?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
ext4: remove /proc tuning knobs
[linux-2.6/cjktty.git] / fs / ext4 / super.c
blob1ec554cc107a2554c55aa4b99cd37d04f8b94b3d
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 (!test_opt(sb, RESERVATION))
820 seq_puts(seq, ",noreservation");
821 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
822 seq_printf(seq, ",commit=%u",
823 (unsigned) (sbi->s_commit_interval / HZ));
824 }
825 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
826 seq_printf(seq, ",min_batch_time=%u",
827 (unsigned) sbi->s_min_batch_time);
828 }
829 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
830 seq_printf(seq, ",max_batch_time=%u",
831 (unsigned) sbi->s_min_batch_time);
832 }
833
834 /*
835 * We're changing the default of barrier mount option, so
836 * let's always display its mount state so it's clear what its
837 * status is.
838 */
839 seq_puts(seq, ",barrier=");
840 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
841 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
842 seq_puts(seq, ",journal_async_commit");
843 if (test_opt(sb, NOBH))
844 seq_puts(seq, ",nobh");
845 if (test_opt(sb, I_VERSION))
846 seq_puts(seq, ",i_version");
847 if (!test_opt(sb, DELALLOC))
848 seq_puts(seq, ",nodelalloc");
849
850
851 if (sbi->s_stripe)
852 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
853 /*
854 * journal mode get enabled in different ways
855 * So just print the value even if we didn't specify it
856 */
857 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
858 seq_puts(seq, ",data=journal");
859 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
860 seq_puts(seq, ",data=ordered");
861 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
862 seq_puts(seq, ",data=writeback");
863
864 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
865 seq_printf(seq, ",inode_readahead_blks=%u",
866 sbi->s_inode_readahead_blks);
867
868 if (test_opt(sb, DATA_ERR_ABORT))
869 seq_puts(seq, ",data_err=abort");
870
871 ext4_show_quota_options(seq, sb);
872 return 0;
873 }
874
875
876 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
877 u64 ino, u32 generation)
878 {
879 struct inode *inode;
880
881 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
882 return ERR_PTR(-ESTALE);
883 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
884 return ERR_PTR(-ESTALE);
885
886 /* iget isn't really right if the inode is currently unallocated!!
887 *
888 * ext4_read_inode will return a bad_inode if the inode had been
889 * deleted, so we should be safe.
890 *
891 * Currently we don't know the generation for parent directory, so
892 * a generation of 0 means "accept any"
893 */
894 inode = ext4_iget(sb, ino);
895 if (IS_ERR(inode))
896 return ERR_CAST(inode);
897 if (generation && inode->i_generation != generation) {
898 iput(inode);
899 return ERR_PTR(-ESTALE);
900 }
901
902 return inode;
903 }
904
905 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
906 int fh_len, int fh_type)
907 {
908 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
909 ext4_nfs_get_inode);
910 }
911
912 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
913 int fh_len, int fh_type)
914 {
915 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
916 ext4_nfs_get_inode);
917 }
918
919 /*
920 * Try to release metadata pages (indirect blocks, directories) which are
921 * mapped via the block device. Since these pages could have journal heads
922 * which would prevent try_to_free_buffers() from freeing them, we must use
923 * jbd2 layer's try_to_free_buffers() function to release them.
924 */
925 static int bdev_try_to_free_page(struct super_block *sb, struct page *page, gfp_t wait)
926 {
927 journal_t *journal = EXT4_SB(sb)->s_journal;
928
929 WARN_ON(PageChecked(page));
930 if (!page_has_buffers(page))
931 return 0;
932 if (journal)
933 return jbd2_journal_try_to_free_buffers(journal, page,
934 wait & ~__GFP_WAIT);
935 return try_to_free_buffers(page);
936 }
937
938 #ifdef CONFIG_QUOTA
939 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
940 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
941
942 static int ext4_write_dquot(struct dquot *dquot);
943 static int ext4_acquire_dquot(struct dquot *dquot);
944 static int ext4_release_dquot(struct dquot *dquot);
945 static int ext4_mark_dquot_dirty(struct dquot *dquot);
946 static int ext4_write_info(struct super_block *sb, int type);
947 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
948 char *path, int remount);
949 static int ext4_quota_on_mount(struct super_block *sb, int type);
950 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
951 size_t len, loff_t off);
952 static ssize_t ext4_quota_write(struct super_block *sb, int type,
953 const char *data, size_t len, loff_t off);
954
955 static struct dquot_operations ext4_quota_operations = {
956 .initialize = dquot_initialize,
957 .drop = dquot_drop,
958 .alloc_space = dquot_alloc_space,
959 .reserve_space = dquot_reserve_space,
960 .claim_space = dquot_claim_space,
961 .release_rsv = dquot_release_reserved_space,
962 .get_reserved_space = ext4_get_reserved_space,
963 .alloc_inode = dquot_alloc_inode,
964 .free_space = dquot_free_space,
965 .free_inode = dquot_free_inode,
966 .transfer = dquot_transfer,
967 .write_dquot = ext4_write_dquot,
968 .acquire_dquot = ext4_acquire_dquot,
969 .release_dquot = ext4_release_dquot,
970 .mark_dirty = ext4_mark_dquot_dirty,
971 .write_info = ext4_write_info,
972 .alloc_dquot = dquot_alloc,
973 .destroy_dquot = dquot_destroy,
974 };
975
976 static struct quotactl_ops ext4_qctl_operations = {
977 .quota_on = ext4_quota_on,
978 .quota_off = vfs_quota_off,
979 .quota_sync = vfs_quota_sync,
980 .get_info = vfs_get_dqinfo,
981 .set_info = vfs_set_dqinfo,
982 .get_dqblk = vfs_get_dqblk,
983 .set_dqblk = vfs_set_dqblk
984 };
985 #endif
986
987 static const struct super_operations ext4_sops = {
988 .alloc_inode = ext4_alloc_inode,
989 .destroy_inode = ext4_destroy_inode,
990 .write_inode = ext4_write_inode,
991 .dirty_inode = ext4_dirty_inode,
992 .delete_inode = ext4_delete_inode,
993 .put_super = ext4_put_super,
994 .write_super = ext4_write_super,
995 .sync_fs = ext4_sync_fs,
996 .freeze_fs = ext4_freeze,
997 .unfreeze_fs = ext4_unfreeze,
998 .statfs = ext4_statfs,
999 .remount_fs = ext4_remount,
1000 .clear_inode = ext4_clear_inode,
1001 .show_options = ext4_show_options,
1002 #ifdef CONFIG_QUOTA
1003 .quota_read = ext4_quota_read,
1004 .quota_write = ext4_quota_write,
1005 #endif
1006 .bdev_try_to_free_page = bdev_try_to_free_page,
1007 };
1008
1009 static const struct export_operations ext4_export_ops = {
1010 .fh_to_dentry = ext4_fh_to_dentry,
1011 .fh_to_parent = ext4_fh_to_parent,
1012 .get_parent = ext4_get_parent,
1013 };
1014
1015 enum {
1016 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1017 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1018 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1019 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1020 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
1021 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1022 Opt_journal_update, Opt_journal_dev,
1023 Opt_journal_checksum, Opt_journal_async_commit,
1024 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1025 Opt_data_err_abort, Opt_data_err_ignore,
1026 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1027 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
1028 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
1029 Opt_grpquota, Opt_i_version,
1030 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1031 Opt_inode_readahead_blks, Opt_journal_ioprio
1032 };
1033
1034 static const match_table_t tokens = {
1035 {Opt_bsd_df, "bsddf"},
1036 {Opt_minix_df, "minixdf"},
1037 {Opt_grpid, "grpid"},
1038 {Opt_grpid, "bsdgroups"},
1039 {Opt_nogrpid, "nogrpid"},
1040 {Opt_nogrpid, "sysvgroups"},
1041 {Opt_resgid, "resgid=%u"},
1042 {Opt_resuid, "resuid=%u"},
1043 {Opt_sb, "sb=%u"},
1044 {Opt_err_cont, "errors=continue"},
1045 {Opt_err_panic, "errors=panic"},
1046 {Opt_err_ro, "errors=remount-ro"},
1047 {Opt_nouid32, "nouid32"},
1048 {Opt_debug, "debug"},
1049 {Opt_oldalloc, "oldalloc"},
1050 {Opt_orlov, "orlov"},
1051 {Opt_user_xattr, "user_xattr"},
1052 {Opt_nouser_xattr, "nouser_xattr"},
1053 {Opt_acl, "acl"},
1054 {Opt_noacl, "noacl"},
1055 {Opt_reservation, "reservation"},
1056 {Opt_noreservation, "noreservation"},
1057 {Opt_noload, "noload"},
1058 {Opt_nobh, "nobh"},
1059 {Opt_bh, "bh"},
1060 {Opt_commit, "commit=%u"},
1061 {Opt_min_batch_time, "min_batch_time=%u"},
1062 {Opt_max_batch_time, "max_batch_time=%u"},
1063 {Opt_journal_update, "journal=update"},
1064 {Opt_journal_dev, "journal_dev=%u"},
1065 {Opt_journal_checksum, "journal_checksum"},
1066 {Opt_journal_async_commit, "journal_async_commit"},
1067 {Opt_abort, "abort"},
1068 {Opt_data_journal, "data=journal"},
1069 {Opt_data_ordered, "data=ordered"},
1070 {Opt_data_writeback, "data=writeback"},
1071 {Opt_data_err_abort, "data_err=abort"},
1072 {Opt_data_err_ignore, "data_err=ignore"},
1073 {Opt_offusrjquota, "usrjquota="},
1074 {Opt_usrjquota, "usrjquota=%s"},
1075 {Opt_offgrpjquota, "grpjquota="},
1076 {Opt_grpjquota, "grpjquota=%s"},
1077 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1078 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1079 {Opt_grpquota, "grpquota"},
1080 {Opt_noquota, "noquota"},
1081 {Opt_quota, "quota"},
1082 {Opt_usrquota, "usrquota"},
1083 {Opt_barrier, "barrier=%u"},
1084 {Opt_i_version, "i_version"},
1085 {Opt_stripe, "stripe=%u"},
1086 {Opt_resize, "resize"},
1087 {Opt_delalloc, "delalloc"},
1088 {Opt_nodelalloc, "nodelalloc"},
1089 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1090 {Opt_journal_ioprio, "journal_ioprio=%u"},
1091 {Opt_err, NULL},
1092 };
1093
1094 static ext4_fsblk_t get_sb_block(void **data)
1095 {
1096 ext4_fsblk_t sb_block;
1097 char *options = (char *) *data;
1098
1099 if (!options || strncmp(options, "sb=", 3) != 0)
1100 return 1; /* Default location */
1101 options += 3;
1102 /*todo: use simple_strtoll with >32bit ext4 */
1103 sb_block = simple_strtoul(options, &options, 0);
1104 if (*options && *options != ',') {
1105 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1106 (char *) *data);
1107 return 1;
1108 }
1109 if (*options == ',')
1110 options++;
1111 *data = (void *) options;
1112 return sb_block;
1113 }
1114
1115 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1116
1117 static int parse_options(char *options, struct super_block *sb,
1118 unsigned long *journal_devnum,
1119 unsigned int *journal_ioprio,
1120 ext4_fsblk_t *n_blocks_count, int is_remount)
1121 {
1122 struct ext4_sb_info *sbi = EXT4_SB(sb);
1123 char *p;
1124 substring_t args[MAX_OPT_ARGS];
1125 int data_opt = 0;
1126 int option;
1127 #ifdef CONFIG_QUOTA
1128 int qtype, qfmt;
1129 char *qname;
1130 #endif
1131
1132 if (!options)
1133 return 1;
1134
1135 while ((p = strsep(&options, ",")) != NULL) {
1136 int token;
1137 if (!*p)
1138 continue;
1139
1140 token = match_token(p, tokens, args);
1141 switch (token) {
1142 case Opt_bsd_df:
1143 clear_opt(sbi->s_mount_opt, MINIX_DF);
1144 break;
1145 case Opt_minix_df:
1146 set_opt(sbi->s_mount_opt, MINIX_DF);
1147 break;
1148 case Opt_grpid:
1149 set_opt(sbi->s_mount_opt, GRPID);
1150 break;
1151 case Opt_nogrpid:
1152 clear_opt(sbi->s_mount_opt, GRPID);
1153 break;
1154 case Opt_resuid:
1155 if (match_int(&args[0], &option))
1156 return 0;
1157 sbi->s_resuid = option;
1158 break;
1159 case Opt_resgid:
1160 if (match_int(&args[0], &option))
1161 return 0;
1162 sbi->s_resgid = option;
1163 break;
1164 case Opt_sb:
1165 /* handled by get_sb_block() instead of here */
1166 /* *sb_block = match_int(&args[0]); */
1167 break;
1168 case Opt_err_panic:
1169 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1170 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1171 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1172 break;
1173 case Opt_err_ro:
1174 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1175 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1176 set_opt(sbi->s_mount_opt, ERRORS_RO);
1177 break;
1178 case Opt_err_cont:
1179 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1180 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1181 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1182 break;
1183 case Opt_nouid32:
1184 set_opt(sbi->s_mount_opt, NO_UID32);
1185 break;
1186 case Opt_debug:
1187 set_opt(sbi->s_mount_opt, DEBUG);
1188 break;
1189 case Opt_oldalloc:
1190 set_opt(sbi->s_mount_opt, OLDALLOC);
1191 break;
1192 case Opt_orlov:
1193 clear_opt(sbi->s_mount_opt, OLDALLOC);
1194 break;
1195 #ifdef CONFIG_EXT4_FS_XATTR
1196 case Opt_user_xattr:
1197 set_opt(sbi->s_mount_opt, XATTR_USER);
1198 break;
1199 case Opt_nouser_xattr:
1200 clear_opt(sbi->s_mount_opt, XATTR_USER);
1201 break;
1202 #else
1203 case Opt_user_xattr:
1204 case Opt_nouser_xattr:
1205 printk(KERN_ERR "EXT4 (no)user_xattr options "
1206 "not supported\n");
1207 break;
1208 #endif
1209 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1210 case Opt_acl:
1211 set_opt(sbi->s_mount_opt, POSIX_ACL);
1212 break;
1213 case Opt_noacl:
1214 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1215 break;
1216 #else
1217 case Opt_acl:
1218 case Opt_noacl:
1219 printk(KERN_ERR "EXT4 (no)acl options "
1220 "not supported\n");
1221 break;
1222 #endif
1223 case Opt_reservation:
1224 set_opt(sbi->s_mount_opt, RESERVATION);
1225 break;
1226 case Opt_noreservation:
1227 clear_opt(sbi->s_mount_opt, RESERVATION);
1228 break;
1229 case Opt_journal_update:
1230 /* @@@ FIXME */
1231 /* Eventually we will want to be able to create
1232 a journal file here. For now, only allow the
1233 user to specify an existing inode to be the
1234 journal file. */
1235 if (is_remount) {
1236 printk(KERN_ERR "EXT4-fs: cannot specify "
1237 "journal on remount\n");
1238 return 0;
1239 }
1240 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1241 break;
1242 case Opt_journal_dev:
1243 if (is_remount) {
1244 printk(KERN_ERR "EXT4-fs: cannot specify "
1245 "journal on remount\n");
1246 return 0;
1247 }
1248 if (match_int(&args[0], &option))
1249 return 0;
1250 *journal_devnum = option;
1251 break;
1252 case Opt_journal_checksum:
1253 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1254 break;
1255 case Opt_journal_async_commit:
1256 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1257 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1258 break;
1259 case Opt_noload:
1260 set_opt(sbi->s_mount_opt, NOLOAD);
1261 break;
1262 case Opt_commit:
1263 if (match_int(&args[0], &option))
1264 return 0;
1265 if (option < 0)
1266 return 0;
1267 if (option == 0)
1268 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1269 sbi->s_commit_interval = HZ * option;
1270 break;
1271 case Opt_max_batch_time:
1272 if (match_int(&args[0], &option))
1273 return 0;
1274 if (option < 0)
1275 return 0;
1276 if (option == 0)
1277 option = EXT4_DEF_MAX_BATCH_TIME;
1278 sbi->s_max_batch_time = option;
1279 break;
1280 case Opt_min_batch_time:
1281 if (match_int(&args[0], &option))
1282 return 0;
1283 if (option < 0)
1284 return 0;
1285 sbi->s_min_batch_time = option;
1286 break;
1287 case Opt_data_journal:
1288 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1289 goto datacheck;
1290 case Opt_data_ordered:
1291 data_opt = EXT4_MOUNT_ORDERED_DATA;
1292 goto datacheck;
1293 case Opt_data_writeback:
1294 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1295 datacheck:
1296 if (is_remount) {
1297 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1298 != data_opt) {
1299 printk(KERN_ERR
1300 "EXT4-fs: cannot change data "
1301 "mode on remount\n");
1302 return 0;
1303 }
1304 } else {
1305 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1306 sbi->s_mount_opt |= data_opt;
1307 }
1308 break;
1309 case Opt_data_err_abort:
1310 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1311 break;
1312 case Opt_data_err_ignore:
1313 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1314 break;
1315 #ifdef CONFIG_QUOTA
1316 case Opt_usrjquota:
1317 qtype = USRQUOTA;
1318 goto set_qf_name;
1319 case Opt_grpjquota:
1320 qtype = GRPQUOTA;
1321 set_qf_name:
1322 if (sb_any_quota_loaded(sb) &&
1323 !sbi->s_qf_names[qtype]) {
1324 printk(KERN_ERR
1325 "EXT4-fs: Cannot change journaled "
1326 "quota options when quota turned on.\n");
1327 return 0;
1328 }
1329 qname = match_strdup(&args[0]);
1330 if (!qname) {
1331 printk(KERN_ERR
1332 "EXT4-fs: not enough memory for "
1333 "storing quotafile name.\n");
1334 return 0;
1335 }
1336 if (sbi->s_qf_names[qtype] &&
1337 strcmp(sbi->s_qf_names[qtype], qname)) {
1338 printk(KERN_ERR
1339 "EXT4-fs: %s quota file already "
1340 "specified.\n", QTYPE2NAME(qtype));
1341 kfree(qname);
1342 return 0;
1343 }
1344 sbi->s_qf_names[qtype] = qname;
1345 if (strchr(sbi->s_qf_names[qtype], '/')) {
1346 printk(KERN_ERR
1347 "EXT4-fs: quotafile must be on "
1348 "filesystem root.\n");
1349 kfree(sbi->s_qf_names[qtype]);
1350 sbi->s_qf_names[qtype] = NULL;
1351 return 0;
1352 }
1353 set_opt(sbi->s_mount_opt, QUOTA);
1354 break;
1355 case Opt_offusrjquota:
1356 qtype = USRQUOTA;
1357 goto clear_qf_name;
1358 case Opt_offgrpjquota:
1359 qtype = GRPQUOTA;
1360 clear_qf_name:
1361 if (sb_any_quota_loaded(sb) &&
1362 sbi->s_qf_names[qtype]) {
1363 printk(KERN_ERR "EXT4-fs: Cannot change "
1364 "journaled quota options when "
1365 "quota turned on.\n");
1366 return 0;
1367 }
1368 /*
1369 * The space will be released later when all options
1370 * are confirmed to be correct
1371 */
1372 sbi->s_qf_names[qtype] = NULL;
1373 break;
1374 case Opt_jqfmt_vfsold:
1375 qfmt = QFMT_VFS_OLD;
1376 goto set_qf_format;
1377 case Opt_jqfmt_vfsv0:
1378 qfmt = QFMT_VFS_V0;
1379 set_qf_format:
1380 if (sb_any_quota_loaded(sb) &&
1381 sbi->s_jquota_fmt != qfmt) {
1382 printk(KERN_ERR "EXT4-fs: Cannot change "
1383 "journaled quota options when "
1384 "quota turned on.\n");
1385 return 0;
1386 }
1387 sbi->s_jquota_fmt = qfmt;
1388 break;
1389 case Opt_quota:
1390 case Opt_usrquota:
1391 set_opt(sbi->s_mount_opt, QUOTA);
1392 set_opt(sbi->s_mount_opt, USRQUOTA);
1393 break;
1394 case Opt_grpquota:
1395 set_opt(sbi->s_mount_opt, QUOTA);
1396 set_opt(sbi->s_mount_opt, GRPQUOTA);
1397 break;
1398 case Opt_noquota:
1399 if (sb_any_quota_loaded(sb)) {
1400 printk(KERN_ERR "EXT4-fs: Cannot change quota "
1401 "options when quota turned on.\n");
1402 return 0;
1403 }
1404 clear_opt(sbi->s_mount_opt, QUOTA);
1405 clear_opt(sbi->s_mount_opt, USRQUOTA);
1406 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1407 break;
1408 #else
1409 case Opt_quota:
1410 case Opt_usrquota:
1411 case Opt_grpquota:
1412 printk(KERN_ERR
1413 "EXT4-fs: quota options not supported.\n");
1414 break;
1415 case Opt_usrjquota:
1416 case Opt_grpjquota:
1417 case Opt_offusrjquota:
1418 case Opt_offgrpjquota:
1419 case Opt_jqfmt_vfsold:
1420 case Opt_jqfmt_vfsv0:
1421 printk(KERN_ERR
1422 "EXT4-fs: journaled quota options not "
1423 "supported.\n");
1424 break;
1425 case Opt_noquota:
1426 break;
1427 #endif
1428 case Opt_abort:
1429 set_opt(sbi->s_mount_opt, ABORT);
1430 break;
1431 case Opt_barrier:
1432 if (match_int(&args[0], &option))
1433 return 0;
1434 if (option)
1435 set_opt(sbi->s_mount_opt, BARRIER);
1436 else
1437 clear_opt(sbi->s_mount_opt, BARRIER);
1438 break;
1439 case Opt_ignore:
1440 break;
1441 case Opt_resize:
1442 if (!is_remount) {
1443 printk("EXT4-fs: resize option only available "
1444 "for remount\n");
1445 return 0;
1446 }
1447 if (match_int(&args[0], &option) != 0)
1448 return 0;
1449 *n_blocks_count = option;
1450 break;
1451 case Opt_nobh:
1452 set_opt(sbi->s_mount_opt, NOBH);
1453 break;
1454 case Opt_bh:
1455 clear_opt(sbi->s_mount_opt, NOBH);
1456 break;
1457 case Opt_i_version:
1458 set_opt(sbi->s_mount_opt, I_VERSION);
1459 sb->s_flags |= MS_I_VERSION;
1460 break;
1461 case Opt_nodelalloc:
1462 clear_opt(sbi->s_mount_opt, DELALLOC);
1463 break;
1464 case Opt_stripe:
1465 if (match_int(&args[0], &option))
1466 return 0;
1467 if (option < 0)
1468 return 0;
1469 sbi->s_stripe = option;
1470 break;
1471 case Opt_delalloc:
1472 set_opt(sbi->s_mount_opt, DELALLOC);
1473 break;
1474 case Opt_inode_readahead_blks:
1475 if (match_int(&args[0], &option))
1476 return 0;
1477 if (option < 0 || option > (1 << 30))
1478 return 0;
1479 if (option & (option - 1)) {
1480 printk(KERN_ERR "EXT4-fs: inode_readahead_blks"
1481 " must be a power of 2\n");
1482 return 0;
1483 }
1484 sbi->s_inode_readahead_blks = option;
1485 break;
1486 case Opt_journal_ioprio:
1487 if (match_int(&args[0], &option))
1488 return 0;
1489 if (option < 0 || option > 7)
1490 break;
1491 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1492 option);
1493 break;
1494 default:
1495 printk(KERN_ERR
1496 "EXT4-fs: Unrecognized mount option \"%s\" "
1497 "or missing value\n", p);
1498 return 0;
1499 }
1500 }
1501 #ifdef CONFIG_QUOTA
1502 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1503 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1504 sbi->s_qf_names[USRQUOTA])
1505 clear_opt(sbi->s_mount_opt, USRQUOTA);
1506
1507 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1508 sbi->s_qf_names[GRPQUOTA])
1509 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1510
1511 if ((sbi->s_qf_names[USRQUOTA] &&
1512 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1513 (sbi->s_qf_names[GRPQUOTA] &&
1514 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1515 printk(KERN_ERR "EXT4-fs: old and new quota "
1516 "format mixing.\n");
1517 return 0;
1518 }
1519
1520 if (!sbi->s_jquota_fmt) {
1521 printk(KERN_ERR "EXT4-fs: journaled quota format "
1522 "not specified.\n");
1523 return 0;
1524 }
1525 } else {
1526 if (sbi->s_jquota_fmt) {
1527 printk(KERN_ERR "EXT4-fs: journaled quota format "
1528 "specified with no journaling "
1529 "enabled.\n");
1530 return 0;
1531 }
1532 }
1533 #endif
1534 return 1;
1535 }
1536
1537 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1538 int read_only)
1539 {
1540 struct ext4_sb_info *sbi = EXT4_SB(sb);
1541 int res = 0;
1542
1543 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1544 printk(KERN_ERR "EXT4-fs warning: revision level too high, "
1545 "forcing read-only mode\n");
1546 res = MS_RDONLY;
1547 }
1548 if (read_only)
1549 return res;
1550 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1551 printk(KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1552 "running e2fsck is recommended\n");
1553 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1554 printk(KERN_WARNING
1555 "EXT4-fs warning: mounting fs with errors, "
1556 "running e2fsck is recommended\n");
1557 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1558 le16_to_cpu(es->s_mnt_count) >=
1559 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1560 printk(KERN_WARNING
1561 "EXT4-fs warning: maximal mount count reached, "
1562 "running e2fsck is recommended\n");
1563 else if (le32_to_cpu(es->s_checkinterval) &&
1564 (le32_to_cpu(es->s_lastcheck) +
1565 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1566 printk(KERN_WARNING
1567 "EXT4-fs warning: checktime reached, "
1568 "running e2fsck is recommended\n");
1569 if (!sbi->s_journal)
1570 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1571 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1572 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1573 le16_add_cpu(&es->s_mnt_count, 1);
1574 es->s_mtime = cpu_to_le32(get_seconds());
1575 ext4_update_dynamic_rev(sb);
1576 if (sbi->s_journal)
1577 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1578
1579 ext4_commit_super(sb, es, 1);
1580 if (test_opt(sb, DEBUG))
1581 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1582 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1583 sb->s_blocksize,
1584 sbi->s_groups_count,
1585 EXT4_BLOCKS_PER_GROUP(sb),
1586 EXT4_INODES_PER_GROUP(sb),
1587 sbi->s_mount_opt);
1588
1589 if (EXT4_SB(sb)->s_journal) {
1590 printk(KERN_INFO "EXT4 FS on %s, %s journal on %s\n",
1591 sb->s_id, EXT4_SB(sb)->s_journal->j_inode ? "internal" :
1592 "external", EXT4_SB(sb)->s_journal->j_devname);
1593 } else {
1594 printk(KERN_INFO "EXT4 FS on %s, no journal\n", sb->s_id);
1595 }
1596 return res;
1597 }
1598
1599 static int ext4_fill_flex_info(struct super_block *sb)
1600 {
1601 struct ext4_sb_info *sbi = EXT4_SB(sb);
1602 struct ext4_group_desc *gdp = NULL;
1603 struct buffer_head *bh;
1604 ext4_group_t flex_group_count;
1605 ext4_group_t flex_group;
1606 int groups_per_flex = 0;
1607 int i;
1608
1609 if (!sbi->s_es->s_log_groups_per_flex) {
1610 sbi->s_log_groups_per_flex = 0;
1611 return 1;
1612 }
1613
1614 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1615 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1616
1617 /* We allocate both existing and potentially added groups */
1618 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1619 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1620 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1621 sbi->s_flex_groups = kzalloc(flex_group_count *
1622 sizeof(struct flex_groups), GFP_KERNEL);
1623 if (sbi->s_flex_groups == NULL) {
1624 printk(KERN_ERR "EXT4-fs: not enough memory for "
1625 "%u flex groups\n", flex_group_count);
1626 goto failed;
1627 }
1628
1629 for (i = 0; i < sbi->s_groups_count; i++) {
1630 gdp = ext4_get_group_desc(sb, i, &bh);
1631
1632 flex_group = ext4_flex_group(sbi, i);
1633 sbi->s_flex_groups[flex_group].free_inodes +=
1634 ext4_free_inodes_count(sb, gdp);
1635 sbi->s_flex_groups[flex_group].free_blocks +=
1636 ext4_free_blks_count(sb, gdp);
1637 }
1638
1639 return 1;
1640 failed:
1641 return 0;
1642 }
1643
1644 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1645 struct ext4_group_desc *gdp)
1646 {
1647 __u16 crc = 0;
1648
1649 if (sbi->s_es->s_feature_ro_compat &
1650 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1651 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1652 __le32 le_group = cpu_to_le32(block_group);
1653
1654 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1655 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1656 crc = crc16(crc, (__u8 *)gdp, offset);
1657 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1658 /* for checksum of struct ext4_group_desc do the rest...*/
1659 if ((sbi->s_es->s_feature_incompat &
1660 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1661 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1662 crc = crc16(crc, (__u8 *)gdp + offset,
1663 le16_to_cpu(sbi->s_es->s_desc_size) -
1664 offset);
1665 }
1666
1667 return cpu_to_le16(crc);
1668 }
1669
1670 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1671 struct ext4_group_desc *gdp)
1672 {
1673 if ((sbi->s_es->s_feature_ro_compat &
1674 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1675 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1676 return 0;
1677
1678 return 1;
1679 }
1680
1681 /* Called at mount-time, super-block is locked */
1682 static int ext4_check_descriptors(struct super_block *sb)
1683 {
1684 struct ext4_sb_info *sbi = EXT4_SB(sb);
1685 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1686 ext4_fsblk_t last_block;
1687 ext4_fsblk_t block_bitmap;
1688 ext4_fsblk_t inode_bitmap;
1689 ext4_fsblk_t inode_table;
1690 int flexbg_flag = 0;
1691 ext4_group_t i;
1692
1693 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1694 flexbg_flag = 1;
1695
1696 ext4_debug("Checking group descriptors");
1697
1698 for (i = 0; i < sbi->s_groups_count; i++) {
1699 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1700
1701 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1702 last_block = ext4_blocks_count(sbi->s_es) - 1;
1703 else
1704 last_block = first_block +
1705 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1706
1707 block_bitmap = ext4_block_bitmap(sb, gdp);
1708 if (block_bitmap < first_block || block_bitmap > last_block) {
1709 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1710 "Block bitmap for group %u not in group "
1711 "(block %llu)!\n", i, block_bitmap);
1712 return 0;
1713 }
1714 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1715 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1716 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1717 "Inode bitmap for group %u not in group "
1718 "(block %llu)!\n", i, inode_bitmap);
1719 return 0;
1720 }
1721 inode_table = ext4_inode_table(sb, gdp);
1722 if (inode_table < first_block ||
1723 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1724 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1725 "Inode table for group %u not in group "
1726 "(block %llu)!\n", i, inode_table);
1727 return 0;
1728 }
1729 spin_lock(sb_bgl_lock(sbi, i));
1730 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1731 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1732 "Checksum for group %u failed (%u!=%u)\n",
1733 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1734 gdp)), le16_to_cpu(gdp->bg_checksum));
1735 if (!(sb->s_flags & MS_RDONLY)) {
1736 spin_unlock(sb_bgl_lock(sbi, i));
1737 return 0;
1738 }
1739 }
1740 spin_unlock(sb_bgl_lock(sbi, i));
1741 if (!flexbg_flag)
1742 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1743 }
1744
1745 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1746 sbi->s_es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
1747 return 1;
1748 }
1749
1750 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1751 * the superblock) which were deleted from all directories, but held open by
1752 * a process at the time of a crash. We walk the list and try to delete these
1753 * inodes at recovery time (only with a read-write filesystem).
1754 *
1755 * In order to keep the orphan inode chain consistent during traversal (in
1756 * case of crash during recovery), we link each inode into the superblock
1757 * orphan list_head and handle it the same way as an inode deletion during
1758 * normal operation (which journals the operations for us).
1759 *
1760 * We only do an iget() and an iput() on each inode, which is very safe if we
1761 * accidentally point at an in-use or already deleted inode. The worst that
1762 * can happen in this case is that we get a "bit already cleared" message from
1763 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1764 * e2fsck was run on this filesystem, and it must have already done the orphan
1765 * inode cleanup for us, so we can safely abort without any further action.
1766 */
1767 static void ext4_orphan_cleanup(struct super_block *sb,
1768 struct ext4_super_block *es)
1769 {
1770 unsigned int s_flags = sb->s_flags;
1771 int nr_orphans = 0, nr_truncates = 0;
1772 #ifdef CONFIG_QUOTA
1773 int i;
1774 #endif
1775 if (!es->s_last_orphan) {
1776 jbd_debug(4, "no orphan inodes to clean up\n");
1777 return;
1778 }
1779
1780 if (bdev_read_only(sb->s_bdev)) {
1781 printk(KERN_ERR "EXT4-fs: write access "
1782 "unavailable, skipping orphan cleanup.\n");
1783 return;
1784 }
1785
1786 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1787 if (es->s_last_orphan)
1788 jbd_debug(1, "Errors on filesystem, "
1789 "clearing orphan list.\n");
1790 es->s_last_orphan = 0;
1791 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1792 return;
1793 }
1794
1795 if (s_flags & MS_RDONLY) {
1796 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
1797 sb->s_id);
1798 sb->s_flags &= ~MS_RDONLY;
1799 }
1800 #ifdef CONFIG_QUOTA
1801 /* Needed for iput() to work correctly and not trash data */
1802 sb->s_flags |= MS_ACTIVE;
1803 /* Turn on quotas so that they are updated correctly */
1804 for (i = 0; i < MAXQUOTAS; i++) {
1805 if (EXT4_SB(sb)->s_qf_names[i]) {
1806 int ret = ext4_quota_on_mount(sb, i);
1807 if (ret < 0)
1808 printk(KERN_ERR
1809 "EXT4-fs: Cannot turn on journaled "
1810 "quota: error %d\n", ret);
1811 }
1812 }
1813 #endif
1814
1815 while (es->s_last_orphan) {
1816 struct inode *inode;
1817
1818 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1819 if (IS_ERR(inode)) {
1820 es->s_last_orphan = 0;
1821 break;
1822 }
1823
1824 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1825 vfs_dq_init(inode);
1826 if (inode->i_nlink) {
1827 printk(KERN_DEBUG
1828 "%s: truncating inode %lu to %lld bytes\n",
1829 __func__, inode->i_ino, inode->i_size);
1830 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1831 inode->i_ino, inode->i_size);
1832 ext4_truncate(inode);
1833 nr_truncates++;
1834 } else {
1835 printk(KERN_DEBUG
1836 "%s: deleting unreferenced inode %lu\n",
1837 __func__, inode->i_ino);
1838 jbd_debug(2, "deleting unreferenced inode %lu\n",
1839 inode->i_ino);
1840 nr_orphans++;
1841 }
1842 iput(inode); /* The delete magic happens here! */
1843 }
1844
1845 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1846
1847 if (nr_orphans)
1848 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
1849 sb->s_id, PLURAL(nr_orphans));
1850 if (nr_truncates)
1851 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
1852 sb->s_id, PLURAL(nr_truncates));
1853 #ifdef CONFIG_QUOTA
1854 /* Turn quotas off */
1855 for (i = 0; i < MAXQUOTAS; i++) {
1856 if (sb_dqopt(sb)->files[i])
1857 vfs_quota_off(sb, i, 0);
1858 }
1859 #endif
1860 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1861 }
1862 /*
1863 * Maximal extent format file size.
1864 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1865 * extent format containers, within a sector_t, and within i_blocks
1866 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1867 * so that won't be a limiting factor.
1868 *
1869 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1870 */
1871 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1872 {
1873 loff_t res;
1874 loff_t upper_limit = MAX_LFS_FILESIZE;
1875
1876 /* small i_blocks in vfs inode? */
1877 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1878 /*
1879 * CONFIG_LBD is not enabled implies the inode
1880 * i_block represent total blocks in 512 bytes
1881 * 32 == size of vfs inode i_blocks * 8
1882 */
1883 upper_limit = (1LL << 32) - 1;
1884
1885 /* total blocks in file system block size */
1886 upper_limit >>= (blkbits - 9);
1887 upper_limit <<= blkbits;
1888 }
1889
1890 /* 32-bit extent-start container, ee_block */
1891 res = 1LL << 32;
1892 res <<= blkbits;
1893 res -= 1;
1894
1895 /* Sanity check against vm- & vfs- imposed limits */
1896 if (res > upper_limit)
1897 res = upper_limit;
1898
1899 return res;
1900 }
1901
1902 /*
1903 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1904 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1905 * We need to be 1 filesystem block less than the 2^48 sector limit.
1906 */
1907 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
1908 {
1909 loff_t res = EXT4_NDIR_BLOCKS;
1910 int meta_blocks;
1911 loff_t upper_limit;
1912 /* This is calculated to be the largest file size for a
1913 * dense, bitmapped file such that the total number of
1914 * sectors in the file, including data and all indirect blocks,
1915 * does not exceed 2^48 -1
1916 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1917 * total number of 512 bytes blocks of the file
1918 */
1919
1920 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1921 /*
1922 * !has_huge_files or CONFIG_LBD is not enabled
1923 * implies the inode i_block represent total blocks in
1924 * 512 bytes 32 == size of vfs inode i_blocks * 8
1925 */
1926 upper_limit = (1LL << 32) - 1;
1927
1928 /* total blocks in file system block size */
1929 upper_limit >>= (bits - 9);
1930
1931 } else {
1932 /*
1933 * We use 48 bit ext4_inode i_blocks
1934 * With EXT4_HUGE_FILE_FL set the i_blocks
1935 * represent total number of blocks in
1936 * file system block size
1937 */
1938 upper_limit = (1LL << 48) - 1;
1939
1940 }
1941
1942 /* indirect blocks */
1943 meta_blocks = 1;
1944 /* double indirect blocks */
1945 meta_blocks += 1 + (1LL << (bits-2));
1946 /* tripple indirect blocks */
1947 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1948
1949 upper_limit -= meta_blocks;
1950 upper_limit <<= bits;
1951
1952 res += 1LL << (bits-2);
1953 res += 1LL << (2*(bits-2));
1954 res += 1LL << (3*(bits-2));
1955 res <<= bits;
1956 if (res > upper_limit)
1957 res = upper_limit;
1958
1959 if (res > MAX_LFS_FILESIZE)
1960 res = MAX_LFS_FILESIZE;
1961
1962 return res;
1963 }
1964
1965 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
1966 ext4_fsblk_t logical_sb_block, int nr)
1967 {
1968 struct ext4_sb_info *sbi = EXT4_SB(sb);
1969 ext4_group_t bg, first_meta_bg;
1970 int has_super = 0;
1971
1972 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1973
1974 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
1975 nr < first_meta_bg)
1976 return logical_sb_block + nr + 1;
1977 bg = sbi->s_desc_per_block * nr;
1978 if (ext4_bg_has_super(sb, bg))
1979 has_super = 1;
1980 return (has_super + ext4_group_first_block_no(sb, bg));
1981 }
1982
1983 /**
1984 * ext4_get_stripe_size: Get the stripe size.
1985 * @sbi: In memory super block info
1986 *
1987 * If we have specified it via mount option, then
1988 * use the mount option value. If the value specified at mount time is
1989 * greater than the blocks per group use the super block value.
1990 * If the super block value is greater than blocks per group return 0.
1991 * Allocator needs it be less than blocks per group.
1992 *
1993 */
1994 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
1995 {
1996 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
1997 unsigned long stripe_width =
1998 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
1999
2000 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2001 return sbi->s_stripe;
2002
2003 if (stripe_width <= sbi->s_blocks_per_group)
2004 return stripe_width;
2005
2006 if (stride <= sbi->s_blocks_per_group)
2007 return stride;
2008
2009 return 0;
2010 }
2011
2012 /* sysfs supprt */
2013
2014 struct ext4_attr {
2015 struct attribute attr;
2016 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2017 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2018 const char *, size_t);
2019 int offset;
2020 };
2021
2022 static int parse_strtoul(const char *buf,
2023 unsigned long max, unsigned long *value)
2024 {
2025 char *endp;
2026
2027 while (*buf && isspace(*buf))
2028 buf++;
2029 *value = simple_strtoul(buf, &endp, 0);
2030 while (*endp && isspace(*endp))
2031 endp++;
2032 if (*endp || *value > max)
2033 return -EINVAL;
2034
2035 return 0;
2036 }
2037
2038 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2039 struct ext4_sb_info *sbi,
2040 char *buf)
2041 {
2042 return snprintf(buf, PAGE_SIZE, "%llu\n",
2043 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2044 }
2045
2046 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2047 struct ext4_sb_info *sbi, char *buf)
2048 {
2049 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2050
2051 return snprintf(buf, PAGE_SIZE, "%lu\n",
2052 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2053 sbi->s_sectors_written_start) >> 1);
2054 }
2055
2056 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2057 struct ext4_sb_info *sbi, char *buf)
2058 {
2059 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2060
2061 return snprintf(buf, PAGE_SIZE, "%llu\n",
2062 sbi->s_kbytes_written +
2063 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2064 EXT4_SB(sb)->s_sectors_written_start) >> 1));
2065 }
2066
2067 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2068 struct ext4_sb_info *sbi,
2069 const char *buf, size_t count)
2070 {
2071 unsigned long t;
2072
2073 if (parse_strtoul(buf, 0x40000000, &t))
2074 return -EINVAL;
2075
2076 /* inode_readahead_blks must be a power of 2 */
2077 if (t & (t-1))
2078 return -EINVAL;
2079
2080 sbi->s_inode_readahead_blks = t;
2081 return count;
2082 }
2083
2084 static ssize_t sbi_ui_show(struct ext4_attr *a,
2085 struct ext4_sb_info *sbi, char *buf)
2086 {
2087 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2088
2089 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2090 }
2091
2092 static ssize_t sbi_ui_store(struct ext4_attr *a,
2093 struct ext4_sb_info *sbi,
2094 const char *buf, size_t count)
2095 {
2096 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2097 unsigned long t;
2098
2099 if (parse_strtoul(buf, 0xffffffff, &t))
2100 return -EINVAL;
2101 *ui = t;
2102 return count;
2103 }
2104
2105 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2106 static struct ext4_attr ext4_attr_##_name = { \
2107 .attr = {.name = __stringify(_name), .mode = _mode }, \
2108 .show = _show, \
2109 .store = _store, \
2110 .offset = offsetof(struct ext4_sb_info, _elname), \
2111 }
2112 #define EXT4_ATTR(name, mode, show, store) \
2113 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2114
2115 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2116 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2117 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2118 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2119 #define ATTR_LIST(name) &ext4_attr_##name.attr
2120
2121 EXT4_RO_ATTR(delayed_allocation_blocks);
2122 EXT4_RO_ATTR(session_write_kbytes);
2123 EXT4_RO_ATTR(lifetime_write_kbytes);
2124 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2125 inode_readahead_blks_store, s_inode_readahead_blks);
2126 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2127 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2128 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2129 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2130 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2131 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2132
2133 static struct attribute *ext4_attrs[] = {
2134 ATTR_LIST(delayed_allocation_blocks),
2135 ATTR_LIST(session_write_kbytes),
2136 ATTR_LIST(lifetime_write_kbytes),
2137 ATTR_LIST(inode_readahead_blks),
2138 ATTR_LIST(mb_stats),
2139 ATTR_LIST(mb_max_to_scan),
2140 ATTR_LIST(mb_min_to_scan),
2141 ATTR_LIST(mb_order2_req),
2142 ATTR_LIST(mb_stream_req),
2143 ATTR_LIST(mb_group_prealloc),
2144 NULL,
2145 };
2146
2147 static ssize_t ext4_attr_show(struct kobject *kobj,
2148 struct attribute *attr, char *buf)
2149 {
2150 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2151 s_kobj);
2152 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2153
2154 return a->show ? a->show(a, sbi, buf) : 0;
2155 }
2156
2157 static ssize_t ext4_attr_store(struct kobject *kobj,
2158 struct attribute *attr,
2159 const char *buf, size_t len)
2160 {
2161 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2162 s_kobj);
2163 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2164
2165 return a->store ? a->store(a, sbi, buf, len) : 0;
2166 }
2167
2168 static void ext4_sb_release(struct kobject *kobj)
2169 {
2170 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2171 s_kobj);
2172 complete(&sbi->s_kobj_unregister);
2173 }
2174
2175
2176 static struct sysfs_ops ext4_attr_ops = {
2177 .show = ext4_attr_show,
2178 .store = ext4_attr_store,
2179 };
2180
2181 static struct kobj_type ext4_ktype = {
2182 .default_attrs = ext4_attrs,
2183 .sysfs_ops = &ext4_attr_ops,
2184 .release = ext4_sb_release,
2185 };
2186
2187 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2188 __releases(kernel_lock)
2189 __acquires(kernel_lock)
2190
2191 {
2192 struct buffer_head *bh;
2193 struct ext4_super_block *es = NULL;
2194 struct ext4_sb_info *sbi;
2195 ext4_fsblk_t block;
2196 ext4_fsblk_t sb_block = get_sb_block(&data);
2197 ext4_fsblk_t logical_sb_block;
2198 unsigned long offset = 0;
2199 unsigned long journal_devnum = 0;
2200 unsigned long def_mount_opts;
2201 struct inode *root;
2202 char *cp;
2203 const char *descr;
2204 int ret = -EINVAL;
2205 int blocksize;
2206 unsigned int db_count;
2207 unsigned int i;
2208 int needs_recovery, has_huge_files;
2209 int features;
2210 __u64 blocks_count;
2211 int err;
2212 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2213
2214 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2215 if (!sbi)
2216 return -ENOMEM;
2217
2218 sbi->s_blockgroup_lock =
2219 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2220 if (!sbi->s_blockgroup_lock) {
2221 kfree(sbi);
2222 return -ENOMEM;
2223 }
2224 sb->s_fs_info = sbi;
2225 sbi->s_mount_opt = 0;
2226 sbi->s_resuid = EXT4_DEF_RESUID;
2227 sbi->s_resgid = EXT4_DEF_RESGID;
2228 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2229 sbi->s_sb_block = sb_block;
2230 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2231 sectors[1]);
2232
2233 unlock_kernel();
2234
2235 /* Cleanup superblock name */
2236 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2237 *cp = '!';
2238
2239 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2240 if (!blocksize) {
2241 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
2242 goto out_fail;
2243 }
2244
2245 /*
2246 * The ext4 superblock will not be buffer aligned for other than 1kB
2247 * block sizes. We need to calculate the offset from buffer start.
2248 */
2249 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2250 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2251 offset = do_div(logical_sb_block, blocksize);
2252 } else {
2253 logical_sb_block = sb_block;
2254 }
2255
2256 if (!(bh = sb_bread(sb, logical_sb_block))) {
2257 printk(KERN_ERR "EXT4-fs: unable to read superblock\n");
2258 goto out_fail;
2259 }
2260 /*
2261 * Note: s_es must be initialized as soon as possible because
2262 * some ext4 macro-instructions depend on its value
2263 */
2264 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2265 sbi->s_es = es;
2266 sb->s_magic = le16_to_cpu(es->s_magic);
2267 if (sb->s_magic != EXT4_SUPER_MAGIC)
2268 goto cantfind_ext4;
2269 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2270
2271 /* Set defaults before we parse the mount options */
2272 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2273 if (def_mount_opts & EXT4_DEFM_DEBUG)
2274 set_opt(sbi->s_mount_opt, DEBUG);
2275 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
2276 set_opt(sbi->s_mount_opt, GRPID);
2277 if (def_mount_opts & EXT4_DEFM_UID16)
2278 set_opt(sbi->s_mount_opt, NO_UID32);
2279 #ifdef CONFIG_EXT4_FS_XATTR
2280 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2281 set_opt(sbi->s_mount_opt, XATTR_USER);
2282 #endif
2283 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2284 if (def_mount_opts & EXT4_DEFM_ACL)
2285 set_opt(sbi->s_mount_opt, POSIX_ACL);
2286 #endif
2287 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2288 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2289 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2290 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2291 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2292 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2293
2294 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2295 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2296 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2297 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2298 else
2299 set_opt(sbi->s_mount_opt, ERRORS_RO);
2300
2301 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2302 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2303 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2304 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2305 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2306
2307 set_opt(sbi->s_mount_opt, RESERVATION);
2308 set_opt(sbi->s_mount_opt, BARRIER);
2309
2310 /*
2311 * enable delayed allocation by default
2312 * Use -o nodelalloc to turn it off
2313 */
2314 set_opt(sbi->s_mount_opt, DELALLOC);
2315
2316
2317 if (!parse_options((char *) data, sb, &journal_devnum,
2318 &journal_ioprio, NULL, 0))
2319 goto failed_mount;
2320
2321 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2322 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2323
2324 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2325 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2326 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2327 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2328 printk(KERN_WARNING
2329 "EXT4-fs warning: feature flags set on rev 0 fs, "
2330 "running e2fsck is recommended\n");
2331
2332 /*
2333 * Check feature flags regardless of the revision level, since we
2334 * previously didn't change the revision level when setting the flags,
2335 * so there is a chance incompat flags are set on a rev 0 filesystem.
2336 */
2337 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
2338 if (features) {
2339 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
2340 "unsupported optional features (%x).\n", sb->s_id,
2341 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2342 ~EXT4_FEATURE_INCOMPAT_SUPP));
2343 goto failed_mount;
2344 }
2345 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
2346 if (!(sb->s_flags & MS_RDONLY) && features) {
2347 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
2348 "unsupported optional features (%x).\n", sb->s_id,
2349 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2350 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2351 goto failed_mount;
2352 }
2353 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2354 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2355 if (has_huge_files) {
2356 /*
2357 * Large file size enabled file system can only be
2358 * mount if kernel is build with CONFIG_LBD
2359 */
2360 if (sizeof(root->i_blocks) < sizeof(u64) &&
2361 !(sb->s_flags & MS_RDONLY)) {
2362 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2363 "files cannot be mounted read-write "
2364 "without CONFIG_LBD.\n", sb->s_id);
2365 goto failed_mount;
2366 }
2367 }
2368 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2369
2370 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2371 blocksize > EXT4_MAX_BLOCK_SIZE) {
2372 printk(KERN_ERR
2373 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
2374 blocksize, sb->s_id);
2375 goto failed_mount;
2376 }
2377
2378 if (sb->s_blocksize != blocksize) {
2379
2380 /* Validate the filesystem blocksize */
2381 if (!sb_set_blocksize(sb, blocksize)) {
2382 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2383 blocksize);
2384 goto failed_mount;
2385 }
2386
2387 brelse(bh);
2388 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2389 offset = do_div(logical_sb_block, blocksize);
2390 bh = sb_bread(sb, logical_sb_block);
2391 if (!bh) {
2392 printk(KERN_ERR
2393 "EXT4-fs: Can't read superblock on 2nd try.\n");
2394 goto failed_mount;
2395 }
2396 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2397 sbi->s_es = es;
2398 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2399 printk(KERN_ERR
2400 "EXT4-fs: Magic mismatch, very weird !\n");
2401 goto failed_mount;
2402 }
2403 }
2404
2405 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2406 has_huge_files);
2407 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2408
2409 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2410 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2411 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2412 } else {
2413 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2414 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2415 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2416 (!is_power_of_2(sbi->s_inode_size)) ||
2417 (sbi->s_inode_size > blocksize)) {
2418 printk(KERN_ERR
2419 "EXT4-fs: unsupported inode size: %d\n",
2420 sbi->s_inode_size);
2421 goto failed_mount;
2422 }
2423 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2424 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2425 }
2426 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2427 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2428 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2429 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2430 !is_power_of_2(sbi->s_desc_size)) {
2431 printk(KERN_ERR
2432 "EXT4-fs: unsupported descriptor size %lu\n",
2433 sbi->s_desc_size);
2434 goto failed_mount;
2435 }
2436 } else
2437 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2438 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2439 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2440 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2441 goto cantfind_ext4;
2442 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2443 if (sbi->s_inodes_per_block == 0)
2444 goto cantfind_ext4;
2445 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2446 sbi->s_inodes_per_block;
2447 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2448 sbi->s_sbh = bh;
2449 sbi->s_mount_state = le16_to_cpu(es->s_state);
2450 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2451 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2452 for (i = 0; i < 4; i++)
2453 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2454 sbi->s_def_hash_version = es->s_def_hash_version;
2455 i = le32_to_cpu(es->s_flags);
2456 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2457 sbi->s_hash_unsigned = 3;
2458 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2459 #ifdef __CHAR_UNSIGNED__
2460 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2461 sbi->s_hash_unsigned = 3;
2462 #else
2463 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2464 #endif
2465 sb->s_dirt = 1;
2466 }
2467
2468 if (sbi->s_blocks_per_group > blocksize * 8) {
2469 printk(KERN_ERR
2470 "EXT4-fs: #blocks per group too big: %lu\n",
2471 sbi->s_blocks_per_group);
2472 goto failed_mount;
2473 }
2474 if (sbi->s_inodes_per_group > blocksize * 8) {
2475 printk(KERN_ERR
2476 "EXT4-fs: #inodes per group too big: %lu\n",
2477 sbi->s_inodes_per_group);
2478 goto failed_mount;
2479 }
2480
2481 if (ext4_blocks_count(es) >
2482 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
2483 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
2484 " too large to mount safely\n", sb->s_id);
2485 if (sizeof(sector_t) < 8)
2486 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
2487 "enabled\n");
2488 goto failed_mount;
2489 }
2490
2491 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2492 goto cantfind_ext4;
2493
2494 /*
2495 * It makes no sense for the first data block to be beyond the end
2496 * of the filesystem.
2497 */
2498 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2499 printk(KERN_WARNING "EXT4-fs: bad geometry: first data"
2500 "block %u is beyond end of filesystem (%llu)\n",
2501 le32_to_cpu(es->s_first_data_block),
2502 ext4_blocks_count(es));
2503 goto failed_mount;
2504 }
2505 blocks_count = (ext4_blocks_count(es) -
2506 le32_to_cpu(es->s_first_data_block) +
2507 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2508 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2509 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2510 printk(KERN_WARNING "EXT4-fs: groups count too large: %u "
2511 "(block count %llu, first data block %u, "
2512 "blocks per group %lu)\n", sbi->s_groups_count,
2513 ext4_blocks_count(es),
2514 le32_to_cpu(es->s_first_data_block),
2515 EXT4_BLOCKS_PER_GROUP(sb));
2516 goto failed_mount;
2517 }
2518 sbi->s_groups_count = blocks_count;
2519 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2520 EXT4_DESC_PER_BLOCK(sb);
2521 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2522 GFP_KERNEL);
2523 if (sbi->s_group_desc == NULL) {
2524 printk(KERN_ERR "EXT4-fs: not enough memory\n");
2525 goto failed_mount;
2526 }
2527
2528 #ifdef CONFIG_PROC_FS
2529 if (ext4_proc_root)
2530 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2531 #endif
2532
2533 bgl_lock_init(sbi->s_blockgroup_lock);
2534
2535 for (i = 0; i < db_count; i++) {
2536 block = descriptor_loc(sb, logical_sb_block, i);
2537 sbi->s_group_desc[i] = sb_bread(sb, block);
2538 if (!sbi->s_group_desc[i]) {
2539 printk(KERN_ERR "EXT4-fs: "
2540 "can't read group descriptor %d\n", i);
2541 db_count = i;
2542 goto failed_mount2;
2543 }
2544 }
2545 if (!ext4_check_descriptors(sb)) {
2546 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
2547 goto failed_mount2;
2548 }
2549 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2550 if (!ext4_fill_flex_info(sb)) {
2551 printk(KERN_ERR
2552 "EXT4-fs: unable to initialize "
2553 "flex_bg meta info!\n");
2554 goto failed_mount2;
2555 }
2556
2557 sbi->s_gdb_count = db_count;
2558 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2559 spin_lock_init(&sbi->s_next_gen_lock);
2560
2561 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2562 ext4_count_free_blocks(sb));
2563 if (!err) {
2564 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2565 ext4_count_free_inodes(sb));
2566 }
2567 if (!err) {
2568 err = percpu_counter_init(&sbi->s_dirs_counter,
2569 ext4_count_dirs(sb));
2570 }
2571 if (!err) {
2572 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2573 }
2574 if (err) {
2575 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2576 goto failed_mount3;
2577 }
2578
2579 sbi->s_stripe = ext4_get_stripe_size(sbi);
2580
2581 /*
2582 * set up enough so that it can read an inode
2583 */
2584 sb->s_op = &ext4_sops;
2585 sb->s_export_op = &ext4_export_ops;
2586 sb->s_xattr = ext4_xattr_handlers;
2587 #ifdef CONFIG_QUOTA
2588 sb->s_qcop = &ext4_qctl_operations;
2589 sb->dq_op = &ext4_quota_operations;
2590 #endif
2591 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2592
2593 sb->s_root = NULL;
2594
2595 needs_recovery = (es->s_last_orphan != 0 ||
2596 EXT4_HAS_INCOMPAT_FEATURE(sb,
2597 EXT4_FEATURE_INCOMPAT_RECOVER));
2598
2599 /*
2600 * The first inode we look at is the journal inode. Don't try
2601 * root first: it may be modified in the journal!
2602 */
2603 if (!test_opt(sb, NOLOAD) &&
2604 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2605 if (ext4_load_journal(sb, es, journal_devnum))
2606 goto failed_mount3;
2607 if (!(sb->s_flags & MS_RDONLY) &&
2608 EXT4_SB(sb)->s_journal->j_failed_commit) {
2609 printk(KERN_CRIT "EXT4-fs error (device %s): "
2610 "ext4_fill_super: Journal transaction "
2611 "%u is corrupt\n", sb->s_id,
2612 EXT4_SB(sb)->s_journal->j_failed_commit);
2613 if (test_opt(sb, ERRORS_RO)) {
2614 printk(KERN_CRIT
2615 "Mounting filesystem read-only\n");
2616 sb->s_flags |= MS_RDONLY;
2617 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2618 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2619 }
2620 if (test_opt(sb, ERRORS_PANIC)) {
2621 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2622 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2623 ext4_commit_super(sb, es, 1);
2624 goto failed_mount4;
2625 }
2626 }
2627 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2628 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2629 printk(KERN_ERR "EXT4-fs: required journal recovery "
2630 "suppressed and not mounted read-only\n");
2631 goto failed_mount4;
2632 } else {
2633 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2634 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2635 sbi->s_journal = NULL;
2636 needs_recovery = 0;
2637 goto no_journal;
2638 }
2639
2640 if (ext4_blocks_count(es) > 0xffffffffULL &&
2641 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2642 JBD2_FEATURE_INCOMPAT_64BIT)) {
2643 printk(KERN_ERR "EXT4-fs: Failed to set 64-bit journal feature\n");
2644 goto failed_mount4;
2645 }
2646
2647 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2648 jbd2_journal_set_features(sbi->s_journal,
2649 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2650 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2651 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2652 jbd2_journal_set_features(sbi->s_journal,
2653 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2654 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2655 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2656 } else {
2657 jbd2_journal_clear_features(sbi->s_journal,
2658 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2659 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2660 }
2661
2662 /* We have now updated the journal if required, so we can
2663 * validate the data journaling mode. */
2664 switch (test_opt(sb, DATA_FLAGS)) {
2665 case 0:
2666 /* No mode set, assume a default based on the journal
2667 * capabilities: ORDERED_DATA if the journal can
2668 * cope, else JOURNAL_DATA
2669 */
2670 if (jbd2_journal_check_available_features
2671 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2672 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2673 else
2674 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2675 break;
2676
2677 case EXT4_MOUNT_ORDERED_DATA:
2678 case EXT4_MOUNT_WRITEBACK_DATA:
2679 if (!jbd2_journal_check_available_features
2680 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2681 printk(KERN_ERR "EXT4-fs: Journal does not support "
2682 "requested data journaling mode\n");
2683 goto failed_mount4;
2684 }
2685 default:
2686 break;
2687 }
2688 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2689
2690 no_journal:
2691
2692 if (test_opt(sb, NOBH)) {
2693 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2694 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
2695 "its supported only with writeback mode\n");
2696 clear_opt(sbi->s_mount_opt, NOBH);
2697 }
2698 }
2699 /*
2700 * The jbd2_journal_load will have done any necessary log recovery,
2701 * so we can safely mount the rest of the filesystem now.
2702 */
2703
2704 root = ext4_iget(sb, EXT4_ROOT_INO);
2705 if (IS_ERR(root)) {
2706 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
2707 ret = PTR_ERR(root);
2708 goto failed_mount4;
2709 }
2710 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2711 iput(root);
2712 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
2713 goto failed_mount4;
2714 }
2715 sb->s_root = d_alloc_root(root);
2716 if (!sb->s_root) {
2717 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2718 iput(root);
2719 ret = -ENOMEM;
2720 goto failed_mount4;
2721 }
2722
2723 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2724
2725 /* determine the minimum size of new large inodes, if present */
2726 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2727 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2728 EXT4_GOOD_OLD_INODE_SIZE;
2729 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2730 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2731 if (sbi->s_want_extra_isize <
2732 le16_to_cpu(es->s_want_extra_isize))
2733 sbi->s_want_extra_isize =
2734 le16_to_cpu(es->s_want_extra_isize);
2735 if (sbi->s_want_extra_isize <
2736 le16_to_cpu(es->s_min_extra_isize))
2737 sbi->s_want_extra_isize =
2738 le16_to_cpu(es->s_min_extra_isize);
2739 }
2740 }
2741 /* Check if enough inode space is available */
2742 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2743 sbi->s_inode_size) {
2744 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2745 EXT4_GOOD_OLD_INODE_SIZE;
2746 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2747 "available.\n");
2748 }
2749
2750 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2751 printk(KERN_WARNING "EXT4-fs: Ignoring delalloc option - "
2752 "requested data journaling mode\n");
2753 clear_opt(sbi->s_mount_opt, DELALLOC);
2754 } else if (test_opt(sb, DELALLOC))
2755 printk(KERN_INFO "EXT4-fs: delayed allocation enabled\n");
2756
2757 ext4_ext_init(sb);
2758 err = ext4_mb_init(sb, needs_recovery);
2759 if (err) {
2760 printk(KERN_ERR "EXT4-fs: failed to initalize mballoc (%d)\n",
2761 err);
2762 goto failed_mount4;
2763 }
2764
2765 sbi->s_kobj.kset = ext4_kset;
2766 init_completion(&sbi->s_kobj_unregister);
2767 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2768 "%s", sb->s_id);
2769 if (err) {
2770 ext4_mb_release(sb);
2771 ext4_ext_release(sb);
2772 goto failed_mount4;
2773 };
2774
2775 /*
2776 * akpm: core read_super() calls in here with the superblock locked.
2777 * That deadlocks, because orphan cleanup needs to lock the superblock
2778 * in numerous places. Here we just pop the lock - it's relatively
2779 * harmless, because we are now ready to accept write_super() requests,
2780 * and aviro says that's the only reason for hanging onto the
2781 * superblock lock.
2782 */
2783 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2784 ext4_orphan_cleanup(sb, es);
2785 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2786 if (needs_recovery) {
2787 printk(KERN_INFO "EXT4-fs: recovery complete.\n");
2788 ext4_mark_recovery_complete(sb, es);
2789 }
2790 if (EXT4_SB(sb)->s_journal) {
2791 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2792 descr = " journalled data mode";
2793 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2794 descr = " ordered data mode";
2795 else
2796 descr = " writeback data mode";
2797 } else
2798 descr = "out journal";
2799
2800 printk(KERN_INFO "EXT4-fs: mounted filesystem %s with%s\n",
2801 sb->s_id, descr);
2802
2803 lock_kernel();
2804 return 0;
2805
2806 cantfind_ext4:
2807 if (!silent)
2808 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
2809 sb->s_id);
2810 goto failed_mount;
2811
2812 failed_mount4:
2813 printk(KERN_ERR "EXT4-fs (device %s): mount failed\n", sb->s_id);
2814 if (sbi->s_journal) {
2815 jbd2_journal_destroy(sbi->s_journal);
2816 sbi->s_journal = NULL;
2817 }
2818 failed_mount3:
2819 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2820 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2821 percpu_counter_destroy(&sbi->s_dirs_counter);
2822 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2823 failed_mount2:
2824 for (i = 0; i < db_count; i++)
2825 brelse(sbi->s_group_desc[i]);
2826 kfree(sbi->s_group_desc);
2827 failed_mount:
2828 if (sbi->s_proc) {
2829 remove_proc_entry(sb->s_id, ext4_proc_root);
2830 }
2831 #ifdef CONFIG_QUOTA
2832 for (i = 0; i < MAXQUOTAS; i++)
2833 kfree(sbi->s_qf_names[i]);
2834 #endif
2835 ext4_blkdev_remove(sbi);
2836 brelse(bh);
2837 out_fail:
2838 sb->s_fs_info = NULL;
2839 kfree(sbi);
2840 lock_kernel();
2841 return ret;
2842 }
2843
2844 /*
2845 * Setup any per-fs journal parameters now. We'll do this both on
2846 * initial mount, once the journal has been initialised but before we've
2847 * done any recovery; and again on any subsequent remount.
2848 */
2849 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2850 {
2851 struct ext4_sb_info *sbi = EXT4_SB(sb);
2852
2853 journal->j_commit_interval = sbi->s_commit_interval;
2854 journal->j_min_batch_time = sbi->s_min_batch_time;
2855 journal->j_max_batch_time = sbi->s_max_batch_time;
2856
2857 spin_lock(&journal->j_state_lock);
2858 if (test_opt(sb, BARRIER))
2859 journal->j_flags |= JBD2_BARRIER;
2860 else
2861 journal->j_flags &= ~JBD2_BARRIER;
2862 if (test_opt(sb, DATA_ERR_ABORT))
2863 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2864 else
2865 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
2866 spin_unlock(&journal->j_state_lock);
2867 }
2868
2869 static journal_t *ext4_get_journal(struct super_block *sb,
2870 unsigned int journal_inum)
2871 {
2872 struct inode *journal_inode;
2873 journal_t *journal;
2874
2875 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2876
2877 /* First, test for the existence of a valid inode on disk. Bad
2878 * things happen if we iget() an unused inode, as the subsequent
2879 * iput() will try to delete it. */
2880
2881 journal_inode = ext4_iget(sb, journal_inum);
2882 if (IS_ERR(journal_inode)) {
2883 printk(KERN_ERR "EXT4-fs: no journal found.\n");
2884 return NULL;
2885 }
2886 if (!journal_inode->i_nlink) {
2887 make_bad_inode(journal_inode);
2888 iput(journal_inode);
2889 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
2890 return NULL;
2891 }
2892
2893 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
2894 journal_inode, journal_inode->i_size);
2895 if (!S_ISREG(journal_inode->i_mode)) {
2896 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
2897 iput(journal_inode);
2898 return NULL;
2899 }
2900
2901 journal = jbd2_journal_init_inode(journal_inode);
2902 if (!journal) {
2903 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
2904 iput(journal_inode);
2905 return NULL;
2906 }
2907 journal->j_private = sb;
2908 ext4_init_journal_params(sb, journal);
2909 return journal;
2910 }
2911
2912 static journal_t *ext4_get_dev_journal(struct super_block *sb,
2913 dev_t j_dev)
2914 {
2915 struct buffer_head *bh;
2916 journal_t *journal;
2917 ext4_fsblk_t start;
2918 ext4_fsblk_t len;
2919 int hblock, blocksize;
2920 ext4_fsblk_t sb_block;
2921 unsigned long offset;
2922 struct ext4_super_block *es;
2923 struct block_device *bdev;
2924
2925 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2926
2927 bdev = ext4_blkdev_get(j_dev);
2928 if (bdev == NULL)
2929 return NULL;
2930
2931 if (bd_claim(bdev, sb)) {
2932 printk(KERN_ERR
2933 "EXT4-fs: failed to claim external journal device.\n");
2934 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
2935 return NULL;
2936 }
2937
2938 blocksize = sb->s_blocksize;
2939 hblock = bdev_hardsect_size(bdev);
2940 if (blocksize < hblock) {
2941 printk(KERN_ERR
2942 "EXT4-fs: blocksize too small for journal device.\n");
2943 goto out_bdev;
2944 }
2945
2946 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2947 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
2948 set_blocksize(bdev, blocksize);
2949 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2950 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
2951 "external journal\n");
2952 goto out_bdev;
2953 }
2954
2955 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2956 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
2957 !(le32_to_cpu(es->s_feature_incompat) &
2958 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2959 printk(KERN_ERR "EXT4-fs: external journal has "
2960 "bad superblock\n");
2961 brelse(bh);
2962 goto out_bdev;
2963 }
2964
2965 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2966 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
2967 brelse(bh);
2968 goto out_bdev;
2969 }
2970
2971 len = ext4_blocks_count(es);
2972 start = sb_block + 1;
2973 brelse(bh); /* we're done with the superblock */
2974
2975 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
2976 start, len, blocksize);
2977 if (!journal) {
2978 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
2979 goto out_bdev;
2980 }
2981 journal->j_private = sb;
2982 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2983 wait_on_buffer(journal->j_sb_buffer);
2984 if (!buffer_uptodate(journal->j_sb_buffer)) {
2985 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
2986 goto out_journal;
2987 }
2988 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2989 printk(KERN_ERR "EXT4-fs: External journal has more than one "
2990 "user (unsupported) - %d\n",
2991 be32_to_cpu(journal->j_superblock->s_nr_users));
2992 goto out_journal;
2993 }
2994 EXT4_SB(sb)->journal_bdev = bdev;
2995 ext4_init_journal_params(sb, journal);
2996 return journal;
2997 out_journal:
2998 jbd2_journal_destroy(journal);
2999 out_bdev:
3000 ext4_blkdev_put(bdev);
3001 return NULL;
3002 }
3003
3004 static int ext4_load_journal(struct super_block *sb,
3005 struct ext4_super_block *es,
3006 unsigned long journal_devnum)
3007 {
3008 journal_t *journal;
3009 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3010 dev_t journal_dev;
3011 int err = 0;
3012 int really_read_only;
3013
3014 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3015
3016 if (journal_devnum &&
3017 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3018 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
3019 "numbers have changed\n");
3020 journal_dev = new_decode_dev(journal_devnum);
3021 } else
3022 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3023
3024 really_read_only = bdev_read_only(sb->s_bdev);
3025
3026 /*
3027 * Are we loading a blank journal or performing recovery after a
3028 * crash? For recovery, we need to check in advance whether we
3029 * can get read-write access to the device.
3030 */
3031
3032 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3033 if (sb->s_flags & MS_RDONLY) {
3034 printk(KERN_INFO "EXT4-fs: INFO: recovery "
3035 "required on readonly filesystem.\n");
3036 if (really_read_only) {
3037 printk(KERN_ERR "EXT4-fs: write access "
3038 "unavailable, cannot proceed.\n");
3039 return -EROFS;
3040 }
3041 printk(KERN_INFO "EXT4-fs: write access will "
3042 "be enabled during recovery.\n");
3043 }
3044 }
3045
3046 if (journal_inum && journal_dev) {
3047 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
3048 "and inode journals!\n");
3049 return -EINVAL;
3050 }
3051
3052 if (journal_inum) {
3053 if (!(journal = ext4_get_journal(sb, journal_inum)))
3054 return -EINVAL;
3055 } else {
3056 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3057 return -EINVAL;
3058 }
3059
3060 if (journal->j_flags & JBD2_BARRIER)
3061 printk(KERN_INFO "EXT4-fs: barriers enabled\n");
3062 else
3063 printk(KERN_INFO "EXT4-fs: barriers disabled\n");
3064
3065 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3066 err = jbd2_journal_update_format(journal);
3067 if (err) {
3068 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
3069 jbd2_journal_destroy(journal);
3070 return err;
3071 }
3072 }
3073
3074 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3075 err = jbd2_journal_wipe(journal, !really_read_only);
3076 if (!err)
3077 err = jbd2_journal_load(journal);
3078
3079 if (err) {
3080 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
3081 jbd2_journal_destroy(journal);
3082 return err;
3083 }
3084
3085 EXT4_SB(sb)->s_journal = journal;
3086 ext4_clear_journal_err(sb, es);
3087
3088 if (journal_devnum &&
3089 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3090 es->s_journal_dev = cpu_to_le32(journal_devnum);
3091 sb->s_dirt = 1;
3092
3093 /* Make sure we flush the recovery flag to disk. */
3094 ext4_commit_super(sb, es, 1);
3095 }
3096
3097 return 0;
3098 }
3099
3100 static int ext4_commit_super(struct super_block *sb,
3101 struct ext4_super_block *es, int sync)
3102 {
3103 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3104 int error = 0;
3105
3106 if (!sbh)
3107 return error;
3108 if (buffer_write_io_error(sbh)) {
3109 /*
3110 * Oh, dear. A previous attempt to write the
3111 * superblock failed. This could happen because the
3112 * USB device was yanked out. Or it could happen to
3113 * be a transient write error and maybe the block will
3114 * be remapped. Nothing we can do but to retry the
3115 * write and hope for the best.
3116 */
3117 printk(KERN_ERR "EXT4-fs: previous I/O error to "
3118 "superblock detected for %s.\n", sb->s_id);
3119 clear_buffer_write_io_error(sbh);
3120 set_buffer_uptodate(sbh);
3121 }
3122 es->s_wtime = cpu_to_le32(get_seconds());
3123 es->s_kbytes_written =
3124 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3125 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3126 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3127 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3128 &EXT4_SB(sb)->s_freeblocks_counter));
3129 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3130 &EXT4_SB(sb)->s_freeinodes_counter));
3131
3132 BUFFER_TRACE(sbh, "marking dirty");
3133 mark_buffer_dirty(sbh);
3134 if (sync) {
3135 error = sync_dirty_buffer(sbh);
3136 if (error)
3137 return error;
3138
3139 error = buffer_write_io_error(sbh);
3140 if (error) {
3141 printk(KERN_ERR "EXT4-fs: I/O error while writing "
3142 "superblock for %s.\n", sb->s_id);
3143 clear_buffer_write_io_error(sbh);
3144 set_buffer_uptodate(sbh);
3145 }
3146 }
3147 return error;
3148 }
3149
3150
3151 /*
3152 * Have we just finished recovery? If so, and if we are mounting (or
3153 * remounting) the filesystem readonly, then we will end up with a
3154 * consistent fs on disk. Record that fact.
3155 */
3156 static void ext4_mark_recovery_complete(struct super_block *sb,
3157 struct ext4_super_block *es)
3158 {
3159 journal_t *journal = EXT4_SB(sb)->s_journal;
3160
3161 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3162 BUG_ON(journal != NULL);
3163 return;
3164 }
3165 jbd2_journal_lock_updates(journal);
3166 if (jbd2_journal_flush(journal) < 0)
3167 goto out;
3168
3169 lock_super(sb);
3170 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3171 sb->s_flags & MS_RDONLY) {
3172 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3173 sb->s_dirt = 0;
3174 ext4_commit_super(sb, es, 1);
3175 }
3176 unlock_super(sb);
3177
3178 out:
3179 jbd2_journal_unlock_updates(journal);
3180 }
3181
3182 /*
3183 * If we are mounting (or read-write remounting) a filesystem whose journal
3184 * has recorded an error from a previous lifetime, move that error to the
3185 * main filesystem now.
3186 */
3187 static void ext4_clear_journal_err(struct super_block *sb,
3188 struct ext4_super_block *es)
3189 {
3190 journal_t *journal;
3191 int j_errno;
3192 const char *errstr;
3193
3194 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3195
3196 journal = EXT4_SB(sb)->s_journal;
3197
3198 /*
3199 * Now check for any error status which may have been recorded in the
3200 * journal by a prior ext4_error() or ext4_abort()
3201 */
3202
3203 j_errno = jbd2_journal_errno(journal);
3204 if (j_errno) {
3205 char nbuf[16];
3206
3207 errstr = ext4_decode_error(sb, j_errno, nbuf);
3208 ext4_warning(sb, __func__, "Filesystem error recorded "
3209 "from previous mount: %s", errstr);
3210 ext4_warning(sb, __func__, "Marking fs in need of "
3211 "filesystem check.");
3212
3213 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3214 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3215 ext4_commit_super(sb, es, 1);
3216
3217 jbd2_journal_clear_err(journal);
3218 }
3219 }
3220
3221 /*
3222 * Force the running and committing transactions to commit,
3223 * and wait on the commit.
3224 */
3225 int ext4_force_commit(struct super_block *sb)
3226 {
3227 journal_t *journal;
3228 int ret = 0;
3229
3230 if (sb->s_flags & MS_RDONLY)
3231 return 0;
3232
3233 journal = EXT4_SB(sb)->s_journal;
3234 if (journal) {
3235 sb->s_dirt = 0;
3236 ret = ext4_journal_force_commit(journal);
3237 }
3238
3239 return ret;
3240 }
3241
3242 /*
3243 * Ext4 always journals updates to the superblock itself, so we don't
3244 * have to propagate any other updates to the superblock on disk at this
3245 * point. (We can probably nuke this function altogether, and remove
3246 * any mention to sb->s_dirt in all of fs/ext4; eventual cleanup...)
3247 */
3248 static void ext4_write_super(struct super_block *sb)
3249 {
3250 if (EXT4_SB(sb)->s_journal) {
3251 if (mutex_trylock(&sb->s_lock) != 0)
3252 BUG();
3253 sb->s_dirt = 0;
3254 } else {
3255 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
3256 }
3257 }
3258
3259 static int ext4_sync_fs(struct super_block *sb, int wait)
3260 {
3261 int ret = 0;
3262 tid_t target;
3263
3264 trace_mark(ext4_sync_fs, "dev %s wait %d", sb->s_id, wait);
3265 sb->s_dirt = 0;
3266 if (EXT4_SB(sb)->s_journal) {
3267 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal,
3268 &target)) {
3269 if (wait)
3270 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal,
3271 target);
3272 }
3273 } else {
3274 ext4_commit_super(sb, EXT4_SB(sb)->s_es, wait);
3275 }
3276 return ret;
3277 }
3278
3279 /*
3280 * LVM calls this function before a (read-only) snapshot is created. This
3281 * gives us a chance to flush the journal completely and mark the fs clean.
3282 */
3283 static int ext4_freeze(struct super_block *sb)
3284 {
3285 int error = 0;
3286 journal_t *journal;
3287 sb->s_dirt = 0;
3288
3289 if (!(sb->s_flags & MS_RDONLY)) {
3290 journal = EXT4_SB(sb)->s_journal;
3291
3292 if (journal) {
3293 /* Now we set up the journal barrier. */
3294 jbd2_journal_lock_updates(journal);
3295
3296 /*
3297 * We don't want to clear needs_recovery flag when we
3298 * failed to flush the journal.
3299 */
3300 error = jbd2_journal_flush(journal);
3301 if (error < 0)
3302 goto out;
3303 }
3304
3305 /* Journal blocked and flushed, clear needs_recovery flag. */
3306 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3307 error = ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
3308 if (error)
3309 goto out;
3310 }
3311 return 0;
3312 out:
3313 jbd2_journal_unlock_updates(journal);
3314 return error;
3315 }
3316
3317 /*
3318 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3319 * flag here, even though the filesystem is not technically dirty yet.
3320 */
3321 static int ext4_unfreeze(struct super_block *sb)
3322 {
3323 if (EXT4_SB(sb)->s_journal && !(sb->s_flags & MS_RDONLY)) {
3324 lock_super(sb);
3325 /* Reser the needs_recovery flag before the fs is unlocked. */
3326 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3327 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
3328 unlock_super(sb);
3329 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3330 }
3331 return 0;
3332 }
3333
3334 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3335 {
3336 struct ext4_super_block *es;
3337 struct ext4_sb_info *sbi = EXT4_SB(sb);
3338 ext4_fsblk_t n_blocks_count = 0;
3339 unsigned long old_sb_flags;
3340 struct ext4_mount_options old_opts;
3341 ext4_group_t g;
3342 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3343 int err;
3344 #ifdef CONFIG_QUOTA
3345 int i;
3346 #endif
3347
3348 /* Store the original options */
3349 old_sb_flags = sb->s_flags;
3350 old_opts.s_mount_opt = sbi->s_mount_opt;
3351 old_opts.s_resuid = sbi->s_resuid;
3352 old_opts.s_resgid = sbi->s_resgid;
3353 old_opts.s_commit_interval = sbi->s_commit_interval;
3354 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3355 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3356 #ifdef CONFIG_QUOTA
3357 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3358 for (i = 0; i < MAXQUOTAS; i++)
3359 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3360 #endif
3361 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3362 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3363
3364 /*
3365 * Allow the "check" option to be passed as a remount option.
3366 */
3367 if (!parse_options(data, sb, NULL, &journal_ioprio,
3368 &n_blocks_count, 1)) {
3369 err = -EINVAL;
3370 goto restore_opts;
3371 }
3372
3373 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
3374 ext4_abort(sb, __func__, "Abort forced by user");
3375
3376 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3377 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3378
3379 es = sbi->s_es;
3380
3381 if (sbi->s_journal) {
3382 ext4_init_journal_params(sb, sbi->s_journal);
3383 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3384 }
3385
3386 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3387 n_blocks_count > ext4_blocks_count(es)) {
3388 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
3389 err = -EROFS;
3390 goto restore_opts;
3391 }
3392
3393 if (*flags & MS_RDONLY) {
3394 /*
3395 * First of all, the unconditional stuff we have to do
3396 * to disable replay of the journal when we next remount
3397 */
3398 sb->s_flags |= MS_RDONLY;
3399
3400 /*
3401 * OK, test if we are remounting a valid rw partition
3402 * readonly, and if so set the rdonly flag and then
3403 * mark the partition as valid again.
3404 */
3405 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3406 (sbi->s_mount_state & EXT4_VALID_FS))
3407 es->s_state = cpu_to_le16(sbi->s_mount_state);
3408
3409 /*
3410 * We have to unlock super so that we can wait for
3411 * transactions.
3412 */
3413 if (sbi->s_journal) {
3414 unlock_super(sb);
3415 ext4_mark_recovery_complete(sb, es);
3416 lock_super(sb);
3417 }
3418 } else {
3419 int ret;
3420 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3421 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
3422 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3423 "remount RDWR because of unsupported "
3424 "optional features (%x).\n", sb->s_id,
3425 (le32_to_cpu(sbi->s_es->s_feature_ro_compat) &
3426 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3427 err = -EROFS;
3428 goto restore_opts;
3429 }
3430
3431 /*
3432 * Make sure the group descriptor checksums
3433 * are sane. If they aren't, refuse to
3434 * remount r/w.
3435 */
3436 for (g = 0; g < sbi->s_groups_count; g++) {
3437 struct ext4_group_desc *gdp =
3438 ext4_get_group_desc(sb, g, NULL);
3439
3440 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3441 printk(KERN_ERR
3442 "EXT4-fs: ext4_remount: "
3443 "Checksum for group %u failed (%u!=%u)\n",
3444 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3445 le16_to_cpu(gdp->bg_checksum));
3446 err = -EINVAL;
3447 goto restore_opts;
3448 }
3449 }
3450
3451 /*
3452 * If we have an unprocessed orphan list hanging
3453 * around from a previously readonly bdev mount,
3454 * require a full umount/remount for now.
3455 */
3456 if (es->s_last_orphan) {
3457 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3458 "remount RDWR because of unprocessed "
3459 "orphan inode list. Please "
3460 "umount/remount instead.\n",
3461 sb->s_id);
3462 err = -EINVAL;
3463 goto restore_opts;
3464 }
3465
3466 /*
3467 * Mounting a RDONLY partition read-write, so reread
3468 * and store the current valid flag. (It may have
3469 * been changed by e2fsck since we originally mounted
3470 * the partition.)
3471 */
3472 if (sbi->s_journal)
3473 ext4_clear_journal_err(sb, es);
3474 sbi->s_mount_state = le16_to_cpu(es->s_state);
3475 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3476 goto restore_opts;
3477 if (!ext4_setup_super(sb, es, 0))
3478 sb->s_flags &= ~MS_RDONLY;
3479 }
3480 }
3481 if (sbi->s_journal == NULL)
3482 ext4_commit_super(sb, es, 1);
3483
3484 #ifdef CONFIG_QUOTA
3485 /* Release old quota file names */
3486 for (i = 0; i < MAXQUOTAS; i++)
3487 if (old_opts.s_qf_names[i] &&
3488 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3489 kfree(old_opts.s_qf_names[i]);
3490 #endif
3491 return 0;
3492 restore_opts:
3493 sb->s_flags = old_sb_flags;
3494 sbi->s_mount_opt = old_opts.s_mount_opt;
3495 sbi->s_resuid = old_opts.s_resuid;
3496 sbi->s_resgid = old_opts.s_resgid;
3497 sbi->s_commit_interval = old_opts.s_commit_interval;
3498 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3499 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3500 #ifdef CONFIG_QUOTA
3501 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3502 for (i = 0; i < MAXQUOTAS; i++) {
3503 if (sbi->s_qf_names[i] &&
3504 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3505 kfree(sbi->s_qf_names[i]);
3506 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3507 }
3508 #endif
3509 return err;
3510 }
3511
3512 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3513 {
3514 struct super_block *sb = dentry->d_sb;
3515 struct ext4_sb_info *sbi = EXT4_SB(sb);
3516 struct ext4_super_block *es = sbi->s_es;
3517 u64 fsid;
3518
3519 if (test_opt(sb, MINIX_DF)) {
3520 sbi->s_overhead_last = 0;
3521 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3522 ext4_group_t ngroups = sbi->s_groups_count, i;
3523 ext4_fsblk_t overhead = 0;
3524 smp_rmb();
3525
3526 /*
3527 * Compute the overhead (FS structures). This is constant
3528 * for a given filesystem unless the number of block groups
3529 * changes so we cache the previous value until it does.
3530 */
3531
3532 /*
3533 * All of the blocks before first_data_block are
3534 * overhead
3535 */
3536 overhead = le32_to_cpu(es->s_first_data_block);
3537
3538 /*
3539 * Add the overhead attributed to the superblock and
3540 * block group descriptors. If the sparse superblocks
3541 * feature is turned on, then not all groups have this.
3542 */
3543 for (i = 0; i < ngroups; i++) {
3544 overhead += ext4_bg_has_super(sb, i) +
3545 ext4_bg_num_gdb(sb, i);
3546 cond_resched();
3547 }
3548
3549 /*
3550 * Every block group has an inode bitmap, a block
3551 * bitmap, and an inode table.
3552 */
3553 overhead += ngroups * (2 + sbi->s_itb_per_group);
3554 sbi->s_overhead_last = overhead;
3555 smp_wmb();
3556 sbi->s_blocks_last = ext4_blocks_count(es);
3557 }
3558
3559 buf->f_type = EXT4_SUPER_MAGIC;
3560 buf->f_bsize = sb->s_blocksize;
3561 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3562 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3563 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3564 ext4_free_blocks_count_set(es, buf->f_bfree);
3565 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3566 if (buf->f_bfree < ext4_r_blocks_count(es))
3567 buf->f_bavail = 0;
3568 buf->f_files = le32_to_cpu(es->s_inodes_count);
3569 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3570 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
3571 buf->f_namelen = EXT4_NAME_LEN;
3572 fsid = le64_to_cpup((void *)es->s_uuid) ^
3573 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3574 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3575 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3576 return 0;
3577 }
3578
3579 /* Helper function for writing quotas on sync - we need to start transaction before quota file
3580 * is locked for write. Otherwise the are possible deadlocks:
3581 * Process 1 Process 2
3582 * ext4_create() quota_sync()
3583 * jbd2_journal_start() write_dquot()
3584 * vfs_dq_init() down(dqio_mutex)
3585 * down(dqio_mutex) jbd2_journal_start()
3586 *
3587 */
3588
3589 #ifdef CONFIG_QUOTA
3590
3591 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3592 {
3593 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3594 }
3595
3596 static int ext4_write_dquot(struct dquot *dquot)
3597 {
3598 int ret, err;
3599 handle_t *handle;
3600 struct inode *inode;
3601
3602 inode = dquot_to_inode(dquot);
3603 handle = ext4_journal_start(inode,
3604 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3605 if (IS_ERR(handle))
3606 return PTR_ERR(handle);
3607 ret = dquot_commit(dquot);
3608 err = ext4_journal_stop(handle);
3609 if (!ret)
3610 ret = err;
3611 return ret;
3612 }
3613
3614 static int ext4_acquire_dquot(struct dquot *dquot)
3615 {
3616 int ret, err;
3617 handle_t *handle;
3618
3619 handle = ext4_journal_start(dquot_to_inode(dquot),
3620 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3621 if (IS_ERR(handle))
3622 return PTR_ERR(handle);
3623 ret = dquot_acquire(dquot);
3624 err = ext4_journal_stop(handle);
3625 if (!ret)
3626 ret = err;
3627 return ret;
3628 }
3629
3630 static int ext4_release_dquot(struct dquot *dquot)
3631 {
3632 int ret, err;
3633 handle_t *handle;
3634
3635 handle = ext4_journal_start(dquot_to_inode(dquot),
3636 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3637 if (IS_ERR(handle)) {
3638 /* Release dquot anyway to avoid endless cycle in dqput() */
3639 dquot_release(dquot);
3640 return PTR_ERR(handle);
3641 }
3642 ret = dquot_release(dquot);
3643 err = ext4_journal_stop(handle);
3644 if (!ret)
3645 ret = err;
3646 return ret;
3647 }
3648
3649 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3650 {
3651 /* Are we journaling quotas? */
3652 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3653 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3654 dquot_mark_dquot_dirty(dquot);
3655 return ext4_write_dquot(dquot);
3656 } else {
3657 return dquot_mark_dquot_dirty(dquot);
3658 }
3659 }
3660
3661 static int ext4_write_info(struct super_block *sb, int type)
3662 {
3663 int ret, err;
3664 handle_t *handle;
3665
3666 /* Data block + inode block */
3667 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3668 if (IS_ERR(handle))
3669 return PTR_ERR(handle);
3670 ret = dquot_commit_info(sb, type);
3671 err = ext4_journal_stop(handle);
3672 if (!ret)
3673 ret = err;
3674 return ret;
3675 }
3676
3677 /*
3678 * Turn on quotas during mount time - we need to find
3679 * the quota file and such...
3680 */
3681 static int ext4_quota_on_mount(struct super_block *sb, int type)
3682 {
3683 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3684 EXT4_SB(sb)->s_jquota_fmt, type);
3685 }
3686
3687 /*
3688 * Standard function to be called on quota_on
3689 */
3690 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3691 char *name, int remount)
3692 {
3693 int err;
3694 struct path path;
3695
3696 if (!test_opt(sb, QUOTA))
3697 return -EINVAL;
3698 /* When remounting, no checks are needed and in fact, name is NULL */
3699 if (remount)
3700 return vfs_quota_on(sb, type, format_id, name, remount);
3701
3702 err = kern_path(name, LOOKUP_FOLLOW, &path);
3703 if (err)
3704 return err;
3705
3706 /* Quotafile not on the same filesystem? */
3707 if (path.mnt->mnt_sb != sb) {
3708 path_put(&path);
3709 return -EXDEV;
3710 }
3711 /* Journaling quota? */
3712 if (EXT4_SB(sb)->s_qf_names[type]) {
3713 /* Quotafile not in fs root? */
3714 if (path.dentry->d_parent != sb->s_root)
3715 printk(KERN_WARNING
3716 "EXT4-fs: Quota file not on filesystem root. "
3717 "Journaled quota will not work.\n");
3718 }
3719
3720 /*
3721 * When we journal data on quota file, we have to flush journal to see
3722 * all updates to the file when we bypass pagecache...
3723 */
3724 if (EXT4_SB(sb)->s_journal &&
3725 ext4_should_journal_data(path.dentry->d_inode)) {
3726 /*
3727 * We don't need to lock updates but journal_flush() could
3728 * otherwise be livelocked...
3729 */
3730 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3731 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3732 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3733 if (err) {
3734 path_put(&path);
3735 return err;
3736 }
3737 }
3738
3739 err = vfs_quota_on_path(sb, type, format_id, &path);
3740 path_put(&path);
3741 return err;
3742 }
3743
3744 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3745 * acquiring the locks... As quota files are never truncated and quota code
3746 * itself serializes the operations (and noone else should touch the files)
3747 * we don't have to be afraid of races */
3748 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3749 size_t len, loff_t off)
3750 {
3751 struct inode *inode = sb_dqopt(sb)->files[type];
3752 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3753 int err = 0;
3754 int offset = off & (sb->s_blocksize - 1);
3755 int tocopy;
3756 size_t toread;
3757 struct buffer_head *bh;
3758 loff_t i_size = i_size_read(inode);
3759
3760 if (off > i_size)
3761 return 0;
3762 if (off+len > i_size)
3763 len = i_size-off;
3764 toread = len;
3765 while (toread > 0) {
3766 tocopy = sb->s_blocksize - offset < toread ?
3767 sb->s_blocksize - offset : toread;
3768 bh = ext4_bread(NULL, inode, blk, 0, &err);
3769 if (err)
3770 return err;
3771 if (!bh) /* A hole? */
3772 memset(data, 0, tocopy);
3773 else
3774 memcpy(data, bh->b_data+offset, tocopy);
3775 brelse(bh);
3776 offset = 0;
3777 toread -= tocopy;
3778 data += tocopy;
3779 blk++;
3780 }
3781 return len;
3782 }
3783
3784 /* Write to quotafile (we know the transaction is already started and has
3785 * enough credits) */
3786 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3787 const char *data, size_t len, loff_t off)
3788 {
3789 struct inode *inode = sb_dqopt(sb)->files[type];
3790 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3791 int err = 0;
3792 int offset = off & (sb->s_blocksize - 1);
3793 int tocopy;
3794 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3795 size_t towrite = len;
3796 struct buffer_head *bh;
3797 handle_t *handle = journal_current_handle();
3798
3799 if (EXT4_SB(sb)->s_journal && !handle) {
3800 printk(KERN_WARNING "EXT4-fs: Quota write (off=%llu, len=%llu)"
3801 " cancelled because transaction is not started.\n",
3802 (unsigned long long)off, (unsigned long long)len);
3803 return -EIO;
3804 }
3805 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3806 while (towrite > 0) {
3807 tocopy = sb->s_blocksize - offset < towrite ?
3808 sb->s_blocksize - offset : towrite;
3809 bh = ext4_bread(handle, inode, blk, 1, &err);
3810 if (!bh)
3811 goto out;
3812 if (journal_quota) {
3813 err = ext4_journal_get_write_access(handle, bh);
3814 if (err) {
3815 brelse(bh);
3816 goto out;
3817 }
3818 }
3819 lock_buffer(bh);
3820 memcpy(bh->b_data+offset, data, tocopy);
3821 flush_dcache_page(bh->b_page);
3822 unlock_buffer(bh);
3823 if (journal_quota)
3824 err = ext4_handle_dirty_metadata(handle, NULL, bh);
3825 else {
3826 /* Always do at least ordered writes for quotas */
3827 err = ext4_jbd2_file_inode(handle, inode);
3828 mark_buffer_dirty(bh);
3829 }
3830 brelse(bh);
3831 if (err)
3832 goto out;
3833 offset = 0;
3834 towrite -= tocopy;
3835 data += tocopy;
3836 blk++;
3837 }
3838 out:
3839 if (len == towrite) {
3840 mutex_unlock(&inode->i_mutex);
3841 return err;
3842 }
3843 if (inode->i_size < off+len-towrite) {
3844 i_size_write(inode, off+len-towrite);
3845 EXT4_I(inode)->i_disksize = inode->i_size;
3846 }
3847 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3848 ext4_mark_inode_dirty(handle, inode);
3849 mutex_unlock(&inode->i_mutex);
3850 return len - towrite;
3851 }
3852
3853 #endif
3854
3855 static int ext4_get_sb(struct file_system_type *fs_type,
3856 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3857 {
3858 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3859 }
3860
3861 static struct file_system_type ext4_fs_type = {
3862 .owner = THIS_MODULE,
3863 .name = "ext4",
3864 .get_sb = ext4_get_sb,
3865 .kill_sb = kill_block_super,
3866 .fs_flags = FS_REQUIRES_DEV,
3867 };
3868
3869 #ifdef CONFIG_EXT4DEV_COMPAT
3870 static int ext4dev_get_sb(struct file_system_type *fs_type,
3871 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3872 {
3873 printk(KERN_WARNING "EXT4-fs: Update your userspace programs "
3874 "to mount using ext4\n");
3875 printk(KERN_WARNING "EXT4-fs: ext4dev backwards compatibility "
3876 "will go away by 2.6.31\n");
3877 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3878 }
3879
3880 static struct file_system_type ext4dev_fs_type = {
3881 .owner = THIS_MODULE,
3882 .name = "ext4dev",
3883 .get_sb = ext4dev_get_sb,
3884 .kill_sb = kill_block_super,
3885 .fs_flags = FS_REQUIRES_DEV,
3886 };
3887 MODULE_ALIAS("ext4dev");
3888 #endif
3889
3890 static int __init init_ext4_fs(void)
3891 {
3892 int err;
3893
3894 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
3895 if (!ext4_kset)
3896 return -ENOMEM;
3897 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
3898 err = init_ext4_mballoc();
3899 if (err)
3900 return err;
3901
3902 err = init_ext4_xattr();
3903 if (err)
3904 goto out2;
3905 err = init_inodecache();
3906 if (err)
3907 goto out1;
3908 err = register_filesystem(&ext4_fs_type);
3909 if (err)
3910 goto out;
3911 #ifdef CONFIG_EXT4DEV_COMPAT
3912 err = register_filesystem(&ext4dev_fs_type);
3913 if (err) {
3914 unregister_filesystem(&ext4_fs_type);
3915 goto out;
3916 }
3917 #endif
3918 return 0;
3919 out:
3920 destroy_inodecache();
3921 out1:
3922 exit_ext4_xattr();
3923 out2:
3924 exit_ext4_mballoc();
3925 return err;
3926 }
3927
3928 static void __exit exit_ext4_fs(void)
3929 {
3930 unregister_filesystem(&ext4_fs_type);
3931 #ifdef CONFIG_EXT4DEV_COMPAT
3932 unregister_filesystem(&ext4dev_fs_type);
3933 #endif
3934 destroy_inodecache();
3935 exit_ext4_xattr();
3936 exit_ext4_mballoc();
3937 remove_proc_entry("fs/ext4", NULL);
3938 kset_unregister(ext4_kset);
3939 }
3940
3941 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3942 MODULE_DESCRIPTION("Fourth Extended Filesystem");
3943 MODULE_LICENSE("GPL");
3944 module_init(init_ext4_fs)
3945 module_exit(exit_ext4_fs)
CJK support for tty framebuffer vt