?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge branch 'mini2440-dev-unlikely' into mini2440-dev
[linux-2.6/mini2440.git] / fs / ext4 / super.c
blobd4ca92aab51411de0bab1bf8824e9aa8974b8d17
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/vmalloc.h>
24 #include <linux/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/smp_lock.h>
30 #include <linux/buffer_head.h>
31 #include <linux/exportfs.h>
32 #include <linux/vfs.h>
33 #include <linux/random.h>
34 #include <linux/mount.h>
35 #include <linux/namei.h>
36 #include <linux/quotaops.h>
37 #include <linux/seq_file.h>
38 #include <linux/proc_fs.h>
39 #include <linux/ctype.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 "mballoc.h"
49
50 #define CREATE_TRACE_POINTS
51 #include <trace/events/ext4.h>
52
53 struct proc_dir_entry *ext4_proc_root;
54 static struct kset *ext4_kset;
55
56 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
57 unsigned long journal_devnum);
58 static int ext4_commit_super(struct super_block *sb, int sync);
59 static void ext4_mark_recovery_complete(struct super_block *sb,
60 struct ext4_super_block *es);
61 static void ext4_clear_journal_err(struct super_block *sb,
62 struct ext4_super_block *es);
63 static int ext4_sync_fs(struct super_block *sb, int wait);
64 static const char *ext4_decode_error(struct super_block *sb, int errno,
65 char nbuf[16]);
66 static int ext4_remount(struct super_block *sb, int *flags, char *data);
67 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
68 static int ext4_unfreeze(struct super_block *sb);
69 static void ext4_write_super(struct super_block *sb);
70 static int ext4_freeze(struct super_block *sb);
71
72
73 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
74 struct ext4_group_desc *bg)
75 {
76 return le32_to_cpu(bg->bg_block_bitmap_lo) |
77 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
78 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
79 }
80
81 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
82 struct ext4_group_desc *bg)
83 {
84 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
85 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
86 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
87 }
88
89 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
90 struct ext4_group_desc *bg)
91 {
92 return le32_to_cpu(bg->bg_inode_table_lo) |
93 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
94 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
95 }
96
97 __u32 ext4_free_blks_count(struct super_block *sb,
98 struct ext4_group_desc *bg)
99 {
100 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
101 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
102 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
103 }
104
105 __u32 ext4_free_inodes_count(struct super_block *sb,
106 struct ext4_group_desc *bg)
107 {
108 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
109 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
110 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
111 }
112
113 __u32 ext4_used_dirs_count(struct super_block *sb,
114 struct ext4_group_desc *bg)
115 {
116 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
117 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
118 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
119 }
120
121 __u32 ext4_itable_unused_count(struct super_block *sb,
122 struct ext4_group_desc *bg)
123 {
124 return le16_to_cpu(bg->bg_itable_unused_lo) |
125 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
126 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
127 }
128
129 void ext4_block_bitmap_set(struct super_block *sb,
130 struct ext4_group_desc *bg, ext4_fsblk_t blk)
131 {
132 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
133 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
134 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
135 }
136
137 void ext4_inode_bitmap_set(struct super_block *sb,
138 struct ext4_group_desc *bg, ext4_fsblk_t blk)
139 {
140 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
141 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
142 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
143 }
144
145 void ext4_inode_table_set(struct super_block *sb,
146 struct ext4_group_desc *bg, ext4_fsblk_t blk)
147 {
148 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
149 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
150 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
151 }
152
153 void ext4_free_blks_set(struct super_block *sb,
154 struct ext4_group_desc *bg, __u32 count)
155 {
156 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
157 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
158 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
159 }
160
161 void ext4_free_inodes_set(struct super_block *sb,
162 struct ext4_group_desc *bg, __u32 count)
163 {
164 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
165 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
166 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
167 }
168
169 void ext4_used_dirs_set(struct super_block *sb,
170 struct ext4_group_desc *bg, __u32 count)
171 {
172 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
173 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
174 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
175 }
176
177 void ext4_itable_unused_set(struct super_block *sb,
178 struct ext4_group_desc *bg, __u32 count)
179 {
180 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
181 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
182 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
183 }
184
185
186 /* Just increment the non-pointer handle value */
187 static handle_t *ext4_get_nojournal(void)
188 {
189 handle_t *handle = current->journal_info;
190 unsigned long ref_cnt = (unsigned long)handle;
191
192 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
193
194 ref_cnt++;
195 handle = (handle_t *)ref_cnt;
196
197 current->journal_info = handle;
198 return handle;
199 }
200
201
202 /* Decrement the non-pointer handle value */
203 static void ext4_put_nojournal(handle_t *handle)
204 {
205 unsigned long ref_cnt = (unsigned long)handle;
206
207 BUG_ON(ref_cnt == 0);
208
209 ref_cnt--;
210 handle = (handle_t *)ref_cnt;
211
212 current->journal_info = handle;
213 }
214
215 /*
216 * Wrappers for jbd2_journal_start/end.
217 *
218 * The only special thing we need to do here is to make sure that all
219 * journal_end calls result in the superblock being marked dirty, so
220 * that sync() will call the filesystem's write_super callback if
221 * appropriate.
222 */
223 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
224 {
225 journal_t *journal;
226
227 if (sb->s_flags & MS_RDONLY)
228 return ERR_PTR(-EROFS);
229
230 /* Special case here: if the journal has aborted behind our
231 * backs (eg. EIO in the commit thread), then we still need to
232 * take the FS itself readonly cleanly. */
233 journal = EXT4_SB(sb)->s_journal;
234 if (journal) {
235 if (is_journal_aborted(journal)) {
236 ext4_abort(sb, __func__, "Detected aborted journal");
237 return ERR_PTR(-EROFS);
238 }
239 return jbd2_journal_start(journal, nblocks);
240 }
241 return ext4_get_nojournal();
242 }
243
244 /*
245 * The only special thing we need to do here is to make sure that all
246 * jbd2_journal_stop calls result in the superblock being marked dirty, so
247 * that sync() will call the filesystem's write_super callback if
248 * appropriate.
249 */
250 int __ext4_journal_stop(const char *where, handle_t *handle)
251 {
252 struct super_block *sb;
253 int err;
254 int rc;
255
256 if (!ext4_handle_valid(handle)) {
257 ext4_put_nojournal(handle);
258 return 0;
259 }
260 sb = handle->h_transaction->t_journal->j_private;
261 err = handle->h_err;
262 rc = jbd2_journal_stop(handle);
263
264 if (!err)
265 err = rc;
266 if (err)
267 __ext4_std_error(sb, where, err);
268 return err;
269 }
270
271 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
272 struct buffer_head *bh, handle_t *handle, int err)
273 {
274 char nbuf[16];
275 const char *errstr = ext4_decode_error(NULL, err, nbuf);
276
277 BUG_ON(!ext4_handle_valid(handle));
278
279 if (bh)
280 BUFFER_TRACE(bh, "abort");
281
282 if (!handle->h_err)
283 handle->h_err = err;
284
285 if (is_handle_aborted(handle))
286 return;
287
288 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
289 caller, errstr, err_fn);
290
291 jbd2_journal_abort_handle(handle);
292 }
293
294 /* Deal with the reporting of failure conditions on a filesystem such as
295 * inconsistencies detected or read IO failures.
296 *
297 * On ext2, we can store the error state of the filesystem in the
298 * superblock. That is not possible on ext4, because we may have other
299 * write ordering constraints on the superblock which prevent us from
300 * writing it out straight away; and given that the journal is about to
301 * be aborted, we can't rely on the current, or future, transactions to
302 * write out the superblock safely.
303 *
304 * We'll just use the jbd2_journal_abort() error code to record an error in
305 * the journal instead. On recovery, the journal will compain about
306 * that error until we've noted it down and cleared it.
307 */
308
309 static void ext4_handle_error(struct super_block *sb)
310 {
311 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
312
313 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
314 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
315
316 if (sb->s_flags & MS_RDONLY)
317 return;
318
319 if (!test_opt(sb, ERRORS_CONT)) {
320 journal_t *journal = EXT4_SB(sb)->s_journal;
321
322 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
323 if (journal)
324 jbd2_journal_abort(journal, -EIO);
325 }
326 if (test_opt(sb, ERRORS_RO)) {
327 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
328 sb->s_flags |= MS_RDONLY;
329 }
330 ext4_commit_super(sb, 1);
331 if (test_opt(sb, ERRORS_PANIC))
332 panic("EXT4-fs (device %s): panic forced after error\n",
333 sb->s_id);
334 }
335
336 void ext4_error(struct super_block *sb, const char *function,
337 const char *fmt, ...)
338 {
339 va_list args;
340
341 va_start(args, fmt);
342 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
343 vprintk(fmt, args);
344 printk("\n");
345 va_end(args);
346
347 ext4_handle_error(sb);
348 }
349
350 static const char *ext4_decode_error(struct super_block *sb, int errno,
351 char nbuf[16])
352 {
353 char *errstr = NULL;
354
355 switch (errno) {
356 case -EIO:
357 errstr = "IO failure";
358 break;
359 case -ENOMEM:
360 errstr = "Out of memory";
361 break;
362 case -EROFS:
363 if (!sb || (EXT4_SB(sb)->s_journal &&
364 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
365 errstr = "Journal has aborted";
366 else
367 errstr = "Readonly filesystem";
368 break;
369 default:
370 /* If the caller passed in an extra buffer for unknown
371 * errors, textualise them now. Else we just return
372 * NULL. */
373 if (nbuf) {
374 /* Check for truncated error codes... */
375 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
376 errstr = nbuf;
377 }
378 break;
379 }
380
381 return errstr;
382 }
383
384 /* __ext4_std_error decodes expected errors from journaling functions
385 * automatically and invokes the appropriate error response. */
386
387 void __ext4_std_error(struct super_block *sb, const char *function, int errno)
388 {
389 char nbuf[16];
390 const char *errstr;
391
392 /* Special case: if the error is EROFS, and we're not already
393 * inside a transaction, then there's really no point in logging
394 * an error. */
395 if (errno == -EROFS && journal_current_handle() == NULL &&
396 (sb->s_flags & MS_RDONLY))
397 return;
398
399 errstr = ext4_decode_error(sb, errno, nbuf);
400 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
401 sb->s_id, function, errstr);
402
403 ext4_handle_error(sb);
404 }
405
406 /*
407 * ext4_abort is a much stronger failure handler than ext4_error. The
408 * abort function may be used to deal with unrecoverable failures such
409 * as journal IO errors or ENOMEM at a critical moment in log management.
410 *
411 * We unconditionally force the filesystem into an ABORT|READONLY state,
412 * unless the error response on the fs has been set to panic in which
413 * case we take the easy way out and panic immediately.
414 */
415
416 void ext4_abort(struct super_block *sb, const char *function,
417 const char *fmt, ...)
418 {
419 va_list args;
420
421 va_start(args, fmt);
422 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
423 vprintk(fmt, args);
424 printk("\n");
425 va_end(args);
426
427 if (test_opt(sb, ERRORS_PANIC))
428 panic("EXT4-fs panic from previous error\n");
429
430 if (sb->s_flags & MS_RDONLY)
431 return;
432
433 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
434 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
435 sb->s_flags |= MS_RDONLY;
436 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
437 if (EXT4_SB(sb)->s_journal)
438 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
439 }
440
441 void ext4_msg (struct super_block * sb, const char *prefix,
442 const char *fmt, ...)
443 {
444 va_list args;
445
446 va_start(args, fmt);
447 printk("%sEXT4-fs (%s): ", prefix, sb->s_id);
448 vprintk(fmt, args);
449 printk("\n");
450 va_end(args);
451 }
452
453 void ext4_warning(struct super_block *sb, const char *function,
454 const char *fmt, ...)
455 {
456 va_list args;
457
458 va_start(args, fmt);
459 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
460 sb->s_id, function);
461 vprintk(fmt, args);
462 printk("\n");
463 va_end(args);
464 }
465
466 void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp,
467 const char *function, const char *fmt, ...)
468 __releases(bitlock)
469 __acquires(bitlock)
470 {
471 va_list args;
472 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
473
474 va_start(args, fmt);
475 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
476 vprintk(fmt, args);
477 printk("\n");
478 va_end(args);
479
480 if (test_opt(sb, ERRORS_CONT)) {
481 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
482 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
483 ext4_commit_super(sb, 0);
484 return;
485 }
486 ext4_unlock_group(sb, grp);
487 ext4_handle_error(sb);
488 /*
489 * We only get here in the ERRORS_RO case; relocking the group
490 * may be dangerous, but nothing bad will happen since the
491 * filesystem will have already been marked read/only and the
492 * journal has been aborted. We return 1 as a hint to callers
493 * who might what to use the return value from
494 * ext4_grp_locked_error() to distinguish beween the
495 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
496 * aggressively from the ext4 function in question, with a
497 * more appropriate error code.
498 */
499 ext4_lock_group(sb, grp);
500 return;
501 }
502
503 void ext4_update_dynamic_rev(struct super_block *sb)
504 {
505 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
506
507 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
508 return;
509
510 ext4_warning(sb, __func__,
511 "updating to rev %d because of new feature flag, "
512 "running e2fsck is recommended",
513 EXT4_DYNAMIC_REV);
514
515 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
516 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
517 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
518 /* leave es->s_feature_*compat flags alone */
519 /* es->s_uuid will be set by e2fsck if empty */
520
521 /*
522 * The rest of the superblock fields should be zero, and if not it
523 * means they are likely already in use, so leave them alone. We
524 * can leave it up to e2fsck to clean up any inconsistencies there.
525 */
526 }
527
528 /*
529 * Open the external journal device
530 */
531 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
532 {
533 struct block_device *bdev;
534 char b[BDEVNAME_SIZE];
535
536 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
537 if (IS_ERR(bdev))
538 goto fail;
539 return bdev;
540
541 fail:
542 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
543 __bdevname(dev, b), PTR_ERR(bdev));
544 return NULL;
545 }
546
547 /*
548 * Release the journal device
549 */
550 static int ext4_blkdev_put(struct block_device *bdev)
551 {
552 bd_release(bdev);
553 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
554 }
555
556 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
557 {
558 struct block_device *bdev;
559 int ret = -ENODEV;
560
561 bdev = sbi->journal_bdev;
562 if (bdev) {
563 ret = ext4_blkdev_put(bdev);
564 sbi->journal_bdev = NULL;
565 }
566 return ret;
567 }
568
569 static inline struct inode *orphan_list_entry(struct list_head *l)
570 {
571 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
572 }
573
574 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
575 {
576 struct list_head *l;
577
578 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
579 le32_to_cpu(sbi->s_es->s_last_orphan));
580
581 printk(KERN_ERR "sb_info orphan list:\n");
582 list_for_each(l, &sbi->s_orphan) {
583 struct inode *inode = orphan_list_entry(l);
584 printk(KERN_ERR " "
585 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
586 inode->i_sb->s_id, inode->i_ino, inode,
587 inode->i_mode, inode->i_nlink,
588 NEXT_ORPHAN(inode));
589 }
590 }
591
592 static void ext4_put_super(struct super_block *sb)
593 {
594 struct ext4_sb_info *sbi = EXT4_SB(sb);
595 struct ext4_super_block *es = sbi->s_es;
596 int i, err;
597
598 flush_workqueue(sbi->dio_unwritten_wq);
599 destroy_workqueue(sbi->dio_unwritten_wq);
600
601 lock_super(sb);
602 lock_kernel();
603 if (sb->s_dirt)
604 ext4_commit_super(sb, 1);
605
606 ext4_release_system_zone(sb);
607 ext4_mb_release(sb);
608 ext4_ext_release(sb);
609 ext4_xattr_put_super(sb);
610 if (sbi->s_journal) {
611 err = jbd2_journal_destroy(sbi->s_journal);
612 sbi->s_journal = NULL;
613 if (err < 0)
614 ext4_abort(sb, __func__,
615 "Couldn't clean up the journal");
616 }
617 if (!(sb->s_flags & MS_RDONLY)) {
618 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
619 es->s_state = cpu_to_le16(sbi->s_mount_state);
620 ext4_commit_super(sb, 1);
621 }
622 if (sbi->s_proc) {
623 remove_proc_entry(sb->s_id, ext4_proc_root);
624 }
625 kobject_del(&sbi->s_kobj);
626
627 for (i = 0; i < sbi->s_gdb_count; i++)
628 brelse(sbi->s_group_desc[i]);
629 kfree(sbi->s_group_desc);
630 if (is_vmalloc_addr(sbi->s_flex_groups))
631 vfree(sbi->s_flex_groups);
632 else
633 kfree(sbi->s_flex_groups);
634 percpu_counter_destroy(&sbi->s_freeblocks_counter);
635 percpu_counter_destroy(&sbi->s_freeinodes_counter);
636 percpu_counter_destroy(&sbi->s_dirs_counter);
637 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
638 brelse(sbi->s_sbh);
639 #ifdef CONFIG_QUOTA
640 for (i = 0; i < MAXQUOTAS; i++)
641 kfree(sbi->s_qf_names[i]);
642 #endif
643
644 /* Debugging code just in case the in-memory inode orphan list
645 * isn't empty. The on-disk one can be non-empty if we've
646 * detected an error and taken the fs readonly, but the
647 * in-memory list had better be clean by this point. */
648 if (!list_empty(&sbi->s_orphan))
649 dump_orphan_list(sb, sbi);
650 J_ASSERT(list_empty(&sbi->s_orphan));
651
652 invalidate_bdev(sb->s_bdev);
653 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
654 /*
655 * Invalidate the journal device's buffers. We don't want them
656 * floating about in memory - the physical journal device may
657 * hotswapped, and it breaks the `ro-after' testing code.
658 */
659 sync_blockdev(sbi->journal_bdev);
660 invalidate_bdev(sbi->journal_bdev);
661 ext4_blkdev_remove(sbi);
662 }
663 sb->s_fs_info = NULL;
664 /*
665 * Now that we are completely done shutting down the
666 * superblock, we need to actually destroy the kobject.
667 */
668 unlock_kernel();
669 unlock_super(sb);
670 kobject_put(&sbi->s_kobj);
671 wait_for_completion(&sbi->s_kobj_unregister);
672 kfree(sbi->s_blockgroup_lock);
673 kfree(sbi);
674 }
675
676 static struct kmem_cache *ext4_inode_cachep;
677
678 /*
679 * Called inside transaction, so use GFP_NOFS
680 */
681 static struct inode *ext4_alloc_inode(struct super_block *sb)
682 {
683 struct ext4_inode_info *ei;
684
685 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
686 if (!ei)
687 return NULL;
688
689 ei->vfs_inode.i_version = 1;
690 ei->vfs_inode.i_data.writeback_index = 0;
691 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
692 INIT_LIST_HEAD(&ei->i_prealloc_list);
693 spin_lock_init(&ei->i_prealloc_lock);
694 /*
695 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
696 * therefore it can be null here. Don't check it, just initialize
697 * jinode.
698 */
699 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
700 ei->i_reserved_data_blocks = 0;
701 ei->i_reserved_meta_blocks = 0;
702 ei->i_allocated_meta_blocks = 0;
703 ei->i_delalloc_reserved_flag = 0;
704 spin_lock_init(&(ei->i_block_reservation_lock));
705 INIT_LIST_HEAD(&ei->i_aio_dio_complete_list);
706 ei->cur_aio_dio = NULL;
707
708 return &ei->vfs_inode;
709 }
710
711 static void ext4_destroy_inode(struct inode *inode)
712 {
713 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
714 ext4_msg(inode->i_sb, KERN_ERR,
715 "Inode %lu (%p): orphan list check failed!",
716 inode->i_ino, EXT4_I(inode));
717 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
718 EXT4_I(inode), sizeof(struct ext4_inode_info),
719 true);
720 dump_stack();
721 }
722 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
723 }
724
725 static void init_once(void *foo)
726 {
727 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
728
729 INIT_LIST_HEAD(&ei->i_orphan);
730 #ifdef CONFIG_EXT4_FS_XATTR
731 init_rwsem(&ei->xattr_sem);
732 #endif
733 init_rwsem(&ei->i_data_sem);
734 inode_init_once(&ei->vfs_inode);
735 }
736
737 static int init_inodecache(void)
738 {
739 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
740 sizeof(struct ext4_inode_info),
741 0, (SLAB_RECLAIM_ACCOUNT|
742 SLAB_MEM_SPREAD),
743 init_once);
744 if (ext4_inode_cachep == NULL)
745 return -ENOMEM;
746 return 0;
747 }
748
749 static void destroy_inodecache(void)
750 {
751 kmem_cache_destroy(ext4_inode_cachep);
752 }
753
754 static void ext4_clear_inode(struct inode *inode)
755 {
756 ext4_discard_preallocations(inode);
757 if (EXT4_JOURNAL(inode))
758 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
759 &EXT4_I(inode)->jinode);
760 }
761
762 static inline void ext4_show_quota_options(struct seq_file *seq,
763 struct super_block *sb)
764 {
765 #if defined(CONFIG_QUOTA)
766 struct ext4_sb_info *sbi = EXT4_SB(sb);
767
768 if (sbi->s_jquota_fmt)
769 seq_printf(seq, ",jqfmt=%s",
770 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
771
772 if (sbi->s_qf_names[USRQUOTA])
773 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
774
775 if (sbi->s_qf_names[GRPQUOTA])
776 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
777
778 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
779 seq_puts(seq, ",usrquota");
780
781 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
782 seq_puts(seq, ",grpquota");
783 #endif
784 }
785
786 /*
787 * Show an option if
788 * - it's set to a non-default value OR
789 * - if the per-sb default is different from the global default
790 */
791 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
792 {
793 int def_errors;
794 unsigned long def_mount_opts;
795 struct super_block *sb = vfs->mnt_sb;
796 struct ext4_sb_info *sbi = EXT4_SB(sb);
797 struct ext4_super_block *es = sbi->s_es;
798
799 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
800 def_errors = le16_to_cpu(es->s_errors);
801
802 if (sbi->s_sb_block != 1)
803 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
804 if (test_opt(sb, MINIX_DF))
805 seq_puts(seq, ",minixdf");
806 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
807 seq_puts(seq, ",grpid");
808 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
809 seq_puts(seq, ",nogrpid");
810 if (sbi->s_resuid != EXT4_DEF_RESUID ||
811 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
812 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
813 }
814 if (sbi->s_resgid != EXT4_DEF_RESGID ||
815 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
816 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
817 }
818 if (test_opt(sb, ERRORS_RO)) {
819 if (def_errors == EXT4_ERRORS_PANIC ||
820 def_errors == EXT4_ERRORS_CONTINUE) {
821 seq_puts(seq, ",errors=remount-ro");
822 }
823 }
824 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
825 seq_puts(seq, ",errors=continue");
826 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
827 seq_puts(seq, ",errors=panic");
828 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
829 seq_puts(seq, ",nouid32");
830 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
831 seq_puts(seq, ",debug");
832 if (test_opt(sb, OLDALLOC))
833 seq_puts(seq, ",oldalloc");
834 #ifdef CONFIG_EXT4_FS_XATTR
835 if (test_opt(sb, XATTR_USER) &&
836 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
837 seq_puts(seq, ",user_xattr");
838 if (!test_opt(sb, XATTR_USER) &&
839 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
840 seq_puts(seq, ",nouser_xattr");
841 }
842 #endif
843 #ifdef CONFIG_EXT4_FS_POSIX_ACL
844 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
845 seq_puts(seq, ",acl");
846 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
847 seq_puts(seq, ",noacl");
848 #endif
849 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
850 seq_printf(seq, ",commit=%u",
851 (unsigned) (sbi->s_commit_interval / HZ));
852 }
853 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
854 seq_printf(seq, ",min_batch_time=%u",
855 (unsigned) sbi->s_min_batch_time);
856 }
857 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
858 seq_printf(seq, ",max_batch_time=%u",
859 (unsigned) sbi->s_min_batch_time);
860 }
861
862 /*
863 * We're changing the default of barrier mount option, so
864 * let's always display its mount state so it's clear what its
865 * status is.
866 */
867 seq_puts(seq, ",barrier=");
868 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
869 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
870 seq_puts(seq, ",journal_async_commit");
871 if (test_opt(sb, NOBH))
872 seq_puts(seq, ",nobh");
873 if (test_opt(sb, I_VERSION))
874 seq_puts(seq, ",i_version");
875 if (!test_opt(sb, DELALLOC))
876 seq_puts(seq, ",nodelalloc");
877
878
879 if (sbi->s_stripe)
880 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
881 /*
882 * journal mode get enabled in different ways
883 * So just print the value even if we didn't specify it
884 */
885 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
886 seq_puts(seq, ",data=journal");
887 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
888 seq_puts(seq, ",data=ordered");
889 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
890 seq_puts(seq, ",data=writeback");
891
892 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
893 seq_printf(seq, ",inode_readahead_blks=%u",
894 sbi->s_inode_readahead_blks);
895
896 if (test_opt(sb, DATA_ERR_ABORT))
897 seq_puts(seq, ",data_err=abort");
898
899 if (test_opt(sb, NO_AUTO_DA_ALLOC))
900 seq_puts(seq, ",noauto_da_alloc");
901
902 ext4_show_quota_options(seq, sb);
903
904 return 0;
905 }
906
907 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
908 u64 ino, u32 generation)
909 {
910 struct inode *inode;
911
912 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
913 return ERR_PTR(-ESTALE);
914 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
915 return ERR_PTR(-ESTALE);
916
917 /* iget isn't really right if the inode is currently unallocated!!
918 *
919 * ext4_read_inode will return a bad_inode if the inode had been
920 * deleted, so we should be safe.
921 *
922 * Currently we don't know the generation for parent directory, so
923 * a generation of 0 means "accept any"
924 */
925 inode = ext4_iget(sb, ino);
926 if (IS_ERR(inode))
927 return ERR_CAST(inode);
928 if (generation && inode->i_generation != generation) {
929 iput(inode);
930 return ERR_PTR(-ESTALE);
931 }
932
933 return inode;
934 }
935
936 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
937 int fh_len, int fh_type)
938 {
939 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
940 ext4_nfs_get_inode);
941 }
942
943 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
944 int fh_len, int fh_type)
945 {
946 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
947 ext4_nfs_get_inode);
948 }
949
950 /*
951 * Try to release metadata pages (indirect blocks, directories) which are
952 * mapped via the block device. Since these pages could have journal heads
953 * which would prevent try_to_free_buffers() from freeing them, we must use
954 * jbd2 layer's try_to_free_buffers() function to release them.
955 */
956 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
957 gfp_t wait)
958 {
959 journal_t *journal = EXT4_SB(sb)->s_journal;
960
961 WARN_ON(PageChecked(page));
962 if (!page_has_buffers(page))
963 return 0;
964 if (journal)
965 return jbd2_journal_try_to_free_buffers(journal, page,
966 wait & ~__GFP_WAIT);
967 return try_to_free_buffers(page);
968 }
969
970 #ifdef CONFIG_QUOTA
971 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
972 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
973
974 static int ext4_write_dquot(struct dquot *dquot);
975 static int ext4_acquire_dquot(struct dquot *dquot);
976 static int ext4_release_dquot(struct dquot *dquot);
977 static int ext4_mark_dquot_dirty(struct dquot *dquot);
978 static int ext4_write_info(struct super_block *sb, int type);
979 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
980 char *path, int remount);
981 static int ext4_quota_on_mount(struct super_block *sb, int type);
982 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
983 size_t len, loff_t off);
984 static ssize_t ext4_quota_write(struct super_block *sb, int type,
985 const char *data, size_t len, loff_t off);
986
987 static const struct dquot_operations ext4_quota_operations = {
988 .initialize = dquot_initialize,
989 .drop = dquot_drop,
990 .alloc_space = dquot_alloc_space,
991 .reserve_space = dquot_reserve_space,
992 .claim_space = dquot_claim_space,
993 .release_rsv = dquot_release_reserved_space,
994 .get_reserved_space = ext4_get_reserved_space,
995 .alloc_inode = dquot_alloc_inode,
996 .free_space = dquot_free_space,
997 .free_inode = dquot_free_inode,
998 .transfer = dquot_transfer,
999 .write_dquot = ext4_write_dquot,
1000 .acquire_dquot = ext4_acquire_dquot,
1001 .release_dquot = ext4_release_dquot,
1002 .mark_dirty = ext4_mark_dquot_dirty,
1003 .write_info = ext4_write_info,
1004 .alloc_dquot = dquot_alloc,
1005 .destroy_dquot = dquot_destroy,
1006 };
1007
1008 static const struct quotactl_ops ext4_qctl_operations = {
1009 .quota_on = ext4_quota_on,
1010 .quota_off = vfs_quota_off,
1011 .quota_sync = vfs_quota_sync,
1012 .get_info = vfs_get_dqinfo,
1013 .set_info = vfs_set_dqinfo,
1014 .get_dqblk = vfs_get_dqblk,
1015 .set_dqblk = vfs_set_dqblk
1016 };
1017 #endif
1018
1019 static const struct super_operations ext4_sops = {
1020 .alloc_inode = ext4_alloc_inode,
1021 .destroy_inode = ext4_destroy_inode,
1022 .write_inode = ext4_write_inode,
1023 .dirty_inode = ext4_dirty_inode,
1024 .delete_inode = ext4_delete_inode,
1025 .put_super = ext4_put_super,
1026 .sync_fs = ext4_sync_fs,
1027 .freeze_fs = ext4_freeze,
1028 .unfreeze_fs = ext4_unfreeze,
1029 .statfs = ext4_statfs,
1030 .remount_fs = ext4_remount,
1031 .clear_inode = ext4_clear_inode,
1032 .show_options = ext4_show_options,
1033 #ifdef CONFIG_QUOTA
1034 .quota_read = ext4_quota_read,
1035 .quota_write = ext4_quota_write,
1036 #endif
1037 .bdev_try_to_free_page = bdev_try_to_free_page,
1038 };
1039
1040 static const struct super_operations ext4_nojournal_sops = {
1041 .alloc_inode = ext4_alloc_inode,
1042 .destroy_inode = ext4_destroy_inode,
1043 .write_inode = ext4_write_inode,
1044 .dirty_inode = ext4_dirty_inode,
1045 .delete_inode = ext4_delete_inode,
1046 .write_super = ext4_write_super,
1047 .put_super = ext4_put_super,
1048 .statfs = ext4_statfs,
1049 .remount_fs = ext4_remount,
1050 .clear_inode = ext4_clear_inode,
1051 .show_options = ext4_show_options,
1052 #ifdef CONFIG_QUOTA
1053 .quota_read = ext4_quota_read,
1054 .quota_write = ext4_quota_write,
1055 #endif
1056 .bdev_try_to_free_page = bdev_try_to_free_page,
1057 };
1058
1059 static const struct export_operations ext4_export_ops = {
1060 .fh_to_dentry = ext4_fh_to_dentry,
1061 .fh_to_parent = ext4_fh_to_parent,
1062 .get_parent = ext4_get_parent,
1063 };
1064
1065 enum {
1066 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1067 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1068 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1069 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1070 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1071 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1072 Opt_journal_update, Opt_journal_dev,
1073 Opt_journal_checksum, Opt_journal_async_commit,
1074 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1075 Opt_data_err_abort, Opt_data_err_ignore,
1076 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1077 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
1078 Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err, Opt_resize,
1079 Opt_usrquota, Opt_grpquota, Opt_i_version,
1080 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1081 Opt_block_validity, Opt_noblock_validity,
1082 Opt_inode_readahead_blks, Opt_journal_ioprio
1083 };
1084
1085 static const match_table_t tokens = {
1086 {Opt_bsd_df, "bsddf"},
1087 {Opt_minix_df, "minixdf"},
1088 {Opt_grpid, "grpid"},
1089 {Opt_grpid, "bsdgroups"},
1090 {Opt_nogrpid, "nogrpid"},
1091 {Opt_nogrpid, "sysvgroups"},
1092 {Opt_resgid, "resgid=%u"},
1093 {Opt_resuid, "resuid=%u"},
1094 {Opt_sb, "sb=%u"},
1095 {Opt_err_cont, "errors=continue"},
1096 {Opt_err_panic, "errors=panic"},
1097 {Opt_err_ro, "errors=remount-ro"},
1098 {Opt_nouid32, "nouid32"},
1099 {Opt_debug, "debug"},
1100 {Opt_oldalloc, "oldalloc"},
1101 {Opt_orlov, "orlov"},
1102 {Opt_user_xattr, "user_xattr"},
1103 {Opt_nouser_xattr, "nouser_xattr"},
1104 {Opt_acl, "acl"},
1105 {Opt_noacl, "noacl"},
1106 {Opt_noload, "noload"},
1107 {Opt_nobh, "nobh"},
1108 {Opt_bh, "bh"},
1109 {Opt_commit, "commit=%u"},
1110 {Opt_min_batch_time, "min_batch_time=%u"},
1111 {Opt_max_batch_time, "max_batch_time=%u"},
1112 {Opt_journal_update, "journal=update"},
1113 {Opt_journal_dev, "journal_dev=%u"},
1114 {Opt_journal_checksum, "journal_checksum"},
1115 {Opt_journal_async_commit, "journal_async_commit"},
1116 {Opt_abort, "abort"},
1117 {Opt_data_journal, "data=journal"},
1118 {Opt_data_ordered, "data=ordered"},
1119 {Opt_data_writeback, "data=writeback"},
1120 {Opt_data_err_abort, "data_err=abort"},
1121 {Opt_data_err_ignore, "data_err=ignore"},
1122 {Opt_offusrjquota, "usrjquota="},
1123 {Opt_usrjquota, "usrjquota=%s"},
1124 {Opt_offgrpjquota, "grpjquota="},
1125 {Opt_grpjquota, "grpjquota=%s"},
1126 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1127 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1128 {Opt_grpquota, "grpquota"},
1129 {Opt_noquota, "noquota"},
1130 {Opt_quota, "quota"},
1131 {Opt_usrquota, "usrquota"},
1132 {Opt_barrier, "barrier=%u"},
1133 {Opt_barrier, "barrier"},
1134 {Opt_nobarrier, "nobarrier"},
1135 {Opt_i_version, "i_version"},
1136 {Opt_stripe, "stripe=%u"},
1137 {Opt_resize, "resize"},
1138 {Opt_delalloc, "delalloc"},
1139 {Opt_nodelalloc, "nodelalloc"},
1140 {Opt_block_validity, "block_validity"},
1141 {Opt_noblock_validity, "noblock_validity"},
1142 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1143 {Opt_journal_ioprio, "journal_ioprio=%u"},
1144 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1145 {Opt_auto_da_alloc, "auto_da_alloc"},
1146 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1147 {Opt_err, NULL},
1148 };
1149
1150 static ext4_fsblk_t get_sb_block(void **data)
1151 {
1152 ext4_fsblk_t sb_block;
1153 char *options = (char *) *data;
1154
1155 if (!options || strncmp(options, "sb=", 3) != 0)
1156 return 1; /* Default location */
1157
1158 options += 3;
1159 /* TODO: use simple_strtoll with >32bit ext4 */
1160 sb_block = simple_strtoul(options, &options, 0);
1161 if (*options && *options != ',') {
1162 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1163 (char *) *data);
1164 return 1;
1165 }
1166 if (*options == ',')
1167 options++;
1168 *data = (void *) options;
1169
1170 return sb_block;
1171 }
1172
1173 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1174
1175 static int parse_options(char *options, struct super_block *sb,
1176 unsigned long *journal_devnum,
1177 unsigned int *journal_ioprio,
1178 ext4_fsblk_t *n_blocks_count, int is_remount)
1179 {
1180 struct ext4_sb_info *sbi = EXT4_SB(sb);
1181 char *p;
1182 substring_t args[MAX_OPT_ARGS];
1183 int data_opt = 0;
1184 int option;
1185 #ifdef CONFIG_QUOTA
1186 int qtype, qfmt;
1187 char *qname;
1188 #endif
1189
1190 if (!options)
1191 return 1;
1192
1193 while ((p = strsep(&options, ",")) != NULL) {
1194 int token;
1195 if (!*p)
1196 continue;
1197
1198 token = match_token(p, tokens, args);
1199 switch (token) {
1200 case Opt_bsd_df:
1201 clear_opt(sbi->s_mount_opt, MINIX_DF);
1202 break;
1203 case Opt_minix_df:
1204 set_opt(sbi->s_mount_opt, MINIX_DF);
1205 break;
1206 case Opt_grpid:
1207 set_opt(sbi->s_mount_opt, GRPID);
1208 break;
1209 case Opt_nogrpid:
1210 clear_opt(sbi->s_mount_opt, GRPID);
1211 break;
1212 case Opt_resuid:
1213 if (match_int(&args[0], &option))
1214 return 0;
1215 sbi->s_resuid = option;
1216 break;
1217 case Opt_resgid:
1218 if (match_int(&args[0], &option))
1219 return 0;
1220 sbi->s_resgid = option;
1221 break;
1222 case Opt_sb:
1223 /* handled by get_sb_block() instead of here */
1224 /* *sb_block = match_int(&args[0]); */
1225 break;
1226 case Opt_err_panic:
1227 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1228 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1229 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1230 break;
1231 case Opt_err_ro:
1232 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1233 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1234 set_opt(sbi->s_mount_opt, ERRORS_RO);
1235 break;
1236 case Opt_err_cont:
1237 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1238 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1239 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1240 break;
1241 case Opt_nouid32:
1242 set_opt(sbi->s_mount_opt, NO_UID32);
1243 break;
1244 case Opt_debug:
1245 set_opt(sbi->s_mount_opt, DEBUG);
1246 break;
1247 case Opt_oldalloc:
1248 set_opt(sbi->s_mount_opt, OLDALLOC);
1249 break;
1250 case Opt_orlov:
1251 clear_opt(sbi->s_mount_opt, OLDALLOC);
1252 break;
1253 #ifdef CONFIG_EXT4_FS_XATTR
1254 case Opt_user_xattr:
1255 set_opt(sbi->s_mount_opt, XATTR_USER);
1256 break;
1257 case Opt_nouser_xattr:
1258 clear_opt(sbi->s_mount_opt, XATTR_USER);
1259 break;
1260 #else
1261 case Opt_user_xattr:
1262 case Opt_nouser_xattr:
1263 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1264 break;
1265 #endif
1266 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1267 case Opt_acl:
1268 set_opt(sbi->s_mount_opt, POSIX_ACL);
1269 break;
1270 case Opt_noacl:
1271 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1272 break;
1273 #else
1274 case Opt_acl:
1275 case Opt_noacl:
1276 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1277 break;
1278 #endif
1279 case Opt_journal_update:
1280 /* @@@ FIXME */
1281 /* Eventually we will want to be able to create
1282 a journal file here. For now, only allow the
1283 user to specify an existing inode to be the
1284 journal file. */
1285 if (is_remount) {
1286 ext4_msg(sb, KERN_ERR,
1287 "Cannot specify journal on remount");
1288 return 0;
1289 }
1290 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1291 break;
1292 case Opt_journal_dev:
1293 if (is_remount) {
1294 ext4_msg(sb, KERN_ERR,
1295 "Cannot specify journal on remount");
1296 return 0;
1297 }
1298 if (match_int(&args[0], &option))
1299 return 0;
1300 *journal_devnum = option;
1301 break;
1302 case Opt_journal_checksum:
1303 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1304 break;
1305 case Opt_journal_async_commit:
1306 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1307 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1308 break;
1309 case Opt_noload:
1310 set_opt(sbi->s_mount_opt, NOLOAD);
1311 break;
1312 case Opt_commit:
1313 if (match_int(&args[0], &option))
1314 return 0;
1315 if (option < 0)
1316 return 0;
1317 if (option == 0)
1318 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1319 sbi->s_commit_interval = HZ * option;
1320 break;
1321 case Opt_max_batch_time:
1322 if (match_int(&args[0], &option))
1323 return 0;
1324 if (option < 0)
1325 return 0;
1326 if (option == 0)
1327 option = EXT4_DEF_MAX_BATCH_TIME;
1328 sbi->s_max_batch_time = option;
1329 break;
1330 case Opt_min_batch_time:
1331 if (match_int(&args[0], &option))
1332 return 0;
1333 if (option < 0)
1334 return 0;
1335 sbi->s_min_batch_time = option;
1336 break;
1337 case Opt_data_journal:
1338 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1339 goto datacheck;
1340 case Opt_data_ordered:
1341 data_opt = EXT4_MOUNT_ORDERED_DATA;
1342 goto datacheck;
1343 case Opt_data_writeback:
1344 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1345 datacheck:
1346 if (is_remount) {
1347 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1348 != data_opt) {
1349 ext4_msg(sb, KERN_ERR,
1350 "Cannot change data mode on remount");
1351 return 0;
1352 }
1353 } else {
1354 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1355 sbi->s_mount_opt |= data_opt;
1356 }
1357 break;
1358 case Opt_data_err_abort:
1359 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1360 break;
1361 case Opt_data_err_ignore:
1362 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1363 break;
1364 #ifdef CONFIG_QUOTA
1365 case Opt_usrjquota:
1366 qtype = USRQUOTA;
1367 goto set_qf_name;
1368 case Opt_grpjquota:
1369 qtype = GRPQUOTA;
1370 set_qf_name:
1371 if (sb_any_quota_loaded(sb) &&
1372 !sbi->s_qf_names[qtype]) {
1373 ext4_msg(sb, KERN_ERR,
1374 "Cannot change journaled "
1375 "quota options when quota turned on");
1376 return 0;
1377 }
1378 qname = match_strdup(&args[0]);
1379 if (!qname) {
1380 ext4_msg(sb, KERN_ERR,
1381 "Not enough memory for "
1382 "storing quotafile name");
1383 return 0;
1384 }
1385 if (sbi->s_qf_names[qtype] &&
1386 strcmp(sbi->s_qf_names[qtype], qname)) {
1387 ext4_msg(sb, KERN_ERR,
1388 "%s quota file already "
1389 "specified", QTYPE2NAME(qtype));
1390 kfree(qname);
1391 return 0;
1392 }
1393 sbi->s_qf_names[qtype] = qname;
1394 if (strchr(sbi->s_qf_names[qtype], '/')) {
1395 ext4_msg(sb, KERN_ERR,
1396 "quotafile must be on "
1397 "filesystem root");
1398 kfree(sbi->s_qf_names[qtype]);
1399 sbi->s_qf_names[qtype] = NULL;
1400 return 0;
1401 }
1402 set_opt(sbi->s_mount_opt, QUOTA);
1403 break;
1404 case Opt_offusrjquota:
1405 qtype = USRQUOTA;
1406 goto clear_qf_name;
1407 case Opt_offgrpjquota:
1408 qtype = GRPQUOTA;
1409 clear_qf_name:
1410 if (sb_any_quota_loaded(sb) &&
1411 sbi->s_qf_names[qtype]) {
1412 ext4_msg(sb, KERN_ERR, "Cannot change "
1413 "journaled quota options when "
1414 "quota turned on");
1415 return 0;
1416 }
1417 /*
1418 * The space will be released later when all options
1419 * are confirmed to be correct
1420 */
1421 sbi->s_qf_names[qtype] = NULL;
1422 break;
1423 case Opt_jqfmt_vfsold:
1424 qfmt = QFMT_VFS_OLD;
1425 goto set_qf_format;
1426 case Opt_jqfmt_vfsv0:
1427 qfmt = QFMT_VFS_V0;
1428 set_qf_format:
1429 if (sb_any_quota_loaded(sb) &&
1430 sbi->s_jquota_fmt != qfmt) {
1431 ext4_msg(sb, KERN_ERR, "Cannot change "
1432 "journaled quota options when "
1433 "quota turned on");
1434 return 0;
1435 }
1436 sbi->s_jquota_fmt = qfmt;
1437 break;
1438 case Opt_quota:
1439 case Opt_usrquota:
1440 set_opt(sbi->s_mount_opt, QUOTA);
1441 set_opt(sbi->s_mount_opt, USRQUOTA);
1442 break;
1443 case Opt_grpquota:
1444 set_opt(sbi->s_mount_opt, QUOTA);
1445 set_opt(sbi->s_mount_opt, GRPQUOTA);
1446 break;
1447 case Opt_noquota:
1448 if (sb_any_quota_loaded(sb)) {
1449 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1450 "options when quota turned on");
1451 return 0;
1452 }
1453 clear_opt(sbi->s_mount_opt, QUOTA);
1454 clear_opt(sbi->s_mount_opt, USRQUOTA);
1455 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1456 break;
1457 #else
1458 case Opt_quota:
1459 case Opt_usrquota:
1460 case Opt_grpquota:
1461 ext4_msg(sb, KERN_ERR,
1462 "quota options not supported");
1463 break;
1464 case Opt_usrjquota:
1465 case Opt_grpjquota:
1466 case Opt_offusrjquota:
1467 case Opt_offgrpjquota:
1468 case Opt_jqfmt_vfsold:
1469 case Opt_jqfmt_vfsv0:
1470 ext4_msg(sb, KERN_ERR,
1471 "journaled quota options not supported");
1472 break;
1473 case Opt_noquota:
1474 break;
1475 #endif
1476 case Opt_abort:
1477 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1478 break;
1479 case Opt_nobarrier:
1480 clear_opt(sbi->s_mount_opt, BARRIER);
1481 break;
1482 case Opt_barrier:
1483 if (match_int(&args[0], &option)) {
1484 set_opt(sbi->s_mount_opt, BARRIER);
1485 break;
1486 }
1487 if (option)
1488 set_opt(sbi->s_mount_opt, BARRIER);
1489 else
1490 clear_opt(sbi->s_mount_opt, BARRIER);
1491 break;
1492 case Opt_ignore:
1493 break;
1494 case Opt_resize:
1495 if (!is_remount) {
1496 ext4_msg(sb, KERN_ERR,
1497 "resize option only available "
1498 "for remount");
1499 return 0;
1500 }
1501 if (match_int(&args[0], &option) != 0)
1502 return 0;
1503 *n_blocks_count = option;
1504 break;
1505 case Opt_nobh:
1506 set_opt(sbi->s_mount_opt, NOBH);
1507 break;
1508 case Opt_bh:
1509 clear_opt(sbi->s_mount_opt, NOBH);
1510 break;
1511 case Opt_i_version:
1512 set_opt(sbi->s_mount_opt, I_VERSION);
1513 sb->s_flags |= MS_I_VERSION;
1514 break;
1515 case Opt_nodelalloc:
1516 clear_opt(sbi->s_mount_opt, DELALLOC);
1517 break;
1518 case Opt_stripe:
1519 if (match_int(&args[0], &option))
1520 return 0;
1521 if (option < 0)
1522 return 0;
1523 sbi->s_stripe = option;
1524 break;
1525 case Opt_delalloc:
1526 set_opt(sbi->s_mount_opt, DELALLOC);
1527 break;
1528 case Opt_block_validity:
1529 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1530 break;
1531 case Opt_noblock_validity:
1532 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1533 break;
1534 case Opt_inode_readahead_blks:
1535 if (match_int(&args[0], &option))
1536 return 0;
1537 if (option < 0 || option > (1 << 30))
1538 return 0;
1539 if (!is_power_of_2(option)) {
1540 ext4_msg(sb, KERN_ERR,
1541 "EXT4-fs: inode_readahead_blks"
1542 " must be a power of 2");
1543 return 0;
1544 }
1545 sbi->s_inode_readahead_blks = option;
1546 break;
1547 case Opt_journal_ioprio:
1548 if (match_int(&args[0], &option))
1549 return 0;
1550 if (option < 0 || option > 7)
1551 break;
1552 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1553 option);
1554 break;
1555 case Opt_noauto_da_alloc:
1556 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1557 break;
1558 case Opt_auto_da_alloc:
1559 if (match_int(&args[0], &option)) {
1560 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1561 break;
1562 }
1563 if (option)
1564 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1565 else
1566 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1567 break;
1568 default:
1569 ext4_msg(sb, KERN_ERR,
1570 "Unrecognized mount option \"%s\" "
1571 "or missing value", p);
1572 return 0;
1573 }
1574 }
1575 #ifdef CONFIG_QUOTA
1576 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1577 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1578 sbi->s_qf_names[USRQUOTA])
1579 clear_opt(sbi->s_mount_opt, USRQUOTA);
1580
1581 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1582 sbi->s_qf_names[GRPQUOTA])
1583 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1584
1585 if ((sbi->s_qf_names[USRQUOTA] &&
1586 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1587 (sbi->s_qf_names[GRPQUOTA] &&
1588 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1589 ext4_msg(sb, KERN_ERR, "old and new quota "
1590 "format mixing");
1591 return 0;
1592 }
1593
1594 if (!sbi->s_jquota_fmt) {
1595 ext4_msg(sb, KERN_ERR, "journaled quota format "
1596 "not specified");
1597 return 0;
1598 }
1599 } else {
1600 if (sbi->s_jquota_fmt) {
1601 ext4_msg(sb, KERN_ERR, "journaled quota format "
1602 "specified with no journaling "
1603 "enabled");
1604 return 0;
1605 }
1606 }
1607 #endif
1608 return 1;
1609 }
1610
1611 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1612 int read_only)
1613 {
1614 struct ext4_sb_info *sbi = EXT4_SB(sb);
1615 int res = 0;
1616
1617 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1618 ext4_msg(sb, KERN_ERR, "revision level too high, "
1619 "forcing read-only mode");
1620 res = MS_RDONLY;
1621 }
1622 if (read_only)
1623 return res;
1624 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1625 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1626 "running e2fsck is recommended");
1627 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1628 ext4_msg(sb, KERN_WARNING,
1629 "warning: mounting fs with errors, "
1630 "running e2fsck is recommended");
1631 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1632 le16_to_cpu(es->s_mnt_count) >=
1633 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1634 ext4_msg(sb, KERN_WARNING,
1635 "warning: maximal mount count reached, "
1636 "running e2fsck is recommended");
1637 else if (le32_to_cpu(es->s_checkinterval) &&
1638 (le32_to_cpu(es->s_lastcheck) +
1639 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1640 ext4_msg(sb, KERN_WARNING,
1641 "warning: checktime reached, "
1642 "running e2fsck is recommended");
1643 if (!sbi->s_journal)
1644 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1645 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1646 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1647 le16_add_cpu(&es->s_mnt_count, 1);
1648 es->s_mtime = cpu_to_le32(get_seconds());
1649 ext4_update_dynamic_rev(sb);
1650 if (sbi->s_journal)
1651 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1652
1653 ext4_commit_super(sb, 1);
1654 if (test_opt(sb, DEBUG))
1655 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1656 "bpg=%lu, ipg=%lu, mo=%04x]\n",
1657 sb->s_blocksize,
1658 sbi->s_groups_count,
1659 EXT4_BLOCKS_PER_GROUP(sb),
1660 EXT4_INODES_PER_GROUP(sb),
1661 sbi->s_mount_opt);
1662
1663 return res;
1664 }
1665
1666 static int ext4_fill_flex_info(struct super_block *sb)
1667 {
1668 struct ext4_sb_info *sbi = EXT4_SB(sb);
1669 struct ext4_group_desc *gdp = NULL;
1670 ext4_group_t flex_group_count;
1671 ext4_group_t flex_group;
1672 int groups_per_flex = 0;
1673 size_t size;
1674 int i;
1675
1676 if (!sbi->s_es->s_log_groups_per_flex) {
1677 sbi->s_log_groups_per_flex = 0;
1678 return 1;
1679 }
1680
1681 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1682 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1683
1684 /* We allocate both existing and potentially added groups */
1685 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1686 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1687 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1688 size = flex_group_count * sizeof(struct flex_groups);
1689 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1690 if (sbi->s_flex_groups == NULL) {
1691 sbi->s_flex_groups = vmalloc(size);
1692 if (sbi->s_flex_groups)
1693 memset(sbi->s_flex_groups, 0, size);
1694 }
1695 if (sbi->s_flex_groups == NULL) {
1696 ext4_msg(sb, KERN_ERR, "not enough memory for "
1697 "%u flex groups", flex_group_count);
1698 goto failed;
1699 }
1700
1701 for (i = 0; i < sbi->s_groups_count; i++) {
1702 gdp = ext4_get_group_desc(sb, i, NULL);
1703
1704 flex_group = ext4_flex_group(sbi, i);
1705 atomic_add(ext4_free_inodes_count(sb, gdp),
1706 &sbi->s_flex_groups[flex_group].free_inodes);
1707 atomic_add(ext4_free_blks_count(sb, gdp),
1708 &sbi->s_flex_groups[flex_group].free_blocks);
1709 atomic_add(ext4_used_dirs_count(sb, gdp),
1710 &sbi->s_flex_groups[flex_group].used_dirs);
1711 }
1712
1713 return 1;
1714 failed:
1715 return 0;
1716 }
1717
1718 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1719 struct ext4_group_desc *gdp)
1720 {
1721 __u16 crc = 0;
1722
1723 if (sbi->s_es->s_feature_ro_compat &
1724 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1725 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1726 __le32 le_group = cpu_to_le32(block_group);
1727
1728 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1729 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1730 crc = crc16(crc, (__u8 *)gdp, offset);
1731 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1732 /* for checksum of struct ext4_group_desc do the rest...*/
1733 if ((sbi->s_es->s_feature_incompat &
1734 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1735 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1736 crc = crc16(crc, (__u8 *)gdp + offset,
1737 le16_to_cpu(sbi->s_es->s_desc_size) -
1738 offset);
1739 }
1740
1741 return cpu_to_le16(crc);
1742 }
1743
1744 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1745 struct ext4_group_desc *gdp)
1746 {
1747 if ((sbi->s_es->s_feature_ro_compat &
1748 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1749 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1750 return 0;
1751
1752 return 1;
1753 }
1754
1755 /* Called at mount-time, super-block is locked */
1756 static int ext4_check_descriptors(struct super_block *sb)
1757 {
1758 struct ext4_sb_info *sbi = EXT4_SB(sb);
1759 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1760 ext4_fsblk_t last_block;
1761 ext4_fsblk_t block_bitmap;
1762 ext4_fsblk_t inode_bitmap;
1763 ext4_fsblk_t inode_table;
1764 int flexbg_flag = 0;
1765 ext4_group_t i;
1766
1767 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1768 flexbg_flag = 1;
1769
1770 ext4_debug("Checking group descriptors");
1771
1772 for (i = 0; i < sbi->s_groups_count; i++) {
1773 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1774
1775 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1776 last_block = ext4_blocks_count(sbi->s_es) - 1;
1777 else
1778 last_block = first_block +
1779 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1780
1781 block_bitmap = ext4_block_bitmap(sb, gdp);
1782 if (block_bitmap < first_block || block_bitmap > last_block) {
1783 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1784 "Block bitmap for group %u not in group "
1785 "(block %llu)!", i, block_bitmap);
1786 return 0;
1787 }
1788 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1789 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1790 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1791 "Inode bitmap for group %u not in group "
1792 "(block %llu)!", i, inode_bitmap);
1793 return 0;
1794 }
1795 inode_table = ext4_inode_table(sb, gdp);
1796 if (inode_table < first_block ||
1797 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1798 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1799 "Inode table for group %u not in group "
1800 "(block %llu)!", i, inode_table);
1801 return 0;
1802 }
1803 ext4_lock_group(sb, i);
1804 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1805 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1806 "Checksum for group %u failed (%u!=%u)",
1807 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1808 gdp)), le16_to_cpu(gdp->bg_checksum));
1809 if (!(sb->s_flags & MS_RDONLY)) {
1810 ext4_unlock_group(sb, i);
1811 return 0;
1812 }
1813 }
1814 ext4_unlock_group(sb, i);
1815 if (!flexbg_flag)
1816 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1817 }
1818
1819 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1820 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
1821 return 1;
1822 }
1823
1824 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1825 * the superblock) which were deleted from all directories, but held open by
1826 * a process at the time of a crash. We walk the list and try to delete these
1827 * inodes at recovery time (only with a read-write filesystem).
1828 *
1829 * In order to keep the orphan inode chain consistent during traversal (in
1830 * case of crash during recovery), we link each inode into the superblock
1831 * orphan list_head and handle it the same way as an inode deletion during
1832 * normal operation (which journals the operations for us).
1833 *
1834 * We only do an iget() and an iput() on each inode, which is very safe if we
1835 * accidentally point at an in-use or already deleted inode. The worst that
1836 * can happen in this case is that we get a "bit already cleared" message from
1837 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1838 * e2fsck was run on this filesystem, and it must have already done the orphan
1839 * inode cleanup for us, so we can safely abort without any further action.
1840 */
1841 static void ext4_orphan_cleanup(struct super_block *sb,
1842 struct ext4_super_block *es)
1843 {
1844 unsigned int s_flags = sb->s_flags;
1845 int nr_orphans = 0, nr_truncates = 0;
1846 #ifdef CONFIG_QUOTA
1847 int i;
1848 #endif
1849 if (!es->s_last_orphan) {
1850 jbd_debug(4, "no orphan inodes to clean up\n");
1851 return;
1852 }
1853
1854 if (bdev_read_only(sb->s_bdev)) {
1855 ext4_msg(sb, KERN_ERR, "write access "
1856 "unavailable, skipping orphan cleanup");
1857 return;
1858 }
1859
1860 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1861 if (es->s_last_orphan)
1862 jbd_debug(1, "Errors on filesystem, "
1863 "clearing orphan list.\n");
1864 es->s_last_orphan = 0;
1865 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1866 return;
1867 }
1868
1869 if (s_flags & MS_RDONLY) {
1870 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1871 sb->s_flags &= ~MS_RDONLY;
1872 }
1873 #ifdef CONFIG_QUOTA
1874 /* Needed for iput() to work correctly and not trash data */
1875 sb->s_flags |= MS_ACTIVE;
1876 /* Turn on quotas so that they are updated correctly */
1877 for (i = 0; i < MAXQUOTAS; i++) {
1878 if (EXT4_SB(sb)->s_qf_names[i]) {
1879 int ret = ext4_quota_on_mount(sb, i);
1880 if (ret < 0)
1881 ext4_msg(sb, KERN_ERR,
1882 "Cannot turn on journaled "
1883 "quota: error %d", ret);
1884 }
1885 }
1886 #endif
1887
1888 while (es->s_last_orphan) {
1889 struct inode *inode;
1890
1891 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1892 if (IS_ERR(inode)) {
1893 es->s_last_orphan = 0;
1894 break;
1895 }
1896
1897 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1898 vfs_dq_init(inode);
1899 if (inode->i_nlink) {
1900 ext4_msg(sb, KERN_DEBUG,
1901 "%s: truncating inode %lu to %lld bytes",
1902 __func__, inode->i_ino, inode->i_size);
1903 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1904 inode->i_ino, inode->i_size);
1905 ext4_truncate(inode);
1906 nr_truncates++;
1907 } else {
1908 ext4_msg(sb, KERN_DEBUG,
1909 "%s: deleting unreferenced inode %lu",
1910 __func__, inode->i_ino);
1911 jbd_debug(2, "deleting unreferenced inode %lu\n",
1912 inode->i_ino);
1913 nr_orphans++;
1914 }
1915 iput(inode); /* The delete magic happens here! */
1916 }
1917
1918 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1919
1920 if (nr_orphans)
1921 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1922 PLURAL(nr_orphans));
1923 if (nr_truncates)
1924 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1925 PLURAL(nr_truncates));
1926 #ifdef CONFIG_QUOTA
1927 /* Turn quotas off */
1928 for (i = 0; i < MAXQUOTAS; i++) {
1929 if (sb_dqopt(sb)->files[i])
1930 vfs_quota_off(sb, i, 0);
1931 }
1932 #endif
1933 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1934 }
1935
1936 /*
1937 * Maximal extent format file size.
1938 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1939 * extent format containers, within a sector_t, and within i_blocks
1940 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1941 * so that won't be a limiting factor.
1942 *
1943 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1944 */
1945 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1946 {
1947 loff_t res;
1948 loff_t upper_limit = MAX_LFS_FILESIZE;
1949
1950 /* small i_blocks in vfs inode? */
1951 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1952 /*
1953 * CONFIG_LBDAF is not enabled implies the inode
1954 * i_block represent total blocks in 512 bytes
1955 * 32 == size of vfs inode i_blocks * 8
1956 */
1957 upper_limit = (1LL << 32) - 1;
1958
1959 /* total blocks in file system block size */
1960 upper_limit >>= (blkbits - 9);
1961 upper_limit <<= blkbits;
1962 }
1963
1964 /* 32-bit extent-start container, ee_block */
1965 res = 1LL << 32;
1966 res <<= blkbits;
1967 res -= 1;
1968
1969 /* Sanity check against vm- & vfs- imposed limits */
1970 if (res > upper_limit)
1971 res = upper_limit;
1972
1973 return res;
1974 }
1975
1976 /*
1977 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1978 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1979 * We need to be 1 filesystem block less than the 2^48 sector limit.
1980 */
1981 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
1982 {
1983 loff_t res = EXT4_NDIR_BLOCKS;
1984 int meta_blocks;
1985 loff_t upper_limit;
1986 /* This is calculated to be the largest file size for a dense, block
1987 * mapped file such that the file's total number of 512-byte sectors,
1988 * including data and all indirect blocks, does not exceed (2^48 - 1).
1989 *
1990 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
1991 * number of 512-byte sectors of the file.
1992 */
1993
1994 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1995 /*
1996 * !has_huge_files or CONFIG_LBDAF not enabled implies that
1997 * the inode i_block field represents total file blocks in
1998 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
1999 */
2000 upper_limit = (1LL << 32) - 1;
2001
2002 /* total blocks in file system block size */
2003 upper_limit >>= (bits - 9);
2004
2005 } else {
2006 /*
2007 * We use 48 bit ext4_inode i_blocks
2008 * With EXT4_HUGE_FILE_FL set the i_blocks
2009 * represent total number of blocks in
2010 * file system block size
2011 */
2012 upper_limit = (1LL << 48) - 1;
2013
2014 }
2015
2016 /* indirect blocks */
2017 meta_blocks = 1;
2018 /* double indirect blocks */
2019 meta_blocks += 1 + (1LL << (bits-2));
2020 /* tripple indirect blocks */
2021 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2022
2023 upper_limit -= meta_blocks;
2024 upper_limit <<= bits;
2025
2026 res += 1LL << (bits-2);
2027 res += 1LL << (2*(bits-2));
2028 res += 1LL << (3*(bits-2));
2029 res <<= bits;
2030 if (res > upper_limit)
2031 res = upper_limit;
2032
2033 if (res > MAX_LFS_FILESIZE)
2034 res = MAX_LFS_FILESIZE;
2035
2036 return res;
2037 }
2038
2039 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2040 ext4_fsblk_t logical_sb_block, int nr)
2041 {
2042 struct ext4_sb_info *sbi = EXT4_SB(sb);
2043 ext4_group_t bg, first_meta_bg;
2044 int has_super = 0;
2045
2046 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2047
2048 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2049 nr < first_meta_bg)
2050 return logical_sb_block + nr + 1;
2051 bg = sbi->s_desc_per_block * nr;
2052 if (ext4_bg_has_super(sb, bg))
2053 has_super = 1;
2054
2055 return (has_super + ext4_group_first_block_no(sb, bg));
2056 }
2057
2058 /**
2059 * ext4_get_stripe_size: Get the stripe size.
2060 * @sbi: In memory super block info
2061 *
2062 * If we have specified it via mount option, then
2063 * use the mount option value. If the value specified at mount time is
2064 * greater than the blocks per group use the super block value.
2065 * If the super block value is greater than blocks per group return 0.
2066 * Allocator needs it be less than blocks per group.
2067 *
2068 */
2069 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2070 {
2071 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2072 unsigned long stripe_width =
2073 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2074
2075 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2076 return sbi->s_stripe;
2077
2078 if (stripe_width <= sbi->s_blocks_per_group)
2079 return stripe_width;
2080
2081 if (stride <= sbi->s_blocks_per_group)
2082 return stride;
2083
2084 return 0;
2085 }
2086
2087 /* sysfs supprt */
2088
2089 struct ext4_attr {
2090 struct attribute attr;
2091 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2092 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2093 const char *, size_t);
2094 int offset;
2095 };
2096
2097 static int parse_strtoul(const char *buf,
2098 unsigned long max, unsigned long *value)
2099 {
2100 char *endp;
2101
2102 while (*buf && isspace(*buf))
2103 buf++;
2104 *value = simple_strtoul(buf, &endp, 0);
2105 while (*endp && isspace(*endp))
2106 endp++;
2107 if (*endp || *value > max)
2108 return -EINVAL;
2109
2110 return 0;
2111 }
2112
2113 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2114 struct ext4_sb_info *sbi,
2115 char *buf)
2116 {
2117 return snprintf(buf, PAGE_SIZE, "%llu\n",
2118 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2119 }
2120
2121 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2122 struct ext4_sb_info *sbi, char *buf)
2123 {
2124 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2125
2126 return snprintf(buf, PAGE_SIZE, "%lu\n",
2127 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2128 sbi->s_sectors_written_start) >> 1);
2129 }
2130
2131 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2132 struct ext4_sb_info *sbi, char *buf)
2133 {
2134 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2135
2136 return snprintf(buf, PAGE_SIZE, "%llu\n",
2137 sbi->s_kbytes_written +
2138 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2139 EXT4_SB(sb)->s_sectors_written_start) >> 1));
2140 }
2141
2142 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2143 struct ext4_sb_info *sbi,
2144 const char *buf, size_t count)
2145 {
2146 unsigned long t;
2147
2148 if (parse_strtoul(buf, 0x40000000, &t))
2149 return -EINVAL;
2150
2151 if (!is_power_of_2(t))
2152 return -EINVAL;
2153
2154 sbi->s_inode_readahead_blks = t;
2155 return count;
2156 }
2157
2158 static ssize_t sbi_ui_show(struct ext4_attr *a,
2159 struct ext4_sb_info *sbi, char *buf)
2160 {
2161 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2162
2163 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2164 }
2165
2166 static ssize_t sbi_ui_store(struct ext4_attr *a,
2167 struct ext4_sb_info *sbi,
2168 const char *buf, size_t count)
2169 {
2170 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2171 unsigned long t;
2172
2173 if (parse_strtoul(buf, 0xffffffff, &t))
2174 return -EINVAL;
2175 *ui = t;
2176 return count;
2177 }
2178
2179 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2180 static struct ext4_attr ext4_attr_##_name = { \
2181 .attr = {.name = __stringify(_name), .mode = _mode }, \
2182 .show = _show, \
2183 .store = _store, \
2184 .offset = offsetof(struct ext4_sb_info, _elname), \
2185 }
2186 #define EXT4_ATTR(name, mode, show, store) \
2187 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2188
2189 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2190 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2191 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2192 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2193 #define ATTR_LIST(name) &ext4_attr_##name.attr
2194
2195 EXT4_RO_ATTR(delayed_allocation_blocks);
2196 EXT4_RO_ATTR(session_write_kbytes);
2197 EXT4_RO_ATTR(lifetime_write_kbytes);
2198 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2199 inode_readahead_blks_store, s_inode_readahead_blks);
2200 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2201 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2202 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2203 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2204 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2205 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2206 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2207 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2208
2209 static struct attribute *ext4_attrs[] = {
2210 ATTR_LIST(delayed_allocation_blocks),
2211 ATTR_LIST(session_write_kbytes),
2212 ATTR_LIST(lifetime_write_kbytes),
2213 ATTR_LIST(inode_readahead_blks),
2214 ATTR_LIST(inode_goal),
2215 ATTR_LIST(mb_stats),
2216 ATTR_LIST(mb_max_to_scan),
2217 ATTR_LIST(mb_min_to_scan),
2218 ATTR_LIST(mb_order2_req),
2219 ATTR_LIST(mb_stream_req),
2220 ATTR_LIST(mb_group_prealloc),
2221 ATTR_LIST(max_writeback_mb_bump),
2222 NULL,
2223 };
2224
2225 static ssize_t ext4_attr_show(struct kobject *kobj,
2226 struct attribute *attr, char *buf)
2227 {
2228 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2229 s_kobj);
2230 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2231
2232 return a->show ? a->show(a, sbi, buf) : 0;
2233 }
2234
2235 static ssize_t ext4_attr_store(struct kobject *kobj,
2236 struct attribute *attr,
2237 const char *buf, size_t len)
2238 {
2239 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2240 s_kobj);
2241 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2242
2243 return a->store ? a->store(a, sbi, buf, len) : 0;
2244 }
2245
2246 static void ext4_sb_release(struct kobject *kobj)
2247 {
2248 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2249 s_kobj);
2250 complete(&sbi->s_kobj_unregister);
2251 }
2252
2253
2254 static struct sysfs_ops ext4_attr_ops = {
2255 .show = ext4_attr_show,
2256 .store = ext4_attr_store,
2257 };
2258
2259 static struct kobj_type ext4_ktype = {
2260 .default_attrs = ext4_attrs,
2261 .sysfs_ops = &ext4_attr_ops,
2262 .release = ext4_sb_release,
2263 };
2264
2265 /*
2266 * Check whether this filesystem can be mounted based on
2267 * the features present and the RDONLY/RDWR mount requested.
2268 * Returns 1 if this filesystem can be mounted as requested,
2269 * 0 if it cannot be.
2270 */
2271 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2272 {
2273 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2274 ext4_msg(sb, KERN_ERR,
2275 "Couldn't mount because of "
2276 "unsupported optional features (%x)",
2277 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2278 ~EXT4_FEATURE_INCOMPAT_SUPP));
2279 return 0;
2280 }
2281
2282 if (readonly)
2283 return 1;
2284
2285 /* Check that feature set is OK for a read-write mount */
2286 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2287 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2288 "unsupported optional features (%x)",
2289 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2290 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2291 return 0;
2292 }
2293 /*
2294 * Large file size enabled file system can only be mounted
2295 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2296 */
2297 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2298 if (sizeof(blkcnt_t) < sizeof(u64)) {
2299 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2300 "cannot be mounted RDWR without "
2301 "CONFIG_LBDAF");
2302 return 0;
2303 }
2304 }
2305 return 1;
2306 }
2307
2308 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2309 __releases(kernel_lock)
2310 __acquires(kernel_lock)
2311 {
2312 struct buffer_head *bh;
2313 struct ext4_super_block *es = NULL;
2314 struct ext4_sb_info *sbi;
2315 ext4_fsblk_t block;
2316 ext4_fsblk_t sb_block = get_sb_block(&data);
2317 ext4_fsblk_t logical_sb_block;
2318 unsigned long offset = 0;
2319 unsigned long journal_devnum = 0;
2320 unsigned long def_mount_opts;
2321 struct inode *root;
2322 char *cp;
2323 const char *descr;
2324 int ret = -EINVAL;
2325 int blocksize;
2326 unsigned int db_count;
2327 unsigned int i;
2328 int needs_recovery, has_huge_files;
2329 __u64 blocks_count;
2330 int err;
2331 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2332
2333 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2334 if (!sbi)
2335 return -ENOMEM;
2336
2337 sbi->s_blockgroup_lock =
2338 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2339 if (!sbi->s_blockgroup_lock) {
2340 kfree(sbi);
2341 return -ENOMEM;
2342 }
2343 sb->s_fs_info = sbi;
2344 sbi->s_mount_opt = 0;
2345 sbi->s_resuid = EXT4_DEF_RESUID;
2346 sbi->s_resgid = EXT4_DEF_RESGID;
2347 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2348 sbi->s_sb_block = sb_block;
2349 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2350 sectors[1]);
2351
2352 unlock_kernel();
2353
2354 /* Cleanup superblock name */
2355 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2356 *cp = '!';
2357
2358 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2359 if (!blocksize) {
2360 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
2361 goto out_fail;
2362 }
2363
2364 /*
2365 * The ext4 superblock will not be buffer aligned for other than 1kB
2366 * block sizes. We need to calculate the offset from buffer start.
2367 */
2368 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2369 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2370 offset = do_div(logical_sb_block, blocksize);
2371 } else {
2372 logical_sb_block = sb_block;
2373 }
2374
2375 if (!(bh = sb_bread(sb, logical_sb_block))) {
2376 ext4_msg(sb, KERN_ERR, "unable to read superblock");
2377 goto out_fail;
2378 }
2379 /*
2380 * Note: s_es must be initialized as soon as possible because
2381 * some ext4 macro-instructions depend on its value
2382 */
2383 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2384 sbi->s_es = es;
2385 sb->s_magic = le16_to_cpu(es->s_magic);
2386 if (sb->s_magic != EXT4_SUPER_MAGIC)
2387 goto cantfind_ext4;
2388 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2389
2390 /* Set defaults before we parse the mount options */
2391 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2392 if (def_mount_opts & EXT4_DEFM_DEBUG)
2393 set_opt(sbi->s_mount_opt, DEBUG);
2394 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
2395 set_opt(sbi->s_mount_opt, GRPID);
2396 if (def_mount_opts & EXT4_DEFM_UID16)
2397 set_opt(sbi->s_mount_opt, NO_UID32);
2398 #ifdef CONFIG_EXT4_FS_XATTR
2399 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2400 set_opt(sbi->s_mount_opt, XATTR_USER);
2401 #endif
2402 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2403 if (def_mount_opts & EXT4_DEFM_ACL)
2404 set_opt(sbi->s_mount_opt, POSIX_ACL);
2405 #endif
2406 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2407 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2408 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2409 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2410 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2411 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2412
2413 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2414 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2415 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2416 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2417 else
2418 set_opt(sbi->s_mount_opt, ERRORS_RO);
2419
2420 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2421 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2422 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2423 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2424 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2425
2426 set_opt(sbi->s_mount_opt, BARRIER);
2427
2428 /*
2429 * enable delayed allocation by default
2430 * Use -o nodelalloc to turn it off
2431 */
2432 set_opt(sbi->s_mount_opt, DELALLOC);
2433
2434 if (!parse_options((char *) data, sb, &journal_devnum,
2435 &journal_ioprio, NULL, 0))
2436 goto failed_mount;
2437
2438 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2439 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2440
2441 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2442 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2443 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2444 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2445 ext4_msg(sb, KERN_WARNING,
2446 "feature flags set on rev 0 fs, "
2447 "running e2fsck is recommended");
2448
2449 /*
2450 * Check feature flags regardless of the revision level, since we
2451 * previously didn't change the revision level when setting the flags,
2452 * so there is a chance incompat flags are set on a rev 0 filesystem.
2453 */
2454 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
2455 goto failed_mount;
2456
2457 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2458
2459 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2460 blocksize > EXT4_MAX_BLOCK_SIZE) {
2461 ext4_msg(sb, KERN_ERR,
2462 "Unsupported filesystem blocksize %d", blocksize);
2463 goto failed_mount;
2464 }
2465
2466 if (sb->s_blocksize != blocksize) {
2467 /* Validate the filesystem blocksize */
2468 if (!sb_set_blocksize(sb, blocksize)) {
2469 ext4_msg(sb, KERN_ERR, "bad block size %d",
2470 blocksize);
2471 goto failed_mount;
2472 }
2473
2474 brelse(bh);
2475 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2476 offset = do_div(logical_sb_block, blocksize);
2477 bh = sb_bread(sb, logical_sb_block);
2478 if (!bh) {
2479 ext4_msg(sb, KERN_ERR,
2480 "Can't read superblock on 2nd try");
2481 goto failed_mount;
2482 }
2483 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2484 sbi->s_es = es;
2485 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2486 ext4_msg(sb, KERN_ERR,
2487 "Magic mismatch, very weird!");
2488 goto failed_mount;
2489 }
2490 }
2491
2492 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2493 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2494 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2495 has_huge_files);
2496 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2497
2498 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2499 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2500 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2501 } else {
2502 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2503 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2504 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2505 (!is_power_of_2(sbi->s_inode_size)) ||
2506 (sbi->s_inode_size > blocksize)) {
2507 ext4_msg(sb, KERN_ERR,
2508 "unsupported inode size: %d",
2509 sbi->s_inode_size);
2510 goto failed_mount;
2511 }
2512 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2513 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2514 }
2515
2516 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2517 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2518 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2519 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2520 !is_power_of_2(sbi->s_desc_size)) {
2521 ext4_msg(sb, KERN_ERR,
2522 "unsupported descriptor size %lu",
2523 sbi->s_desc_size);
2524 goto failed_mount;
2525 }
2526 } else
2527 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2528
2529 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2530 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2531 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2532 goto cantfind_ext4;
2533
2534 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2535 if (sbi->s_inodes_per_block == 0)
2536 goto cantfind_ext4;
2537 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2538 sbi->s_inodes_per_block;
2539 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2540 sbi->s_sbh = bh;
2541 sbi->s_mount_state = le16_to_cpu(es->s_state);
2542 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2543 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2544
2545 for (i = 0; i < 4; i++)
2546 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2547 sbi->s_def_hash_version = es->s_def_hash_version;
2548 i = le32_to_cpu(es->s_flags);
2549 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2550 sbi->s_hash_unsigned = 3;
2551 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2552 #ifdef __CHAR_UNSIGNED__
2553 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2554 sbi->s_hash_unsigned = 3;
2555 #else
2556 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2557 #endif
2558 sb->s_dirt = 1;
2559 }
2560
2561 if (sbi->s_blocks_per_group > blocksize * 8) {
2562 ext4_msg(sb, KERN_ERR,
2563 "#blocks per group too big: %lu",
2564 sbi->s_blocks_per_group);
2565 goto failed_mount;
2566 }
2567 if (sbi->s_inodes_per_group > blocksize * 8) {
2568 ext4_msg(sb, KERN_ERR,
2569 "#inodes per group too big: %lu",
2570 sbi->s_inodes_per_group);
2571 goto failed_mount;
2572 }
2573
2574 /*
2575 * Test whether we have more sectors than will fit in sector_t,
2576 * and whether the max offset is addressable by the page cache.
2577 */
2578 if ((ext4_blocks_count(es) >
2579 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2580 (ext4_blocks_count(es) >
2581 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
2582 ext4_msg(sb, KERN_ERR, "filesystem"
2583 " too large to mount safely on this system");
2584 if (sizeof(sector_t) < 8)
2585 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
2586 ret = -EFBIG;
2587 goto failed_mount;
2588 }
2589
2590 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2591 goto cantfind_ext4;
2592
2593 /* check blocks count against device size */
2594 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2595 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2596 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2597 "exceeds size of device (%llu blocks)",
2598 ext4_blocks_count(es), blocks_count);
2599 goto failed_mount;
2600 }
2601
2602 /*
2603 * It makes no sense for the first data block to be beyond the end
2604 * of the filesystem.
2605 */
2606 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2607 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2608 "block %u is beyond end of filesystem (%llu)",
2609 le32_to_cpu(es->s_first_data_block),
2610 ext4_blocks_count(es));
2611 goto failed_mount;
2612 }
2613 blocks_count = (ext4_blocks_count(es) -
2614 le32_to_cpu(es->s_first_data_block) +
2615 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2616 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2617 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2618 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
2619 "(block count %llu, first data block %u, "
2620 "blocks per group %lu)", sbi->s_groups_count,
2621 ext4_blocks_count(es),
2622 le32_to_cpu(es->s_first_data_block),
2623 EXT4_BLOCKS_PER_GROUP(sb));
2624 goto failed_mount;
2625 }
2626 sbi->s_groups_count = blocks_count;
2627 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
2628 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
2629 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2630 EXT4_DESC_PER_BLOCK(sb);
2631 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2632 GFP_KERNEL);
2633 if (sbi->s_group_desc == NULL) {
2634 ext4_msg(sb, KERN_ERR, "not enough memory");
2635 goto failed_mount;
2636 }
2637
2638 #ifdef CONFIG_PROC_FS
2639 if (ext4_proc_root)
2640 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2641 #endif
2642
2643 bgl_lock_init(sbi->s_blockgroup_lock);
2644
2645 for (i = 0; i < db_count; i++) {
2646 block = descriptor_loc(sb, logical_sb_block, i);
2647 sbi->s_group_desc[i] = sb_bread(sb, block);
2648 if (!sbi->s_group_desc[i]) {
2649 ext4_msg(sb, KERN_ERR,
2650 "can't read group descriptor %d", i);
2651 db_count = i;
2652 goto failed_mount2;
2653 }
2654 }
2655 if (!ext4_check_descriptors(sb)) {
2656 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
2657 goto failed_mount2;
2658 }
2659 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2660 if (!ext4_fill_flex_info(sb)) {
2661 ext4_msg(sb, KERN_ERR,
2662 "unable to initialize "
2663 "flex_bg meta info!");
2664 goto failed_mount2;
2665 }
2666
2667 sbi->s_gdb_count = db_count;
2668 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2669 spin_lock_init(&sbi->s_next_gen_lock);
2670
2671 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2672 ext4_count_free_blocks(sb));
2673 if (!err) {
2674 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2675 ext4_count_free_inodes(sb));
2676 }
2677 if (!err) {
2678 err = percpu_counter_init(&sbi->s_dirs_counter,
2679 ext4_count_dirs(sb));
2680 }
2681 if (!err) {
2682 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2683 }
2684 if (err) {
2685 ext4_msg(sb, KERN_ERR, "insufficient memory");
2686 goto failed_mount3;
2687 }
2688
2689 sbi->s_stripe = ext4_get_stripe_size(sbi);
2690 sbi->s_max_writeback_mb_bump = 128;
2691
2692 /*
2693 * set up enough so that it can read an inode
2694 */
2695 if (!test_opt(sb, NOLOAD) &&
2696 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2697 sb->s_op = &ext4_sops;
2698 else
2699 sb->s_op = &ext4_nojournal_sops;
2700 sb->s_export_op = &ext4_export_ops;
2701 sb->s_xattr = ext4_xattr_handlers;
2702 #ifdef CONFIG_QUOTA
2703 sb->s_qcop = &ext4_qctl_operations;
2704 sb->dq_op = &ext4_quota_operations;
2705 #endif
2706 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2707 mutex_init(&sbi->s_orphan_lock);
2708 mutex_init(&sbi->s_resize_lock);
2709
2710 sb->s_root = NULL;
2711
2712 needs_recovery = (es->s_last_orphan != 0 ||
2713 EXT4_HAS_INCOMPAT_FEATURE(sb,
2714 EXT4_FEATURE_INCOMPAT_RECOVER));
2715
2716 /*
2717 * The first inode we look at is the journal inode. Don't try
2718 * root first: it may be modified in the journal!
2719 */
2720 if (!test_opt(sb, NOLOAD) &&
2721 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2722 if (ext4_load_journal(sb, es, journal_devnum))
2723 goto failed_mount3;
2724 if (!(sb->s_flags & MS_RDONLY) &&
2725 EXT4_SB(sb)->s_journal->j_failed_commit) {
2726 ext4_msg(sb, KERN_CRIT, "error: "
2727 "ext4_fill_super: Journal transaction "
2728 "%u is corrupt",
2729 EXT4_SB(sb)->s_journal->j_failed_commit);
2730 if (test_opt(sb, ERRORS_RO)) {
2731 ext4_msg(sb, KERN_CRIT,
2732 "Mounting filesystem read-only");
2733 sb->s_flags |= MS_RDONLY;
2734 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2735 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2736 }
2737 if (test_opt(sb, ERRORS_PANIC)) {
2738 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2739 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2740 ext4_commit_super(sb, 1);
2741 goto failed_mount4;
2742 }
2743 }
2744 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2745 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2746 ext4_msg(sb, KERN_ERR, "required journal recovery "
2747 "suppressed and not mounted read-only");
2748 goto failed_mount4;
2749 } else {
2750 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2751 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2752 sbi->s_journal = NULL;
2753 needs_recovery = 0;
2754 goto no_journal;
2755 }
2756
2757 if (ext4_blocks_count(es) > 0xffffffffULL &&
2758 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2759 JBD2_FEATURE_INCOMPAT_64BIT)) {
2760 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
2761 goto failed_mount4;
2762 }
2763
2764 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2765 jbd2_journal_set_features(sbi->s_journal,
2766 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2767 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2768 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2769 jbd2_journal_set_features(sbi->s_journal,
2770 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2771 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2772 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2773 } else {
2774 jbd2_journal_clear_features(sbi->s_journal,
2775 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2776 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2777 }
2778
2779 /* We have now updated the journal if required, so we can
2780 * validate the data journaling mode. */
2781 switch (test_opt(sb, DATA_FLAGS)) {
2782 case 0:
2783 /* No mode set, assume a default based on the journal
2784 * capabilities: ORDERED_DATA if the journal can
2785 * cope, else JOURNAL_DATA
2786 */
2787 if (jbd2_journal_check_available_features
2788 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2789 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2790 else
2791 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2792 break;
2793
2794 case EXT4_MOUNT_ORDERED_DATA:
2795 case EXT4_MOUNT_WRITEBACK_DATA:
2796 if (!jbd2_journal_check_available_features
2797 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2798 ext4_msg(sb, KERN_ERR, "Journal does not support "
2799 "requested data journaling mode");
2800 goto failed_mount4;
2801 }
2802 default:
2803 break;
2804 }
2805 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2806
2807 no_journal:
2808
2809 if (test_opt(sb, NOBH)) {
2810 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2811 ext4_msg(sb, KERN_WARNING, "Ignoring nobh option - "
2812 "its supported only with writeback mode");
2813 clear_opt(sbi->s_mount_opt, NOBH);
2814 }
2815 }
2816 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
2817 if (!EXT4_SB(sb)->dio_unwritten_wq) {
2818 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
2819 goto failed_mount_wq;
2820 }
2821
2822 /*
2823 * The jbd2_journal_load will have done any necessary log recovery,
2824 * so we can safely mount the rest of the filesystem now.
2825 */
2826
2827 root = ext4_iget(sb, EXT4_ROOT_INO);
2828 if (IS_ERR(root)) {
2829 ext4_msg(sb, KERN_ERR, "get root inode failed");
2830 ret = PTR_ERR(root);
2831 goto failed_mount4;
2832 }
2833 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2834 iput(root);
2835 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
2836 goto failed_mount4;
2837 }
2838 sb->s_root = d_alloc_root(root);
2839 if (!sb->s_root) {
2840 ext4_msg(sb, KERN_ERR, "get root dentry failed");
2841 iput(root);
2842 ret = -ENOMEM;
2843 goto failed_mount4;
2844 }
2845
2846 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2847
2848 /* determine the minimum size of new large inodes, if present */
2849 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2850 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2851 EXT4_GOOD_OLD_INODE_SIZE;
2852 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2853 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2854 if (sbi->s_want_extra_isize <
2855 le16_to_cpu(es->s_want_extra_isize))
2856 sbi->s_want_extra_isize =
2857 le16_to_cpu(es->s_want_extra_isize);
2858 if (sbi->s_want_extra_isize <
2859 le16_to_cpu(es->s_min_extra_isize))
2860 sbi->s_want_extra_isize =
2861 le16_to_cpu(es->s_min_extra_isize);
2862 }
2863 }
2864 /* Check if enough inode space is available */
2865 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2866 sbi->s_inode_size) {
2867 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2868 EXT4_GOOD_OLD_INODE_SIZE;
2869 ext4_msg(sb, KERN_INFO, "required extra inode space not"
2870 "available");
2871 }
2872
2873 if (test_opt(sb, DELALLOC) &&
2874 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
2875 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
2876 "requested data journaling mode");
2877 clear_opt(sbi->s_mount_opt, DELALLOC);
2878 }
2879
2880 err = ext4_setup_system_zone(sb);
2881 if (err) {
2882 ext4_msg(sb, KERN_ERR, "failed to initialize system "
2883 "zone (%d)\n", err);
2884 goto failed_mount4;
2885 }
2886
2887 ext4_ext_init(sb);
2888 err = ext4_mb_init(sb, needs_recovery);
2889 if (err) {
2890 ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
2891 err);
2892 goto failed_mount4;
2893 }
2894
2895 sbi->s_kobj.kset = ext4_kset;
2896 init_completion(&sbi->s_kobj_unregister);
2897 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2898 "%s", sb->s_id);
2899 if (err) {
2900 ext4_mb_release(sb);
2901 ext4_ext_release(sb);
2902 goto failed_mount4;
2903 };
2904
2905 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2906 ext4_orphan_cleanup(sb, es);
2907 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2908 if (needs_recovery) {
2909 ext4_msg(sb, KERN_INFO, "recovery complete");
2910 ext4_mark_recovery_complete(sb, es);
2911 }
2912 if (EXT4_SB(sb)->s_journal) {
2913 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2914 descr = " journalled data mode";
2915 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2916 descr = " ordered data mode";
2917 else
2918 descr = " writeback data mode";
2919 } else
2920 descr = "out journal";
2921
2922 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s", descr);
2923
2924 lock_kernel();
2925 return 0;
2926
2927 cantfind_ext4:
2928 if (!silent)
2929 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
2930 goto failed_mount;
2931
2932 failed_mount4:
2933 ext4_msg(sb, KERN_ERR, "mount failed");
2934 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
2935 failed_mount_wq:
2936 ext4_release_system_zone(sb);
2937 if (sbi->s_journal) {
2938 jbd2_journal_destroy(sbi->s_journal);
2939 sbi->s_journal = NULL;
2940 }
2941 failed_mount3:
2942 if (sbi->s_flex_groups) {
2943 if (is_vmalloc_addr(sbi->s_flex_groups))
2944 vfree(sbi->s_flex_groups);
2945 else
2946 kfree(sbi->s_flex_groups);
2947 }
2948 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2949 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2950 percpu_counter_destroy(&sbi->s_dirs_counter);
2951 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2952 failed_mount2:
2953 for (i = 0; i < db_count; i++)
2954 brelse(sbi->s_group_desc[i]);
2955 kfree(sbi->s_group_desc);
2956 failed_mount:
2957 if (sbi->s_proc) {
2958 remove_proc_entry(sb->s_id, ext4_proc_root);
2959 }
2960 #ifdef CONFIG_QUOTA
2961 for (i = 0; i < MAXQUOTAS; i++)
2962 kfree(sbi->s_qf_names[i]);
2963 #endif
2964 ext4_blkdev_remove(sbi);
2965 brelse(bh);
2966 out_fail:
2967 sb->s_fs_info = NULL;
2968 kfree(sbi->s_blockgroup_lock);
2969 kfree(sbi);
2970 lock_kernel();
2971 return ret;
2972 }
2973
2974 /*
2975 * Setup any per-fs journal parameters now. We'll do this both on
2976 * initial mount, once the journal has been initialised but before we've
2977 * done any recovery; and again on any subsequent remount.
2978 */
2979 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2980 {
2981 struct ext4_sb_info *sbi = EXT4_SB(sb);
2982
2983 journal->j_commit_interval = sbi->s_commit_interval;
2984 journal->j_min_batch_time = sbi->s_min_batch_time;
2985 journal->j_max_batch_time = sbi->s_max_batch_time;
2986
2987 spin_lock(&journal->j_state_lock);
2988 if (test_opt(sb, BARRIER))
2989 journal->j_flags |= JBD2_BARRIER;
2990 else
2991 journal->j_flags &= ~JBD2_BARRIER;
2992 if (test_opt(sb, DATA_ERR_ABORT))
2993 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2994 else
2995 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
2996 spin_unlock(&journal->j_state_lock);
2997 }
2998
2999 static journal_t *ext4_get_journal(struct super_block *sb,
3000 unsigned int journal_inum)
3001 {
3002 struct inode *journal_inode;
3003 journal_t *journal;
3004
3005 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3006
3007 /* First, test for the existence of a valid inode on disk. Bad
3008 * things happen if we iget() an unused inode, as the subsequent
3009 * iput() will try to delete it. */
3010
3011 journal_inode = ext4_iget(sb, journal_inum);
3012 if (IS_ERR(journal_inode)) {
3013 ext4_msg(sb, KERN_ERR, "no journal found");
3014 return NULL;
3015 }
3016 if (!journal_inode->i_nlink) {
3017 make_bad_inode(journal_inode);
3018 iput(journal_inode);
3019 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3020 return NULL;
3021 }
3022
3023 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3024 journal_inode, journal_inode->i_size);
3025 if (!S_ISREG(journal_inode->i_mode)) {
3026 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3027 iput(journal_inode);
3028 return NULL;
3029 }
3030
3031 journal = jbd2_journal_init_inode(journal_inode);
3032 if (!journal) {
3033 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3034 iput(journal_inode);
3035 return NULL;
3036 }
3037 journal->j_private = sb;
3038 ext4_init_journal_params(sb, journal);
3039 return journal;
3040 }
3041
3042 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3043 dev_t j_dev)
3044 {
3045 struct buffer_head *bh;
3046 journal_t *journal;
3047 ext4_fsblk_t start;
3048 ext4_fsblk_t len;
3049 int hblock, blocksize;
3050 ext4_fsblk_t sb_block;
3051 unsigned long offset;
3052 struct ext4_super_block *es;
3053 struct block_device *bdev;
3054
3055 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3056
3057 bdev = ext4_blkdev_get(j_dev, sb);
3058 if (bdev == NULL)
3059 return NULL;
3060
3061 if (bd_claim(bdev, sb)) {
3062 ext4_msg(sb, KERN_ERR,
3063 "failed to claim external journal device");
3064 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3065 return NULL;
3066 }
3067
3068 blocksize = sb->s_blocksize;
3069 hblock = bdev_logical_block_size(bdev);
3070 if (blocksize < hblock) {
3071 ext4_msg(sb, KERN_ERR,
3072 "blocksize too small for journal device");
3073 goto out_bdev;
3074 }
3075
3076 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3077 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3078 set_blocksize(bdev, blocksize);
3079 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3080 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3081 "external journal");
3082 goto out_bdev;
3083 }
3084
3085 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3086 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3087 !(le32_to_cpu(es->s_feature_incompat) &
3088 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3089 ext4_msg(sb, KERN_ERR, "external journal has "
3090 "bad superblock");
3091 brelse(bh);
3092 goto out_bdev;
3093 }
3094
3095 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3096 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3097 brelse(bh);
3098 goto out_bdev;
3099 }
3100
3101 len = ext4_blocks_count(es);
3102 start = sb_block + 1;
3103 brelse(bh); /* we're done with the superblock */
3104
3105 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3106 start, len, blocksize);
3107 if (!journal) {
3108 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3109 goto out_bdev;
3110 }
3111 journal->j_private = sb;
3112 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3113 wait_on_buffer(journal->j_sb_buffer);
3114 if (!buffer_uptodate(journal->j_sb_buffer)) {
3115 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3116 goto out_journal;
3117 }
3118 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3119 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3120 "user (unsupported) - %d",
3121 be32_to_cpu(journal->j_superblock->s_nr_users));
3122 goto out_journal;
3123 }
3124 EXT4_SB(sb)->journal_bdev = bdev;
3125 ext4_init_journal_params(sb, journal);
3126 return journal;
3127
3128 out_journal:
3129 jbd2_journal_destroy(journal);
3130 out_bdev:
3131 ext4_blkdev_put(bdev);
3132 return NULL;
3133 }
3134
3135 static int ext4_load_journal(struct super_block *sb,
3136 struct ext4_super_block *es,
3137 unsigned long journal_devnum)
3138 {
3139 journal_t *journal;
3140 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3141 dev_t journal_dev;
3142 int err = 0;
3143 int really_read_only;
3144
3145 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3146
3147 if (journal_devnum &&
3148 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3149 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3150 "numbers have changed");
3151 journal_dev = new_decode_dev(journal_devnum);
3152 } else
3153 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3154
3155 really_read_only = bdev_read_only(sb->s_bdev);
3156
3157 /*
3158 * Are we loading a blank journal or performing recovery after a
3159 * crash? For recovery, we need to check in advance whether we
3160 * can get read-write access to the device.
3161 */
3162 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3163 if (sb->s_flags & MS_RDONLY) {
3164 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3165 "required on readonly filesystem");
3166 if (really_read_only) {
3167 ext4_msg(sb, KERN_ERR, "write access "
3168 "unavailable, cannot proceed");
3169 return -EROFS;
3170 }
3171 ext4_msg(sb, KERN_INFO, "write access will "
3172 "be enabled during recovery");
3173 }
3174 }
3175
3176 if (journal_inum && journal_dev) {
3177 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3178 "and inode journals!");
3179 return -EINVAL;
3180 }
3181
3182 if (journal_inum) {
3183 if (!(journal = ext4_get_journal(sb, journal_inum)))
3184 return -EINVAL;
3185 } else {
3186 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3187 return -EINVAL;
3188 }
3189
3190 if (!(journal->j_flags & JBD2_BARRIER))
3191 ext4_msg(sb, KERN_INFO, "barriers disabled");
3192
3193 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3194 err = jbd2_journal_update_format(journal);
3195 if (err) {
3196 ext4_msg(sb, KERN_ERR, "error updating journal");
3197 jbd2_journal_destroy(journal);
3198 return err;
3199 }
3200 }
3201
3202 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3203 err = jbd2_journal_wipe(journal, !really_read_only);
3204 if (!err)
3205 err = jbd2_journal_load(journal);
3206
3207 if (err) {
3208 ext4_msg(sb, KERN_ERR, "error loading journal");
3209 jbd2_journal_destroy(journal);
3210 return err;
3211 }
3212
3213 EXT4_SB(sb)->s_journal = journal;
3214 ext4_clear_journal_err(sb, es);
3215
3216 if (journal_devnum &&
3217 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3218 es->s_journal_dev = cpu_to_le32(journal_devnum);
3219
3220 /* Make sure we flush the recovery flag to disk. */
3221 ext4_commit_super(sb, 1);
3222 }
3223
3224 return 0;
3225 }
3226
3227 static int ext4_commit_super(struct super_block *sb, int sync)
3228 {
3229 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3230 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3231 int error = 0;
3232
3233 if (!sbh)
3234 return error;
3235 if (buffer_write_io_error(sbh)) {
3236 /*
3237 * Oh, dear. A previous attempt to write the
3238 * superblock failed. This could happen because the
3239 * USB device was yanked out. Or it could happen to
3240 * be a transient write error and maybe the block will
3241 * be remapped. Nothing we can do but to retry the
3242 * write and hope for the best.
3243 */
3244 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3245 "superblock detected");
3246 clear_buffer_write_io_error(sbh);
3247 set_buffer_uptodate(sbh);
3248 }
3249 /*
3250 * If the file system is mounted read-only, don't update the
3251 * superblock write time. This avoids updating the superblock
3252 * write time when we are mounting the root file system
3253 * read/only but we need to replay the journal; at that point,
3254 * for people who are east of GMT and who make their clock
3255 * tick in localtime for Windows bug-for-bug compatibility,
3256 * the clock is set in the future, and this will cause e2fsck
3257 * to complain and force a full file system check.
3258 */
3259 if (!(sb->s_flags & MS_RDONLY))
3260 es->s_wtime = cpu_to_le32(get_seconds());
3261 es->s_kbytes_written =
3262 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3263 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3264 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3265 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3266 &EXT4_SB(sb)->s_freeblocks_counter));
3267 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3268 &EXT4_SB(sb)->s_freeinodes_counter));
3269 sb->s_dirt = 0;
3270 BUFFER_TRACE(sbh, "marking dirty");
3271 mark_buffer_dirty(sbh);
3272 if (sync) {
3273 error = sync_dirty_buffer(sbh);
3274 if (error)
3275 return error;
3276
3277 error = buffer_write_io_error(sbh);
3278 if (error) {
3279 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3280 "superblock");
3281 clear_buffer_write_io_error(sbh);
3282 set_buffer_uptodate(sbh);
3283 }
3284 }
3285 return error;
3286 }
3287
3288 /*
3289 * Have we just finished recovery? If so, and if we are mounting (or
3290 * remounting) the filesystem readonly, then we will end up with a
3291 * consistent fs on disk. Record that fact.
3292 */
3293 static void ext4_mark_recovery_complete(struct super_block *sb,
3294 struct ext4_super_block *es)
3295 {
3296 journal_t *journal = EXT4_SB(sb)->s_journal;
3297
3298 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3299 BUG_ON(journal != NULL);
3300 return;
3301 }
3302 jbd2_journal_lock_updates(journal);
3303 if (jbd2_journal_flush(journal) < 0)
3304 goto out;
3305
3306 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3307 sb->s_flags & MS_RDONLY) {
3308 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3309 ext4_commit_super(sb, 1);
3310 }
3311
3312 out:
3313 jbd2_journal_unlock_updates(journal);
3314 }
3315
3316 /*
3317 * If we are mounting (or read-write remounting) a filesystem whose journal
3318 * has recorded an error from a previous lifetime, move that error to the
3319 * main filesystem now.
3320 */
3321 static void ext4_clear_journal_err(struct super_block *sb,
3322 struct ext4_super_block *es)
3323 {
3324 journal_t *journal;
3325 int j_errno;
3326 const char *errstr;
3327
3328 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3329
3330 journal = EXT4_SB(sb)->s_journal;
3331
3332 /*
3333 * Now check for any error status which may have been recorded in the
3334 * journal by a prior ext4_error() or ext4_abort()
3335 */
3336
3337 j_errno = jbd2_journal_errno(journal);
3338 if (j_errno) {
3339 char nbuf[16];
3340
3341 errstr = ext4_decode_error(sb, j_errno, nbuf);
3342 ext4_warning(sb, __func__, "Filesystem error recorded "
3343 "from previous mount: %s", errstr);
3344 ext4_warning(sb, __func__, "Marking fs in need of "
3345 "filesystem check.");
3346
3347 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3348 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3349 ext4_commit_super(sb, 1);
3350
3351 jbd2_journal_clear_err(journal);
3352 }
3353 }
3354
3355 /*
3356 * Force the running and committing transactions to commit,
3357 * and wait on the commit.
3358 */
3359 int ext4_force_commit(struct super_block *sb)
3360 {
3361 journal_t *journal;
3362 int ret = 0;
3363
3364 if (sb->s_flags & MS_RDONLY)
3365 return 0;
3366
3367 journal = EXT4_SB(sb)->s_journal;
3368 if (journal)
3369 ret = ext4_journal_force_commit(journal);
3370
3371 return ret;
3372 }
3373
3374 static void ext4_write_super(struct super_block *sb)
3375 {
3376 lock_super(sb);
3377 ext4_commit_super(sb, 1);
3378 unlock_super(sb);
3379 }
3380
3381 static int ext4_sync_fs(struct super_block *sb, int wait)
3382 {
3383 int ret = 0;
3384 tid_t target;
3385 struct ext4_sb_info *sbi = EXT4_SB(sb);
3386
3387 trace_ext4_sync_fs(sb, wait);
3388 flush_workqueue(sbi->dio_unwritten_wq);
3389 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
3390 if (wait)
3391 jbd2_log_wait_commit(sbi->s_journal, target);
3392 }
3393 return ret;
3394 }
3395
3396 /*
3397 * LVM calls this function before a (read-only) snapshot is created. This
3398 * gives us a chance to flush the journal completely and mark the fs clean.
3399 */
3400 static int ext4_freeze(struct super_block *sb)
3401 {
3402 int error = 0;
3403 journal_t *journal;
3404
3405 if (sb->s_flags & MS_RDONLY)
3406 return 0;
3407
3408 journal = EXT4_SB(sb)->s_journal;
3409
3410 /* Now we set up the journal barrier. */
3411 jbd2_journal_lock_updates(journal);
3412
3413 /*
3414 * Don't clear the needs_recovery flag if we failed to flush
3415 * the journal.
3416 */
3417 error = jbd2_journal_flush(journal);
3418 if (error < 0) {
3419 out:
3420 jbd2_journal_unlock_updates(journal);
3421 return error;
3422 }
3423
3424 /* Journal blocked and flushed, clear needs_recovery flag. */
3425 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3426 error = ext4_commit_super(sb, 1);
3427 if (error)
3428 goto out;
3429 return 0;
3430 }
3431
3432 /*
3433 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3434 * flag here, even though the filesystem is not technically dirty yet.
3435 */
3436 static int ext4_unfreeze(struct super_block *sb)
3437 {
3438 if (sb->s_flags & MS_RDONLY)
3439 return 0;
3440
3441 lock_super(sb);
3442 /* Reset the needs_recovery flag before the fs is unlocked. */
3443 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3444 ext4_commit_super(sb, 1);
3445 unlock_super(sb);
3446 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3447 return 0;
3448 }
3449
3450 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3451 {
3452 struct ext4_super_block *es;
3453 struct ext4_sb_info *sbi = EXT4_SB(sb);
3454 ext4_fsblk_t n_blocks_count = 0;
3455 unsigned long old_sb_flags;
3456 struct ext4_mount_options old_opts;
3457 ext4_group_t g;
3458 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3459 int err;
3460 #ifdef CONFIG_QUOTA
3461 int i;
3462 #endif
3463
3464 lock_kernel();
3465
3466 /* Store the original options */
3467 lock_super(sb);
3468 old_sb_flags = sb->s_flags;
3469 old_opts.s_mount_opt = sbi->s_mount_opt;
3470 old_opts.s_resuid = sbi->s_resuid;
3471 old_opts.s_resgid = sbi->s_resgid;
3472 old_opts.s_commit_interval = sbi->s_commit_interval;
3473 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3474 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3475 #ifdef CONFIG_QUOTA
3476 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3477 for (i = 0; i < MAXQUOTAS; i++)
3478 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3479 #endif
3480 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3481 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3482
3483 /*
3484 * Allow the "check" option to be passed as a remount option.
3485 */
3486 if (!parse_options(data, sb, NULL, &journal_ioprio,
3487 &n_blocks_count, 1)) {
3488 err = -EINVAL;
3489 goto restore_opts;
3490 }
3491
3492 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
3493 ext4_abort(sb, __func__, "Abort forced by user");
3494
3495 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3496 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3497
3498 es = sbi->s_es;
3499
3500 if (sbi->s_journal) {
3501 ext4_init_journal_params(sb, sbi->s_journal);
3502 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3503 }
3504
3505 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3506 n_blocks_count > ext4_blocks_count(es)) {
3507 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
3508 err = -EROFS;
3509 goto restore_opts;
3510 }
3511
3512 if (*flags & MS_RDONLY) {
3513 /*
3514 * First of all, the unconditional stuff we have to do
3515 * to disable replay of the journal when we next remount
3516 */
3517 sb->s_flags |= MS_RDONLY;
3518
3519 /*
3520 * OK, test if we are remounting a valid rw partition
3521 * readonly, and if so set the rdonly flag and then
3522 * mark the partition as valid again.
3523 */
3524 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3525 (sbi->s_mount_state & EXT4_VALID_FS))
3526 es->s_state = cpu_to_le16(sbi->s_mount_state);
3527
3528 if (sbi->s_journal)
3529 ext4_mark_recovery_complete(sb, es);
3530 } else {
3531 /* Make sure we can mount this feature set readwrite */
3532 if (!ext4_feature_set_ok(sb, 0)) {
3533 err = -EROFS;
3534 goto restore_opts;
3535 }
3536 /*
3537 * Make sure the group descriptor checksums
3538 * are sane. If they aren't, refuse to remount r/w.
3539 */
3540 for (g = 0; g < sbi->s_groups_count; g++) {
3541 struct ext4_group_desc *gdp =
3542 ext4_get_group_desc(sb, g, NULL);
3543
3544 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3545 ext4_msg(sb, KERN_ERR,
3546 "ext4_remount: Checksum for group %u failed (%u!=%u)",
3547 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3548 le16_to_cpu(gdp->bg_checksum));
3549 err = -EINVAL;
3550 goto restore_opts;
3551 }
3552 }
3553
3554 /*
3555 * If we have an unprocessed orphan list hanging
3556 * around from a previously readonly bdev mount,
3557 * require a full umount/remount for now.
3558 */
3559 if (es->s_last_orphan) {
3560 ext4_msg(sb, KERN_WARNING, "Couldn't "
3561 "remount RDWR because of unprocessed "
3562 "orphan inode list. Please "
3563 "umount/remount instead");
3564 err = -EINVAL;
3565 goto restore_opts;
3566 }
3567
3568 /*
3569 * Mounting a RDONLY partition read-write, so reread
3570 * and store the current valid flag. (It may have
3571 * been changed by e2fsck since we originally mounted
3572 * the partition.)
3573 */
3574 if (sbi->s_journal)
3575 ext4_clear_journal_err(sb, es);
3576 sbi->s_mount_state = le16_to_cpu(es->s_state);
3577 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3578 goto restore_opts;
3579 if (!ext4_setup_super(sb, es, 0))
3580 sb->s_flags &= ~MS_RDONLY;
3581 }
3582 }
3583 ext4_setup_system_zone(sb);
3584 if (sbi->s_journal == NULL)
3585 ext4_commit_super(sb, 1);
3586
3587 #ifdef CONFIG_QUOTA
3588 /* Release old quota file names */
3589 for (i = 0; i < MAXQUOTAS; i++)
3590 if (old_opts.s_qf_names[i] &&
3591 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3592 kfree(old_opts.s_qf_names[i]);
3593 #endif
3594 unlock_super(sb);
3595 unlock_kernel();
3596 return 0;
3597
3598 restore_opts:
3599 sb->s_flags = old_sb_flags;
3600 sbi->s_mount_opt = old_opts.s_mount_opt;
3601 sbi->s_resuid = old_opts.s_resuid;
3602 sbi->s_resgid = old_opts.s_resgid;
3603 sbi->s_commit_interval = old_opts.s_commit_interval;
3604 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3605 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3606 #ifdef CONFIG_QUOTA
3607 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3608 for (i = 0; i < MAXQUOTAS; i++) {
3609 if (sbi->s_qf_names[i] &&
3610 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3611 kfree(sbi->s_qf_names[i]);
3612 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3613 }
3614 #endif
3615 unlock_super(sb);
3616 unlock_kernel();
3617 return err;
3618 }
3619
3620 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3621 {
3622 struct super_block *sb = dentry->d_sb;
3623 struct ext4_sb_info *sbi = EXT4_SB(sb);
3624 struct ext4_super_block *es = sbi->s_es;
3625 u64 fsid;
3626
3627 if (test_opt(sb, MINIX_DF)) {
3628 sbi->s_overhead_last = 0;
3629 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3630 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3631 ext4_fsblk_t overhead = 0;
3632
3633 /*
3634 * Compute the overhead (FS structures). This is constant
3635 * for a given filesystem unless the number of block groups
3636 * changes so we cache the previous value until it does.
3637 */
3638
3639 /*
3640 * All of the blocks before first_data_block are
3641 * overhead
3642 */
3643 overhead = le32_to_cpu(es->s_first_data_block);
3644
3645 /*
3646 * Add the overhead attributed to the superblock and
3647 * block group descriptors. If the sparse superblocks
3648 * feature is turned on, then not all groups have this.
3649 */
3650 for (i = 0; i < ngroups; i++) {
3651 overhead += ext4_bg_has_super(sb, i) +
3652 ext4_bg_num_gdb(sb, i);
3653 cond_resched();
3654 }
3655
3656 /*
3657 * Every block group has an inode bitmap, a block
3658 * bitmap, and an inode table.
3659 */
3660 overhead += ngroups * (2 + sbi->s_itb_per_group);
3661 sbi->s_overhead_last = overhead;
3662 smp_wmb();
3663 sbi->s_blocks_last = ext4_blocks_count(es);
3664 }
3665
3666 buf->f_type = EXT4_SUPER_MAGIC;
3667 buf->f_bsize = sb->s_blocksize;
3668 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3669 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3670 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3671 ext4_free_blocks_count_set(es, buf->f_bfree);
3672 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3673 if (buf->f_bfree < ext4_r_blocks_count(es))
3674 buf->f_bavail = 0;
3675 buf->f_files = le32_to_cpu(es->s_inodes_count);
3676 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3677 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
3678 buf->f_namelen = EXT4_NAME_LEN;
3679 fsid = le64_to_cpup((void *)es->s_uuid) ^
3680 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3681 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3682 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3683
3684 return 0;
3685 }
3686
3687 /* Helper function for writing quotas on sync - we need to start transaction
3688 * before quota file is locked for write. Otherwise the are possible deadlocks:
3689 * Process 1 Process 2
3690 * ext4_create() quota_sync()
3691 * jbd2_journal_start() write_dquot()
3692 * vfs_dq_init() down(dqio_mutex)
3693 * down(dqio_mutex) jbd2_journal_start()
3694 *
3695 */
3696
3697 #ifdef CONFIG_QUOTA
3698
3699 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3700 {
3701 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3702 }
3703
3704 static int ext4_write_dquot(struct dquot *dquot)
3705 {
3706 int ret, err;
3707 handle_t *handle;
3708 struct inode *inode;
3709
3710 inode = dquot_to_inode(dquot);
3711 handle = ext4_journal_start(inode,
3712 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3713 if (IS_ERR(handle))
3714 return PTR_ERR(handle);
3715 ret = dquot_commit(dquot);
3716 err = ext4_journal_stop(handle);
3717 if (!ret)
3718 ret = err;
3719 return ret;
3720 }
3721
3722 static int ext4_acquire_dquot(struct dquot *dquot)
3723 {
3724 int ret, err;
3725 handle_t *handle;
3726
3727 handle = ext4_journal_start(dquot_to_inode(dquot),
3728 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3729 if (IS_ERR(handle))
3730 return PTR_ERR(handle);
3731 ret = dquot_acquire(dquot);
3732 err = ext4_journal_stop(handle);
3733 if (!ret)
3734 ret = err;
3735 return ret;
3736 }
3737
3738 static int ext4_release_dquot(struct dquot *dquot)
3739 {
3740 int ret, err;
3741 handle_t *handle;
3742
3743 handle = ext4_journal_start(dquot_to_inode(dquot),
3744 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3745 if (IS_ERR(handle)) {
3746 /* Release dquot anyway to avoid endless cycle in dqput() */
3747 dquot_release(dquot);
3748 return PTR_ERR(handle);
3749 }
3750 ret = dquot_release(dquot);
3751 err = ext4_journal_stop(handle);
3752 if (!ret)
3753 ret = err;
3754 return ret;
3755 }
3756
3757 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3758 {
3759 /* Are we journaling quotas? */
3760 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3761 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3762 dquot_mark_dquot_dirty(dquot);
3763 return ext4_write_dquot(dquot);
3764 } else {
3765 return dquot_mark_dquot_dirty(dquot);
3766 }
3767 }
3768
3769 static int ext4_write_info(struct super_block *sb, int type)
3770 {
3771 int ret, err;
3772 handle_t *handle;
3773
3774 /* Data block + inode block */
3775 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3776 if (IS_ERR(handle))
3777 return PTR_ERR(handle);
3778 ret = dquot_commit_info(sb, type);
3779 err = ext4_journal_stop(handle);
3780 if (!ret)
3781 ret = err;
3782 return ret;
3783 }
3784
3785 /*
3786 * Turn on quotas during mount time - we need to find
3787 * the quota file and such...
3788 */
3789 static int ext4_quota_on_mount(struct super_block *sb, int type)
3790 {
3791 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3792 EXT4_SB(sb)->s_jquota_fmt, type);
3793 }
3794
3795 /*
3796 * Standard function to be called on quota_on
3797 */
3798 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3799 char *name, int remount)
3800 {
3801 int err;
3802 struct path path;
3803
3804 if (!test_opt(sb, QUOTA))
3805 return -EINVAL;
3806 /* When remounting, no checks are needed and in fact, name is NULL */
3807 if (remount)
3808 return vfs_quota_on(sb, type, format_id, name, remount);
3809
3810 err = kern_path(name, LOOKUP_FOLLOW, &path);
3811 if (err)
3812 return err;
3813
3814 /* Quotafile not on the same filesystem? */
3815 if (path.mnt->mnt_sb != sb) {
3816 path_put(&path);
3817 return -EXDEV;
3818 }
3819 /* Journaling quota? */
3820 if (EXT4_SB(sb)->s_qf_names[type]) {
3821 /* Quotafile not in fs root? */
3822 if (path.dentry->d_parent != sb->s_root)
3823 ext4_msg(sb, KERN_WARNING,
3824 "Quota file not on filesystem root. "
3825 "Journaled quota will not work");
3826 }
3827
3828 /*
3829 * When we journal data on quota file, we have to flush journal to see
3830 * all updates to the file when we bypass pagecache...
3831 */
3832 if (EXT4_SB(sb)->s_journal &&
3833 ext4_should_journal_data(path.dentry->d_inode)) {
3834 /*
3835 * We don't need to lock updates but journal_flush() could
3836 * otherwise be livelocked...
3837 */
3838 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3839 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3840 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3841 if (err) {
3842 path_put(&path);
3843 return err;
3844 }
3845 }
3846
3847 err = vfs_quota_on_path(sb, type, format_id, &path);
3848 path_put(&path);
3849 return err;
3850 }
3851
3852 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3853 * acquiring the locks... As quota files are never truncated and quota code
3854 * itself serializes the operations (and noone else should touch the files)
3855 * we don't have to be afraid of races */
3856 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3857 size_t len, loff_t off)
3858 {
3859 struct inode *inode = sb_dqopt(sb)->files[type];
3860 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3861 int err = 0;
3862 int offset = off & (sb->s_blocksize - 1);
3863 int tocopy;
3864 size_t toread;
3865 struct buffer_head *bh;
3866 loff_t i_size = i_size_read(inode);
3867
3868 if (off > i_size)
3869 return 0;
3870 if (off+len > i_size)
3871 len = i_size-off;
3872 toread = len;
3873 while (toread > 0) {
3874 tocopy = sb->s_blocksize - offset < toread ?
3875 sb->s_blocksize - offset : toread;
3876 bh = ext4_bread(NULL, inode, blk, 0, &err);
3877 if (err)
3878 return err;
3879 if (!bh) /* A hole? */
3880 memset(data, 0, tocopy);
3881 else
3882 memcpy(data, bh->b_data+offset, tocopy);
3883 brelse(bh);
3884 offset = 0;
3885 toread -= tocopy;
3886 data += tocopy;
3887 blk++;
3888 }
3889 return len;
3890 }
3891
3892 /* Write to quotafile (we know the transaction is already started and has
3893 * enough credits) */
3894 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3895 const char *data, size_t len, loff_t off)
3896 {
3897 struct inode *inode = sb_dqopt(sb)->files[type];
3898 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3899 int err = 0;
3900 int offset = off & (sb->s_blocksize - 1);
3901 int tocopy;
3902 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3903 size_t towrite = len;
3904 struct buffer_head *bh;
3905 handle_t *handle = journal_current_handle();
3906
3907 if (EXT4_SB(sb)->s_journal && !handle) {
3908 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
3909 " cancelled because transaction is not started",
3910 (unsigned long long)off, (unsigned long long)len);
3911 return -EIO;
3912 }
3913 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3914 while (towrite > 0) {
3915 tocopy = sb->s_blocksize - offset < towrite ?
3916 sb->s_blocksize - offset : towrite;
3917 bh = ext4_bread(handle, inode, blk, 1, &err);
3918 if (!bh)
3919 goto out;
3920 if (journal_quota) {
3921 err = ext4_journal_get_write_access(handle, bh);
3922 if (err) {
3923 brelse(bh);
3924 goto out;
3925 }
3926 }
3927 lock_buffer(bh);
3928 memcpy(bh->b_data+offset, data, tocopy);
3929 flush_dcache_page(bh->b_page);
3930 unlock_buffer(bh);
3931 if (journal_quota)
3932 err = ext4_handle_dirty_metadata(handle, NULL, bh);
3933 else {
3934 /* Always do at least ordered writes for quotas */
3935 err = ext4_jbd2_file_inode(handle, inode);
3936 mark_buffer_dirty(bh);
3937 }
3938 brelse(bh);
3939 if (err)
3940 goto out;
3941 offset = 0;
3942 towrite -= tocopy;
3943 data += tocopy;
3944 blk++;
3945 }
3946 out:
3947 if (len == towrite) {
3948 mutex_unlock(&inode->i_mutex);
3949 return err;
3950 }
3951 if (inode->i_size < off+len-towrite) {
3952 i_size_write(inode, off+len-towrite);
3953 EXT4_I(inode)->i_disksize = inode->i_size;
3954 }
3955 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3956 ext4_mark_inode_dirty(handle, inode);
3957 mutex_unlock(&inode->i_mutex);
3958 return len - towrite;
3959 }
3960
3961 #endif
3962
3963 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
3964 const char *dev_name, void *data, struct vfsmount *mnt)
3965 {
3966 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
3967 }
3968
3969 static struct file_system_type ext4_fs_type = {
3970 .owner = THIS_MODULE,
3971 .name = "ext4",
3972 .get_sb = ext4_get_sb,
3973 .kill_sb = kill_block_super,
3974 .fs_flags = FS_REQUIRES_DEV,
3975 };
3976
3977 static int __init init_ext4_fs(void)
3978 {
3979 int err;
3980
3981 err = init_ext4_system_zone();
3982 if (err)
3983 return err;
3984 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
3985 if (!ext4_kset)
3986 goto out4;
3987 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
3988 err = init_ext4_mballoc();
3989 if (err)
3990 goto out3;
3991
3992 err = init_ext4_xattr();
3993 if (err)
3994 goto out2;
3995 err = init_inodecache();
3996 if (err)
3997 goto out1;
3998 err = register_filesystem(&ext4_fs_type);
3999 if (err)
4000 goto out;
4001 return 0;
4002 out:
4003 destroy_inodecache();
4004 out1:
4005 exit_ext4_xattr();
4006 out2:
4007 exit_ext4_mballoc();
4008 out3:
4009 remove_proc_entry("fs/ext4", NULL);
4010 kset_unregister(ext4_kset);
4011 out4:
4012 exit_ext4_system_zone();
4013 return err;
4014 }
4015
4016 static void __exit exit_ext4_fs(void)
4017 {
4018 unregister_filesystem(&ext4_fs_type);
4019 destroy_inodecache();
4020 exit_ext4_xattr();
4021 exit_ext4_mballoc();
4022 remove_proc_entry("fs/ext4", NULL);
4023 kset_unregister(ext4_kset);
4024 exit_ext4_system_zone();
4025 }
4026
4027 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4028 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4029 MODULE_LICENSE("GPL");
4030 module_init(init_ext4_fs)
4031 module_exit(exit_ext4_fs)
Support for the Chinese Samsung S3C2440 based development boards