch_time=%u"},
1062 {Opt_journal_update, "journal=update"},
1063 {Opt_journal_dev, "journal_dev=%u"},
1064 {Opt_journal_checksum, "journal_checksum"},
1065 {Opt_journal_async_commit, "journal_async_commit"},
1066 {Opt_abort, "abort"},
1067 {Opt_data_journal, "data=journal"},
1068 {Opt_data_ordered, "data=ordered"},
1069 {Opt_data_writeback, "data=writeback"},
1070 {Opt_data_err_abort, "data_err=abort"},
1071 {Opt_data_err_ignore, "data_err=ignore"},
1072 {Opt_offusrjquota, "usrjquota="},
1073 {Opt_usrjquota, "usrjquota=%s"},
1074 {Opt_offgrpjquota, "grpjquota="},
1075 {Opt_grpjquota, "grpjquota=%s"},
1076 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1077 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1078 {Opt_grpquota, "grpquota"},
1079 {Opt_noquota, "noquota"},
1080 {Opt_quota, "quota"},
1081 {Opt_usrquota, "usrquota"},
1082 {Opt_barrier, "barrier=%u"},
1083 {Opt_barrier, "barrier"},
1084 {Opt_nobarrier, "nobarrier"},
1085 {Opt_i_version, "i_version"},
1086 {Opt_stripe, "stripe=%u"},
1087 {Opt_resize, "resize"},
1088 {Opt_delalloc, "delalloc"},
1089 {Opt_nodelalloc, "nodelalloc"},
1090 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1091 {Opt_journal_ioprio, "journal_ioprio=%u"},
1092 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1093 {Opt_auto_da_alloc, "auto_da_alloc"},
1094 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1095 {Opt_err, NULL},
1096 };
1097
1098 static ext4_fsblk_t get_sb_block(void **data)
1099 {
1100 ext4_fsblk_t sb_block;
1101 char *options = (char *) *data;
1102
1103 if (!options || strncmp(options, "sb=", 3) != 0)
1104 return 1; /* Default location */
1105 options += 3;
1106 /*todo: use simple_strtoll with >32bit ext4 */
1107 sb_block = simple_strtoul(options, &options, 0);
1108 if (*options && *options != ',') {
1109 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1110 (char *) *data);
1111 return 1;
1112 }
1113 if (*options == ',')
1114 options++;
1115 *data = (void *) options;
1116 return sb_block;
1117 }
1118
1119 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1120
1121 static int parse_options(char *options, struct super_block *sb,
1122 unsigned long *journal_devnum,
1123 unsigned int *journal_ioprio,
1124 ext4_fsblk_t *n_blocks_count, int is_remount)
1125 {
1126 struct ext4_sb_info *sbi = EXT4_SB(sb);
1127 char *p;
1128 substring_t args[MAX_OPT_ARGS];
1129 int data_opt = 0;
1130 int option;
1131 #ifdef CONFIG_QUOTA
1132 int qtype, qfmt;
1133 char *qname;
1134 #endif
1135
1136 if (!options)
1137 return 1;
1138
1139 while ((p = strsep(&options, ",")) != NULL) {
1140 int token;
1141 if (!*p)
1142 continue;
1143
1144 token = match_token(p, tokens, args);
1145 switch (token) {
1146 case Opt_bsd_df:
1147 clear_opt(sbi->s_mount_opt, MINIX_DF);
1148 break;
1149 case Opt_minix_df:
1150 set_opt(sbi->s_mount_opt, MINIX_DF);
1151 break;
1152 case Opt_grpid:
1153 set_opt(sbi->s_mount_opt, GRPID);
1154 break;
1155 case Opt_nogrpid:
1156 clear_opt(sbi->s_mount_opt, GRPID);
1157 break;
1158 case Opt_resuid:
1159 if (match_int(&args[0], &option))
1160 return 0;
1161 sbi->s_resuid = option;
1162 break;
1163 case Opt_resgid:
1164 if (match_int(&args[0], &option))
1165 return 0;
1166 sbi->s_resgid = option;
1167 break;
1168 case Opt_sb:
1169 /* handled by get_sb_block() instead of here */
1170 /* *sb_block = match_int(&args[0]); */
1171 break;
1172 case Opt_err_panic:
1173 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1174 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1175 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1176 break;
1177 case Opt_err_ro:
1178 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1179 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1180 set_opt(sbi->s_mount_opt, ERRORS_RO);
1181 break;
1182 case Opt_err_cont:
1183 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1184 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1185 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1186 break;
1187 case Opt_nouid32:
1188 set_opt(sbi->s_mount_opt, NO_UID32);
1189 break;
1190 case Opt_debug:
1191 set_opt(sbi->s_mount_opt, DEBUG);
1192 break;
1193 case Opt_oldalloc:
1194 set_opt(sbi->s_mount_opt, OLDALLOC);
1195 break;
1196 case Opt_orlov:
1197 clear_opt(sbi->s_mount_opt, OLDALLOC);
1198 break;
1199 #ifdef CONFIG_EXT4_FS_XATTR
1200 case Opt_user_xattr:
1201 set_opt(sbi->s_mount_opt, XATTR_USER);
1202 break;
1203 case Opt_nouser_xattr:
1204 clear_opt(sbi->s_mount_opt, XATTR_USER);
1205 break;
1206 #else
1207 case Opt_user_xattr:
1208 case Opt_nouser_xattr:
1209 printk(KERN_ERR "EXT4 (no)user_xattr options "
1210 "not supported\n");
1211 break;
1212 #endif
1213 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1214 case Opt_acl:
1215 set_opt(sbi->s_mount_opt, POSIX_ACL);
1216 break;
1217 case Opt_noacl:
1218 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1219 break;
1220 #else
1221 case Opt_acl:
1222 case Opt_noacl:
1223 printk(KERN_ERR "EXT4 (no)acl options "
1224 "not supported\n");
1225 break;
1226 #endif
1227 case Opt_journal_update:
1228 /* @@@ FIXME */
1229 /* Eventually we will want to be able to create
1230 a journal file here. For now, only allow the
1231 user to specify an existing inode to be the
1232 journal file. */
1233 if (is_remount) {
1234 printk(KERN_ERR "EXT4-fs: cannot specify "
1235 "journal on remount\n");
1236 return 0;
1237 }
1238 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1239 break;
1240 case Opt_journal_dev:
1241 if (is_remount) {
1242 printk(KERN_ERR "EXT4-fs: cannot specify "
1243 "journal on remount\n");
1244 return 0;
1245 }
1246 if (match_int(&args[0], &option))
1247 return 0;
1248 *journal_devnum = option;
1249 break;
1250 case Opt_journal_checksum:
1251 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1252 break;
1253 case Opt_journal_async_commit:
1254 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1255 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1256 break;
1257 case Opt_noload:
1258 set_opt(sbi->s_mount_opt, NOLOAD);
1259 break;
1260 case Opt_commit:
1261 if (match_int(&args[0], &option))
1262 return 0;
1263 if (option < 0)
1264 return 0;
1265 if (option == 0)
1266 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1267 sbi->s_commit_interval = HZ * option;
1268 break;
1269 case Opt_max_batch_time:
1270 if (match_int(&args[0], &option))
1271 return 0;
1272 if (option < 0)
1273 return 0;
1274 if (option == 0)
1275 option = EXT4_DEF_MAX_BATCH_TIME;
1276 sbi->s_max_batch_time = option;
1277 break;
1278 case Opt_min_batch_time:
1279 if (match_int(&args[0], &option))
1280 return 0;
1281 if (option < 0)
1282 return 0;
1283 sbi->s_min_batch_time = option;
1284 break;
1285 case Opt_data_journal:
1286 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1287 goto datacheck;
1288 case Opt_data_ordered:
1289 data_opt = EXT4_MOUNT_ORDERED_DATA;
1290 goto datacheck;
1291 case Opt_data_writeback:
1292 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1293 datacheck:
1294 if (is_remount) {
1295 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1296 != data_opt) {
1297 printk(KERN_ERR
1298 "EXT4-fs: cannot change data "
1299 "mode on remount\n");
1300 return 0;
1301 }
1302 } else {
1303 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1304 sbi->s_mount_opt |= data_opt;
1305 }
1306 break;
1307 case Opt_data_err_abort:
1308 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1309 break;
1310 case Opt_data_err_ignore:
1311 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1312 break;
1313 #ifdef CONFIG_QUOTA
1314 case Opt_usrjquota:
1315 qtype = USRQUOTA;
1316 goto set_qf_name;
1317 case Opt_grpjquota:
1318 qtype = GRPQUOTA;
1319 set_qf_name:
1320 if (sb_any_quota_loaded(sb) &&
1321 !sbi->s_qf_names[qtype]) {
1322 printk(KERN_ERR
1323 "EXT4-fs: Cannot change journaled "
1324 "quota options when quota turned on.\n");
1325 return 0;
1326 }
1327 qname = match_strdup(&args[0]);
1328 if (!qname) {
1329 printk(KERN_ERR
1330 "EXT4-fs: not enough memory for "
1331 "storing quotafile name.\n");
1332 return 0;
1333 }
1334 if (sbi->s_qf_names[qtype] &&
1335 strcmp(sbi->s_qf_names[qtype], qname)) {
1336 printk(KERN_ERR
1337 "EXT4-fs: %s quota file already "
1338 "specified.\n", QTYPE2NAME(qtype));
1339 kfree(qname);
1340 return 0;
1341 }
1342 sbi->s_qf_names[qtype] = qname;
1343 if (strchr(sbi->s_qf_names[qtype], '/')) {
1344 printk(KERN_ERR
1345 "EXT4-fs: quotafile must be on "
1346 "filesystem root.\n");
1347 kfree(sbi->s_qf_names[qtype]);
1348 sbi->s_qf_names[qtype] = NULL;
1349 return 0;
1350 }
1351 set_opt(sbi->s_mount_opt, QUOTA);
1352 break;
1353 case Opt_offusrjquota:
1354 qtype = USRQUOTA;
1355 goto clear_qf_name;
1356 case Opt_offgrpjquota:
1357 qtype = GRPQUOTA;
1358 clear_qf_name:
1359 if (sb_any_quota_loaded(sb) &&
1360 sbi->s_qf_names[qtype]) {
1361 printk(KERN_ERR "EXT4-fs: Cannot change "
1362 "journaled quota options when "
1363 "quota turned on.\n");
1364 return 0;
1365 }
1366 /*
1367 * The space will be released later when all options
1368 * are confirmed to be correct
1369 */
1370 sbi->s_qf_names[qtype] = NULL;
1371 break;
1372 case Opt_jqfmt_vfsold:
1373 qfmt = QFMT_VFS_OLD;
1374 goto set_qf_format;
1375 case Opt_jqfmt_vfsv0:
1376 qfmt = QFMT_VFS_V0;
1377 set_qf_format:
1378 if (sb_any_quota_loaded(sb) &&
1379 sbi->s_jquota_fmt != qfmt) {
1380 printk(KERN_ERR "EXT4-fs: Cannot change "
1381 "journaled quota options when "
1382 "quota turned on.\n");
1383 return 0;
1384 }
1385 sbi->s_jquota_fmt = qfmt;
1386 break;
1387 case Opt_quota:
1388 case Opt_usrquota:
1389 set_opt(sbi->s_mount_opt, QUOTA);
1390 set_opt(sbi->s_mount_opt, USRQUOTA);
1391 break;
1392 case Opt_grpquota:
1393 set_opt(sbi->s_mount_opt, QUOTA);
1394 set_opt(sbi->s_mount_opt, GRPQUOTA);
1395 break;
1396 case Opt_noquota:
1397 if (sb_any_quota_loaded(sb)) {
1398 printk(KERN_ERR "EXT4-fs: Cannot change quota "
1399 "options when quota turned on.\n");
1400 return 0;
1401 }
1402 clear_opt(sbi->s_mount_opt, QUOTA);
1403 clear_opt(sbi->s_mount_opt, USRQUOTA);
1404 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1405 break;
1406 #else
1407 case Opt_quota:
1408 case Opt_usrquota:
1409 case Opt_grpquota:
1410 printk(KERN_ERR
1411 "EXT4-fs: quota options not supported.\n");
1412 break;
1413 case Opt_usrjquota:
1414 case Opt_grpjquota:
1415 case Opt_offusrjquota:
1416 case Opt_offgrpjquota:
1417 case Opt_jqfmt_vfsold:
1418 case Opt_jqfmt_vfsv0:
1419 printk(KERN_ERR
1420 "EXT4-fs: journaled quota options not "
1421 "supported.\n");
1422 break;
1423 case Opt_noquota:
1424 break;
1425 #endif
1426 case Opt_abort:
1427 set_opt(sbi->s_mount_opt, ABORT);
1428 break;
1429 case Opt_nobarrier:
1430 clear_opt(sbi->s_mount_opt, BARRIER);
1431 break;
1432 case Opt_barrier:
1433 if (match_int(&args[0], &option)) {
1434 set_opt(sbi->s_mount_opt, BARRIER);
1435 break;
1436 }
1437 if (option)
1438 set_opt(sbi->s_mount_opt, BARRIER);
1439 else
1440 clear_opt(sbi->s_mount_opt, BARRIER);
1441 break;
1442 case Opt_ignore:
1443 break;
1444 case Opt_resize:
1445 if (!is_remount) {
1446 printk("EXT4-fs: resize option only available "
1447 "for remount\n");
1448 return 0;
1449 }
1450 if (match_int(&args[0], &option) != 0)
1451 return 0;
1452 *n_blocks_count = option;
1453 break;
1454 case Opt_nobh:
1455 set_opt(sbi->s_mount_opt, NOBH);
1456 break;
1457 case Opt_bh:
1458 clear_opt(sbi->s_mount_opt, NOBH);
1459 break;
1460 case Opt_i_version:
1461 set_opt(sbi->s_mount_opt, I_VERSION);
1462 sb->s_flags |= MS_I_VERSION;
1463 break;
1464 case Opt_nodelalloc:
1465 clear_opt(sbi->s_mount_opt, DELALLOC);
1466 break;
1467 case Opt_stripe:
1468 if (match_int(&args[0], &option))
1469 return 0;
1470 if (option < 0)
1471 return 0;
1472 sbi->s_stripe = option;
1473 break;
1474 case Opt_delalloc:
1475 set_opt(sbi->s_mount_opt, DELALLOC);
1476 break;
1477 case Opt_inode_readahead_blks:
1478 if (match_int(&args[0], &option))
1479 return 0;
1480 if (option < 0 || option > (1 << 30))
1481 return 0;
1482 if (option & (option - 1)) {
1483 printk(KERN_ERR "EXT4-fs: inode_readahead_blks"
1484 " must be a power of 2\n");
1485 return 0;
1486 }
1487 sbi->s_inode_readahead_blks = option;
1488 break;
1489 case Opt_journal_ioprio:
1490 if (match_int(&args[0], &option))
1491 return 0;
1492 if (option < 0 || option > 7)
1493 break;
1494 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1495 option);
1496 break;
1497 case Opt_noauto_da_alloc:
1498 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1499 break;
1500 case Opt_auto_da_alloc:
1501 if (match_int(&args[0], &option)) {
1502 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1503 break;
1504 }
1505 if (option)
1506 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1507 else
1508 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1509 break;
1510 default:
1511 printk(KERN_ERR
1512 "EXT4-fs: Unrecognized mount option \"%s\" "
1513 "or missing value\n", p);
1514 return 0;
1515 }
1516 }
1517 #ifdef CONFIG_QUOTA
1518 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1519 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1520 sbi->s_qf_names[USRQUOTA])
1521 clear_opt(sbi->s_mount_opt, USRQUOTA);
1522
1523 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1524 sbi->s_qf_names[GRPQUOTA])
1525 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1526
1527 if ((sbi->s_qf_names[USRQUOTA] &&
1528 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1529 (sbi->s_qf_names[GRPQUOTA] &&
1530 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1531 printk(KERN_ERR "EXT4-fs: old and new quota "
1532 "format mixing.\n");
1533 return 0;
1534 }
1535
1536 if (!sbi->s_jquota_fmt) {
1537 printk(KERN_ERR "EXT4-fs: journaled quota format "
1538 "not specified.\n");
1539 return 0;
1540 }
1541 } else {
1542 if (sbi->s_jquota_fmt) {
1543 printk(KERN_ERR "EXT4-fs: journaled quota format "
1544 "specified with no journaling "
1545 "enabled.\n");
1546 return 0;
1547 }
1548 }
1549 #endif
1550 return 1;
1551 }
1552
1553 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1554 int read_only)
1555 {
1556 struct ext4_sb_info *sbi = EXT4_SB(sb);
1557 int res = 0;
1558
1559 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1560 printk(KERN_ERR "EXT4-fs warning: revision level too high, "
1561 "forcing read-only mode\n");
1562 res = MS_RDONLY;
1563 }
1564 if (read_only)
1565 return res;
1566 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1567 printk(KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1568 "running e2fsck is recommended\n");
1569 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1570 printk(KERN_WARNING
1571 "EXT4-fs warning: mounting fs with errors, "
1572 "running e2fsck is recommended\n");
1573 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1574 le16_to_cpu(es->s_mnt_count) >=
1575 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1576 printk(KERN_WARNING
1577 "EXT4-fs warning: maximal mount count reached, "
1578 "running e2fsck is recommended\n");
1579 else if (le32_to_cpu(es->s_checkinterval) &&
1580 (le32_to_cpu(es->s_lastcheck) +
1581 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1582 printk(KERN_WARNING
1583 "EXT4-fs warning: checktime reached, "
1584 "running e2fsck is recommended\n");
1585 if (!sbi->s_journal)
1586 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1587 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1588 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1589 le16_add_cpu(&es->s_mnt_count, 1);
1590 es->s_mtime = cpu_to_le32(get_seconds());
1591 ext4_update_dynamic_rev(sb);
1592 if (sbi->s_journal)
1593 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1594
1595 ext4_commit_super(sb, es, 1);
1596 if (test_opt(sb, DEBUG))
1597 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1598 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1599 sb->s_blocksize,
1600 sbi->s_groups_count,
1601 EXT4_BLOCKS_PER_GROUP(sb),
1602 EXT4_INODES_PER_GROUP(sb),
1603 sbi->s_mount_opt);
1604
1605 if (EXT4_SB(sb)->s_journal) {
1606 printk(KERN_INFO "EXT4 FS on %s, %s journal on %s\n",
1607 sb->s_id, EXT4_SB(sb)->s_journal->j_inode ? "internal" :
1608 "external", EXT4_SB(sb)->s_journal->j_devname);
1609 } else {
1610 printk(KERN_INFO "EXT4 FS on %s, no journal\n", sb->s_id);
1611 }
1612 return res;
1613 }
1614
1615 static int ext4_fill_flex_info(struct super_block *sb)
1616 {
1617 struct ext4_sb_info *sbi = EXT4_SB(sb);
1618 struct ext4_group_desc *gdp = NULL;
1619 struct buffer_head *bh;
1620 ext4_group_t flex_group_count;
1621 ext4_group_t flex_group;
1622 int groups_per_flex = 0;
1623 int i;
1624
1625 if (!sbi->s_es->s_log_groups_per_flex) {
1626 sbi->s_log_groups_per_flex = 0;
1627 return 1;
1628 }
1629
1630 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1631 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1632
1633 /* We allocate both existing and potentially added groups */
1634 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1635 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1636 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1637 sbi->s_flex_groups = kzalloc(flex_group_count *
1638 sizeof(struct flex_groups), GFP_KERNEL);
1639 if (sbi->s_flex_groups == NULL) {
1640 printk(KERN_ERR "EXT4-fs: not enough memory for "
1641 "%u flex groups\n", flex_group_count);
1642 goto failed;
1643 }
1644
1645 for (i = 0; i < sbi->s_groups_count; i++) {
1646 gdp = ext4_get_group_desc(sb, i, &bh);
1647
1648 flex_group = ext4_flex_group(sbi, i);
1649 atomic_set(&sbi->s_flex_groups[flex_group].free_inodes,
1650 ext4_free_inodes_count(sb, gdp));
1651 atomic_set(&sbi->s_flex_groups[flex_group].free_blocks,
1652 ext4_free_blks_count(sb, gdp));
1653 atomic_set(&sbi->s_flex_groups[flex_group].used_dirs,
1654 ext4_used_dirs_count(sb, gdp));
1655 }
1656
1657 return 1;
1658 failed:
1659 return 0;
1660 }
1661
1662 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1663 struct ext4_group_desc *gdp)
1664 {
1665 __u16 crc = 0;
1666
1667 if (sbi->s_es->s_feature_ro_compat &
1668 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1669 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1670 __le32 le_group = cpu_to_le32(block_group);
1671
1672 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1673 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1674 crc = crc16(crc, (__u8 *)gdp, offset);
1675 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1676 /* for checksum of struct ext4_group_desc do the rest...*/
1677 if ((sbi->s_es->s_feature_incompat &
1678 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1679 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1680 crc = crc16(crc, (__u8 *)gdp + offset,
1681 le16_to_cpu(sbi->s_es->s_desc_size) -
1682 offset);
1683 }
1684
1685 return cpu_to_le16(crc);
1686 }
1687
1688 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1689 struct ext4_group_desc *gdp)
1690 {
1691 if ((sbi->s_es->s_feature_ro_compat &
1692 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1693 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1694 return 0;
1695
1696 return 1;
1697 }
1698
1699 /* Called at mount-time, super-block is locked */
1700 static int ext4_check_descriptors(struct super_block *sb)
1701 {
1702 struct ext4_sb_info *sbi = EXT4_SB(sb);
1703 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1704 ext4_fsblk_t last_block;
1705 ext4_fsblk_t block_bitmap;
1706 ext4_fsblk_t inode_bitmap;
1707 ext4_fsblk_t inode_table;
1708 int flexbg_flag = 0;
1709 ext4_group_t i;
1710
1711 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1712 flexbg_flag = 1;
1713
1714 ext4_debug("Checking group descriptors");
1715
1716 for (i = 0; i < sbi->s_groups_count; i++) {
1717 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1718
1719 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1720 last_block = ext4_blocks_count(sbi->s_es) - 1;
1721 else
1722 last_block = first_block +
1723 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1724
1725 block_bitmap = ext4_block_bitmap(sb, gdp);
1726 if (block_bitmap < first_block || block_bitmap > last_block) {
1727 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1728 "Block bitmap for group %u not in group "
1729 "(block %llu)!\n", i, block_bitmap);
1730 return 0;
1731 }
1732 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1733 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1734 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1735 "Inode bitmap for group %u not in group "
1736 "(block %llu)!\n", i, inode_bitmap);
1737 return 0;
1738 }
1739 inode_table = ext4_inode_table(sb, gdp);
1740 if (inode_table < first_block ||
1741 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1742 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1743 "Inode table for group %u not in group "
1744 "(block %llu)!\n", i, inode_table);
1745 return 0;
1746 }
1747 spin_lock(sb_bgl_lock(sbi, i));
1748 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1749 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1750 "Checksum for group %u failed (%u!=%u)\n",
1751 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1752 gdp)), le16_to_cpu(gdp->bg_checksum));
1753 if (!(sb->s_flags & MS_RDONLY)) {
1754 spin_unlock(sb_bgl_lock(sbi, i));
1755 return 0;
1756 }
1757 }
1758 spin_unlock(sb_bgl_lock(sbi, i));
1759 if (!flexbg_flag)
1760 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1761 }
1762
1763 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1764 sbi->s_es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
1765 return 1;
1766 }
1767
1768 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1769 * the superblock) which were deleted from all directories, but held open by
1770 * a process at the time of a crash. We walk the list and try to delete these
1771 * inodes at recovery time (only with a read-write filesystem).
1772 *
1773 * In order to keep the orphan inode chain consistent during traversal (in
1774 * case of crash during recovery), we link each inode into the superblock
1775 * orphan list_head and handle it the same way as an inode deletion during
1776 * normal operation (which journals the operations for us).
1777 *
1778 * We only do an iget() and an iput() on each inode, which is very safe if we
1779 * accidentally point at an in-use or already deleted inode. The worst that
1780 * can happen in this case is that we get a "bit already cleared" message from
1781 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1782 * e2fsck was run on this filesystem, and it must have already done the orphan
1783 * inode cleanup for us, so we can safely abort without any further action.
1784 */
1785 static void ext4_orphan_cleanup(struct super_block *sb,
1786 struct ext4_super_block *es)
1787 {
1788 unsigned int s_flags = sb->s_flags;
1789 int nr_orphans = 0, nr_truncates = 0;
1790 #ifdef CONFIG_QUOTA
1791 int i;
1792 #endif
1793 if (!es->s_last_orphan) {
1794 jbd_debug(4, "no orphan inodes to clean up\n");
1795 return;
1796 }
1797
1798 if (bdev_read_only(sb->s_bdev)) {
1799 printk(KERN_ERR "EXT4-fs: write access "
1800 "unavailable, skipping orphan cleanup.\n");
1801 return;
1802 }
1803
1804 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1805 if (es->s_last_orphan)
1806 jbd_debug(1, "Errors on filesystem, "
1807 "clearing orphan list.\n");
1808 es->s_last_orphan = 0;
1809 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1810 return;
1811 }
1812
1813 if (s_flags & MS_RDONLY) {
1814 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
1815 sb->s_id);
1816 sb->s_flags &= ~MS_RDONLY;
1817 }
1818 #ifdef CONFIG_QUOTA
1819 /* Needed for iput() to work correctly and not trash data */
1820 sb->s_flags |= MS_ACTIVE;
1821 /* Turn on quotas so that they are updated correctly */
1822 for (i = 0; i < MAXQUOTAS; i++) {
1823 if (EXT4_SB(sb)->s_qf_names[i]) {
1824 int ret = ext4_quota_on_mount(sb, i);
1825 if (ret < 0)
1826 printk(KERN_ERR
1827 "EXT4-fs: Cannot turn on journaled "
1828 "quota: error %d\n", ret);
1829 }
1830 }
1831 #endif
1832
1833 while (es->s_last_orphan) {
1834 struct inode *inode;
1835
1836 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1837 if (IS_ERR(inode)) {
1838 es->s_last_orphan = 0;
1839 break;
1840 }
1841
1842 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1843 vfs_dq_init(inode);
1844 if (inode->i_nlink) {
1845 printk(KERN_DEBUG
1846 "%s: truncating inode %lu to %lld bytes\n",
1847 __func__, inode->i_ino, inode->i_size);
1848 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1849 inode->i_ino, inode->i_size);
1850 ext4_truncate(inode);
1851 nr_truncates++;
1852 } else {
1853 printk(KERN_DEBUG
1854 "%s: deleting unreferenced inode %lu\n",
1855 __func__, inode->i_ino);
1856 jbd_debug(2, "deleting unreferenced inode %lu\n",
1857 inode->i_ino);
1858 nr_orphans++;
1859 }
1860 iput(inode); /* The delete magic happens here! */
1861 }
1862
1863 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1864
1865 if (nr_orphans)
1866 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
1867 sb->s_id, PLURAL(nr_orphans));
1868 if (nr_truncates)
1869 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
1870 sb->s_id, PLURAL(nr_truncates));
1871 #ifdef CONFIG_QUOTA
1872 /* Turn quotas off */
1873 for (i = 0; i < MAXQUOTAS; i++) {
1874 if (sb_dqopt(sb)->files[i])
1875 vfs_quota_off(sb, i, 0);
1876 }
1877 #endif
1878 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1879 }
1880 /*
1881 * Maximal extent format file size.
1882 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1883 * extent format containers, within a sector_t, and within i_blocks
1884 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1885 * so that won't be a limiting factor.
1886 *
1887 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1888 */
1889 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1890 {
1891 loff_t res;
1892 loff_t upper_limit = MAX_LFS_FILESIZE;
1893
1894 /* small i_blocks in vfs inode? */
1895 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1896 /*
1897 * CONFIG_LBD is not enabled implies the inode
1898 * i_block represent total blocks in 512 bytes
1899 * 32 == size of vfs inode i_blocks * 8
1900 */
1901 upper_limit = (1LL << 32) - 1;
1902
1903 /* total blocks in file system block size */
1904 upper_limit >>= (blkbits - 9);
1905 upper_limit <<= blkbits;
1906 }
1907
1908 /* 32-bit extent-start container, ee_block */
1909 res = 1LL << 32;
1910 res <<= blkbits;
1911 res -= 1;
1912
1913 /* Sanity check against vm- & vfs- imposed limits */
1914 if (res > upper_limit)
1915 res = upper_limit;
1916
1917 return res;
1918 }
1919
1920 /*
1921 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1922 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1923 * We need to be 1 filesystem block less than the 2^48 sector limit.
1924 */
1925 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
1926 {
1927 loff_t res = EXT4_NDIR_BLOCKS;
1928 int meta_blocks;
1929 loff_t upper_limit;
1930 /* This is calculated to be the largest file size for a
1931 * dense, bitmapped file such that the total number of
1932 * sectors in the file, including data and all indirect blocks,
1933 * does not exceed 2^48 -1
1934 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1935 * total number of 512 bytes blocks of the file
1936 */
1937
1938 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1939 /*
1940 * !has_huge_files or CONFIG_LBD is not enabled
1941 * implies the inode i_block represent total blocks in
1942 * 512 bytes 32 == size of vfs inode i_blocks * 8
1943 */
1944 upper_limit = (1LL << 32) - 1;
1945
1946 /* total blocks in file system block size */
1947 upper_limit >>= (bits - 9);
1948
1949 } else {
1950 /*
1951 * We use 48 bit ext4_inode i_blocks
1952 * With EXT4_HUGE_FILE_FL set the i_blocks
1953 * represent total number of blocks in
1954 * file system block size
1955 */
1956 upper_limit = (1LL << 48) - 1;
1957
1958 }
1959
1960 /* indirect blocks */
1961 meta_blocks = 1;
1962 /* double indirect blocks */
1963 meta_blocks += 1 + (1LL << (bits-2));
1964 /* tripple indirect blocks */
1965 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1966
1967 upper_limit -= meta_blocks;
1968 upper_limit <<= bits;
1969
1970 res += 1LL << (bits-2);
1971 res += 1LL << (2*(bits-2));
1972 res += 1LL << (3*(bits-2));
1973 res <<= bits;
1974 if (res > upper_limit)
1975 res = upper_limit;
1976
1977 if (res > MAX_LFS_FILESIZE)
1978 res = MAX_LFS_FILESIZE;
1979
1980 return res;
1981 }
1982
1983 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
1984 ext4_fsblk_t logical_sb_block, int nr)
1985 {
1986 struct ext4_sb_info *sbi = EXT4_SB(sb);
1987 ext4_group_t bg, first_meta_bg;
1988 int has_super = 0;
1989
1990 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1991
1992 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
1993 nr < first_meta_bg)
1994 return logical_sb_block + nr + 1;
1995 bg = sbi->s_desc_per_block * nr;
1996 if (ext4_bg_has_super(sb, bg))
1997 has_super = 1;
1998 return (has_super + ext4_group_first_block_no(sb, bg));
1999 }
2000
2001 /**
2002 * ext4_get_stripe_size: Get the stripe size.
2003 * @sbi: In memory super block info
2004 *
2005 * If we have specified it via mount option, then
2006 * use the mount option value. If the value specified at mount time is
2007 * greater than the blocks per group use the super block value.
2008 * If the super block value is greater than blocks per group return 0.
2009 * Allocator needs it be less than blocks per group.
2010 *
2011 */
2012 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2013 {
2014 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2015 unsigned long stripe_width =
2016 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2017
2018 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2019 return sbi->s_stripe;
2020
2021 if (stripe_width <= sbi->s_blocks_per_group)
2022 return stripe_width;
2023
2024 if (stride <= sbi->s_blocks_per_group)
2025 return stride;
2026
2027 return 0;
2028 }
2029
2030 /* sysfs supprt */
2031
2032 struct ext4_attr {
2033 struct attribute attr;
2034 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2035 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2036 const char *, size_t);
2037 int offset;
2038 };
2039
2040 static int parse_strtoul(const char *buf,
2041 unsigned long max, unsigned long *value)
2042 {
2043 char *endp;
2044
2045 while (*buf && isspace(*buf))
2046 buf++;
2047 *value = simple_strtoul(buf, &endp, 0);
2048 while (*endp && isspace(*endp))
2049 endp++;
2050 if (*endp || *value > max)
2051 return -EINVAL;
2052
2053 return 0;
2054 }
2055
2056 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2057 struct ext4_sb_info *sbi,
2058 char *buf)
2059 {
2060 return snprintf(buf, PAGE_SIZE, "%llu\n",
2061 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2062 }
2063
2064 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2065 struct ext4_sb_info *sbi, char *buf)
2066 {
2067 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2068
2069 return snprintf(buf, PAGE_SIZE, "%lu\n",
2070 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2071 sbi->s_sectors_written_start) >> 1);
2072 }
2073
2074 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2075 struct ext4_sb_info *sbi, char *buf)
2076 {
2077 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2078
2079 return snprintf(buf, PAGE_SIZE, "%llu\n",
2080 sbi->s_kbytes_written +
2081 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2082 EXT4_SB(sb)->s_sectors_written_start) >> 1));
2083 }
2084
2085 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2086 struct ext4_sb_info *sbi,
2087 const char *buf, size_t count)
2088 {
2089 unsigned long t;
2090
2091 if (parse_strtoul(buf, 0x40000000, &t))
2092 return -EINVAL;
2093
2094 /* inode_readahead_blks must be a power of 2 */
2095 if (t & (t-1))
2096 return -EINVAL;
2097
2098 sbi->s_inode_readahead_blks = t;
2099 return count;
2100 }
2101
2102 static ssize_t sbi_ui_show(struct ext4_attr *a,
2103 struct ext4_sb_info *sbi, char *buf)
2104 {
2105 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2106
2107 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2108 }
2109
2110 static ssize_t sbi_ui_store(struct ext4_attr *a,
2111 struct ext4_sb_info *sbi,
2112 const char *buf, size_t count)
2113 {
2114 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2115 unsigned long t;
2116
2117 if (parse_strtoul(buf, 0xffffffff, &t))
2118 return -EINVAL;
2119 *ui = t;
2120 return count;
2121 }
2122
2123 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2124 static struct ext4_attr ext4_attr_##_name = { \
2125 .attr = {.name = __stringify(_name), .mode = _mode }, \
2126 .show = _show, \
2127 .store = _store, \
2128 .offset = offsetof(struct ext4_sb_info, _elname), \
2129 }
2130 #define EXT4_ATTR(name, mode, show, store) \
2131 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2132
2133 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2134 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2135 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2136 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2137 #define ATTR_LIST(name) &ext4_attr_##name.attr
2138
2139 EXT4_RO_ATTR(delayed_allocation_blocks);
2140 EXT4_RO_ATTR(session_write_kbytes);
2141 EXT4_RO_ATTR(lifetime_write_kbytes);
2142 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2143 inode_readahead_blks_store, s_inode_readahead_blks);
2144 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2145 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2146 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2147 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2148 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2149 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2150
2151 static struct attribute *ext4_attrs[] = {
2152 ATTR_LIST(delayed_allocation_blocks),
2153 ATTR_LIST(session_write_kbytes),
2154 ATTR_LIST(lifetime_write_kbytes),
2155 ATTR_LIST(inode_readahead_blks),
2156 ATTR_LIST(mb_stats),
2157 ATTR_LIST(mb_max_to_scan),
2158 ATTR_LIST(mb_min_to_scan),
2159 ATTR_LIST(mb_order2_req),
2160 ATTR_LIST(mb_stream_req),
2161 ATTR_LIST(mb_group_prealloc),
2162 NULL,
2163 };
2164
2165 static ssize_t ext4_attr_show(struct kobject *kobj,
2166 struct attribute *attr, char *buf)
2167 {
2168 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2169 s_kobj);
2170 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2171
2172 return a->show ? a->show(a, sbi, buf) : 0;
2173 }
2174
2175 static ssize_t ext4_attr_store(struct kobject *kobj,
2176 struct attribute *attr,
2177 const char *buf, size_t len)
2178 {
2179 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2180 s_kobj);
2181 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2182
2183 return a->store ? a->store(a, sbi, buf, len) : 0;
2184 }
2185
2186 static void ext4_sb_release(struct kobject *kobj)
2187 {
2188 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2189 s_kobj);
2190 complete(&sbi->s_kobj_unregister);
2191 }
2192
2193
2194 static struct sysfs_ops ext4_attr_ops = {
2195 .show = ext4_attr_show,
2196 .store = ext4_attr_store,
2197 };
2198
2199 static struct kobj_type ext4_ktype = {
2200 .default_attrs = ext4_attrs,
2201 .sysfs_ops = &ext4_attr_ops,
2202 .release = ext4_sb_release,
2203 };
2204
2205 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2206 __releases(kernel_lock)
2207 __acquires(kernel_lock)
2208
2209 {
2210 struct buffer_head *bh;
2211 struct ext4_super_block *es = NULL;
2212 struct ext4_sb_info *sbi;
2213 ext4_fsblk_t block;
2214 ext4_fsblk_t sb_block = get_sb_block(&data);
2215 ext4_fsblk_t logical_sb_block;
2216 unsigned long offset = 0;
2217 unsigned long journal_devnum = 0;
2218 unsigned long def_mount_opts;
2219 struct inode *root;
2220 char *cp;
2221 const char *descr;
2222 int ret = -EINVAL;
2223 int blocksize;
2224 unsigned int db_count;
2225 unsigned int i;
2226 int needs_recovery, has_huge_files;
2227 int features;
2228 __u64 blocks_count;
2229 int err;
2230 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2231
2232 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2233 if (!sbi)
2234 return -ENOMEM;
2235
2236 sbi->s_blockgroup_lock =
2237 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2238 if (!sbi->s_blockgroup_lock) {
2239 kfree(sbi);
2240 return -ENOMEM;
2241 }
2242 sb->s_fs_info = sbi;
2243 sbi->s_mount_opt = 0;
2244 sbi->s_resuid = EXT4_DEF_RESUID;
2245 sbi->s_resgid = EXT4_DEF_RESGID;
2246 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2247 sbi->s_sb_block = sb_block;
2248 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2249 sectors[1]);
2250
2251 unlock_kernel();
2252
2253 /* Cleanup superblock name */
2254 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2255 *cp = '!';
2256
2257 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2258 if (!blocksize) {
2259 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
2260 goto out_fail;
2261 }
2262
2263 /*
2264 * The ext4 superblock will not be buffer aligned for other than 1kB
2265 * block sizes. We need to calculate the offset from buffer start.
2266 */
2267 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2268 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2269 offset = do_div(logical_sb_block, blocksize);
2270 } else {
2271 logical_sb_block = sb_block;
2272 }
2273
2274 if (!(bh = sb_bread(sb, logical_sb_block))) {
2275 printk(KERN_ERR "EXT4-fs: unable to read superblock\n");
2276 goto out_fail;
2277 }
2278 /*
2279 * Note: s_es must be initialized as soon as possible because
2280 * some ext4 macro-instructions depend on its value
2281 */
2282 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2283 sbi->s_es = es;
2284 sb->s_magic = le16_to_cpu(es->s_magic);
2285 if (sb->s_magic != EXT4_SUPER_MAGIC)
2286 goto cantfind_ext4;
2287 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2288
2289 /* Set defaults before we parse the mount options */
2290 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2291 if (def_mount_opts & EXT4_DEFM_DEBUG)
2292 set_opt(sbi->s_mount_opt, DEBUG);
2293 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
2294 set_opt(sbi->s_mount_opt, GRPID);
2295 if (def_mount_opts & EXT4_DEFM_UID16)
2296 set_opt(sbi->s_mount_opt, NO_UID32);
2297 #ifdef CONFIG_EXT4_FS_XATTR
2298 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2299 set_opt(sbi->s_mount_opt, XATTR_USER);
2300 #endif
2301 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2302 if (def_mount_opts & EXT4_DEFM_ACL)
2303 set_opt(sbi->s_mount_opt, POSIX_ACL);
2304 #endif
2305 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2306 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2307 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2308 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2309 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2310 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2311
2312 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2313 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2314 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2315 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2316 else
2317 set_opt(sbi->s_mount_opt, ERRORS_RO);
2318
2319 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2320 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2321 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2322 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2323 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2324
2325 set_opt(sbi->s_mount_opt, BARRIER);
2326
2327 /*
2328 * enable delayed allocation by default
2329 * Use -o nodelalloc to turn it off
2330 */
2331 set_opt(sbi->s_mount_opt, DELALLOC);
2332
2333
2334 if (!parse_options((char *) data, sb, &journal_devnum,
2335 &journal_ioprio, NULL, 0))
2336 goto failed_mount;
2337
2338 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2339 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2340
2341 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2342 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2343 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2344 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2345 printk(KERN_WARNING
2346 "EXT4-fs warning: feature flags set on rev 0 fs, "
2347 "running e2fsck is recommended\n");
2348
2349 /*
2350 * Check feature flags regardless of the revision level, since we
2351 * previously didn't change the revision level when setting the flags,
2352 * so there is a chance incompat flags are set on a rev 0 filesystem.
2353 */
2354 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
2355 if (features) {
2356 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
2357 "unsupported optional features (%x).\n", sb->s_id,
2358 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2359 ~EXT4_FEATURE_INCOMPAT_SUPP));
2360 goto failed_mount;
2361 }
2362 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
2363 if (!(sb->s_flags & MS_RDONLY) && features) {
2364 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
2365 "unsupported optional features (%x).\n", sb->s_id,
2366 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2367 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2368 goto failed_mount;
2369 }
2370 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2371 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2372 if (has_huge_files) {
2373 /*
2374 * Large file size enabled file system can only be
2375 * mount if kernel is build with CONFIG_LBD
2376 */
2377 if (sizeof(root->i_blocks) < sizeof(u64) &&
2378 !(sb->s_flags & MS_RDONLY)) {
2379 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2380 "files cannot be mounted read-write "
2381 "without CONFIG_LBD.\n", sb->s_id);
2382 goto failed_mount;
2383 }
2384 }
2385 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2386
2387 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2388 blocksize > EXT4_MAX_BLOCK_SIZE) {
2389 printk(KERN_ERR
2390 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
2391 blocksize, sb->s_id);
2392 goto failed_mount;
2393 }
2394
2395 if (sb->s_blocksize != blocksize) {
2396
2397 /* Validate the filesystem blocksize */
2398 if (!sb_set_blocksize(sb, blocksize)) {
2399 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2400 blocksize);
2401 goto failed_mount;
2402 }
2403
2404 brelse(bh);
2405 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2406 offset = do_div(logical_sb_block, blocksize);
2407 bh = sb_bread(sb, logical_sb_block);
2408 if (!bh) {
2409 printk(KERN_ERR
2410 "EXT4-fs: Can't read superblock on 2nd try.\n");
2411 goto failed_mount;
2412 }
2413 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2414 sbi->s_es = es;
2415 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2416 printk(KERN_ERR
2417 "EXT4-fs: Magic mismatch, very weird !\n");
2418 goto failed_mount;
2419 }
2420 }
2421
2422 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2423 has_huge_files);
2424 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2425
2426 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2427 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2428 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2429 } else {
2430 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2431 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2432 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2433 (!is_power_of_2(sbi->s_inode_size)) ||
2434 (sbi->s_inode_size > blocksize)) {
2435 printk(KERN_ERR
2436 "EXT4-fs: unsupported inode size: %d\n",
2437 sbi->s_inode_size);
2438 goto failed_mount;
2439 }
2440 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2441 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2442 }
2443 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2444 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2445 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2446 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2447 !is_power_of_2(sbi->s_desc_size)) {
2448 printk(KERN_ERR
2449 "EXT4-fs: unsupported descriptor size %lu\n",
2450 sbi->s_desc_size);
2451 goto failed_mount;
2452 }
2453 } else
2454 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2455 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2456 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2457 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2458 goto cantfind_ext4;
2459 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2460 if (sbi->s_inodes_per_block == 0)
2461 goto cantfind_ext4;
2462 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2463 sbi->s_inodes_per_block;
2464 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2465 sbi->s_sbh = bh;
2466 sbi->s_mount_state = le16_to_cpu(es->s_state);
2467 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2468 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2469 for (i = 0; i < 4; i++)
2470 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2471 sbi->s_def_hash_version = es->s_def_hash_version;
2472 i = le32_to_cpu(es->s_flags);
2473 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2474 sbi->s_hash_unsigned = 3;
2475 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2476 #ifdef __CHAR_UNSIGNED__
2477 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2478 sbi->s_hash_unsigned = 3;
2479 #else
2480 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2481 #endif
2482 sb->s_dirt = 1;
2483 }
2484
2485 if (sbi->s_blocks_per_group > blocksize * 8) {
2486 printk(KERN_ERR
2487 "EXT4-fs: #blocks per group too big: %lu\n",
2488 sbi->s_blocks_per_group);
2489 goto failed_mount;
2490 }
2491 if (sbi->s_inodes_per_group > blocksize * 8) {
2492 printk(KERN_ERR
2493 "EXT4-fs: #inodes per group too big: %lu\n",
2494 sbi->s_inodes_per_group);
2495 goto failed_mount;
2496 }
2497
2498 if (ext4_blocks_count(es) >
2499 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
2500 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
2501 " too large to mount safely\n", sb->s_id);
2502 if (sizeof(sector_t) < 8)
2503 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
2504 "enabled\n");
2505 goto failed_mount;
2506 }
2507
2508 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2509 goto cantfind_ext4;
2510
2511 /* check blocks count against device size */
2512 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2513 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2514 printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu "
2515 "exceeds size of device (%llu blocks)\n",
2516 ext4_blocks_count(es), blocks_count);
2517 goto failed_mount;
2518 }
2519
2520 /*
2521 * It makes no sense for the first data block to be beyond the end
2522 * of the filesystem.
2523 */
2524 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2525 printk(KERN_WARNING "EXT4-fs: bad geometry: first data"
2526 "block %u is beyond end of filesystem (%llu)\n",
2527 le32_to_cpu(es->s_first_data_block),
2528 ext4_blocks_count(es));
2529 goto failed_mount;
2530 }
2531 blocks_count = (ext4_blocks_count(es) -
2532 le32_to_cpu(es->s_first_data_block) +
2533 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2534 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2535 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2536 printk(KERN_WARNING "EXT4-fs: groups count too large: %u "
2537 "(block count %llu, first data block %u, "
2538 "blocks per group %lu)\n", sbi->s_groups_count,
2539 ext4_blocks_count(es),
2540 le32_to_cpu(es->s_first_data_block),
2541 EXT4_BLOCKS_PER_GROUP(sb));
2542 goto failed_mount;
2543 }
2544 sbi->s_groups_count = blocks_count;
2545 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2546 EXT4_DESC_PER_BLOCK(sb);
2547 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2548 GFP_KERNEL);
2549 if (sbi->s_group_desc == NULL) {
2550 printk(KERN_ERR "EXT4-fs: not enough memory\n");
2551 goto failed_mount;
2552 }
2553
2554 #ifdef CONFIG_PROC_FS
2555 if (ext4_proc_root)
2556 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2557 #endif
2558
2559 bgl_lock_init(sbi->s_blockgroup_lock);
2560
2561 for (i = 0; i < db_count; i++) {
2562 block = descriptor_loc(sb, logical_sb_block, i);
2563 sbi->s_group_desc[i] = sb_bread(sb, block);
2564 if (!sbi->s_group_desc[i]) {
2565 printk(KERN_ERR "EXT4-fs: "
2566 "can't read group descriptor %d\n", i);
2567 db_count = i;
2568 goto failed_mount2;
2569 }
2570 }
2571 if (!ext4_check_descriptors(sb)) {
2572 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
2573 goto failed_mount2;
2574 }
2575 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2576 if (!ext4_fill_flex_info(sb)) {
2577 printk(KERN_ERR
2578 "EXT4-fs: unable to initialize "
2579 "flex_bg meta info!\n");
2580 goto failed_mount2;
2581 }
2582
2583 sbi->s_gdb_count = db_count;
2584 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2585 spin_lock_init(&sbi->s_next_gen_lock);
2586
2587 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2588 ext4_count_free_blocks(sb));
2589 if (!err) {
2590 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2591 ext4_count_free_inodes(sb));
2592 }
2593 if (!err) {
2594 err = percpu_counter_init(&sbi->s_dirs_counter,
2595 ext4_count_dirs(sb));
2596 }
2597 if (!err) {
2598 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2599 }
2600 if (err) {
2601 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2602 goto failed_mount3;
2603 }
2604
2605 sbi->s_stripe = ext4_get_stripe_size(sbi);
2606
2607 /*
2608 * set up enough so that it can read an inode
2609 */
2610 sb->s_op = &ext4_sops;
2611 sb->s_export_op = &ext4_export_ops;
2612 sb->s_xattr = ext4_xattr_handlers;
2613 #ifdef CONFIG_QUOTA
2614 sb->s_qcop = &ext4_qctl_operations;
2615 sb->dq_op = &ext4_quota_operations;
2616 #endif
2617 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2618
2619 sb->s_root = NULL;
2620
2621 needs_recovery = (es->s_last_orphan != 0 ||
2622 EXT4_HAS_INCOMPAT_FEATURE(sb,
2623 EXT4_FEATURE_INCOMPAT_RECOVER));
2624
2625 /*
2626 * The first inode we look at is the journal inode. Don't try
2627 * root first: it may be modified in the journal!
2628 */
2629 if (!test_opt(sb, NOLOAD) &&
2630 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2631 if (ext4_load_journal(sb, es, journal_devnum))
2632 goto failed_mount3;
2633 if (!(sb->s_flags & MS_RDONLY) &&
2634 EXT4_SB(sb)->s_journal->j_failed_commit) {
2635 printk(KERN_CRIT "EXT4-fs error (device %s): "
2636 "ext4_fill_super: Journal transaction "
2637 "%u is corrupt\n", sb->s_id,
2638 EXT4_SB(sb)->s_journal->j_failed_commit);
2639 if (test_opt(sb, ERRORS_RO)) {
2640 printk(KERN_CRIT
2641 "Mounting filesystem read-only\n");
2642 sb->s_flags |= MS_RDONLY;
2643 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2644 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2645 }
2646 if (test_opt(sb, ERRORS_PANIC)) {
2647 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2648 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2649 ext4_commit_super(sb, es, 1);
2650 goto failed_mount4;
2651 }
2652 }
2653 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2654 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2655 printk(KERN_ERR "EXT4-fs: required journal recovery "
2656 "suppressed and not mounted read-only\n");
2657 goto failed_mount4;
2658 } else {
2659 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2660 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2661 sbi->s_journal = NULL;
2662 needs_recovery = 0;
2663 goto no_journal;
2664 }
2665
2666 if (ext4_blocks_count(es) > 0xffffffffULL &&
2667 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2668 JBD2_FEATURE_INCOMPAT_64BIT)) {
2669 printk(KERN_ERR "EXT4-fs: Failed to set 64-bit journal feature\n");
2670 goto failed_mount4;
2671 }
2672
2673 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2674 jbd2_journal_set_features(sbi->s_journal,
2675 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2676 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2677 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2678 jbd2_journal_set_features(sbi->s_journal,
2679 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2680 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2681 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2682 } else {
2683 jbd2_journal_clear_features(sbi->s_journal,
2684 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2685 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2686 }
2687
2688 /* We have now updated the journal if required, so we can
2689 * validate the data journaling mode. */
2690 switch (test_opt(sb, DATA_FLAGS)) {
2691 case 0:
2692 /* No mode set, assume a default based on the journal
2693 * capabilities: ORDERED_DATA if the journal can
2694 * cope, else JOURNAL_DATA
2695 */
2696 if (jbd2_journal_check_available_features
2697 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2698 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2699 else
2700 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2701 break;
2702
2703 case EXT4_MOUNT_ORDERED_DATA:
2704 case EXT4_MOUNT_WRITEBACK_DATA:
2705 if (!jbd2_journal_check_available_features
2706 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2707 printk(KERN_ERR "EXT4-fs: Journal does not support "
2708 "requested data journaling mode\n");
2709 goto failed_mount4;
2710 }
2711 default:
2712 break;
2713 }
2714 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2715
2716 no_journal:
2717
2718 if (test_opt(sb, NOBH)) {
2719 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2720 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
2721 "its supported only with writeback mode\n");
2722 clear_opt(sbi->s_mount_opt, NOBH);
2723 }
2724 }
2725 /*
2726 * The jbd2_journal_load will have done any necessary log recovery,
2727 * so we can safely mount the rest of the filesystem now.
2728 */
2729
2730 root = ext4_iget(sb, EXT4_ROOT_INO);
2731 if (IS_ERR(root)) {
2732 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
2733 ret = PTR_ERR(root);
2734 goto failed_mount4;
2735 }
2736 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2737 iput(root);
2738 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
2739 goto failed_mount4;
2740 }
2741 sb->s_root = d_alloc_root(root);
2742 if (!sb->s_root) {
2743 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2744 iput(root);
2745 ret = -ENOMEM;
2746 goto failed_mount4;
2747 }
2748
2749 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2750
2751 /* determine the minimum size of new large inodes, if present */
2752 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2753 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2754 EXT4_GOOD_OLD_INODE_SIZE;
2755 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2756 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2757 if (sbi->s_want_extra_isize <
2758 le16_to_cpu(es->s_want_extra_isize))
2759 sbi->s_want_extra_isize =
2760 le16_to_cpu(es->s_want_extra_isize);
2761 if (sbi->s_want_extra_isize <
2762 le16_to_cpu(es->s_min_extra_isize))
2763 sbi->s_want_extra_isize =
2764 le16_to_cpu(es->s_min_extra_isize);
2765 }
2766 }
2767 /* Check if enough inode space is available */
2768 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2769 sbi->s_inode_size) {
2770 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2771 EXT4_GOOD_OLD_INODE_SIZE;
2772 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2773 "available.\n");
2774 }
2775
2776 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2777 printk(KERN_WARNING "EXT4-fs: Ignoring delalloc option - "
2778 "requested data journaling mode\n");
2779 clear_opt(sbi->s_mount_opt, DELALLOC);
2780 } else if (test_opt(sb, DELALLOC))
2781 printk(KERN_INFO "EXT4-fs: delayed allocation enabled\n");
2782
2783 ext4_ext_init(sb);
2784 err = ext4_mb_init(sb, needs_recovery);
2785 if (err) {
2786 printk(KERN_ERR "EXT4-fs: failed to initalize mballoc (%d)\n",
2787 err);
2788 goto failed_mount4;
2789 }
2790
2791 sbi->s_kobj.kset = ext4_kset;
2792 init_completion(&sbi->s_kobj_unregister);
2793 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2794 "%s", sb->s_id);
2795 if (err) {
2796 ext4_mb_release(sb);
2797 ext4_ext_release(sb);
2798 goto failed_mount4;
2799 };
2800
2801 /*
2802 * akpm: core read_super() calls in here with the superblock locked.
2803 * That deadlocks, because orphan cleanup needs to lock the superblock
2804 * in numerous places. Here we just pop the lock - it's relatively
2805 * harmless, because we are now ready to accept write_super() requests,
2806 * and aviro says that's the only reason for hanging onto the
2807 * superblock lock.
2808 */
2809 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2810 ext4_orphan_cleanup(sb, es);
2811 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2812 if (needs_recovery) {
2813 printk(KERN_INFO "EXT4-fs: recovery complete.\n");
2814 ext4_mark_recovery_complete(sb, es);
2815 }
2816 if (EXT4_SB(sb)->s_journal) {
2817 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2818 descr = " journalled data mode";
2819 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2820 descr = " ordered data mode";
2821 else
2822 descr = " writeback data mode";
2823 } else
2824 descr = "out journal";
2825
2826 printk(KERN_INFO "EXT4-fs: mounted filesystem %s with%s\n",
2827 sb->s_id, descr);
2828
2829 lock_kernel();
2830 return 0;
2831
2832 cantfind_ext4:
2833 if (!silent)
2834 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
2835 sb->s_id);
2836 goto failed_mount;
2837
2838 failed_mount4:
2839 printk(KERN_ERR "EXT4-fs (device %s): mount failed\n", sb->s_id);
2840 if (sbi->s_journal) {
2841 jbd2_journal_destroy(sbi->s_journal);
2842 sbi->s_journal = NULL;
2843 }
2844 failed_mount3:
2845 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2846 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2847 percpu_counter_destroy(&sbi->s_dirs_counter);
2848 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2849 failed_mount2:
2850 for (i = 0; i < db_count; i++)
2851 brelse(sbi->s_group_desc[i]);
2852 kfree(sbi->s_group_desc);
2853 failed_mount:
2854 if (sbi->s_proc) {
2855 remove_proc_entry(sb->s_id, ext4_proc_root);
2856 }
2857 #ifdef CONFIG_QUOTA
2858 for (i = 0; i < MAXQUOTAS; i++)
2859 kfree(sbi->s_qf_names[i]);
2860 #endif
2861 ext4_blkdev_remove(sbi);
2862 brelse(bh);
2863 out_fail:
2864 sb->s_fs_info = NULL;
2865 kfree(sbi);
2866 lock_kernel();
2867 return ret;
2868 }
2869
2870 /*
2871 * Setup any per-fs journal parameters now. We'll do this both on
2872 * initial mount, once the journal has been initialised but before we've
2873 * done any recovery; and again on any subsequent remount.
2874 */
2875 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2876 {
2877 struct ext4_sb_info *sbi = EXT4_SB(sb);
2878
2879 journal->j_commit_interval = sbi->s_commit_interval;
2880 journal->j_min_batch_time = sbi->s_min_batch_time;
2881 journal->j_max_batch_time = sbi->s_max_batch_time;
2882
2883 spin_lock(&journal->j_state_lock);
2884 if (test_opt(sb, BARRIER))
2885 journal->j_flags |= JBD2_BARRIER;
2886 else
2887 journal->j_flags &= ~JBD2_BARRIER;
2888 if (test_opt(sb, DATA_ERR_ABORT))
2889 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2890 else
2891 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
2892 spin_unlock(&journal->j_state_lock);
2893 }
2894
2895 static journal_t *ext4_get_journal(struct super_block *sb,
2896 unsigned int journal_inum)
2897 {
2898 struct inode *journal_inode;
2899 journal_t *journal;
2900
2901 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2902
2903 /* First, test for the existence of a valid inode on disk. Bad
2904 * things happen if we iget() an unused inode, as the subsequent
2905 * iput() will try to delete it. */
2906
2907 journal_inode = ext4_iget(sb, journal_inum);
2908 if (IS_ERR(journal_inode)) {
2909 printk(KERN_ERR "EXT4-fs: no journal found.\n");
2910 return NULL;
2911 }
2912 if (!journal_inode->i_nlink) {
2913 make_bad_inode(journal_inode);
2914 iput(journal_inode);
2915 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
2916 return NULL;
2917 }
2918
2919 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
2920 journal_inode, journal_inode->i_size);
2921 if (!S_ISREG(journal_inode->i_mode)) {
2922 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
2923 iput(journal_inode);
2924 return NULL;
2925 }
2926
2927 journal = jbd2_journal_init_inode(journal_inode);
2928 if (!journal) {
2929 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
2930 iput(journal_inode);
2931 return NULL;
2932 }
2933 journal->j_private = sb;
2934 ext4_init_journal_params(sb, journal);
2935 return journal;
2936 }
2937
2938 static journal_t *ext4_get_dev_journal(struct super_block *sb,
2939 dev_t j_dev)
2940 {
2941 struct buffer_head *bh;
2942 journal_t *journal;
2943 ext4_fsblk_t start;
2944 ext4_fsblk_t len;
2945 int hblock, blocksize;
2946 ext4_fsblk_t sb_block;
2947 unsigned long offset;
2948 struct ext4_super_block *es;
2949 struct block_device *bdev;
2950
2951 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2952
2953 bdev = ext4_blkdev_get(j_dev);
2954 if (bdev == NULL)
2955 return NULL;
2956
2957 if (bd_claim(bdev, sb)) {
2958 printk(KERN_ERR
2959 "EXT4-fs: failed to claim external journal device.\n");
2960 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
2961 return NULL;
2962 }
2963
2964 blocksize = sb->s_blocksize;
2965 hblock = bdev_hardsect_size(bdev);
2966 if (blocksize < hblock) {
2967 printk(KERN_ERR
2968 "EXT4-fs: blocksize too small for journal device.\n");
2969 goto out_bdev;
2970 }
2971
2972 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2973 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
2974 set_blocksize(bdev, blocksize);
2975 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2976 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
2977 "external journal\n");
2978 goto out_bdev;
2979 }
2980
2981 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2982 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
2983 !(le32_to_cpu(es->s_feature_incompat) &
2984 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2985 printk(KERN_ERR "EXT4-fs: external journal has "
2986 "bad superblock\n");
2987 brelse(bh);
2988 goto out_bdev;
2989 }
2990
2991 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2992 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
2993 brelse(bh);
2994 goto out_bdev;
2995 }
2996
2997 len = ext4_blocks_count(es);
2998 start = sb_block + 1;
2999 brelse(bh); /* we're done with the superblock */
3000
3001 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3002 start, len, blocksize);
3003 if (!journal) {
3004 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
3005 goto out_bdev;
3006 }
3007 journal->j_private = sb;
3008 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3009 wait_on_buffer(journal->j_sb_buffer);
3010 if (!buffer_uptodate(journal->j_sb_buffer)) {
3011 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
3012 goto out_journal;
3013 }
3014 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3015 printk(KERN_ERR "EXT4-fs: External journal has more than one "
3016 "user (unsupported) - %d\n",
3017 be32_to_cpu(journal->j_superblock->s_nr_users));
3018 goto out_journal;
3019 }
3020 EXT4_SB(sb)->journal_bdev = bdev;
3021 ext4_init_journal_params(sb, journal);
3022 return journal;
3023 out_journal:
3024 jbd2_journal_destroy(journal);
3025 out_bdev:
3026 ext4_blkdev_put(bdev);
3027 return NULL;
3028 }
3029
3030 static int ext4_load_journal(struct super_block *sb,
3031 struct ext4_super_block *es,
3032 unsigned long journal_devnum)
3033 {
3034 journal_t *journal;
3035 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3036 dev_t journal_dev;
3037 int err = 0;
3038 int really_read_only;
3039
3040 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3041
3042 if (journal_devnum &&
3043 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3044 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
3045 "numbers have changed\n");
3046 journal_dev = new_decode_dev(journal_devnum);
3047 } else
3048 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3049
3050 really_read_only = bdev_read_only(sb->s_bdev);
3051
3052 /*
3053 * Are we loading a blank journal or performing recovery after a
3054 * crash? For recovery, we need to check in advance whether we
3055 * can get read-write access to the device.
3056 */
3057
3058 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3059 if (sb->s_flags & MS_RDONLY) {
3060 printk(KERN_INFO "EXT4-fs: INFO: recovery "
3061 "required on readonly filesystem.\n");
3062 if (really_read_only) {
3063 printk(KERN_ERR "EXT4-fs: write access "
3064 "unavailable, cannot proceed.\n");
3065 return -EROFS;
3066 }
3067 printk(KERN_INFO "EXT4-fs: write access will "
3068 "be enabled during recovery.\n");
3069 }
3070 }
3071
3072 if (journal_inum && journal_dev) {
3073 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
3074 "and inode journals!\n");
3075 return -EINVAL;
3076 }
3077
3078 if (journal_inum) {
3079 if (!(journal = ext4_get_journal(sb, journal_inum)))
3080 return -EINVAL;
3081 } else {
3082 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3083 return -EINVAL;
3084 }
3085
3086 if (journal->j_flags & JBD2_BARRIER)
3087 printk(KERN_INFO "EXT4-fs: barriers enabled\n");
3088 else
3089 printk(KERN_INFO "EXT4-fs: barriers disabled\n");
3090
3091 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3092 err = jbd2_journal_update_format(journal);
3093 if (err) {
3094 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
3095 jbd2_journal_destroy(journal);
3096 return err;
3097 }
3098 }
3099
3100 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3101 err = jbd2_journal_wipe(journal, !really_read_only);
3102 if (!err)
3103 err = jbd2_journal_load(journal);
3104
3105 if (err) {
3106 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
3107 jbd2_journal_destroy(journal);
3108 return err;
3109 }
3110
3111 EXT4_SB(sb)->s_journal = journal;
3112 ext4_clear_journal_err(sb, es);
3113
3114 if (journal_devnum &&
3115 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3116 es->s_journal_dev = cpu_to_le32(journal_devnum);
3117 sb->s_dirt = 1;
3118
3119 /* Make sure we flush the recovery flag to disk. */
3120 ext4_commit_super(sb, es, 1);
3121 }
3122
3123 return 0;
3124 }
3125
3126 static int ext4_commit_super(struct super_block *sb,
3127 struct ext4_super_block *es, int sync)
3128 {
3129 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3130 int error = 0;
3131
3132 if (!sbh)
3133 return error;
3134 if (buffer_write_io_error(sbh)) {
3135 /*
3136 * Oh, dear. A previous attempt to write the
3137 * superblock failed. This could happen because the
3138 * USB device was yanked out. Or it could happen to
3139 * be a transient write error and maybe the block will
3140 * be remapped. Nothing we can do but to retry the
3141 * write and hope for the best.
3142 */
3143 printk(KERN_ERR "EXT4-fs: previous I/O error to "
3144 "superblock detected for %s.\n", sb->s_id);
3145 clear_buffer_write_io_error(sbh);
3146 set_buffer_uptodate(sbh);
3147 }
3148 es->s_wtime = cpu_to_le32(get_seconds());
3149 es->s_kbytes_written =
3150 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3151 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3152 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3153 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3154 &EXT4_SB(sb)->s_freeblocks_counter));
3155 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3156 &EXT4_SB(sb)->s_freeinodes_counter));
3157
3158 BUFFER_TRACE(sbh, "marking dirty");
3159 mark_buffer_dirty(sbh);
3160 if (sync) {
3161 error = sync_dirty_buffer(sbh);
3162 if (error)
3163 return error;
3164
3165 error = buffer_write_io_error(sbh);
3166 if (error) {
3167 printk(KERN_ERR "EXT4-fs: I/O error while writing "
3168 "superblock for %s.\n", sb->s_id);
3169 clear_buffer_write_io_error(sbh);
3170 set_buffer_uptodate(sbh);
3171 }
3172 }
3173 return error;
3174 }
3175
3176
3177 /*
3178 * Have we just finished recovery? If so, and if we are mounting (or
3179 * remounting) the filesystem readonly, then we will end up with a
3180 * consistent fs on disk. Record that fact.
3181 */
3182 static void ext4_mark_recovery_complete(struct super_block *sb,
3183 struct ext4_super_block *es)
3184 {
3185 journal_t *journal = EXT4_SB(sb)->s_journal;
3186
3187 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3188 BUG_ON(journal != NULL);
3189 return;
3190 }
3191 jbd2_journal_lock_updates(journal);
3192 if (jbd2_journal_flush(journal) < 0)
3193 goto out;
3194
3195 lock_super(sb);
3196 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3197 sb->s_flags & MS_RDONLY) {
3198 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3199 sb->s_dirt = 0;
3200 ext4_commit_super(sb, es, 1);
3201 }
3202 unlock_super(sb);
3203
3204 out:
3205 jbd2_journal_unlock_updates(journal);
3206 }
3207
3208 /*
3209 * If we are mounting (or read-write remounting) a filesystem whose journal
3210 * has recorded an error from a previous lifetime, move that error to the
3211 * main filesystem now.
3212 */
3213 static void ext4_clear_journal_err(struct super_block *sb,
3214 struct ext4_super_block *es)
3215 {
3216 journal_t *journal;
3217 int j_errno;
3218 const char *errstr;
3219
3220 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3221
3222 journal = EXT4_SB(sb)->s_journal;
3223
3224 /*
3225 * Now check for any error status which may have been recorded in the
3226 * journal by a prior ext4_error() or ext4_abort()
3227 */
3228
3229 j_errno = jbd2_journal_errno(journal);
3230 if (j_errno) {
3231 char nbuf[16];
3232
3233 errstr = ext4_decode_error(sb, j_errno, nbuf);
3234 ext4_warning(sb, __func__, "Filesystem error recorded "
3235 "from previous mount: %s", errstr);
3236 ext4_warning(sb, __func__, "Marking fs in need of "
3237 "filesystem check.");
3238
3239 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3240 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3241 ext4_commit_super(sb, es, 1);
3242
3243 jbd2_journal_clear_err(journal);
3244 }
3245 }
3246
3247 /*
3248 * Force the running and committing transactions to commit,
3249 * and wait on the commit.
3250 */
3251 int ext4_force_commit(struct super_block *sb)
3252 {
3253 journal_t *journal;
3254 int ret = 0;
3255
3256 if (sb->s_flags & MS_RDONLY)
3257 return 0;
3258
3259 journal = EXT4_SB(sb)->s_journal;
3260 if (journal) {
3261 sb->s_dirt = 0;
3262 ret = ext4_journal_force_commit(journal);
3263 }
3264
3265 return ret;
3266 }
3267
3268 /*
3269 * Ext4 always journals updates to the superblock itself, so we don't
3270 * have to propagate any other updates to the superblock on disk at this
3271 * point. (We can probably nuke this function altogether, and remove
3272 * any mention to sb->s_dirt in all of fs/ext4; eventual cleanup...)
3273 */
3274 static void ext4_write_super(struct super_block *sb)
3275 {
3276 if (EXT4_SB(sb)->s_journal) {
3277 if (mutex_trylock(&sb->s_lock) != 0)
3278 BUG();
3279 sb->s_dirt = 0;
3280 } else {
3281 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
3282 }
3283 }
3284
3285 static int ext4_sync_fs(struct super_block *sb, int wait)
3286 {
3287 int ret = 0;
3288 tid_t target;
3289
3290 trace_mark(ext4_sync_fs, "dev %s wait %d", sb->s_id, wait);
3291 sb->s_dirt = 0;
3292 if (EXT4_SB(sb)->s_journal) {
3293 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal,
3294 &target)) {
3295 if (wait)
3296 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal,
3297 target);
3298 }
3299 } else {
3300 ext4_commit_super(sb, EXT4_SB(sb)->s_es, wait);
3301 }
3302 return ret;
3303 }
3304
3305 /*
3306 * LVM calls this function before a (read-only) snapshot is created. This
3307 * gives us a chance to flush the journal completely and mark the fs clean.
3308 */
3309 static int ext4_freeze(struct super_block *sb)
3310 {
3311 int error = 0;
3312 journal_t *journal;
3313 sb->s_dirt = 0;
3314
3315 if (!(sb->s_flags & MS_RDONLY)) {
3316 journal = EXT4_SB(sb)->s_journal;
3317
3318 if (journal) {
3319 /* Now we set up the journal barrier. */
3320 jbd2_journal_lock_updates(journal);
3321
3322 /*
3323 * We don't want to clear needs_recovery flag when we
3324 * failed to flush the journal.
3325 */
3326 error = jbd2_journal_flush(journal);
3327 if (error < 0)
3328 goto out;
3329 }
3330
3331 /* Journal blocked and flushed, clear needs_recovery flag. */
3332 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3333 error = ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
3334 if (error)
3335 goto out;
3336 }
3337 return 0;
3338 out:
3339 jbd2_journal_unlock_updates(journal);
3340 return error;
3341 }
3342
3343 /*
3344 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3345 * flag here, even though the filesystem is not technically dirty yet.
3346 */
3347 static int ext4_unfreeze(struct super_block *sb)
3348 {
3349 if (EXT4_SB(sb)->s_journal && !(sb->s_flags & MS_RDONLY)) {
3350 lock_super(sb);
3351 /* Reser the needs_recovery flag before the fs is unlocked. */
3352 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3353 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
3354 unlock_super(sb);
3355 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3356 }
3357 return 0;
3358 }
3359
3360 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3361 {
3362 struct ext4_super_block *es;
3363 struct ext4_sb_info *sbi = EXT4_SB(sb);
3364 ext4_fsblk_t n_blocks_count = 0;
3365 unsigned long old_sb_flags;
3366 struct ext4_mount_options old_opts;
3367 ext4_group_t g;
3368 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3369 int err;
3370 #ifdef CONFIG_QUOTA
3371 int i;
3372 #endif
3373
3374 /* Store the original options */
3375 old_sb_flags = sb->s_flags;
3376 old_opts.s_mount_opt = sbi->s_mount_opt;
3377 old_opts.s_resuid = sbi->s_resuid;
3378 old_opts.s_resgid = sbi->s_resgid;
3379 old_opts.s_commit_interval = sbi->s_commit_interval;
3380 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3381 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3382 #ifdef CONFIG_QUOTA
3383 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3384 for (i = 0; i < MAXQUOTAS; i++)
3385 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3386 #endif
3387 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3388 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3389
3390 /*
3391 * Allow the "check" option to be passed as a remount option.
3392 */
3393 if (!parse_options(data, sb, NULL, &journal_ioprio,
3394 &n_blocks_count, 1)) {
3395 err = -EINVAL;
3396 goto restore_opts;
3397 }
3398
3399 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
3400 ext4_abort(sb, __func__, "Abort forced by user");
3401
3402 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3403 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3404
3405 es = sbi->s_es;
3406
3407 if (sbi->s_journal) {
3408 ext4_init_journal_params(sb, sbi->s_journal);
3409 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3410 }
3411
3412 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3413 n_blocks_count > ext4_blocks_count(es)) {
3414 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
3415 err = -EROFS;
3416 goto restore_opts;
3417 }
3418
3419 if (*flags & MS_RDONLY) {
3420 /*
3421 * First of all, the unconditional stuff we have to do
3422 * to disable replay of the journal when we next remount
3423 */
3424 sb->s_flags |= MS_RDONLY;
3425
3426 /*
3427 * OK, test if we are remounting a valid rw partition
3428 * readonly, and if so set the rdonly flag and then
3429 * mark the partition as valid again.
3430 */
3431 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3432 (sbi->s_mount_state & EXT4_VALID_FS))
3433 es->s_state = cpu_to_le16(sbi->s_mount_state);
3434
3435 /*
3436 * We have to unlock super so that we can wait for
3437 * transactions.
3438 */
3439 if (sbi->s_journal) {
3440 unlock_super(sb);
3441 ext4_mark_recovery_complete(sb, es);
3442 lock_super(sb);
3443 }
3444 } else {
3445 int ret;
3446 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3447 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
3448 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3449 "remount RDWR because of unsupported "
3450 "optional features (%x).\n", sb->s_id,
3451 (le32_to_cpu(sbi->s_es->s_feature_ro_compat) &
3452 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3453 err = -EROFS;
3454 goto restore_opts;
3455 }
3456
3457 /*
3458 * Make sure the group descriptor checksums
3459 * are sane. If they aren't, refuse to
3460 * remount r/w.
3461 */
3462 for (g = 0; g < sbi->s_groups_count; g++) {
3463 struct ext4_group_desc *gdp =
3464 ext4_get_group_desc(sb, g, NULL);
3465
3466 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3467 printk(KERN_ERR
3468 "EXT4-fs: ext4_remount: "
3469 "Checksum for group %u failed (%u!=%u)\n",
3470 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3471 le16_to_cpu(gdp->bg_checksum));
3472 err = -EINVAL;
3473 goto restore_opts;
3474 }
3475 }
3476
3477 /*
3478 * If we have an unprocessed orphan list hanging
3479 * around from a previously readonly bdev mount,
3480 * require a full umount/remount for now.
3481 */
3482 if (es->s_last_orphan) {
3483 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3484 "remount RDWR because of unprocessed "
3485 "orphan inode list. Please "
3486 "umount/remount instead.\n",
3487 sb->s_id);
3488 err = -EINVAL;
3489 goto restore_opts;
3490 }
3491
3492 /*
3493 * Mounting a RDONLY partition read-write, so reread
3494 * and store the current valid flag. (It may have
3495 * been changed by e2fsck since we originally mounted
3496 * the partition.)
3497 */
3498 if (sbi->s_journal)
3499 ext4_clear_journal_err(sb, es);
3500 sbi->s_mount_state = le16_to_cpu(es->s_state);
3501 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3502 goto restore_opts;
3503 if (!ext4_setup_super(sb, es, 0))
3504 sb->s_flags &= ~MS_RDONLY;
3505 }
3506 }
3507 if (sbi->s_journal == NULL)
3508 ext4_commit_super(sb, es, 1);
3509
3510 #ifdef CONFIG_QUOTA
3511 /* Release old quota file names */
3512 for (i = 0; i < MAXQUOTAS; i++)
3513 if (old_opts.s_qf_names[i] &&
3514 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3515 kfree(old_opts.s_qf_names[i]);
3516 #endif
3517 return 0;
3518 restore_opts:
3519 sb->s_flags = old_sb_flags;
3520 sbi->s_mount_opt = old_opts.s_mount_opt;
3521 sbi->s_resuid = old_opts.s_resuid;
3522 sbi->s_resgid = old_opts.s_resgid;
3523 sbi->s_commit_interval = old_opts.s_commit_interval;
3524 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3525 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3526 #ifdef CONFIG_QUOTA
3527 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3528 for (i = 0; i < MAXQUOTAS; i++) {
3529 if (sbi->s_qf_names[i] &&
3530 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3531 kfree(sbi->s_qf_names[i]);
3532 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3533 }
3534 #endif
3535 return err;
3536 }
3537
3538 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3539 {
3540 struct super_block *sb = dentry->d_sb;
3541 struct ext4_sb_info *sbi = EXT4_SB(sb);
3542 struct ext4_super_block *es = sbi->s_es;
3543 u64 fsid;
3544
3545 if (test_opt(sb, MINIX_DF)) {
3546 sbi->s_overhead_last = 0;
3547 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3548 ext4_group_t ngroups = sbi->s_groups_count, i;
3549 ext4_fsblk_t overhead = 0;
3550 smp_rmb();
3551
3552 /*
3553 * Compute the overhead (FS structures). This is constant
3554 * for a given filesystem unless the number of block groups
3555 * changes so we cache the previous value until it does.
3556 */
3557
3558 /*
3559 * All of the blocks before first_data_block are
3560 * overhead
3561 */
3562 overhead = le32_to_cpu(es->s_first_data_block);
3563
3564 /*
3565 * Add the overhead attributed to the superblock and
3566 * block group descriptors. If the sparse superblocks
3567 * feature is turned on, then not all groups have this.
3568 */
3569 for (i = 0; i < ngroups; i++) {
3570 overhead += ext4_bg_has_super(sb, i) +
3571 ext4_bg_num_gdb(sb, i);
3572 cond_resched();
3573 }
3574
3575 /*
3576 * Every block group has an inode bitmap, a block
3577 * bitmap, and an inode table.
3578 */
3579 overhead += ngroups * (2 + sbi->s_itb_per_group);
3580 sbi->s_overhead_last = overhead;
3581 smp_wmb();
3582 sbi->s_blocks_last = ext4_blocks_count(es);
3583 }
3584
3585 buf->f_type = EXT4_SUPER_MAGIC;
3586 buf->f_bsize = sb->s_blocksize;
3587 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3588 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3589 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3590 ext4_free_blocks_count_set(es, buf->f_bfree);
3591 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3592 if (buf->f_bfree < ext4_r_blocks_count(es))
3593 buf->f_bavail = 0;
3594 buf->f_files = le32_to_cpu(es->s_inodes_count);
3595 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3596 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
3597 buf->f_namelen = EXT4_NAME_LEN;
3598 fsid = le64_to_cpup((void *)es->s_uuid) ^
3599 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3600 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3601 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3602 return 0;
3603 }
3604
3605 /* Helper function for writing quotas on sync - we need to start transaction before quota file
3606 * is locked for write. Otherwise the are possible deadlocks:
3607 * Process 1 Process 2
3608 * ext4_create() quota_sync()
3609 * jbd2_journal_start() write_dquot()
3610 * vfs_dq_init() down(dqio_mutex)
3611 * down(dqio_mutex) jbd2_journal_start()
3612 *
3613 */
3614
3615 #ifdef CONFIG_QUOTA
3616
3617 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3618 {
3619 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3620 }
3621
3622 static int ext4_write_dquot(struct dquot *dquot)
3623 {
3624 int ret, err;
3625 handle_t *handle;
3626 struct inode *inode;
3627
3628 inode = dquot_to_inode(dquot);
3629 handle = ext4_journal_start(inode,
3630 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3631 if (IS_ERR(handle))
3632 return PTR_ERR(handle);
3633 ret = dquot_commit(dquot);
3634 err = ext4_journal_stop(handle);
3635 if (!ret)
3636 ret = err;
3637 return ret;
3638 }
3639
3640 static int ext4_acquire_dquot(struct dquot *dquot)
3641 {
3642 int ret, err;
3643 handle_t *handle;
3644
3645 handle = ext4_journal_start(dquot_to_inode(dquot),
3646 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3647 if (IS_ERR(handle))
3648 return PTR_ERR(handle);
3649 ret = dquot_acquire(dquot);
3650 err = ext4_journal_stop(handle);
3651 if (!ret)
3652 ret = err;
3653 return ret;
3654 }
3655
3656 static int ext4_release_dquot(struct dquot *dquot)
3657 {
3658 int ret, err;
3659 handle_t *handle;
3660
3661 handle = ext4_journal_start(dquot_to_inode(dquot),
3662 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3663 if (IS_ERR(handle)) {
3664 /* Release dquot anyway to avoid endless cycle in dqput() */
3665 dquot_release(dquot);
3666 return PTR_ERR(handle);
3667 }
3668 ret = dquot_release(dquot);
3669 err = ext4_journal_stop(handle);
3670 if (!ret)
3671 ret = err;
3672 return ret;
3673 }
3674
3675 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3676 {
3677 /* Are we journaling quotas? */
3678 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3679 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3680 dquot_mark_dquot_dirty(dquot);
3681 return ext4_write_dquot(dquot);
3682 } else {
3683 return dquot_mark_dquot_dirty(dquot);
3684 }
3685 }
3686
3687 static int ext4_write_info(struct super_block *sb, int type)
3688 {
3689 int ret, err;
3690 handle_t *handle;
3691
3692 /* Data block + inode block */
3693 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3694 if (IS_ERR(handle))
3695 return PTR_ERR(handle);
3696 ret = dquot_commit_info(sb, type);
3697 err = ext4_journal_stop(handle);
3698 if (!ret)
3699 ret = err;
3700 return ret;
3701 }
3702
3703 /*
3704 * Turn on quotas during mount time - we need to find
3705 * the quota file and such...
3706 */
3707 static int ext4_quota_on_mount(struct super_block *sb, int type)
3708 {
3709 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3710 EXT4_SB(sb)->s_jquota_fmt, type);
3711 }
3712
3713 /*
3714 * Standard function to be called on quota_on
3715 */
3716 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3717 char *name, int remount)
3718 {
3719 int err;
3720 struct path path;
3721
3722 if (!test_opt(sb, QUOTA))
3723 return -EINVAL;
3724 /* When remounting, no checks are needed and in fact, name is NULL */
3725 if (remount)
3726 return vfs_quota_on(sb, type, format_id, name, remount);
3727
3728 err = kern_path(name, LOOKUP_FOLLOW, &path);
3729 if (err)
3730 return err;
3731
3732 /* Quotafile not on the same filesystem? */
3733 if (path.mnt->mnt_sb != sb) {
3734 path_put(&path);
3735 return -EXDEV;
3736 }
3737 /* Journaling quota? */
3738 if (EXT4_SB(sb)->s_qf_names[type]) {
3739 /* Quotafile not in fs root? */
3740 if (path.dentry->d_parent != sb->s_root)
3741 printk(KERN_WARNING
3742 "EXT4-fs: Quota file not on filesystem root. "
3743 "Journaled quota will not work.\n");
3744 }
3745
3746 /*
3747 * When we journal data on quota file, we have to flush journal to see
3748 * all updates to the file when we bypass pagecache...
3749 */
3750 if (EXT4_SB(sb)->s_journal &&
3751 ext4_should_journal_data(path.dentry->d_inode)) {
3752 /*
3753 * We don't need to lock updates but journal_flush() could
3754 * otherwise be livelocked...
3755 */
3756 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3757 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3758 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3759 if (err) {
3760 path_put(&path);
3761 return err;
3762 }
3763 }
3764
3765 err = vfs_quota_on_path(sb, type, format_id, &path);
3766 path_put(&path);
3767 return err;
3768 }
3769
3770 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3771 * acquiring the locks... As quota files are never truncated and quota code
3772 * itself serializes the operations (and noone else should touch the files)
3773 * we don't have to be afraid of races */
3774 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3775 size_t len, loff_t off)
3776 {
3777 struct inode *inode = sb_dqopt(sb)->files[type];
3778 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3779 int err = 0;
3780 int offset = off & (sb->s_blocksize - 1);
3781 int tocopy;
3782 size_t toread;
3783 struct buffer_head *bh;
3784 loff_t i_size = i_size_read(inode);
3785
3786 if (off > i_size)
3787 return 0;
3788 if (off+len > i_size)
3789 len = i_size-off;
3790 toread = len;
3791 while (toread > 0) {
3792 tocopy = sb->s_blocksize - offset < toread ?
3793 sb->s_blocksize - offset : toread;
3794 bh = ext4_bread(NULL, inode, blk, 0, &err);
3795 if (err)
3796 return err;
3797 if (!bh) /* A hole? */
3798 memset(data, 0, tocopy);
3799 else
3800 memcpy(data, bh->b_data+offset, tocopy);
3801 brelse(bh);
3802 offset = 0;
3803 toread -= tocopy;
3804 data += tocopy;
3805 blk++;
3806 }
3807 return len;
3808 }
3809
3810 /* Write to quotafile (we know the transaction is already started and has
3811 * enough credits) */
3812 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3813 const char *data, size_t len, loff_t off)
3814 {
3815 struct inode *inode = sb_dqopt(sb)->files[type];
3816 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3817 int err = 0;
3818 int offset = off & (sb->s_blocksize - 1);
3819 int tocopy;
3820 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3821 size_t towrite = len;
3822 struct buffer_head *bh;
3823 handle_t *handle = journal_current_handle();
3824
3825 if (EXT4_SB(sb)->s_journal && !handle) {
3826 printk(KERN_WARNING "EXT4-fs: Quota write (off=%llu, len=%llu)"
3827 " cancelled because transaction is not started.\n",
3828 (unsigned long long)off, (unsigned long long)len);
3829 return -EIO;
3830 }
3831 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3832 while (towrite > 0) {
3833 tocopy = sb->s_blocksize - offset < towrite ?
3834 sb->s_blocksize - offset : towrite;
3835 bh = ext4_bread(handle, inode, blk, 1, &err);
3836 if (!bh)
3837 goto out;
3838 if (journal_quota) {
3839 err = ext4_journal_get_write_access(handle, bh);
3840 if (err) {
3841 brelse(bh);
3842 goto out;
3843 }
3844 }
3845 lock_buffer(bh);
3846 memcpy(bh->b_data+offset, data, tocopy);
3847 flush_dcache_page(bh->b_page);
3848 unlock_buffer(bh);
3849 if (journal_quota)
3850 err = ext4_handle_dirty_metadata(handle, NULL, bh);
3851 else {
3852 /* Always do at least ordered writes for quotas */
3853 err = ext4_jbd2_file_inode(handle, inode);
3854 mark_buffer_dirty(bh);
3855 }
3856 brelse(bh);
3857 if (err)
3858 goto out;
3859 offset = 0;
3860 towrite -= tocopy;
3861 data += tocopy;
3862 blk++;
3863 }
3864 out:
3865 if (len == towrite) {
3866 mutex_unlock(&inode->i_mutex);
3867 return err;
3868 }
3869 if (inode->i_size < off+len-towrite) {
3870 i_size_write(inode, off+len-towrite);
3871 EXT4_I(inode)->i_disksize = inode->i_size;
3872 }
3873 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3874 ext4_mark_inode_dirty(handle, inode);
3875 mutex_unlock(&inode->i_mutex);
3876 return len - towrite;
3877 }
3878
3879 #endif
3880
3881 static int ext4_get_sb(struct file_system_type *fs_type,
3882 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3883 {
3884 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3885 }
3886
3887 static struct file_system_type ext4_fs_type = {
3888 .owner = THIS_MODULE,
3889 .name = "ext4",
3890 .get_sb = ext4_get_sb,
3891 .kill_sb = kill_block_super,
3892 .fs_flags = FS_REQUIRES_DEV,
3893 };
3894
3895 #ifdef CONFIG_EXT4DEV_COMPAT
3896 static int ext4dev_get_sb(struct file_system_type *fs_type,
3897 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3898 {
3899 printk(KERN_WARNING "EXT4-fs: Update your userspace programs "
3900 "to mount using ext4\n");
3901 printk(KERN_WARNING "EXT4-fs: ext4dev backwards compatibility "
3902 "will go away by 2.6.31\n");
3903 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3904 }
3905
3906 static struct file_system_type ext4dev_fs_type = {
3907 .owner = THIS_MODULE,
3908 .name = "ext4dev",
3909 .get_sb = ext4dev_get_sb,
3910 .kill_sb = kill_block_super,
3911 .fs_flags = FS_REQUIRES_DEV,
3912 };
3913 MODULE_ALIAS("ext4dev");
3914 #endif
3915
3916 static int __init init_ext4_fs(void)
3917 {
3918 int err;
3919
3920 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
3921 if (!ext4_kset)
3922 return -ENOMEM;
3923 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
3924 err = init_ext4_mballoc();
3925 if (err)
3926 return err;
3927
3928 err = init_ext4_xattr();
3929 if (err)
3930 goto out2;
3931 err = init_inodecache();
3932 if (err)
3933 goto out1;
3934 err = register_filesystem(&ext4_fs_type);
3935 if (err)
3936 goto out;
3937 #ifdef CONFIG_EXT4DEV_COMPAT
3938 err = register_filesystem(&ext4dev_fs_type);
3939 if (err) {
3940 unregister_filesystem(&ext4_fs_type);
3941 goto out;
3942 }
3943 #endif
3944 return 0;
3945 out:
3946 destroy_inodecache();
3947 out1:
3948 exit_ext4_xattr();
3949 out2:
3950 exit_ext4_mballoc();
3951 return err;
3952 }
3953
3954 static void __exit exit_ext4_fs(void)
3955 {
3956 unregister_filesystem(&ext4_fs_type);
3957 #ifdef CONFIG_EXT4DEV_COMPAT
3958 unregister_filesystem(&ext4dev_fs_type);
3959 #endif
3960 destroy_inodecache();
3961 exit_ext4_xattr();
3962 exit_ext4_mballoc();
3963 remove_proc_entry("fs/ext4", NULL);
3964 kset_unregister(ext4_kset);
3965 }
3966
3967 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3968 MODULE_DESCRIPTION("Fourth Extended Filesystem");
3969 MODULE_LICENSE("GPL");
3970 module_init(init_ext4_fs)
3971 module_exit(exit_ext4_fs)
Support for the Chinese Samsung S3C2440 based development boards