?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
ccwgroup: move attributes to attribute group
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / fs / ext4 / super.c
blobf27e045df7a6e502c9d6d92389a8062eeb8a0e24
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 vfs_check_frozen(sb, SB_FREEZE_TRANS);
231 /* Special case here: if the journal has aborted behind our
232 * backs (eg. EIO in the commit thread), then we still need to
233 * take the FS itself readonly cleanly. */
234 journal = EXT4_SB(sb)->s_journal;
235 if (journal) {
236 if (is_journal_aborted(journal)) {
237 ext4_abort(sb, __func__, "Detected aborted journal");
238 return ERR_PTR(-EROFS);
239 }
240 return jbd2_journal_start(journal, nblocks);
241 }
242 return ext4_get_nojournal();
243 }
244
245 /*
246 * The only special thing we need to do here is to make sure that all
247 * jbd2_journal_stop calls result in the superblock being marked dirty, so
248 * that sync() will call the filesystem's write_super callback if
249 * appropriate.
250 */
251 int __ext4_journal_stop(const char *where, handle_t *handle)
252 {
253 struct super_block *sb;
254 int err;
255 int rc;
256
257 if (!ext4_handle_valid(handle)) {
258 ext4_put_nojournal(handle);
259 return 0;
260 }
261 sb = handle->h_transaction->t_journal->j_private;
262 err = handle->h_err;
263 rc = jbd2_journal_stop(handle);
264
265 if (!err)
266 err = rc;
267 if (err)
268 __ext4_std_error(sb, where, err);
269 return err;
270 }
271
272 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
273 struct buffer_head *bh, handle_t *handle, int err)
274 {
275 char nbuf[16];
276 const char *errstr = ext4_decode_error(NULL, err, nbuf);
277
278 BUG_ON(!ext4_handle_valid(handle));
279
280 if (bh)
281 BUFFER_TRACE(bh, "abort");
282
283 if (!handle->h_err)
284 handle->h_err = err;
285
286 if (is_handle_aborted(handle))
287 return;
288
289 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
290 caller, errstr, err_fn);
291
292 jbd2_journal_abort_handle(handle);
293 }
294
295 /* Deal with the reporting of failure conditions on a filesystem such as
296 * inconsistencies detected or read IO failures.
297 *
298 * On ext2, we can store the error state of the filesystem in the
299 * superblock. That is not possible on ext4, because we may have other
300 * write ordering constraints on the superblock which prevent us from
301 * writing it out straight away; and given that the journal is about to
302 * be aborted, we can't rely on the current, or future, transactions to
303 * write out the superblock safely.
304 *
305 * We'll just use the jbd2_journal_abort() error code to record an error in
306 * the journal instead. On recovery, the journal will compain about
307 * that error until we've noted it down and cleared it.
308 */
309
310 static void ext4_handle_error(struct super_block *sb)
311 {
312 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
313
314 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
315 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
316
317 if (sb->s_flags & MS_RDONLY)
318 return;
319
320 if (!test_opt(sb, ERRORS_CONT)) {
321 journal_t *journal = EXT4_SB(sb)->s_journal;
322
323 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
324 if (journal)
325 jbd2_journal_abort(journal, -EIO);
326 }
327 if (test_opt(sb, ERRORS_RO)) {
328 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
329 sb->s_flags |= MS_RDONLY;
330 }
331 ext4_commit_super(sb, 1);
332 if (test_opt(sb, ERRORS_PANIC))
333 panic("EXT4-fs (device %s): panic forced after error\n",
334 sb->s_id);
335 }
336
337 void ext4_error(struct super_block *sb, const char *function,
338 const char *fmt, ...)
339 {
340 va_list args;
341
342 va_start(args, fmt);
343 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
344 vprintk(fmt, args);
345 printk("\n");
346 va_end(args);
347
348 ext4_handle_error(sb);
349 }
350
351 static const char *ext4_decode_error(struct super_block *sb, int errno,
352 char nbuf[16])
353 {
354 char *errstr = NULL;
355
356 switch (errno) {
357 case -EIO:
358 errstr = "IO failure";
359 break;
360 case -ENOMEM:
361 errstr = "Out of memory";
362 break;
363 case -EROFS:
364 if (!sb || (EXT4_SB(sb)->s_journal &&
365 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
366 errstr = "Journal has aborted";
367 else
368 errstr = "Readonly filesystem";
369 break;
370 default:
371 /* If the caller passed in an extra buffer for unknown
372 * errors, textualise them now. Else we just return
373 * NULL. */
374 if (nbuf) {
375 /* Check for truncated error codes... */
376 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
377 errstr = nbuf;
378 }
379 break;
380 }
381
382 return errstr;
383 }
384
385 /* __ext4_std_error decodes expected errors from journaling functions
386 * automatically and invokes the appropriate error response. */
387
388 void __ext4_std_error(struct super_block *sb, const char *function, int errno)
389 {
390 char nbuf[16];
391 const char *errstr;
392
393 /* Special case: if the error is EROFS, and we're not already
394 * inside a transaction, then there's really no point in logging
395 * an error. */
396 if (errno == -EROFS && journal_current_handle() == NULL &&
397 (sb->s_flags & MS_RDONLY))
398 return;
399
400 errstr = ext4_decode_error(sb, errno, nbuf);
401 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
402 sb->s_id, function, errstr);
403
404 ext4_handle_error(sb);
405 }
406
407 /*
408 * ext4_abort is a much stronger failure handler than ext4_error. The
409 * abort function may be used to deal with unrecoverable failures such
410 * as journal IO errors or ENOMEM at a critical moment in log management.
411 *
412 * We unconditionally force the filesystem into an ABORT|READONLY state,
413 * unless the error response on the fs has been set to panic in which
414 * case we take the easy way out and panic immediately.
415 */
416
417 void ext4_abort(struct super_block *sb, const char *function,
418 const char *fmt, ...)
419 {
420 va_list args;
421
422 va_start(args, fmt);
423 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
424 vprintk(fmt, args);
425 printk("\n");
426 va_end(args);
427
428 if (test_opt(sb, ERRORS_PANIC))
429 panic("EXT4-fs panic from previous error\n");
430
431 if (sb->s_flags & MS_RDONLY)
432 return;
433
434 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
435 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
436 sb->s_flags |= MS_RDONLY;
437 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
438 if (EXT4_SB(sb)->s_journal)
439 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
440 }
441
442 void ext4_msg (struct super_block * sb, const char *prefix,
443 const char *fmt, ...)
444 {
445 va_list args;
446
447 va_start(args, fmt);
448 printk("%sEXT4-fs (%s): ", prefix, sb->s_id);
449 vprintk(fmt, args);
450 printk("\n");
451 va_end(args);
452 }
453
454 void ext4_warning(struct super_block *sb, const char *function,
455 const char *fmt, ...)
456 {
457 va_list args;
458
459 va_start(args, fmt);
460 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
461 sb->s_id, function);
462 vprintk(fmt, args);
463 printk("\n");
464 va_end(args);
465 }
466
467 void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp,
468 const char *function, const char *fmt, ...)
469 __releases(bitlock)
470 __acquires(bitlock)
471 {
472 va_list args;
473 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
474
475 va_start(args, fmt);
476 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
477 vprintk(fmt, args);
478 printk("\n");
479 va_end(args);
480
481 if (test_opt(sb, ERRORS_CONT)) {
482 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
483 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
484 ext4_commit_super(sb, 0);
485 return;
486 }
487 ext4_unlock_group(sb, grp);
488 ext4_handle_error(sb);
489 /*
490 * We only get here in the ERRORS_RO case; relocking the group
491 * may be dangerous, but nothing bad will happen since the
492 * filesystem will have already been marked read/only and the
493 * journal has been aborted. We return 1 as a hint to callers
494 * who might what to use the return value from
495 * ext4_grp_locked_error() to distinguish beween the
496 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
497 * aggressively from the ext4 function in question, with a
498 * more appropriate error code.
499 */
500 ext4_lock_group(sb, grp);
501 return;
502 }
503
504 void ext4_update_dynamic_rev(struct super_block *sb)
505 {
506 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
507
508 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
509 return;
510
511 ext4_warning(sb, __func__,
512 "updating to rev %d because of new feature flag, "
513 "running e2fsck is recommended",
514 EXT4_DYNAMIC_REV);
515
516 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
517 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
518 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
519 /* leave es->s_feature_*compat flags alone */
520 /* es->s_uuid will be set by e2fsck if empty */
521
522 /*
523 * The rest of the superblock fields should be zero, and if not it
524 * means they are likely already in use, so leave them alone. We
525 * can leave it up to e2fsck to clean up any inconsistencies there.
526 */
527 }
528
529 /*
530 * Open the external journal device
531 */
532 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
533 {
534 struct block_device *bdev;
535 char b[BDEVNAME_SIZE];
536
537 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
538 if (IS_ERR(bdev))
539 goto fail;
540 return bdev;
541
542 fail:
543 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
544 __bdevname(dev, b), PTR_ERR(bdev));
545 return NULL;
546 }
547
548 /*
549 * Release the journal device
550 */
551 static int ext4_blkdev_put(struct block_device *bdev)
552 {
553 bd_release(bdev);
554 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
555 }
556
557 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
558 {
559 struct block_device *bdev;
560 int ret = -ENODEV;
561
562 bdev = sbi->journal_bdev;
563 if (bdev) {
564 ret = ext4_blkdev_put(bdev);
565 sbi->journal_bdev = NULL;
566 }
567 return ret;
568 }
569
570 static inline struct inode *orphan_list_entry(struct list_head *l)
571 {
572 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
573 }
574
575 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
576 {
577 struct list_head *l;
578
579 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
580 le32_to_cpu(sbi->s_es->s_last_orphan));
581
582 printk(KERN_ERR "sb_info orphan list:\n");
583 list_for_each(l, &sbi->s_orphan) {
584 struct inode *inode = orphan_list_entry(l);
585 printk(KERN_ERR " "
586 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
587 inode->i_sb->s_id, inode->i_ino, inode,
588 inode->i_mode, inode->i_nlink,
589 NEXT_ORPHAN(inode));
590 }
591 }
592
593 static void ext4_put_super(struct super_block *sb)
594 {
595 struct ext4_sb_info *sbi = EXT4_SB(sb);
596 struct ext4_super_block *es = sbi->s_es;
597 int i, err;
598
599 flush_workqueue(sbi->dio_unwritten_wq);
600 destroy_workqueue(sbi->dio_unwritten_wq);
601
602 lock_super(sb);
603 lock_kernel();
604 if (sb->s_dirt)
605 ext4_commit_super(sb, 1);
606
607 if (sbi->s_journal) {
608 err = jbd2_journal_destroy(sbi->s_journal);
609 sbi->s_journal = NULL;
610 if (err < 0)
611 ext4_abort(sb, __func__,
612 "Couldn't clean up the journal");
613 }
614
615 ext4_release_system_zone(sb);
616 ext4_mb_release(sb);
617 ext4_ext_release(sb);
618 ext4_xattr_put_super(sb);
619
620 if (!(sb->s_flags & MS_RDONLY)) {
621 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
622 es->s_state = cpu_to_le16(sbi->s_mount_state);
623 ext4_commit_super(sb, 1);
624 }
625 if (sbi->s_proc) {
626 remove_proc_entry(sb->s_id, ext4_proc_root);
627 }
628 kobject_del(&sbi->s_kobj);
629
630 for (i = 0; i < sbi->s_gdb_count; i++)
631 brelse(sbi->s_group_desc[i]);
632 kfree(sbi->s_group_desc);
633 if (is_vmalloc_addr(sbi->s_flex_groups))
634 vfree(sbi->s_flex_groups);
635 else
636 kfree(sbi->s_flex_groups);
637 percpu_counter_destroy(&sbi->s_freeblocks_counter);
638 percpu_counter_destroy(&sbi->s_freeinodes_counter);
639 percpu_counter_destroy(&sbi->s_dirs_counter);
640 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
641 brelse(sbi->s_sbh);
642 #ifdef CONFIG_QUOTA
643 for (i = 0; i < MAXQUOTAS; i++)
644 kfree(sbi->s_qf_names[i]);
645 #endif
646
647 /* Debugging code just in case the in-memory inode orphan list
648 * isn't empty. The on-disk one can be non-empty if we've
649 * detected an error and taken the fs readonly, but the
650 * in-memory list had better be clean by this point. */
651 if (!list_empty(&sbi->s_orphan))
652 dump_orphan_list(sb, sbi);
653 J_ASSERT(list_empty(&sbi->s_orphan));
654
655 invalidate_bdev(sb->s_bdev);
656 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
657 /*
658 * Invalidate the journal device's buffers. We don't want them
659 * floating about in memory - the physical journal device may
660 * hotswapped, and it breaks the `ro-after' testing code.
661 */
662 sync_blockdev(sbi->journal_bdev);
663 invalidate_bdev(sbi->journal_bdev);
664 ext4_blkdev_remove(sbi);
665 }
666 sb->s_fs_info = NULL;
667 /*
668 * Now that we are completely done shutting down the
669 * superblock, we need to actually destroy the kobject.
670 */
671 unlock_kernel();
672 unlock_super(sb);
673 kobject_put(&sbi->s_kobj);
674 wait_for_completion(&sbi->s_kobj_unregister);
675 kfree(sbi->s_blockgroup_lock);
676 kfree(sbi);
677 }
678
679 static struct kmem_cache *ext4_inode_cachep;
680
681 /*
682 * Called inside transaction, so use GFP_NOFS
683 */
684 static struct inode *ext4_alloc_inode(struct super_block *sb)
685 {
686 struct ext4_inode_info *ei;
687
688 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
689 if (!ei)
690 return NULL;
691
692 ei->vfs_inode.i_version = 1;
693 ei->vfs_inode.i_data.writeback_index = 0;
694 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
695 INIT_LIST_HEAD(&ei->i_prealloc_list);
696 spin_lock_init(&ei->i_prealloc_lock);
697 /*
698 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
699 * therefore it can be null here. Don't check it, just initialize
700 * jinode.
701 */
702 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
703 ei->i_reserved_data_blocks = 0;
704 ei->i_reserved_meta_blocks = 0;
705 ei->i_allocated_meta_blocks = 0;
706 ei->i_da_metadata_calc_len = 0;
707 ei->i_delalloc_reserved_flag = 0;
708 spin_lock_init(&(ei->i_block_reservation_lock));
709 #ifdef CONFIG_QUOTA
710 ei->i_reserved_quota = 0;
711 #endif
712 INIT_LIST_HEAD(&ei->i_aio_dio_complete_list);
713 ei->cur_aio_dio = NULL;
714 ei->i_sync_tid = 0;
715 ei->i_datasync_tid = 0;
716
717 return &ei->vfs_inode;
718 }
719
720 static void ext4_destroy_inode(struct inode *inode)
721 {
722 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
723 ext4_msg(inode->i_sb, KERN_ERR,
724 "Inode %lu (%p): orphan list check failed!",
725 inode->i_ino, EXT4_I(inode));
726 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
727 EXT4_I(inode), sizeof(struct ext4_inode_info),
728 true);
729 dump_stack();
730 }
731 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
732 }
733
734 static void init_once(void *foo)
735 {
736 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
737
738 INIT_LIST_HEAD(&ei->i_orphan);
739 #ifdef CONFIG_EXT4_FS_XATTR
740 init_rwsem(&ei->xattr_sem);
741 #endif
742 init_rwsem(&ei->i_data_sem);
743 inode_init_once(&ei->vfs_inode);
744 }
745
746 static int init_inodecache(void)
747 {
748 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
749 sizeof(struct ext4_inode_info),
750 0, (SLAB_RECLAIM_ACCOUNT|
751 SLAB_MEM_SPREAD),
752 init_once);
753 if (ext4_inode_cachep == NULL)
754 return -ENOMEM;
755 return 0;
756 }
757
758 static void destroy_inodecache(void)
759 {
760 kmem_cache_destroy(ext4_inode_cachep);
761 }
762
763 static void ext4_clear_inode(struct inode *inode)
764 {
765 ext4_discard_preallocations(inode);
766 if (EXT4_JOURNAL(inode))
767 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
768 &EXT4_I(inode)->jinode);
769 }
770
771 static inline void ext4_show_quota_options(struct seq_file *seq,
772 struct super_block *sb)
773 {
774 #if defined(CONFIG_QUOTA)
775 struct ext4_sb_info *sbi = EXT4_SB(sb);
776
777 if (sbi->s_jquota_fmt)
778 seq_printf(seq, ",jqfmt=%s",
779 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
780
781 if (sbi->s_qf_names[USRQUOTA])
782 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
783
784 if (sbi->s_qf_names[GRPQUOTA])
785 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
786
787 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
788 seq_puts(seq, ",usrquota");
789
790 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
791 seq_puts(seq, ",grpquota");
792 #endif
793 }
794
795 /*
796 * Show an option if
797 * - it's set to a non-default value OR
798 * - if the per-sb default is different from the global default
799 */
800 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
801 {
802 int def_errors;
803 unsigned long def_mount_opts;
804 struct super_block *sb = vfs->mnt_sb;
805 struct ext4_sb_info *sbi = EXT4_SB(sb);
806 struct ext4_super_block *es = sbi->s_es;
807
808 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
809 def_errors = le16_to_cpu(es->s_errors);
810
811 if (sbi->s_sb_block != 1)
812 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
813 if (test_opt(sb, MINIX_DF))
814 seq_puts(seq, ",minixdf");
815 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
816 seq_puts(seq, ",grpid");
817 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
818 seq_puts(seq, ",nogrpid");
819 if (sbi->s_resuid != EXT4_DEF_RESUID ||
820 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
821 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
822 }
823 if (sbi->s_resgid != EXT4_DEF_RESGID ||
824 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
825 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
826 }
827 if (test_opt(sb, ERRORS_RO)) {
828 if (def_errors == EXT4_ERRORS_PANIC ||
829 def_errors == EXT4_ERRORS_CONTINUE) {
830 seq_puts(seq, ",errors=remount-ro");
831 }
832 }
833 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
834 seq_puts(seq, ",errors=continue");
835 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
836 seq_puts(seq, ",errors=panic");
837 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
838 seq_puts(seq, ",nouid32");
839 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
840 seq_puts(seq, ",debug");
841 if (test_opt(sb, OLDALLOC))
842 seq_puts(seq, ",oldalloc");
843 #ifdef CONFIG_EXT4_FS_XATTR
844 if (test_opt(sb, XATTR_USER) &&
845 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
846 seq_puts(seq, ",user_xattr");
847 if (!test_opt(sb, XATTR_USER) &&
848 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
849 seq_puts(seq, ",nouser_xattr");
850 }
851 #endif
852 #ifdef CONFIG_EXT4_FS_POSIX_ACL
853 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
854 seq_puts(seq, ",acl");
855 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
856 seq_puts(seq, ",noacl");
857 #endif
858 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
859 seq_printf(seq, ",commit=%u",
860 (unsigned) (sbi->s_commit_interval / HZ));
861 }
862 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
863 seq_printf(seq, ",min_batch_time=%u",
864 (unsigned) sbi->s_min_batch_time);
865 }
866 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
867 seq_printf(seq, ",max_batch_time=%u",
868 (unsigned) sbi->s_min_batch_time);
869 }
870
871 /*
872 * We're changing the default of barrier mount option, so
873 * let's always display its mount state so it's clear what its
874 * status is.
875 */
876 seq_puts(seq, ",barrier=");
877 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
878 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
879 seq_puts(seq, ",journal_async_commit");
880 else if (test_opt(sb, JOURNAL_CHECKSUM))
881 seq_puts(seq, ",journal_checksum");
882 if (test_opt(sb, NOBH))
883 seq_puts(seq, ",nobh");
884 if (test_opt(sb, I_VERSION))
885 seq_puts(seq, ",i_version");
886 if (!test_opt(sb, DELALLOC))
887 seq_puts(seq, ",nodelalloc");
888
889
890 if (sbi->s_stripe)
891 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
892 /*
893 * journal mode get enabled in different ways
894 * So just print the value even if we didn't specify it
895 */
896 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
897 seq_puts(seq, ",data=journal");
898 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
899 seq_puts(seq, ",data=ordered");
900 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
901 seq_puts(seq, ",data=writeback");
902
903 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
904 seq_printf(seq, ",inode_readahead_blks=%u",
905 sbi->s_inode_readahead_blks);
906
907 if (test_opt(sb, DATA_ERR_ABORT))
908 seq_puts(seq, ",data_err=abort");
909
910 if (test_opt(sb, NO_AUTO_DA_ALLOC))
911 seq_puts(seq, ",noauto_da_alloc");
912
913 if (test_opt(sb, DISCARD))
914 seq_puts(seq, ",discard");
915
916 if (test_opt(sb, NOLOAD))
917 seq_puts(seq, ",norecovery");
918
919 ext4_show_quota_options(seq, sb);
920
921 return 0;
922 }
923
924 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
925 u64 ino, u32 generation)
926 {
927 struct inode *inode;
928
929 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
930 return ERR_PTR(-ESTALE);
931 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
932 return ERR_PTR(-ESTALE);
933
934 /* iget isn't really right if the inode is currently unallocated!!
935 *
936 * ext4_read_inode will return a bad_inode if the inode had been
937 * deleted, so we should be safe.
938 *
939 * Currently we don't know the generation for parent directory, so
940 * a generation of 0 means "accept any"
941 */
942 inode = ext4_iget(sb, ino);
943 if (IS_ERR(inode))
944 return ERR_CAST(inode);
945 if (generation && inode->i_generation != generation) {
946 iput(inode);
947 return ERR_PTR(-ESTALE);
948 }
949
950 return inode;
951 }
952
953 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
954 int fh_len, int fh_type)
955 {
956 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
957 ext4_nfs_get_inode);
958 }
959
960 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
961 int fh_len, int fh_type)
962 {
963 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
964 ext4_nfs_get_inode);
965 }
966
967 /*
968 * Try to release metadata pages (indirect blocks, directories) which are
969 * mapped via the block device. Since these pages could have journal heads
970 * which would prevent try_to_free_buffers() from freeing them, we must use
971 * jbd2 layer's try_to_free_buffers() function to release them.
972 */
973 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
974 gfp_t wait)
975 {
976 journal_t *journal = EXT4_SB(sb)->s_journal;
977
978 WARN_ON(PageChecked(page));
979 if (!page_has_buffers(page))
980 return 0;
981 if (journal)
982 return jbd2_journal_try_to_free_buffers(journal, page,
983 wait & ~__GFP_WAIT);
984 return try_to_free_buffers(page);
985 }
986
987 #ifdef CONFIG_QUOTA
988 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
989 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
990
991 static int ext4_write_dquot(struct dquot *dquot);
992 static int ext4_acquire_dquot(struct dquot *dquot);
993 static int ext4_release_dquot(struct dquot *dquot);
994 static int ext4_mark_dquot_dirty(struct dquot *dquot);
995 static int ext4_write_info(struct super_block *sb, int type);
996 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
997 char *path, int remount);
998 static int ext4_quota_on_mount(struct super_block *sb, int type);
999 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1000 size_t len, loff_t off);
1001 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1002 const char *data, size_t len, loff_t off);
1003
1004 static const struct dquot_operations ext4_quota_operations = {
1005 .initialize = dquot_initialize,
1006 .drop = dquot_drop,
1007 .alloc_space = dquot_alloc_space,
1008 .reserve_space = dquot_reserve_space,
1009 .claim_space = dquot_claim_space,
1010 .release_rsv = dquot_release_reserved_space,
1011 #ifdef CONFIG_QUOTA
1012 .get_reserved_space = ext4_get_reserved_space,
1013 #endif
1014 .alloc_inode = dquot_alloc_inode,
1015 .free_space = dquot_free_space,
1016 .free_inode = dquot_free_inode,
1017 .transfer = dquot_transfer,
1018 .write_dquot = ext4_write_dquot,
1019 .acquire_dquot = ext4_acquire_dquot,
1020 .release_dquot = ext4_release_dquot,
1021 .mark_dirty = ext4_mark_dquot_dirty,
1022 .write_info = ext4_write_info,
1023 .alloc_dquot = dquot_alloc,
1024 .destroy_dquot = dquot_destroy,
1025 };
1026
1027 static const struct quotactl_ops ext4_qctl_operations = {
1028 .quota_on = ext4_quota_on,
1029 .quota_off = vfs_quota_off,
1030 .quota_sync = vfs_quota_sync,
1031 .get_info = vfs_get_dqinfo,
1032 .set_info = vfs_set_dqinfo,
1033 .get_dqblk = vfs_get_dqblk,
1034 .set_dqblk = vfs_set_dqblk
1035 };
1036 #endif
1037
1038 static const struct super_operations ext4_sops = {
1039 .alloc_inode = ext4_alloc_inode,
1040 .destroy_inode = ext4_destroy_inode,
1041 .write_inode = ext4_write_inode,
1042 .dirty_inode = ext4_dirty_inode,
1043 .delete_inode = ext4_delete_inode,
1044 .put_super = ext4_put_super,
1045 .sync_fs = ext4_sync_fs,
1046 .freeze_fs = ext4_freeze,
1047 .unfreeze_fs = ext4_unfreeze,
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 super_operations ext4_nojournal_sops = {
1060 .alloc_inode = ext4_alloc_inode,
1061 .destroy_inode = ext4_destroy_inode,
1062 .write_inode = ext4_write_inode,
1063 .dirty_inode = ext4_dirty_inode,
1064 .delete_inode = ext4_delete_inode,
1065 .write_super = ext4_write_super,
1066 .put_super = ext4_put_super,
1067 .statfs = ext4_statfs,
1068 .remount_fs = ext4_remount,
1069 .clear_inode = ext4_clear_inode,
1070 .show_options = ext4_show_options,
1071 #ifdef CONFIG_QUOTA
1072 .quota_read = ext4_quota_read,
1073 .quota_write = ext4_quota_write,
1074 #endif
1075 .bdev_try_to_free_page = bdev_try_to_free_page,
1076 };
1077
1078 static const struct export_operations ext4_export_ops = {
1079 .fh_to_dentry = ext4_fh_to_dentry,
1080 .fh_to_parent = ext4_fh_to_parent,
1081 .get_parent = ext4_get_parent,
1082 };
1083
1084 enum {
1085 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1086 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1087 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1088 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1089 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1090 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1091 Opt_journal_update, Opt_journal_dev,
1092 Opt_journal_checksum, Opt_journal_async_commit,
1093 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1094 Opt_data_err_abort, Opt_data_err_ignore,
1095 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1096 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
1097 Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err, Opt_resize,
1098 Opt_usrquota, Opt_grpquota, Opt_i_version,
1099 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1100 Opt_block_validity, Opt_noblock_validity,
1101 Opt_inode_readahead_blks, Opt_journal_ioprio,
1102 Opt_discard, Opt_nodiscard,
1103 };
1104
1105 static const match_table_t tokens = {
1106 {Opt_bsd_df, "bsddf"},
1107 {Opt_minix_df, "minixdf"},
1108 {Opt_grpid, "grpid"},
1109 {Opt_grpid, "bsdgroups"},
1110 {Opt_nogrpid, "nogrpid"},
1111 {Opt_nogrpid, "sysvgroups"},
1112 {Opt_resgid, "resgid=%u"},
1113 {Opt_resuid, "resuid=%u"},
1114 {Opt_sb, "sb=%u"},
1115 {Opt_err_cont, "errors=continue"},
1116 {Opt_err_panic, "errors=panic"},
1117 {Opt_err_ro, "errors=remount-ro"},
1118 {Opt_nouid32, "nouid32"},
1119 {Opt_debug, "debug"},
1120 {Opt_oldalloc, "oldalloc"},
1121 {Opt_orlov, "orlov"},
1122 {Opt_user_xattr, "user_xattr"},
1123 {Opt_nouser_xattr, "nouser_xattr"},
1124 {Opt_acl, "acl"},
1125 {Opt_noacl, "noacl"},
1126 {Opt_noload, "noload"},
1127 {Opt_noload, "norecovery"},
1128 {Opt_nobh, "nobh"},
1129 {Opt_bh, "bh"},
1130 {Opt_commit, "commit=%u"},
1131 {Opt_min_batch_time, "min_batch_time=%u"},
1132 {Opt_max_batch_time, "max_batch_time=%u"},
1133 {Opt_journal_update, "journal=update"},
1134 {Opt_journal_dev, "journal_dev=%u"},
1135 {Opt_journal_checksum, "journal_checksum"},
1136 {Opt_journal_async_commit, "journal_async_commit"},
1137 {Opt_abort, "abort"},
1138 {Opt_data_journal, "data=journal"},
1139 {Opt_data_ordered, "data=ordered"},
1140 {Opt_data_writeback, "data=writeback"},
1141 {Opt_data_err_abort, "data_err=abort"},
1142 {Opt_data_err_ignore, "data_err=ignore"},
1143 {Opt_offusrjquota, "usrjquota="},
1144 {Opt_usrjquota, "usrjquota=%s"},
1145 {Opt_offgrpjquota, "grpjquota="},
1146 {Opt_grpjquota, "grpjquota=%s"},
1147 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1148 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1149 {Opt_grpquota, "grpquota"},
1150 {Opt_noquota, "noquota"},
1151 {Opt_quota, "quota"},
1152 {Opt_usrquota, "usrquota"},
1153 {Opt_barrier, "barrier=%u"},
1154 {Opt_barrier, "barrier"},
1155 {Opt_nobarrier, "nobarrier"},
1156 {Opt_i_version, "i_version"},
1157 {Opt_stripe, "stripe=%u"},
1158 {Opt_resize, "resize"},
1159 {Opt_delalloc, "delalloc"},
1160 {Opt_nodelalloc, "nodelalloc"},
1161 {Opt_block_validity, "block_validity"},
1162 {Opt_noblock_validity, "noblock_validity"},
1163 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1164 {Opt_journal_ioprio, "journal_ioprio=%u"},
1165 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1166 {Opt_auto_da_alloc, "auto_da_alloc"},
1167 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1168 {Opt_discard, "discard"},
1169 {Opt_nodiscard, "nodiscard"},
1170 {Opt_err, NULL},
1171 };
1172
1173 static ext4_fsblk_t get_sb_block(void **data)
1174 {
1175 ext4_fsblk_t sb_block;
1176 char *options = (char *) *data;
1177
1178 if (!options || strncmp(options, "sb=", 3) != 0)
1179 return 1; /* Default location */
1180
1181 options += 3;
1182 /* TODO: use simple_strtoll with >32bit ext4 */
1183 sb_block = simple_strtoul(options, &options, 0);
1184 if (*options && *options != ',') {
1185 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1186 (char *) *data);
1187 return 1;
1188 }
1189 if (*options == ',')
1190 options++;
1191 *data = (void *) options;
1192
1193 return sb_block;
1194 }
1195
1196 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1197
1198 static int parse_options(char *options, struct super_block *sb,
1199 unsigned long *journal_devnum,
1200 unsigned int *journal_ioprio,
1201 ext4_fsblk_t *n_blocks_count, int is_remount)
1202 {
1203 struct ext4_sb_info *sbi = EXT4_SB(sb);
1204 char *p;
1205 substring_t args[MAX_OPT_ARGS];
1206 int data_opt = 0;
1207 int option;
1208 #ifdef CONFIG_QUOTA
1209 int qtype, qfmt;
1210 char *qname;
1211 #endif
1212
1213 if (!options)
1214 return 1;
1215
1216 while ((p = strsep(&options, ",")) != NULL) {
1217 int token;
1218 if (!*p)
1219 continue;
1220
1221 /*
1222 * Initialize args struct so we know whether arg was
1223 * found; some options take optional arguments.
1224 */
1225 args[0].to = args[0].from = 0;
1226 token = match_token(p, tokens, args);
1227 switch (token) {
1228 case Opt_bsd_df:
1229 clear_opt(sbi->s_mount_opt, MINIX_DF);
1230 break;
1231 case Opt_minix_df:
1232 set_opt(sbi->s_mount_opt, MINIX_DF);
1233 break;
1234 case Opt_grpid:
1235 set_opt(sbi->s_mount_opt, GRPID);
1236 break;
1237 case Opt_nogrpid:
1238 clear_opt(sbi->s_mount_opt, GRPID);
1239 break;
1240 case Opt_resuid:
1241 if (match_int(&args[0], &option))
1242 return 0;
1243 sbi->s_resuid = option;
1244 break;
1245 case Opt_resgid:
1246 if (match_int(&args[0], &option))
1247 return 0;
1248 sbi->s_resgid = option;
1249 break;
1250 case Opt_sb:
1251 /* handled by get_sb_block() instead of here */
1252 /* *sb_block = match_int(&args[0]); */
1253 break;
1254 case Opt_err_panic:
1255 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1256 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1257 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1258 break;
1259 case Opt_err_ro:
1260 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1261 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1262 set_opt(sbi->s_mount_opt, ERRORS_RO);
1263 break;
1264 case Opt_err_cont:
1265 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1266 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1267 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1268 break;
1269 case Opt_nouid32:
1270 set_opt(sbi->s_mount_opt, NO_UID32);
1271 break;
1272 case Opt_debug:
1273 set_opt(sbi->s_mount_opt, DEBUG);
1274 break;
1275 case Opt_oldalloc:
1276 set_opt(sbi->s_mount_opt, OLDALLOC);
1277 break;
1278 case Opt_orlov:
1279 clear_opt(sbi->s_mount_opt, OLDALLOC);
1280 break;
1281 #ifdef CONFIG_EXT4_FS_XATTR
1282 case Opt_user_xattr:
1283 set_opt(sbi->s_mount_opt, XATTR_USER);
1284 break;
1285 case Opt_nouser_xattr:
1286 clear_opt(sbi->s_mount_opt, XATTR_USER);
1287 break;
1288 #else
1289 case Opt_user_xattr:
1290 case Opt_nouser_xattr:
1291 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1292 break;
1293 #endif
1294 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1295 case Opt_acl:
1296 set_opt(sbi->s_mount_opt, POSIX_ACL);
1297 break;
1298 case Opt_noacl:
1299 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1300 break;
1301 #else
1302 case Opt_acl:
1303 case Opt_noacl:
1304 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1305 break;
1306 #endif
1307 case Opt_journal_update:
1308 /* @@@ FIXME */
1309 /* Eventually we will want to be able to create
1310 a journal file here. For now, only allow the
1311 user to specify an existing inode to be the
1312 journal file. */
1313 if (is_remount) {
1314 ext4_msg(sb, KERN_ERR,
1315 "Cannot specify journal on remount");
1316 return 0;
1317 }
1318 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1319 break;
1320 case Opt_journal_dev:
1321 if (is_remount) {
1322 ext4_msg(sb, KERN_ERR,
1323 "Cannot specify journal on remount");
1324 return 0;
1325 }
1326 if (match_int(&args[0], &option))
1327 return 0;
1328 *journal_devnum = option;
1329 break;
1330 case Opt_journal_checksum:
1331 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1332 break;
1333 case Opt_journal_async_commit:
1334 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1335 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1336 break;
1337 case Opt_noload:
1338 set_opt(sbi->s_mount_opt, NOLOAD);
1339 break;
1340 case Opt_commit:
1341 if (match_int(&args[0], &option))
1342 return 0;
1343 if (option < 0)
1344 return 0;
1345 if (option == 0)
1346 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1347 sbi->s_commit_interval = HZ * option;
1348 break;
1349 case Opt_max_batch_time:
1350 if (match_int(&args[0], &option))
1351 return 0;
1352 if (option < 0)
1353 return 0;
1354 if (option == 0)
1355 option = EXT4_DEF_MAX_BATCH_TIME;
1356 sbi->s_max_batch_time = option;
1357 break;
1358 case Opt_min_batch_time:
1359 if (match_int(&args[0], &option))
1360 return 0;
1361 if (option < 0)
1362 return 0;
1363 sbi->s_min_batch_time = option;
1364 break;
1365 case Opt_data_journal:
1366 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1367 goto datacheck;
1368 case Opt_data_ordered:
1369 data_opt = EXT4_MOUNT_ORDERED_DATA;
1370 goto datacheck;
1371 case Opt_data_writeback:
1372 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1373 datacheck:
1374 if (is_remount) {
1375 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1376 != data_opt) {
1377 ext4_msg(sb, KERN_ERR,
1378 "Cannot change data mode on remount");
1379 return 0;
1380 }
1381 } else {
1382 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1383 sbi->s_mount_opt |= data_opt;
1384 }
1385 break;
1386 case Opt_data_err_abort:
1387 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1388 break;
1389 case Opt_data_err_ignore:
1390 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1391 break;
1392 #ifdef CONFIG_QUOTA
1393 case Opt_usrjquota:
1394 qtype = USRQUOTA;
1395 goto set_qf_name;
1396 case Opt_grpjquota:
1397 qtype = GRPQUOTA;
1398 set_qf_name:
1399 if (sb_any_quota_loaded(sb) &&
1400 !sbi->s_qf_names[qtype]) {
1401 ext4_msg(sb, KERN_ERR,
1402 "Cannot change journaled "
1403 "quota options when quota turned on");
1404 return 0;
1405 }
1406 qname = match_strdup(&args[0]);
1407 if (!qname) {
1408 ext4_msg(sb, KERN_ERR,
1409 "Not enough memory for "
1410 "storing quotafile name");
1411 return 0;
1412 }
1413 if (sbi->s_qf_names[qtype] &&
1414 strcmp(sbi->s_qf_names[qtype], qname)) {
1415 ext4_msg(sb, KERN_ERR,
1416 "%s quota file already "
1417 "specified", QTYPE2NAME(qtype));
1418 kfree(qname);
1419 return 0;
1420 }
1421 sbi->s_qf_names[qtype] = qname;
1422 if (strchr(sbi->s_qf_names[qtype], '/')) {
1423 ext4_msg(sb, KERN_ERR,
1424 "quotafile must be on "
1425 "filesystem root");
1426 kfree(sbi->s_qf_names[qtype]);
1427 sbi->s_qf_names[qtype] = NULL;
1428 return 0;
1429 }
1430 set_opt(sbi->s_mount_opt, QUOTA);
1431 break;
1432 case Opt_offusrjquota:
1433 qtype = USRQUOTA;
1434 goto clear_qf_name;
1435 case Opt_offgrpjquota:
1436 qtype = GRPQUOTA;
1437 clear_qf_name:
1438 if (sb_any_quota_loaded(sb) &&
1439 sbi->s_qf_names[qtype]) {
1440 ext4_msg(sb, KERN_ERR, "Cannot change "
1441 "journaled quota options when "
1442 "quota turned on");
1443 return 0;
1444 }
1445 /*
1446 * The space will be released later when all options
1447 * are confirmed to be correct
1448 */
1449 sbi->s_qf_names[qtype] = NULL;
1450 break;
1451 case Opt_jqfmt_vfsold:
1452 qfmt = QFMT_VFS_OLD;
1453 goto set_qf_format;
1454 case Opt_jqfmt_vfsv0:
1455 qfmt = QFMT_VFS_V0;
1456 set_qf_format:
1457 if (sb_any_quota_loaded(sb) &&
1458 sbi->s_jquota_fmt != qfmt) {
1459 ext4_msg(sb, KERN_ERR, "Cannot change "
1460 "journaled quota options when "
1461 "quota turned on");
1462 return 0;
1463 }
1464 sbi->s_jquota_fmt = qfmt;
1465 break;
1466 case Opt_quota:
1467 case Opt_usrquota:
1468 set_opt(sbi->s_mount_opt, QUOTA);
1469 set_opt(sbi->s_mount_opt, USRQUOTA);
1470 break;
1471 case Opt_grpquota:
1472 set_opt(sbi->s_mount_opt, QUOTA);
1473 set_opt(sbi->s_mount_opt, GRPQUOTA);
1474 break;
1475 case Opt_noquota:
1476 if (sb_any_quota_loaded(sb)) {
1477 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1478 "options when quota turned on");
1479 return 0;
1480 }
1481 clear_opt(sbi->s_mount_opt, QUOTA);
1482 clear_opt(sbi->s_mount_opt, USRQUOTA);
1483 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1484 break;
1485 #else
1486 case Opt_quota:
1487 case Opt_usrquota:
1488 case Opt_grpquota:
1489 ext4_msg(sb, KERN_ERR,
1490 "quota options not supported");
1491 break;
1492 case Opt_usrjquota:
1493 case Opt_grpjquota:
1494 case Opt_offusrjquota:
1495 case Opt_offgrpjquota:
1496 case Opt_jqfmt_vfsold:
1497 case Opt_jqfmt_vfsv0:
1498 ext4_msg(sb, KERN_ERR,
1499 "journaled quota options not supported");
1500 break;
1501 case Opt_noquota:
1502 break;
1503 #endif
1504 case Opt_abort:
1505 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1506 break;
1507 case Opt_nobarrier:
1508 clear_opt(sbi->s_mount_opt, BARRIER);
1509 break;
1510 case Opt_barrier:
1511 if (args[0].from) {
1512 if (match_int(&args[0], &option))
1513 return 0;
1514 } else
1515 option = 1; /* No argument, default to 1 */
1516 if (option)
1517 set_opt(sbi->s_mount_opt, BARRIER);
1518 else
1519 clear_opt(sbi->s_mount_opt, BARRIER);
1520 break;
1521 case Opt_ignore:
1522 break;
1523 case Opt_resize:
1524 if (!is_remount) {
1525 ext4_msg(sb, KERN_ERR,
1526 "resize option only available "
1527 "for remount");
1528 return 0;
1529 }
1530 if (match_int(&args[0], &option) != 0)
1531 return 0;
1532 *n_blocks_count = option;
1533 break;
1534 case Opt_nobh:
1535 set_opt(sbi->s_mount_opt, NOBH);
1536 break;
1537 case Opt_bh:
1538 clear_opt(sbi->s_mount_opt, NOBH);
1539 break;
1540 case Opt_i_version:
1541 set_opt(sbi->s_mount_opt, I_VERSION);
1542 sb->s_flags |= MS_I_VERSION;
1543 break;
1544 case Opt_nodelalloc:
1545 clear_opt(sbi->s_mount_opt, DELALLOC);
1546 break;
1547 case Opt_stripe:
1548 if (match_int(&args[0], &option))
1549 return 0;
1550 if (option < 0)
1551 return 0;
1552 sbi->s_stripe = option;
1553 break;
1554 case Opt_delalloc:
1555 set_opt(sbi->s_mount_opt, DELALLOC);
1556 break;
1557 case Opt_block_validity:
1558 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1559 break;
1560 case Opt_noblock_validity:
1561 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1562 break;
1563 case Opt_inode_readahead_blks:
1564 if (match_int(&args[0], &option))
1565 return 0;
1566 if (option < 0 || option > (1 << 30))
1567 return 0;
1568 if (!is_power_of_2(option)) {
1569 ext4_msg(sb, KERN_ERR,
1570 "EXT4-fs: inode_readahead_blks"
1571 " must be a power of 2");
1572 return 0;
1573 }
1574 sbi->s_inode_readahead_blks = option;
1575 break;
1576 case Opt_journal_ioprio:
1577 if (match_int(&args[0], &option))
1578 return 0;
1579 if (option < 0 || option > 7)
1580 break;
1581 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1582 option);
1583 break;
1584 case Opt_noauto_da_alloc:
1585 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1586 break;
1587 case Opt_auto_da_alloc:
1588 if (args[0].from) {
1589 if (match_int(&args[0], &option))
1590 return 0;
1591 } else
1592 option = 1; /* No argument, default to 1 */
1593 if (option)
1594 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1595 else
1596 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1597 break;
1598 case Opt_discard:
1599 set_opt(sbi->s_mount_opt, DISCARD);
1600 break;
1601 case Opt_nodiscard:
1602 clear_opt(sbi->s_mount_opt, DISCARD);
1603 break;
1604 default:
1605 ext4_msg(sb, KERN_ERR,
1606 "Unrecognized mount option \"%s\" "
1607 "or missing value", p);
1608 return 0;
1609 }
1610 }
1611 #ifdef CONFIG_QUOTA
1612 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1613 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1614 sbi->s_qf_names[USRQUOTA])
1615 clear_opt(sbi->s_mount_opt, USRQUOTA);
1616
1617 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1618 sbi->s_qf_names[GRPQUOTA])
1619 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1620
1621 if ((sbi->s_qf_names[USRQUOTA] &&
1622 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1623 (sbi->s_qf_names[GRPQUOTA] &&
1624 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1625 ext4_msg(sb, KERN_ERR, "old and new quota "
1626 "format mixing");
1627 return 0;
1628 }
1629
1630 if (!sbi->s_jquota_fmt) {
1631 ext4_msg(sb, KERN_ERR, "journaled quota format "
1632 "not specified");
1633 return 0;
1634 }
1635 } else {
1636 if (sbi->s_jquota_fmt) {
1637 ext4_msg(sb, KERN_ERR, "journaled quota format "
1638 "specified with no journaling "
1639 "enabled");
1640 return 0;
1641 }
1642 }
1643 #endif
1644 return 1;
1645 }
1646
1647 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1648 int read_only)
1649 {
1650 struct ext4_sb_info *sbi = EXT4_SB(sb);
1651 int res = 0;
1652
1653 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1654 ext4_msg(sb, KERN_ERR, "revision level too high, "
1655 "forcing read-only mode");
1656 res = MS_RDONLY;
1657 }
1658 if (read_only)
1659 return res;
1660 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1661 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1662 "running e2fsck is recommended");
1663 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1664 ext4_msg(sb, KERN_WARNING,
1665 "warning: mounting fs with errors, "
1666 "running e2fsck is recommended");
1667 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1668 le16_to_cpu(es->s_mnt_count) >=
1669 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1670 ext4_msg(sb, KERN_WARNING,
1671 "warning: maximal mount count reached, "
1672 "running e2fsck is recommended");
1673 else if (le32_to_cpu(es->s_checkinterval) &&
1674 (le32_to_cpu(es->s_lastcheck) +
1675 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1676 ext4_msg(sb, KERN_WARNING,
1677 "warning: checktime reached, "
1678 "running e2fsck is recommended");
1679 if (!sbi->s_journal)
1680 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1681 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1682 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1683 le16_add_cpu(&es->s_mnt_count, 1);
1684 es->s_mtime = cpu_to_le32(get_seconds());
1685 ext4_update_dynamic_rev(sb);
1686 if (sbi->s_journal)
1687 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1688
1689 ext4_commit_super(sb, 1);
1690 if (test_opt(sb, DEBUG))
1691 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1692 "bpg=%lu, ipg=%lu, mo=%04x]\n",
1693 sb->s_blocksize,
1694 sbi->s_groups_count,
1695 EXT4_BLOCKS_PER_GROUP(sb),
1696 EXT4_INODES_PER_GROUP(sb),
1697 sbi->s_mount_opt);
1698
1699 return res;
1700 }
1701
1702 static int ext4_fill_flex_info(struct super_block *sb)
1703 {
1704 struct ext4_sb_info *sbi = EXT4_SB(sb);
1705 struct ext4_group_desc *gdp = NULL;
1706 ext4_group_t flex_group_count;
1707 ext4_group_t flex_group;
1708 int groups_per_flex = 0;
1709 size_t size;
1710 int i;
1711
1712 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1713 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1714
1715 if (groups_per_flex < 2) {
1716 sbi->s_log_groups_per_flex = 0;
1717 return 1;
1718 }
1719
1720 /* We allocate both existing and potentially added groups */
1721 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1722 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1723 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1724 size = flex_group_count * sizeof(struct flex_groups);
1725 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1726 if (sbi->s_flex_groups == NULL) {
1727 sbi->s_flex_groups = vmalloc(size);
1728 if (sbi->s_flex_groups)
1729 memset(sbi->s_flex_groups, 0, size);
1730 }
1731 if (sbi->s_flex_groups == NULL) {
1732 ext4_msg(sb, KERN_ERR, "not enough memory for "
1733 "%u flex groups", flex_group_count);
1734 goto failed;
1735 }
1736
1737 for (i = 0; i < sbi->s_groups_count; i++) {
1738 gdp = ext4_get_group_desc(sb, i, NULL);
1739
1740 flex_group = ext4_flex_group(sbi, i);
1741 atomic_add(ext4_free_inodes_count(sb, gdp),
1742 &sbi->s_flex_groups[flex_group].free_inodes);
1743 atomic_add(ext4_free_blks_count(sb, gdp),
1744 &sbi->s_flex_groups[flex_group].free_blocks);
1745 atomic_add(ext4_used_dirs_count(sb, gdp),
1746 &sbi->s_flex_groups[flex_group].used_dirs);
1747 }
1748
1749 return 1;
1750 failed:
1751 return 0;
1752 }
1753
1754 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1755 struct ext4_group_desc *gdp)
1756 {
1757 __u16 crc = 0;
1758
1759 if (sbi->s_es->s_feature_ro_compat &
1760 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1761 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1762 __le32 le_group = cpu_to_le32(block_group);
1763
1764 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1765 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1766 crc = crc16(crc, (__u8 *)gdp, offset);
1767 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1768 /* for checksum of struct ext4_group_desc do the rest...*/
1769 if ((sbi->s_es->s_feature_incompat &
1770 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1771 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1772 crc = crc16(crc, (__u8 *)gdp + offset,
1773 le16_to_cpu(sbi->s_es->s_desc_size) -
1774 offset);
1775 }
1776
1777 return cpu_to_le16(crc);
1778 }
1779
1780 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1781 struct ext4_group_desc *gdp)
1782 {
1783 if ((sbi->s_es->s_feature_ro_compat &
1784 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1785 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1786 return 0;
1787
1788 return 1;
1789 }
1790
1791 /* Called at mount-time, super-block is locked */
1792 static int ext4_check_descriptors(struct super_block *sb)
1793 {
1794 struct ext4_sb_info *sbi = EXT4_SB(sb);
1795 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1796 ext4_fsblk_t last_block;
1797 ext4_fsblk_t block_bitmap;
1798 ext4_fsblk_t inode_bitmap;
1799 ext4_fsblk_t inode_table;
1800 int flexbg_flag = 0;
1801 ext4_group_t i;
1802
1803 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1804 flexbg_flag = 1;
1805
1806 ext4_debug("Checking group descriptors");
1807
1808 for (i = 0; i < sbi->s_groups_count; i++) {
1809 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1810
1811 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1812 last_block = ext4_blocks_count(sbi->s_es) - 1;
1813 else
1814 last_block = first_block +
1815 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1816
1817 block_bitmap = ext4_block_bitmap(sb, gdp);
1818 if (block_bitmap < first_block || block_bitmap > last_block) {
1819 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1820 "Block bitmap for group %u not in group "
1821 "(block %llu)!", i, block_bitmap);
1822 return 0;
1823 }
1824 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1825 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1826 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1827 "Inode bitmap for group %u not in group "
1828 "(block %llu)!", i, inode_bitmap);
1829 return 0;
1830 }
1831 inode_table = ext4_inode_table(sb, gdp);
1832 if (inode_table < first_block ||
1833 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1834 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1835 "Inode table for group %u not in group "
1836 "(block %llu)!", i, inode_table);
1837 return 0;
1838 }
1839 ext4_lock_group(sb, i);
1840 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1841 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1842 "Checksum for group %u failed (%u!=%u)",
1843 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1844 gdp)), le16_to_cpu(gdp->bg_checksum));
1845 if (!(sb->s_flags & MS_RDONLY)) {
1846 ext4_unlock_group(sb, i);
1847 return 0;
1848 }
1849 }
1850 ext4_unlock_group(sb, i);
1851 if (!flexbg_flag)
1852 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1853 }
1854
1855 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1856 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
1857 return 1;
1858 }
1859
1860 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1861 * the superblock) which were deleted from all directories, but held open by
1862 * a process at the time of a crash. We walk the list and try to delete these
1863 * inodes at recovery time (only with a read-write filesystem).
1864 *
1865 * In order to keep the orphan inode chain consistent during traversal (in
1866 * case of crash during recovery), we link each inode into the superblock
1867 * orphan list_head and handle it the same way as an inode deletion during
1868 * normal operation (which journals the operations for us).
1869 *
1870 * We only do an iget() and an iput() on each inode, which is very safe if we
1871 * accidentally point at an in-use or already deleted inode. The worst that
1872 * can happen in this case is that we get a "bit already cleared" message from
1873 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1874 * e2fsck was run on this filesystem, and it must have already done the orphan
1875 * inode cleanup for us, so we can safely abort without any further action.
1876 */
1877 static void ext4_orphan_cleanup(struct super_block *sb,
1878 struct ext4_super_block *es)
1879 {
1880 unsigned int s_flags = sb->s_flags;
1881 int nr_orphans = 0, nr_truncates = 0;
1882 #ifdef CONFIG_QUOTA
1883 int i;
1884 #endif
1885 if (!es->s_last_orphan) {
1886 jbd_debug(4, "no orphan inodes to clean up\n");
1887 return;
1888 }
1889
1890 if (bdev_read_only(sb->s_bdev)) {
1891 ext4_msg(sb, KERN_ERR, "write access "
1892 "unavailable, skipping orphan cleanup");
1893 return;
1894 }
1895
1896 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1897 if (es->s_last_orphan)
1898 jbd_debug(1, "Errors on filesystem, "
1899 "clearing orphan list.\n");
1900 es->s_last_orphan = 0;
1901 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1902 return;
1903 }
1904
1905 if (s_flags & MS_RDONLY) {
1906 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1907 sb->s_flags &= ~MS_RDONLY;
1908 }
1909 #ifdef CONFIG_QUOTA
1910 /* Needed for iput() to work correctly and not trash data */
1911 sb->s_flags |= MS_ACTIVE;
1912 /* Turn on quotas so that they are updated correctly */
1913 for (i = 0; i < MAXQUOTAS; i++) {
1914 if (EXT4_SB(sb)->s_qf_names[i]) {
1915 int ret = ext4_quota_on_mount(sb, i);
1916 if (ret < 0)
1917 ext4_msg(sb, KERN_ERR,
1918 "Cannot turn on journaled "
1919 "quota: error %d", ret);
1920 }
1921 }
1922 #endif
1923
1924 while (es->s_last_orphan) {
1925 struct inode *inode;
1926
1927 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1928 if (IS_ERR(inode)) {
1929 es->s_last_orphan = 0;
1930 break;
1931 }
1932
1933 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1934 vfs_dq_init(inode);
1935 if (inode->i_nlink) {
1936 ext4_msg(sb, KERN_DEBUG,
1937 "%s: truncating inode %lu to %lld bytes",
1938 __func__, inode->i_ino, inode->i_size);
1939 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1940 inode->i_ino, inode->i_size);
1941 ext4_truncate(inode);
1942 nr_truncates++;
1943 } else {
1944 ext4_msg(sb, KERN_DEBUG,
1945 "%s: deleting unreferenced inode %lu",
1946 __func__, inode->i_ino);
1947 jbd_debug(2, "deleting unreferenced inode %lu\n",
1948 inode->i_ino);
1949 nr_orphans++;
1950 }
1951 iput(inode); /* The delete magic happens here! */
1952 }
1953
1954 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1955
1956 if (nr_orphans)
1957 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1958 PLURAL(nr_orphans));
1959 if (nr_truncates)
1960 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1961 PLURAL(nr_truncates));
1962 #ifdef CONFIG_QUOTA
1963 /* Turn quotas off */
1964 for (i = 0; i < MAXQUOTAS; i++) {
1965 if (sb_dqopt(sb)->files[i])
1966 vfs_quota_off(sb, i, 0);
1967 }
1968 #endif
1969 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1970 }
1971
1972 /*
1973 * Maximal extent format file size.
1974 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1975 * extent format containers, within a sector_t, and within i_blocks
1976 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1977 * so that won't be a limiting factor.
1978 *
1979 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1980 */
1981 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1982 {
1983 loff_t res;
1984 loff_t upper_limit = MAX_LFS_FILESIZE;
1985
1986 /* small i_blocks in vfs inode? */
1987 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1988 /*
1989 * CONFIG_LBDAF is not enabled implies the inode
1990 * i_block represent total blocks in 512 bytes
1991 * 32 == size of vfs inode i_blocks * 8
1992 */
1993 upper_limit = (1LL << 32) - 1;
1994
1995 /* total blocks in file system block size */
1996 upper_limit >>= (blkbits - 9);
1997 upper_limit <<= blkbits;
1998 }
1999
2000 /* 32-bit extent-start container, ee_block */
2001 res = 1LL << 32;
2002 res <<= blkbits;
2003 res -= 1;
2004
2005 /* Sanity check against vm- & vfs- imposed limits */
2006 if (res > upper_limit)
2007 res = upper_limit;
2008
2009 return res;
2010 }
2011
2012 /*
2013 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2014 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2015 * We need to be 1 filesystem block less than the 2^48 sector limit.
2016 */
2017 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2018 {
2019 loff_t res = EXT4_NDIR_BLOCKS;
2020 int meta_blocks;
2021 loff_t upper_limit;
2022 /* This is calculated to be the largest file size for a dense, block
2023 * mapped file such that the file's total number of 512-byte sectors,
2024 * including data and all indirect blocks, does not exceed (2^48 - 1).
2025 *
2026 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2027 * number of 512-byte sectors of the file.
2028 */
2029
2030 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2031 /*
2032 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2033 * the inode i_block field represents total file blocks in
2034 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2035 */
2036 upper_limit = (1LL << 32) - 1;
2037
2038 /* total blocks in file system block size */
2039 upper_limit >>= (bits - 9);
2040
2041 } else {
2042 /*
2043 * We use 48 bit ext4_inode i_blocks
2044 * With EXT4_HUGE_FILE_FL set the i_blocks
2045 * represent total number of blocks in
2046 * file system block size
2047 */
2048 upper_limit = (1LL << 48) - 1;
2049
2050 }
2051
2052 /* indirect blocks */
2053 meta_blocks = 1;
2054 /* double indirect blocks */
2055 meta_blocks += 1 + (1LL << (bits-2));
2056 /* tripple indirect blocks */
2057 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2058
2059 upper_limit -= meta_blocks;
2060 upper_limit <<= bits;
2061
2062 res += 1LL << (bits-2);
2063 res += 1LL << (2*(bits-2));
2064 res += 1LL << (3*(bits-2));
2065 res <<= bits;
2066 if (res > upper_limit)
2067 res = upper_limit;
2068
2069 if (res > MAX_LFS_FILESIZE)
2070 res = MAX_LFS_FILESIZE;
2071
2072 return res;
2073 }
2074
2075 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2076 ext4_fsblk_t logical_sb_block, int nr)
2077 {
2078 struct ext4_sb_info *sbi = EXT4_SB(sb);
2079 ext4_group_t bg, first_meta_bg;
2080 int has_super = 0;
2081
2082 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2083
2084 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2085 nr < first_meta_bg)
2086 return logical_sb_block + nr + 1;
2087 bg = sbi->s_desc_per_block * nr;
2088 if (ext4_bg_has_super(sb, bg))
2089 has_super = 1;
2090
2091 return (has_super + ext4_group_first_block_no(sb, bg));
2092 }
2093
2094 /**
2095 * ext4_get_stripe_size: Get the stripe size.
2096 * @sbi: In memory super block info
2097 *
2098 * If we have specified it via mount option, then
2099 * use the mount option value. If the value specified at mount time is
2100 * greater than the blocks per group use the super block value.
2101 * If the super block value is greater than blocks per group return 0.
2102 * Allocator needs it be less than blocks per group.
2103 *
2104 */
2105 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2106 {
2107 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2108 unsigned long stripe_width =
2109 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2110
2111 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2112 return sbi->s_stripe;
2113
2114 if (stripe_width <= sbi->s_blocks_per_group)
2115 return stripe_width;
2116
2117 if (stride <= sbi->s_blocks_per_group)
2118 return stride;
2119
2120 return 0;
2121 }
2122
2123 /* sysfs supprt */
2124
2125 struct ext4_attr {
2126 struct attribute attr;
2127 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2128 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2129 const char *, size_t);
2130 int offset;
2131 };
2132
2133 static int parse_strtoul(const char *buf,
2134 unsigned long max, unsigned long *value)
2135 {
2136 char *endp;
2137
2138 while (*buf && isspace(*buf))
2139 buf++;
2140 *value = simple_strtoul(buf, &endp, 0);
2141 while (*endp && isspace(*endp))
2142 endp++;
2143 if (*endp || *value > max)
2144 return -EINVAL;
2145
2146 return 0;
2147 }
2148
2149 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2150 struct ext4_sb_info *sbi,
2151 char *buf)
2152 {
2153 return snprintf(buf, PAGE_SIZE, "%llu\n",
2154 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2155 }
2156
2157 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2158 struct ext4_sb_info *sbi, char *buf)
2159 {
2160 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2161
2162 return snprintf(buf, PAGE_SIZE, "%lu\n",
2163 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2164 sbi->s_sectors_written_start) >> 1);
2165 }
2166
2167 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2168 struct ext4_sb_info *sbi, char *buf)
2169 {
2170 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2171
2172 return snprintf(buf, PAGE_SIZE, "%llu\n",
2173 sbi->s_kbytes_written +
2174 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2175 EXT4_SB(sb)->s_sectors_written_start) >> 1));
2176 }
2177
2178 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2179 struct ext4_sb_info *sbi,
2180 const char *buf, size_t count)
2181 {
2182 unsigned long t;
2183
2184 if (parse_strtoul(buf, 0x40000000, &t))
2185 return -EINVAL;
2186
2187 if (!is_power_of_2(t))
2188 return -EINVAL;
2189
2190 sbi->s_inode_readahead_blks = t;
2191 return count;
2192 }
2193
2194 static ssize_t sbi_ui_show(struct ext4_attr *a,
2195 struct ext4_sb_info *sbi, char *buf)
2196 {
2197 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2198
2199 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2200 }
2201
2202 static ssize_t sbi_ui_store(struct ext4_attr *a,
2203 struct ext4_sb_info *sbi,
2204 const char *buf, size_t count)
2205 {
2206 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2207 unsigned long t;
2208
2209 if (parse_strtoul(buf, 0xffffffff, &t))
2210 return -EINVAL;
2211 *ui = t;
2212 return count;
2213 }
2214
2215 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2216 static struct ext4_attr ext4_attr_##_name = { \
2217 .attr = {.name = __stringify(_name), .mode = _mode }, \
2218 .show = _show, \
2219 .store = _store, \
2220 .offset = offsetof(struct ext4_sb_info, _elname), \
2221 }
2222 #define EXT4_ATTR(name, mode, show, store) \
2223 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2224
2225 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2226 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2227 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2228 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2229 #define ATTR_LIST(name) &ext4_attr_##name.attr
2230
2231 EXT4_RO_ATTR(delayed_allocation_blocks);
2232 EXT4_RO_ATTR(session_write_kbytes);
2233 EXT4_RO_ATTR(lifetime_write_kbytes);
2234 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2235 inode_readahead_blks_store, s_inode_readahead_blks);
2236 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2237 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2238 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2239 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2240 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2241 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2242 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2243 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2244
2245 static struct attribute *ext4_attrs[] = {
2246 ATTR_LIST(delayed_allocation_blocks),
2247 ATTR_LIST(session_write_kbytes),
2248 ATTR_LIST(lifetime_write_kbytes),
2249 ATTR_LIST(inode_readahead_blks),
2250 ATTR_LIST(inode_goal),
2251 ATTR_LIST(mb_stats),
2252 ATTR_LIST(mb_max_to_scan),
2253 ATTR_LIST(mb_min_to_scan),
2254 ATTR_LIST(mb_order2_req),
2255 ATTR_LIST(mb_stream_req),
2256 ATTR_LIST(mb_group_prealloc),
2257 ATTR_LIST(max_writeback_mb_bump),
2258 NULL,
2259 };
2260
2261 static ssize_t ext4_attr_show(struct kobject *kobj,
2262 struct attribute *attr, char *buf)
2263 {
2264 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2265 s_kobj);
2266 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2267
2268 return a->show ? a->show(a, sbi, buf) : 0;
2269 }
2270
2271 static ssize_t ext4_attr_store(struct kobject *kobj,
2272 struct attribute *attr,
2273 const char *buf, size_t len)
2274 {
2275 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2276 s_kobj);
2277 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2278
2279 return a->store ? a->store(a, sbi, buf, len) : 0;
2280 }
2281
2282 static void ext4_sb_release(struct kobject *kobj)
2283 {
2284 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2285 s_kobj);
2286 complete(&sbi->s_kobj_unregister);
2287 }
2288
2289
2290 static struct sysfs_ops ext4_attr_ops = {
2291 .show = ext4_attr_show,
2292 .store = ext4_attr_store,
2293 };
2294
2295 static struct kobj_type ext4_ktype = {
2296 .default_attrs = ext4_attrs,
2297 .sysfs_ops = &ext4_attr_ops,
2298 .release = ext4_sb_release,
2299 };
2300
2301 /*
2302 * Check whether this filesystem can be mounted based on
2303 * the features present and the RDONLY/RDWR mount requested.
2304 * Returns 1 if this filesystem can be mounted as requested,
2305 * 0 if it cannot be.
2306 */
2307 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2308 {
2309 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2310 ext4_msg(sb, KERN_ERR,
2311 "Couldn't mount because of "
2312 "unsupported optional features (%x)",
2313 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2314 ~EXT4_FEATURE_INCOMPAT_SUPP));
2315 return 0;
2316 }
2317
2318 if (readonly)
2319 return 1;
2320
2321 /* Check that feature set is OK for a read-write mount */
2322 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2323 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2324 "unsupported optional features (%x)",
2325 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2326 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2327 return 0;
2328 }
2329 /*
2330 * Large file size enabled file system can only be mounted
2331 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2332 */
2333 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2334 if (sizeof(blkcnt_t) < sizeof(u64)) {
2335 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2336 "cannot be mounted RDWR without "
2337 "CONFIG_LBDAF");
2338 return 0;
2339 }
2340 }
2341 return 1;
2342 }
2343
2344 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2345 __releases(kernel_lock)
2346 __acquires(kernel_lock)
2347 {
2348 struct buffer_head *bh;
2349 struct ext4_super_block *es = NULL;
2350 struct ext4_sb_info *sbi;
2351 ext4_fsblk_t block;
2352 ext4_fsblk_t sb_block = get_sb_block(&data);
2353 ext4_fsblk_t logical_sb_block;
2354 unsigned long offset = 0;
2355 unsigned long journal_devnum = 0;
2356 unsigned long def_mount_opts;
2357 struct inode *root;
2358 char *cp;
2359 const char *descr;
2360 int ret = -EINVAL;
2361 int blocksize;
2362 unsigned int db_count;
2363 unsigned int i;
2364 int needs_recovery, has_huge_files;
2365 __u64 blocks_count;
2366 int err;
2367 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2368
2369 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2370 if (!sbi)
2371 return -ENOMEM;
2372
2373 sbi->s_blockgroup_lock =
2374 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2375 if (!sbi->s_blockgroup_lock) {
2376 kfree(sbi);
2377 return -ENOMEM;
2378 }
2379 sb->s_fs_info = sbi;
2380 sbi->s_mount_opt = 0;
2381 sbi->s_resuid = EXT4_DEF_RESUID;
2382 sbi->s_resgid = EXT4_DEF_RESGID;
2383 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2384 sbi->s_sb_block = sb_block;
2385 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2386 sectors[1]);
2387
2388 unlock_kernel();
2389
2390 /* Cleanup superblock name */
2391 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2392 *cp = '!';
2393
2394 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2395 if (!blocksize) {
2396 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
2397 goto out_fail;
2398 }
2399
2400 /*
2401 * The ext4 superblock will not be buffer aligned for other than 1kB
2402 * block sizes. We need to calculate the offset from buffer start.
2403 */
2404 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2405 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2406 offset = do_div(logical_sb_block, blocksize);
2407 } else {
2408 logical_sb_block = sb_block;
2409 }
2410
2411 if (!(bh = sb_bread(sb, logical_sb_block))) {
2412 ext4_msg(sb, KERN_ERR, "unable to read superblock");
2413 goto out_fail;
2414 }
2415 /*
2416 * Note: s_es must be initialized as soon as possible because
2417 * some ext4 macro-instructions depend on its value
2418 */
2419 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2420 sbi->s_es = es;
2421 sb->s_magic = le16_to_cpu(es->s_magic);
2422 if (sb->s_magic != EXT4_SUPER_MAGIC)
2423 goto cantfind_ext4;
2424 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2425
2426 /* Set defaults before we parse the mount options */
2427 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2428 if (def_mount_opts & EXT4_DEFM_DEBUG)
2429 set_opt(sbi->s_mount_opt, DEBUG);
2430 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
2431 set_opt(sbi->s_mount_opt, GRPID);
2432 if (def_mount_opts & EXT4_DEFM_UID16)
2433 set_opt(sbi->s_mount_opt, NO_UID32);
2434 #ifdef CONFIG_EXT4_FS_XATTR
2435 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2436 set_opt(sbi->s_mount_opt, XATTR_USER);
2437 #endif
2438 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2439 if (def_mount_opts & EXT4_DEFM_ACL)
2440 set_opt(sbi->s_mount_opt, POSIX_ACL);
2441 #endif
2442 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2443 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2444 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2445 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2446 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2447 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2448
2449 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2450 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2451 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2452 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2453 else
2454 set_opt(sbi->s_mount_opt, ERRORS_RO);
2455
2456 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2457 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2458 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2459 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2460 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2461
2462 set_opt(sbi->s_mount_opt, BARRIER);
2463
2464 /*
2465 * enable delayed allocation by default
2466 * Use -o nodelalloc to turn it off
2467 */
2468 set_opt(sbi->s_mount_opt, DELALLOC);
2469
2470 if (!parse_options((char *) data, sb, &journal_devnum,
2471 &journal_ioprio, NULL, 0))
2472 goto failed_mount;
2473
2474 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2475 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2476
2477 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2478 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2479 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2480 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2481 ext4_msg(sb, KERN_WARNING,
2482 "feature flags set on rev 0 fs, "
2483 "running e2fsck is recommended");
2484
2485 /*
2486 * Check feature flags regardless of the revision level, since we
2487 * previously didn't change the revision level when setting the flags,
2488 * so there is a chance incompat flags are set on a rev 0 filesystem.
2489 */
2490 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
2491 goto failed_mount;
2492
2493 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2494
2495 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2496 blocksize > EXT4_MAX_BLOCK_SIZE) {
2497 ext4_msg(sb, KERN_ERR,
2498 "Unsupported filesystem blocksize %d", blocksize);
2499 goto failed_mount;
2500 }
2501
2502 if (sb->s_blocksize != blocksize) {
2503 /* Validate the filesystem blocksize */
2504 if (!sb_set_blocksize(sb, blocksize)) {
2505 ext4_msg(sb, KERN_ERR, "bad block size %d",
2506 blocksize);
2507 goto failed_mount;
2508 }
2509
2510 brelse(bh);
2511 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2512 offset = do_div(logical_sb_block, blocksize);
2513 bh = sb_bread(sb, logical_sb_block);
2514 if (!bh) {
2515 ext4_msg(sb, KERN_ERR,
2516 "Can't read superblock on 2nd try");
2517 goto failed_mount;
2518 }
2519 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2520 sbi->s_es = es;
2521 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2522 ext4_msg(sb, KERN_ERR,
2523 "Magic mismatch, very weird!");
2524 goto failed_mount;
2525 }
2526 }
2527
2528 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2529 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2530 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2531 has_huge_files);
2532 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2533
2534 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2535 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2536 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2537 } else {
2538 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2539 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2540 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2541 (!is_power_of_2(sbi->s_inode_size)) ||
2542 (sbi->s_inode_size > blocksize)) {
2543 ext4_msg(sb, KERN_ERR,
2544 "unsupported inode size: %d",
2545 sbi->s_inode_size);
2546 goto failed_mount;
2547 }
2548 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2549 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2550 }
2551
2552 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2553 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2554 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2555 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2556 !is_power_of_2(sbi->s_desc_size)) {
2557 ext4_msg(sb, KERN_ERR,
2558 "unsupported descriptor size %lu",
2559 sbi->s_desc_size);
2560 goto failed_mount;
2561 }
2562 } else
2563 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2564
2565 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2566 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2567 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2568 goto cantfind_ext4;
2569
2570 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2571 if (sbi->s_inodes_per_block == 0)
2572 goto cantfind_ext4;
2573 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2574 sbi->s_inodes_per_block;
2575 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2576 sbi->s_sbh = bh;
2577 sbi->s_mount_state = le16_to_cpu(es->s_state);
2578 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2579 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2580
2581 for (i = 0; i < 4; i++)
2582 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2583 sbi->s_def_hash_version = es->s_def_hash_version;
2584 i = le32_to_cpu(es->s_flags);
2585 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2586 sbi->s_hash_unsigned = 3;
2587 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2588 #ifdef __CHAR_UNSIGNED__
2589 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2590 sbi->s_hash_unsigned = 3;
2591 #else
2592 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2593 #endif
2594 sb->s_dirt = 1;
2595 }
2596
2597 if (sbi->s_blocks_per_group > blocksize * 8) {
2598 ext4_msg(sb, KERN_ERR,
2599 "#blocks per group too big: %lu",
2600 sbi->s_blocks_per_group);
2601 goto failed_mount;
2602 }
2603 if (sbi->s_inodes_per_group > blocksize * 8) {
2604 ext4_msg(sb, KERN_ERR,
2605 "#inodes per group too big: %lu",
2606 sbi->s_inodes_per_group);
2607 goto failed_mount;
2608 }
2609
2610 /*
2611 * Test whether we have more sectors than will fit in sector_t,
2612 * and whether the max offset is addressable by the page cache.
2613 */
2614 if ((ext4_blocks_count(es) >
2615 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2616 (ext4_blocks_count(es) >
2617 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
2618 ext4_msg(sb, KERN_ERR, "filesystem"
2619 " too large to mount safely on this system");
2620 if (sizeof(sector_t) < 8)
2621 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
2622 ret = -EFBIG;
2623 goto failed_mount;
2624 }
2625
2626 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2627 goto cantfind_ext4;
2628
2629 /* check blocks count against device size */
2630 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2631 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2632 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2633 "exceeds size of device (%llu blocks)",
2634 ext4_blocks_count(es), blocks_count);
2635 goto failed_mount;
2636 }
2637
2638 /*
2639 * It makes no sense for the first data block to be beyond the end
2640 * of the filesystem.
2641 */
2642 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2643 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2644 "block %u is beyond end of filesystem (%llu)",
2645 le32_to_cpu(es->s_first_data_block),
2646 ext4_blocks_count(es));
2647 goto failed_mount;
2648 }
2649 blocks_count = (ext4_blocks_count(es) -
2650 le32_to_cpu(es->s_first_data_block) +
2651 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2652 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2653 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2654 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
2655 "(block count %llu, first data block %u, "
2656 "blocks per group %lu)", sbi->s_groups_count,
2657 ext4_blocks_count(es),
2658 le32_to_cpu(es->s_first_data_block),
2659 EXT4_BLOCKS_PER_GROUP(sb));
2660 goto failed_mount;
2661 }
2662 sbi->s_groups_count = blocks_count;
2663 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
2664 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
2665 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2666 EXT4_DESC_PER_BLOCK(sb);
2667 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2668 GFP_KERNEL);
2669 if (sbi->s_group_desc == NULL) {
2670 ext4_msg(sb, KERN_ERR, "not enough memory");
2671 goto failed_mount;
2672 }
2673
2674 #ifdef CONFIG_PROC_FS
2675 if (ext4_proc_root)
2676 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2677 #endif
2678
2679 bgl_lock_init(sbi->s_blockgroup_lock);
2680
2681 for (i = 0; i < db_count; i++) {
2682 block = descriptor_loc(sb, logical_sb_block, i);
2683 sbi->s_group_desc[i] = sb_bread(sb, block);
2684 if (!sbi->s_group_desc[i]) {
2685 ext4_msg(sb, KERN_ERR,
2686 "can't read group descriptor %d", i);
2687 db_count = i;
2688 goto failed_mount2;
2689 }
2690 }
2691 if (!ext4_check_descriptors(sb)) {
2692 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
2693 goto failed_mount2;
2694 }
2695 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2696 if (!ext4_fill_flex_info(sb)) {
2697 ext4_msg(sb, KERN_ERR,
2698 "unable to initialize "
2699 "flex_bg meta info!");
2700 goto failed_mount2;
2701 }
2702
2703 sbi->s_gdb_count = db_count;
2704 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2705 spin_lock_init(&sbi->s_next_gen_lock);
2706
2707 sbi->s_stripe = ext4_get_stripe_size(sbi);
2708 sbi->s_max_writeback_mb_bump = 128;
2709
2710 /*
2711 * set up enough so that it can read an inode
2712 */
2713 if (!test_opt(sb, NOLOAD) &&
2714 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2715 sb->s_op = &ext4_sops;
2716 else
2717 sb->s_op = &ext4_nojournal_sops;
2718 sb->s_export_op = &ext4_export_ops;
2719 sb->s_xattr = ext4_xattr_handlers;
2720 #ifdef CONFIG_QUOTA
2721 sb->s_qcop = &ext4_qctl_operations;
2722 sb->dq_op = &ext4_quota_operations;
2723 #endif
2724 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2725 mutex_init(&sbi->s_orphan_lock);
2726 mutex_init(&sbi->s_resize_lock);
2727
2728 sb->s_root = NULL;
2729
2730 needs_recovery = (es->s_last_orphan != 0 ||
2731 EXT4_HAS_INCOMPAT_FEATURE(sb,
2732 EXT4_FEATURE_INCOMPAT_RECOVER));
2733
2734 /*
2735 * The first inode we look at is the journal inode. Don't try
2736 * root first: it may be modified in the journal!
2737 */
2738 if (!test_opt(sb, NOLOAD) &&
2739 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2740 if (ext4_load_journal(sb, es, journal_devnum))
2741 goto failed_mount3;
2742 if (!(sb->s_flags & MS_RDONLY) &&
2743 EXT4_SB(sb)->s_journal->j_failed_commit) {
2744 ext4_msg(sb, KERN_CRIT, "error: "
2745 "ext4_fill_super: Journal transaction "
2746 "%u is corrupt",
2747 EXT4_SB(sb)->s_journal->j_failed_commit);
2748 if (test_opt(sb, ERRORS_RO)) {
2749 ext4_msg(sb, KERN_CRIT,
2750 "Mounting filesystem read-only");
2751 sb->s_flags |= MS_RDONLY;
2752 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2753 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2754 }
2755 if (test_opt(sb, ERRORS_PANIC)) {
2756 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2757 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2758 ext4_commit_super(sb, 1);
2759 goto failed_mount4;
2760 }
2761 }
2762 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2763 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2764 ext4_msg(sb, KERN_ERR, "required journal recovery "
2765 "suppressed and not mounted read-only");
2766 goto failed_mount4;
2767 } else {
2768 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2769 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2770 sbi->s_journal = NULL;
2771 needs_recovery = 0;
2772 goto no_journal;
2773 }
2774
2775 if (ext4_blocks_count(es) > 0xffffffffULL &&
2776 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2777 JBD2_FEATURE_INCOMPAT_64BIT)) {
2778 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
2779 goto failed_mount4;
2780 }
2781
2782 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2783 jbd2_journal_set_features(sbi->s_journal,
2784 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2785 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2786 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2787 jbd2_journal_set_features(sbi->s_journal,
2788 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2789 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2790 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2791 } else {
2792 jbd2_journal_clear_features(sbi->s_journal,
2793 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2794 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2795 }
2796
2797 /* We have now updated the journal if required, so we can
2798 * validate the data journaling mode. */
2799 switch (test_opt(sb, DATA_FLAGS)) {
2800 case 0:
2801 /* No mode set, assume a default based on the journal
2802 * capabilities: ORDERED_DATA if the journal can
2803 * cope, else JOURNAL_DATA
2804 */
2805 if (jbd2_journal_check_available_features
2806 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2807 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2808 else
2809 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2810 break;
2811
2812 case EXT4_MOUNT_ORDERED_DATA:
2813 case EXT4_MOUNT_WRITEBACK_DATA:
2814 if (!jbd2_journal_check_available_features
2815 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2816 ext4_msg(sb, KERN_ERR, "Journal does not support "
2817 "requested data journaling mode");
2818 goto failed_mount4;
2819 }
2820 default:
2821 break;
2822 }
2823 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2824
2825 no_journal:
2826 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2827 ext4_count_free_blocks(sb));
2828 if (!err)
2829 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2830 ext4_count_free_inodes(sb));
2831 if (!err)
2832 err = percpu_counter_init(&sbi->s_dirs_counter,
2833 ext4_count_dirs(sb));
2834 if (!err)
2835 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2836 if (err) {
2837 ext4_msg(sb, KERN_ERR, "insufficient memory");
2838 goto failed_mount_wq;
2839 }
2840 if (test_opt(sb, NOBH)) {
2841 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2842 ext4_msg(sb, KERN_WARNING, "Ignoring nobh option - "
2843 "its supported only with writeback mode");
2844 clear_opt(sbi->s_mount_opt, NOBH);
2845 }
2846 }
2847 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
2848 if (!EXT4_SB(sb)->dio_unwritten_wq) {
2849 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
2850 goto failed_mount_wq;
2851 }
2852
2853 /*
2854 * The jbd2_journal_load will have done any necessary log recovery,
2855 * so we can safely mount the rest of the filesystem now.
2856 */
2857
2858 root = ext4_iget(sb, EXT4_ROOT_INO);
2859 if (IS_ERR(root)) {
2860 ext4_msg(sb, KERN_ERR, "get root inode failed");
2861 ret = PTR_ERR(root);
2862 goto failed_mount4;
2863 }
2864 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2865 iput(root);
2866 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
2867 goto failed_mount4;
2868 }
2869 sb->s_root = d_alloc_root(root);
2870 if (!sb->s_root) {
2871 ext4_msg(sb, KERN_ERR, "get root dentry failed");
2872 iput(root);
2873 ret = -ENOMEM;
2874 goto failed_mount4;
2875 }
2876
2877 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2878
2879 /* determine the minimum size of new large inodes, if present */
2880 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2881 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2882 EXT4_GOOD_OLD_INODE_SIZE;
2883 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2884 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2885 if (sbi->s_want_extra_isize <
2886 le16_to_cpu(es->s_want_extra_isize))
2887 sbi->s_want_extra_isize =
2888 le16_to_cpu(es->s_want_extra_isize);
2889 if (sbi->s_want_extra_isize <
2890 le16_to_cpu(es->s_min_extra_isize))
2891 sbi->s_want_extra_isize =
2892 le16_to_cpu(es->s_min_extra_isize);
2893 }
2894 }
2895 /* Check if enough inode space is available */
2896 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2897 sbi->s_inode_size) {
2898 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2899 EXT4_GOOD_OLD_INODE_SIZE;
2900 ext4_msg(sb, KERN_INFO, "required extra inode space not"
2901 "available");
2902 }
2903
2904 if (test_opt(sb, DELALLOC) &&
2905 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
2906 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
2907 "requested data journaling mode");
2908 clear_opt(sbi->s_mount_opt, DELALLOC);
2909 }
2910
2911 err = ext4_setup_system_zone(sb);
2912 if (err) {
2913 ext4_msg(sb, KERN_ERR, "failed to initialize system "
2914 "zone (%d)", err);
2915 goto failed_mount4;
2916 }
2917
2918 ext4_ext_init(sb);
2919 err = ext4_mb_init(sb, needs_recovery);
2920 if (err) {
2921 ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
2922 err);
2923 goto failed_mount4;
2924 }
2925
2926 sbi->s_kobj.kset = ext4_kset;
2927 init_completion(&sbi->s_kobj_unregister);
2928 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2929 "%s", sb->s_id);
2930 if (err) {
2931 ext4_mb_release(sb);
2932 ext4_ext_release(sb);
2933 goto failed_mount4;
2934 };
2935
2936 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2937 ext4_orphan_cleanup(sb, es);
2938 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2939 if (needs_recovery) {
2940 ext4_msg(sb, KERN_INFO, "recovery complete");
2941 ext4_mark_recovery_complete(sb, es);
2942 }
2943 if (EXT4_SB(sb)->s_journal) {
2944 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2945 descr = " journalled data mode";
2946 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2947 descr = " ordered data mode";
2948 else
2949 descr = " writeback data mode";
2950 } else
2951 descr = "out journal";
2952
2953 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s", descr);
2954
2955 lock_kernel();
2956 return 0;
2957
2958 cantfind_ext4:
2959 if (!silent)
2960 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
2961 goto failed_mount;
2962
2963 failed_mount4:
2964 ext4_msg(sb, KERN_ERR, "mount failed");
2965 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
2966 failed_mount_wq:
2967 ext4_release_system_zone(sb);
2968 if (sbi->s_journal) {
2969 jbd2_journal_destroy(sbi->s_journal);
2970 sbi->s_journal = NULL;
2971 }
2972 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2973 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2974 percpu_counter_destroy(&sbi->s_dirs_counter);
2975 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2976 failed_mount3:
2977 if (sbi->s_flex_groups) {
2978 if (is_vmalloc_addr(sbi->s_flex_groups))
2979 vfree(sbi->s_flex_groups);
2980 else
2981 kfree(sbi->s_flex_groups);
2982 }
2983 failed_mount2:
2984 for (i = 0; i < db_count; i++)
2985 brelse(sbi->s_group_desc[i]);
2986 kfree(sbi->s_group_desc);
2987 failed_mount:
2988 if (sbi->s_proc) {
2989 remove_proc_entry(sb->s_id, ext4_proc_root);
2990 }
2991 #ifdef CONFIG_QUOTA
2992 for (i = 0; i < MAXQUOTAS; i++)
2993 kfree(sbi->s_qf_names[i]);
2994 #endif
2995 ext4_blkdev_remove(sbi);
2996 brelse(bh);
2997 out_fail:
2998 sb->s_fs_info = NULL;
2999 kfree(sbi->s_blockgroup_lock);
3000 kfree(sbi);
3001 lock_kernel();
3002 return ret;
3003 }
3004
3005 /*
3006 * Setup any per-fs journal parameters now. We'll do this both on
3007 * initial mount, once the journal has been initialised but before we've
3008 * done any recovery; and again on any subsequent remount.
3009 */
3010 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3011 {
3012 struct ext4_sb_info *sbi = EXT4_SB(sb);
3013
3014 journal->j_commit_interval = sbi->s_commit_interval;
3015 journal->j_min_batch_time = sbi->s_min_batch_time;
3016 journal->j_max_batch_time = sbi->s_max_batch_time;
3017
3018 spin_lock(&journal->j_state_lock);
3019 if (test_opt(sb, BARRIER))
3020 journal->j_flags |= JBD2_BARRIER;
3021 else
3022 journal->j_flags &= ~JBD2_BARRIER;
3023 if (test_opt(sb, DATA_ERR_ABORT))
3024 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3025 else
3026 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3027 spin_unlock(&journal->j_state_lock);
3028 }
3029
3030 static journal_t *ext4_get_journal(struct super_block *sb,
3031 unsigned int journal_inum)
3032 {
3033 struct inode *journal_inode;
3034 journal_t *journal;
3035
3036 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3037
3038 /* First, test for the existence of a valid inode on disk. Bad
3039 * things happen if we iget() an unused inode, as the subsequent
3040 * iput() will try to delete it. */
3041
3042 journal_inode = ext4_iget(sb, journal_inum);
3043 if (IS_ERR(journal_inode)) {
3044 ext4_msg(sb, KERN_ERR, "no journal found");
3045 return NULL;
3046 }
3047 if (!journal_inode->i_nlink) {
3048 make_bad_inode(journal_inode);
3049 iput(journal_inode);
3050 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3051 return NULL;
3052 }
3053
3054 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3055 journal_inode, journal_inode->i_size);
3056 if (!S_ISREG(journal_inode->i_mode)) {
3057 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3058 iput(journal_inode);
3059 return NULL;
3060 }
3061
3062 journal = jbd2_journal_init_inode(journal_inode);
3063 if (!journal) {
3064 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3065 iput(journal_inode);
3066 return NULL;
3067 }
3068 journal->j_private = sb;
3069 ext4_init_journal_params(sb, journal);
3070 return journal;
3071 }
3072
3073 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3074 dev_t j_dev)
3075 {
3076 struct buffer_head *bh;
3077 journal_t *journal;
3078 ext4_fsblk_t start;
3079 ext4_fsblk_t len;
3080 int hblock, blocksize;
3081 ext4_fsblk_t sb_block;
3082 unsigned long offset;
3083 struct ext4_super_block *es;
3084 struct block_device *bdev;
3085
3086 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3087
3088 bdev = ext4_blkdev_get(j_dev, sb);
3089 if (bdev == NULL)
3090 return NULL;
3091
3092 if (bd_claim(bdev, sb)) {
3093 ext4_msg(sb, KERN_ERR,
3094 "failed to claim external journal device");
3095 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3096 return NULL;
3097 }
3098
3099 blocksize = sb->s_blocksize;
3100 hblock = bdev_logical_block_size(bdev);
3101 if (blocksize < hblock) {
3102 ext4_msg(sb, KERN_ERR,
3103 "blocksize too small for journal device");
3104 goto out_bdev;
3105 }
3106
3107 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3108 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3109 set_blocksize(bdev, blocksize);
3110 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3111 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3112 "external journal");
3113 goto out_bdev;
3114 }
3115
3116 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3117 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3118 !(le32_to_cpu(es->s_feature_incompat) &
3119 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3120 ext4_msg(sb, KERN_ERR, "external journal has "
3121 "bad superblock");
3122 brelse(bh);
3123 goto out_bdev;
3124 }
3125
3126 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3127 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3128 brelse(bh);
3129 goto out_bdev;
3130 }
3131
3132 len = ext4_blocks_count(es);
3133 start = sb_block + 1;
3134 brelse(bh); /* we're done with the superblock */
3135
3136 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3137 start, len, blocksize);
3138 if (!journal) {
3139 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3140 goto out_bdev;
3141 }
3142 journal->j_private = sb;
3143 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3144 wait_on_buffer(journal->j_sb_buffer);
3145 if (!buffer_uptodate(journal->j_sb_buffer)) {
3146 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3147 goto out_journal;
3148 }
3149 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3150 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3151 "user (unsupported) - %d",
3152 be32_to_cpu(journal->j_superblock->s_nr_users));
3153 goto out_journal;
3154 }
3155 EXT4_SB(sb)->journal_bdev = bdev;
3156 ext4_init_journal_params(sb, journal);
3157 return journal;
3158
3159 out_journal:
3160 jbd2_journal_destroy(journal);
3161 out_bdev:
3162 ext4_blkdev_put(bdev);
3163 return NULL;
3164 }
3165
3166 static int ext4_load_journal(struct super_block *sb,
3167 struct ext4_super_block *es,
3168 unsigned long journal_devnum)
3169 {
3170 journal_t *journal;
3171 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3172 dev_t journal_dev;
3173 int err = 0;
3174 int really_read_only;
3175
3176 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3177
3178 if (journal_devnum &&
3179 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3180 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3181 "numbers have changed");
3182 journal_dev = new_decode_dev(journal_devnum);
3183 } else
3184 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3185
3186 really_read_only = bdev_read_only(sb->s_bdev);
3187
3188 /*
3189 * Are we loading a blank journal or performing recovery after a
3190 * crash? For recovery, we need to check in advance whether we
3191 * can get read-write access to the device.
3192 */
3193 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3194 if (sb->s_flags & MS_RDONLY) {
3195 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3196 "required on readonly filesystem");
3197 if (really_read_only) {
3198 ext4_msg(sb, KERN_ERR, "write access "
3199 "unavailable, cannot proceed");
3200 return -EROFS;
3201 }
3202 ext4_msg(sb, KERN_INFO, "write access will "
3203 "be enabled during recovery");
3204 }
3205 }
3206
3207 if (journal_inum && journal_dev) {
3208 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3209 "and inode journals!");
3210 return -EINVAL;
3211 }
3212
3213 if (journal_inum) {
3214 if (!(journal = ext4_get_journal(sb, journal_inum)))
3215 return -EINVAL;
3216 } else {
3217 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3218 return -EINVAL;
3219 }
3220
3221 if (!(journal->j_flags & JBD2_BARRIER))
3222 ext4_msg(sb, KERN_INFO, "barriers disabled");
3223
3224 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3225 err = jbd2_journal_update_format(journal);
3226 if (err) {
3227 ext4_msg(sb, KERN_ERR, "error updating journal");
3228 jbd2_journal_destroy(journal);
3229 return err;
3230 }
3231 }
3232
3233 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3234 err = jbd2_journal_wipe(journal, !really_read_only);
3235 if (!err)
3236 err = jbd2_journal_load(journal);
3237
3238 if (err) {
3239 ext4_msg(sb, KERN_ERR, "error loading journal");
3240 jbd2_journal_destroy(journal);
3241 return err;
3242 }
3243
3244 EXT4_SB(sb)->s_journal = journal;
3245 ext4_clear_journal_err(sb, es);
3246
3247 if (journal_devnum &&
3248 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3249 es->s_journal_dev = cpu_to_le32(journal_devnum);
3250
3251 /* Make sure we flush the recovery flag to disk. */
3252 ext4_commit_super(sb, 1);
3253 }
3254
3255 return 0;
3256 }
3257
3258 static int ext4_commit_super(struct super_block *sb, int sync)
3259 {
3260 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3261 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3262 int error = 0;
3263
3264 if (!sbh)
3265 return error;
3266 if (buffer_write_io_error(sbh)) {
3267 /*
3268 * Oh, dear. A previous attempt to write the
3269 * superblock failed. This could happen because the
3270 * USB device was yanked out. Or it could happen to
3271 * be a transient write error and maybe the block will
3272 * be remapped. Nothing we can do but to retry the
3273 * write and hope for the best.
3274 */
3275 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3276 "superblock detected");
3277 clear_buffer_write_io_error(sbh);
3278 set_buffer_uptodate(sbh);
3279 }
3280 /*
3281 * If the file system is mounted read-only, don't update the
3282 * superblock write time. This avoids updating the superblock
3283 * write time when we are mounting the root file system
3284 * read/only but we need to replay the journal; at that point,
3285 * for people who are east of GMT and who make their clock
3286 * tick in localtime for Windows bug-for-bug compatibility,
3287 * the clock is set in the future, and this will cause e2fsck
3288 * to complain and force a full file system check.
3289 */
3290 if (!(sb->s_flags & MS_RDONLY))
3291 es->s_wtime = cpu_to_le32(get_seconds());
3292 es->s_kbytes_written =
3293 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3294 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3295 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3296 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3297 &EXT4_SB(sb)->s_freeblocks_counter));
3298 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3299 &EXT4_SB(sb)->s_freeinodes_counter));
3300 sb->s_dirt = 0;
3301 BUFFER_TRACE(sbh, "marking dirty");
3302 mark_buffer_dirty(sbh);
3303 if (sync) {
3304 error = sync_dirty_buffer(sbh);
3305 if (error)
3306 return error;
3307
3308 error = buffer_write_io_error(sbh);
3309 if (error) {
3310 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3311 "superblock");
3312 clear_buffer_write_io_error(sbh);
3313 set_buffer_uptodate(sbh);
3314 }
3315 }
3316 return error;
3317 }
3318
3319 /*
3320 * Have we just finished recovery? If so, and if we are mounting (or
3321 * remounting) the filesystem readonly, then we will end up with a
3322 * consistent fs on disk. Record that fact.
3323 */
3324 static void ext4_mark_recovery_complete(struct super_block *sb,
3325 struct ext4_super_block *es)
3326 {
3327 journal_t *journal = EXT4_SB(sb)->s_journal;
3328
3329 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3330 BUG_ON(journal != NULL);
3331 return;
3332 }
3333 jbd2_journal_lock_updates(journal);
3334 if (jbd2_journal_flush(journal) < 0)
3335 goto out;
3336
3337 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3338 sb->s_flags & MS_RDONLY) {
3339 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3340 ext4_commit_super(sb, 1);
3341 }
3342
3343 out:
3344 jbd2_journal_unlock_updates(journal);
3345 }
3346
3347 /*
3348 * If we are mounting (or read-write remounting) a filesystem whose journal
3349 * has recorded an error from a previous lifetime, move that error to the
3350 * main filesystem now.
3351 */
3352 static void ext4_clear_journal_err(struct super_block *sb,
3353 struct ext4_super_block *es)
3354 {
3355 journal_t *journal;
3356 int j_errno;
3357 const char *errstr;
3358
3359 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3360
3361 journal = EXT4_SB(sb)->s_journal;
3362
3363 /*
3364 * Now check for any error status which may have been recorded in the
3365 * journal by a prior ext4_error() or ext4_abort()
3366 */
3367
3368 j_errno = jbd2_journal_errno(journal);
3369 if (j_errno) {
3370 char nbuf[16];
3371
3372 errstr = ext4_decode_error(sb, j_errno, nbuf);
3373 ext4_warning(sb, __func__, "Filesystem error recorded "
3374 "from previous mount: %s", errstr);
3375 ext4_warning(sb, __func__, "Marking fs in need of "
3376 "filesystem check.");
3377
3378 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3379 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3380 ext4_commit_super(sb, 1);
3381
3382 jbd2_journal_clear_err(journal);
3383 }
3384 }
3385
3386 /*
3387 * Force the running and committing transactions to commit,
3388 * and wait on the commit.
3389 */
3390 int ext4_force_commit(struct super_block *sb)
3391 {
3392 journal_t *journal;
3393 int ret = 0;
3394
3395 if (sb->s_flags & MS_RDONLY)
3396 return 0;
3397
3398 journal = EXT4_SB(sb)->s_journal;
3399 if (journal) {
3400 vfs_check_frozen(sb, SB_FREEZE_TRANS);
3401 ret = ext4_journal_force_commit(journal);
3402 }
3403
3404 return ret;
3405 }
3406
3407 static void ext4_write_super(struct super_block *sb)
3408 {
3409 lock_super(sb);
3410 ext4_commit_super(sb, 1);
3411 unlock_super(sb);
3412 }
3413
3414 static int ext4_sync_fs(struct super_block *sb, int wait)
3415 {
3416 int ret = 0;
3417 tid_t target;
3418 struct ext4_sb_info *sbi = EXT4_SB(sb);
3419
3420 trace_ext4_sync_fs(sb, wait);
3421 flush_workqueue(sbi->dio_unwritten_wq);
3422 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
3423 if (wait)
3424 jbd2_log_wait_commit(sbi->s_journal, target);
3425 }
3426 return ret;
3427 }
3428
3429 /*
3430 * LVM calls this function before a (read-only) snapshot is created. This
3431 * gives us a chance to flush the journal completely and mark the fs clean.
3432 */
3433 static int ext4_freeze(struct super_block *sb)
3434 {
3435 int error = 0;
3436 journal_t *journal;
3437
3438 if (sb->s_flags & MS_RDONLY)
3439 return 0;
3440
3441 journal = EXT4_SB(sb)->s_journal;
3442
3443 /* Now we set up the journal barrier. */
3444 jbd2_journal_lock_updates(journal);
3445
3446 /*
3447 * Don't clear the needs_recovery flag if we failed to flush
3448 * the journal.
3449 */
3450 error = jbd2_journal_flush(journal);
3451 if (error < 0)
3452 goto out;
3453
3454 /* Journal blocked and flushed, clear needs_recovery flag. */
3455 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3456 error = ext4_commit_super(sb, 1);
3457 out:
3458 /* we rely on s_frozen to stop further updates */
3459 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3460 return error;
3461 }
3462
3463 /*
3464 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3465 * flag here, even though the filesystem is not technically dirty yet.
3466 */
3467 static int ext4_unfreeze(struct super_block *sb)
3468 {
3469 if (sb->s_flags & MS_RDONLY)
3470 return 0;
3471
3472 lock_super(sb);
3473 /* Reset the needs_recovery flag before the fs is unlocked. */
3474 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3475 ext4_commit_super(sb, 1);
3476 unlock_super(sb);
3477 return 0;
3478 }
3479
3480 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3481 {
3482 struct ext4_super_block *es;
3483 struct ext4_sb_info *sbi = EXT4_SB(sb);
3484 ext4_fsblk_t n_blocks_count = 0;
3485 unsigned long old_sb_flags;
3486 struct ext4_mount_options old_opts;
3487 ext4_group_t g;
3488 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3489 int err;
3490 #ifdef CONFIG_QUOTA
3491 int i;
3492 #endif
3493
3494 lock_kernel();
3495
3496 /* Store the original options */
3497 lock_super(sb);
3498 old_sb_flags = sb->s_flags;
3499 old_opts.s_mount_opt = sbi->s_mount_opt;
3500 old_opts.s_resuid = sbi->s_resuid;
3501 old_opts.s_resgid = sbi->s_resgid;
3502 old_opts.s_commit_interval = sbi->s_commit_interval;
3503 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3504 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3505 #ifdef CONFIG_QUOTA
3506 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3507 for (i = 0; i < MAXQUOTAS; i++)
3508 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3509 #endif
3510 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3511 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3512
3513 /*
3514 * Allow the "check" option to be passed as a remount option.
3515 */
3516 if (!parse_options(data, sb, NULL, &journal_ioprio,
3517 &n_blocks_count, 1)) {
3518 err = -EINVAL;
3519 goto restore_opts;
3520 }
3521
3522 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
3523 ext4_abort(sb, __func__, "Abort forced by user");
3524
3525 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3526 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3527
3528 es = sbi->s_es;
3529
3530 if (sbi->s_journal) {
3531 ext4_init_journal_params(sb, sbi->s_journal);
3532 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3533 }
3534
3535 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3536 n_blocks_count > ext4_blocks_count(es)) {
3537 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
3538 err = -EROFS;
3539 goto restore_opts;
3540 }
3541
3542 if (*flags & MS_RDONLY) {
3543 /*
3544 * First of all, the unconditional stuff we have to do
3545 * to disable replay of the journal when we next remount
3546 */
3547 sb->s_flags |= MS_RDONLY;
3548
3549 /*
3550 * OK, test if we are remounting a valid rw partition
3551 * readonly, and if so set the rdonly flag and then
3552 * mark the partition as valid again.
3553 */
3554 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3555 (sbi->s_mount_state & EXT4_VALID_FS))
3556 es->s_state = cpu_to_le16(sbi->s_mount_state);
3557
3558 if (sbi->s_journal)
3559 ext4_mark_recovery_complete(sb, es);
3560 } else {
3561 /* Make sure we can mount this feature set readwrite */
3562 if (!ext4_feature_set_ok(sb, 0)) {
3563 err = -EROFS;
3564 goto restore_opts;
3565 }
3566 /*
3567 * Make sure the group descriptor checksums
3568 * are sane. If they aren't, refuse to remount r/w.
3569 */
3570 for (g = 0; g < sbi->s_groups_count; g++) {
3571 struct ext4_group_desc *gdp =
3572 ext4_get_group_desc(sb, g, NULL);
3573
3574 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3575 ext4_msg(sb, KERN_ERR,
3576 "ext4_remount: Checksum for group %u failed (%u!=%u)",
3577 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3578 le16_to_cpu(gdp->bg_checksum));
3579 err = -EINVAL;
3580 goto restore_opts;
3581 }
3582 }
3583
3584 /*
3585 * If we have an unprocessed orphan list hanging
3586 * around from a previously readonly bdev mount,
3587 * require a full umount/remount for now.
3588 */
3589 if (es->s_last_orphan) {
3590 ext4_msg(sb, KERN_WARNING, "Couldn't "
3591 "remount RDWR because of unprocessed "
3592 "orphan inode list. Please "
3593 "umount/remount instead");
3594 err = -EINVAL;
3595 goto restore_opts;
3596 }
3597
3598 /*
3599 * Mounting a RDONLY partition read-write, so reread
3600 * and store the current valid flag. (It may have
3601 * been changed by e2fsck since we originally mounted
3602 * the partition.)
3603 */
3604 if (sbi->s_journal)
3605 ext4_clear_journal_err(sb, es);
3606 sbi->s_mount_state = le16_to_cpu(es->s_state);
3607 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3608 goto restore_opts;
3609 if (!ext4_setup_super(sb, es, 0))
3610 sb->s_flags &= ~MS_RDONLY;
3611 }
3612 }
3613 ext4_setup_system_zone(sb);
3614 if (sbi->s_journal == NULL)
3615 ext4_commit_super(sb, 1);
3616
3617 #ifdef CONFIG_QUOTA
3618 /* Release old quota file names */
3619 for (i = 0; i < MAXQUOTAS; i++)
3620 if (old_opts.s_qf_names[i] &&
3621 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3622 kfree(old_opts.s_qf_names[i]);
3623 #endif
3624 unlock_super(sb);
3625 unlock_kernel();
3626 return 0;
3627
3628 restore_opts:
3629 sb->s_flags = old_sb_flags;
3630 sbi->s_mount_opt = old_opts.s_mount_opt;
3631 sbi->s_resuid = old_opts.s_resuid;
3632 sbi->s_resgid = old_opts.s_resgid;
3633 sbi->s_commit_interval = old_opts.s_commit_interval;
3634 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3635 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3636 #ifdef CONFIG_QUOTA
3637 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3638 for (i = 0; i < MAXQUOTAS; i++) {
3639 if (sbi->s_qf_names[i] &&
3640 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3641 kfree(sbi->s_qf_names[i]);
3642 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3643 }
3644 #endif
3645 unlock_super(sb);
3646 unlock_kernel();
3647 return err;
3648 }
3649
3650 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3651 {
3652 struct super_block *sb = dentry->d_sb;
3653 struct ext4_sb_info *sbi = EXT4_SB(sb);
3654 struct ext4_super_block *es = sbi->s_es;
3655 u64 fsid;
3656
3657 if (test_opt(sb, MINIX_DF)) {
3658 sbi->s_overhead_last = 0;
3659 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3660 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3661 ext4_fsblk_t overhead = 0;
3662
3663 /*
3664 * Compute the overhead (FS structures). This is constant
3665 * for a given filesystem unless the number of block groups
3666 * changes so we cache the previous value until it does.
3667 */
3668
3669 /*
3670 * All of the blocks before first_data_block are
3671 * overhead
3672 */
3673 overhead = le32_to_cpu(es->s_first_data_block);
3674
3675 /*
3676 * Add the overhead attributed to the superblock and
3677 * block group descriptors. If the sparse superblocks
3678 * feature is turned on, then not all groups have this.
3679 */
3680 for (i = 0; i < ngroups; i++) {
3681 overhead += ext4_bg_has_super(sb, i) +
3682 ext4_bg_num_gdb(sb, i);
3683 cond_resched();
3684 }
3685
3686 /*
3687 * Every block group has an inode bitmap, a block
3688 * bitmap, and an inode table.
3689 */
3690 overhead += ngroups * (2 + sbi->s_itb_per_group);
3691 sbi->s_overhead_last = overhead;
3692 smp_wmb();
3693 sbi->s_blocks_last = ext4_blocks_count(es);
3694 }
3695
3696 buf->f_type = EXT4_SUPER_MAGIC;
3697 buf->f_bsize = sb->s_blocksize;
3698 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3699 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3700 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3701 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3702 if (buf->f_bfree < ext4_r_blocks_count(es))
3703 buf->f_bavail = 0;
3704 buf->f_files = le32_to_cpu(es->s_inodes_count);
3705 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3706 buf->f_namelen = EXT4_NAME_LEN;
3707 fsid = le64_to_cpup((void *)es->s_uuid) ^
3708 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3709 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3710 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3711
3712 return 0;
3713 }
3714
3715 /* Helper function for writing quotas on sync - we need to start transaction
3716 * before quota file is locked for write. Otherwise the are possible deadlocks:
3717 * Process 1 Process 2
3718 * ext4_create() quota_sync()
3719 * jbd2_journal_start() write_dquot()
3720 * vfs_dq_init() down(dqio_mutex)
3721 * down(dqio_mutex) jbd2_journal_start()
3722 *
3723 */
3724
3725 #ifdef CONFIG_QUOTA
3726
3727 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3728 {
3729 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3730 }
3731
3732 static int ext4_write_dquot(struct dquot *dquot)
3733 {
3734 int ret, err;
3735 handle_t *handle;
3736 struct inode *inode;
3737
3738 inode = dquot_to_inode(dquot);
3739 handle = ext4_journal_start(inode,
3740 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3741 if (IS_ERR(handle))
3742 return PTR_ERR(handle);
3743 ret = dquot_commit(dquot);
3744 err = ext4_journal_stop(handle);
3745 if (!ret)
3746 ret = err;
3747 return ret;
3748 }
3749
3750 static int ext4_acquire_dquot(struct dquot *dquot)
3751 {
3752 int ret, err;
3753 handle_t *handle;
3754
3755 handle = ext4_journal_start(dquot_to_inode(dquot),
3756 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3757 if (IS_ERR(handle))
3758 return PTR_ERR(handle);
3759 ret = dquot_acquire(dquot);
3760 err = ext4_journal_stop(handle);
3761 if (!ret)
3762 ret = err;
3763 return ret;
3764 }
3765
3766 static int ext4_release_dquot(struct dquot *dquot)
3767 {
3768 int ret, err;
3769 handle_t *handle;
3770
3771 handle = ext4_journal_start(dquot_to_inode(dquot),
3772 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3773 if (IS_ERR(handle)) {
3774 /* Release dquot anyway to avoid endless cycle in dqput() */
3775 dquot_release(dquot);
3776 return PTR_ERR(handle);
3777 }
3778 ret = dquot_release(dquot);
3779 err = ext4_journal_stop(handle);
3780 if (!ret)
3781 ret = err;
3782 return ret;
3783 }
3784
3785 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3786 {
3787 /* Are we journaling quotas? */
3788 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3789 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3790 dquot_mark_dquot_dirty(dquot);
3791 return ext4_write_dquot(dquot);
3792 } else {
3793 return dquot_mark_dquot_dirty(dquot);
3794 }
3795 }
3796
3797 static int ext4_write_info(struct super_block *sb, int type)
3798 {
3799 int ret, err;
3800 handle_t *handle;
3801
3802 /* Data block + inode block */
3803 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3804 if (IS_ERR(handle))
3805 return PTR_ERR(handle);
3806 ret = dquot_commit_info(sb, type);
3807 err = ext4_journal_stop(handle);
3808 if (!ret)
3809 ret = err;
3810 return ret;
3811 }
3812
3813 /*
3814 * Turn on quotas during mount time - we need to find
3815 * the quota file and such...
3816 */
3817 static int ext4_quota_on_mount(struct super_block *sb, int type)
3818 {
3819 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3820 EXT4_SB(sb)->s_jquota_fmt, type);
3821 }
3822
3823 /*
3824 * Standard function to be called on quota_on
3825 */
3826 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3827 char *name, int remount)
3828 {
3829 int err;
3830 struct path path;
3831
3832 if (!test_opt(sb, QUOTA))
3833 return -EINVAL;
3834 /* When remounting, no checks are needed and in fact, name is NULL */
3835 if (remount)
3836 return vfs_quota_on(sb, type, format_id, name, remount);
3837
3838 err = kern_path(name, LOOKUP_FOLLOW, &path);
3839 if (err)
3840 return err;
3841
3842 /* Quotafile not on the same filesystem? */
3843 if (path.mnt->mnt_sb != sb) {
3844 path_put(&path);
3845 return -EXDEV;
3846 }
3847 /* Journaling quota? */
3848 if (EXT4_SB(sb)->s_qf_names[type]) {
3849 /* Quotafile not in fs root? */
3850 if (path.dentry->d_parent != sb->s_root)
3851 ext4_msg(sb, KERN_WARNING,
3852 "Quota file not on filesystem root. "
3853 "Journaled quota will not work");
3854 }
3855
3856 /*
3857 * When we journal data on quota file, we have to flush journal to see
3858 * all updates to the file when we bypass pagecache...
3859 */
3860 if (EXT4_SB(sb)->s_journal &&
3861 ext4_should_journal_data(path.dentry->d_inode)) {
3862 /*
3863 * We don't need to lock updates but journal_flush() could
3864 * otherwise be livelocked...
3865 */
3866 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3867 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3868 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3869 if (err) {
3870 path_put(&path);
3871 return err;
3872 }
3873 }
3874
3875 err = vfs_quota_on_path(sb, type, format_id, &path);
3876 path_put(&path);
3877 return err;
3878 }
3879
3880 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3881 * acquiring the locks... As quota files are never truncated and quota code
3882 * itself serializes the operations (and noone else should touch the files)
3883 * we don't have to be afraid of races */
3884 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3885 size_t len, loff_t off)
3886 {
3887 struct inode *inode = sb_dqopt(sb)->files[type];
3888 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3889 int err = 0;
3890 int offset = off & (sb->s_blocksize - 1);
3891 int tocopy;
3892 size_t toread;
3893 struct buffer_head *bh;
3894 loff_t i_size = i_size_read(inode);
3895
3896 if (off > i_size)
3897 return 0;
3898 if (off+len > i_size)
3899 len = i_size-off;
3900 toread = len;
3901 while (toread > 0) {
3902 tocopy = sb->s_blocksize - offset < toread ?
3903 sb->s_blocksize - offset : toread;
3904 bh = ext4_bread(NULL, inode, blk, 0, &err);
3905 if (err)
3906 return err;
3907 if (!bh) /* A hole? */
3908 memset(data, 0, tocopy);
3909 else
3910 memcpy(data, bh->b_data+offset, tocopy);
3911 brelse(bh);
3912 offset = 0;
3913 toread -= tocopy;
3914 data += tocopy;
3915 blk++;
3916 }
3917 return len;
3918 }
3919
3920 /* Write to quotafile (we know the transaction is already started and has
3921 * enough credits) */
3922 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3923 const char *data, size_t len, loff_t off)
3924 {
3925 struct inode *inode = sb_dqopt(sb)->files[type];
3926 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3927 int err = 0;
3928 int offset = off & (sb->s_blocksize - 1);
3929 int tocopy;
3930 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3931 size_t towrite = len;
3932 struct buffer_head *bh;
3933 handle_t *handle = journal_current_handle();
3934
3935 if (EXT4_SB(sb)->s_journal && !handle) {
3936 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
3937 " cancelled because transaction is not started",
3938 (unsigned long long)off, (unsigned long long)len);
3939 return -EIO;
3940 }
3941 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3942 while (towrite > 0) {
3943 tocopy = sb->s_blocksize - offset < towrite ?
3944 sb->s_blocksize - offset : towrite;
3945 bh = ext4_bread(handle, inode, blk, 1, &err);
3946 if (!bh)
3947 goto out;
3948 if (journal_quota) {
3949 err = ext4_journal_get_write_access(handle, bh);
3950 if (err) {
3951 brelse(bh);
3952 goto out;
3953 }
3954 }
3955 lock_buffer(bh);
3956 memcpy(bh->b_data+offset, data, tocopy);
3957 flush_dcache_page(bh->b_page);
3958 unlock_buffer(bh);
3959 if (journal_quota)
3960 err = ext4_handle_dirty_metadata(handle, NULL, bh);
3961 else {
3962 /* Always do at least ordered writes for quotas */
3963 err = ext4_jbd2_file_inode(handle, inode);
3964 mark_buffer_dirty(bh);
3965 }
3966 brelse(bh);
3967 if (err)
3968 goto out;
3969 offset = 0;
3970 towrite -= tocopy;
3971 data += tocopy;
3972 blk++;
3973 }
3974 out:
3975 if (len == towrite) {
3976 mutex_unlock(&inode->i_mutex);
3977 return err;
3978 }
3979 if (inode->i_size < off+len-towrite) {
3980 i_size_write(inode, off+len-towrite);
3981 EXT4_I(inode)->i_disksize = inode->i_size;
3982 }
3983 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3984 ext4_mark_inode_dirty(handle, inode);
3985 mutex_unlock(&inode->i_mutex);
3986 return len - towrite;
3987 }
3988
3989 #endif
3990
3991 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
3992 const char *dev_name, void *data, struct vfsmount *mnt)
3993 {
3994 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
3995 }
3996
3997 static struct file_system_type ext4_fs_type = {
3998 .owner = THIS_MODULE,
3999 .name = "ext4",
4000 .get_sb = ext4_get_sb,
4001 .kill_sb = kill_block_super,
4002 .fs_flags = FS_REQUIRES_DEV,
4003 };
4004
4005 static int __init init_ext4_fs(void)
4006 {
4007 int err;
4008
4009 ext4_check_flag_values();
4010 err = init_ext4_system_zone();
4011 if (err)
4012 return err;
4013 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4014 if (!ext4_kset)
4015 goto out4;
4016 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
4017 err = init_ext4_mballoc();
4018 if (err)
4019 goto out3;
4020
4021 err = init_ext4_xattr();
4022 if (err)
4023 goto out2;
4024 err = init_inodecache();
4025 if (err)
4026 goto out1;
4027 err = register_filesystem(&ext4_fs_type);
4028 if (err)
4029 goto out;
4030 return 0;
4031 out:
4032 destroy_inodecache();
4033 out1:
4034 exit_ext4_xattr();
4035 out2:
4036 exit_ext4_mballoc();
4037 out3:
4038 remove_proc_entry("fs/ext4", NULL);
4039 kset_unregister(ext4_kset);
4040 out4:
4041 exit_ext4_system_zone();
4042 return err;
4043 }
4044
4045 static void __exit exit_ext4_fs(void)
4046 {
4047 unregister_filesystem(&ext4_fs_type);
4048 destroy_inodecache();
4049 exit_ext4_xattr();
4050 exit_ext4_mballoc();
4051 remove_proc_entry("fs/ext4", NULL);
4052 kset_unregister(ext4_kset);
4053 exit_ext4_system_zone();
4054 }
4055
4056 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4057 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4058 MODULE_LICENSE("GPL");
4059 module_init(init_ext4_fs)
4060 module_exit(exit_ext4_fs)
Linux 2.6 ibm-acpi/thinkpad-acpi driver git tree