?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
ext4: fix undefined behavior in ext4_fill_flex_info()
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / fs / ext4 / super.c
blobf1e7077a06a3dce33213208924d28b7d20f2f9f9
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 unsigned 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 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1714 sbi->s_log_groups_per_flex = 0;
1715 return 1;
1716 }
1717 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1718
1719 /* We allocate both existing and potentially added groups */
1720 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1721 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1722 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1723 size = flex_group_count * sizeof(struct flex_groups);
1724 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1725 if (sbi->s_flex_groups == NULL) {
1726 sbi->s_flex_groups = vmalloc(size);
1727 if (sbi->s_flex_groups)
1728 memset(sbi->s_flex_groups, 0, size);
1729 }
1730 if (sbi->s_flex_groups == NULL) {
1731 ext4_msg(sb, KERN_ERR, "not enough memory for "
1732 "%u flex groups", flex_group_count);
1733 goto failed;
1734 }
1735
1736 for (i = 0; i < sbi->s_groups_count; i++) {
1737 gdp = ext4_get_group_desc(sb, i, NULL);
1738
1739 flex_group = ext4_flex_group(sbi, i);
1740 atomic_add(ext4_free_inodes_count(sb, gdp),
1741 &sbi->s_flex_groups[flex_group].free_inodes);
1742 atomic_add(ext4_free_blks_count(sb, gdp),
1743 &sbi->s_flex_groups[flex_group].free_blocks);
1744 atomic_add(ext4_used_dirs_count(sb, gdp),
1745 &sbi->s_flex_groups[flex_group].used_dirs);
1746 }
1747
1748 return 1;
1749 failed:
1750 return 0;
1751 }
1752
1753 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1754 struct ext4_group_desc *gdp)
1755 {
1756 __u16 crc = 0;
1757
1758 if (sbi->s_es->s_feature_ro_compat &
1759 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1760 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1761 __le32 le_group = cpu_to_le32(block_group);
1762
1763 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1764 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1765 crc = crc16(crc, (__u8 *)gdp, offset);
1766 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1767 /* for checksum of struct ext4_group_desc do the rest...*/
1768 if ((sbi->s_es->s_feature_incompat &
1769 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1770 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1771 crc = crc16(crc, (__u8 *)gdp + offset,
1772 le16_to_cpu(sbi->s_es->s_desc_size) -
1773 offset);
1774 }
1775
1776 return cpu_to_le16(crc);
1777 }
1778
1779 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1780 struct ext4_group_desc *gdp)
1781 {
1782 if ((sbi->s_es->s_feature_ro_compat &
1783 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1784 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1785 return 0;
1786
1787 return 1;
1788 }
1789
1790 /* Called at mount-time, super-block is locked */
1791 static int ext4_check_descriptors(struct super_block *sb)
1792 {
1793 struct ext4_sb_info *sbi = EXT4_SB(sb);
1794 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1795 ext4_fsblk_t last_block;
1796 ext4_fsblk_t block_bitmap;
1797 ext4_fsblk_t inode_bitmap;
1798 ext4_fsblk_t inode_table;
1799 int flexbg_flag = 0;
1800 ext4_group_t i;
1801
1802 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1803 flexbg_flag = 1;
1804
1805 ext4_debug("Checking group descriptors");
1806
1807 for (i = 0; i < sbi->s_groups_count; i++) {
1808 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1809
1810 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1811 last_block = ext4_blocks_count(sbi->s_es) - 1;
1812 else
1813 last_block = first_block +
1814 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1815
1816 block_bitmap = ext4_block_bitmap(sb, gdp);
1817 if (block_bitmap < first_block || block_bitmap > last_block) {
1818 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1819 "Block bitmap for group %u not in group "
1820 "(block %llu)!", i, block_bitmap);
1821 return 0;
1822 }
1823 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1824 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1825 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1826 "Inode bitmap for group %u not in group "
1827 "(block %llu)!", i, inode_bitmap);
1828 return 0;
1829 }
1830 inode_table = ext4_inode_table(sb, gdp);
1831 if (inode_table < first_block ||
1832 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1833 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1834 "Inode table for group %u not in group "
1835 "(block %llu)!", i, inode_table);
1836 return 0;
1837 }
1838 ext4_lock_group(sb, i);
1839 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1840 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1841 "Checksum for group %u failed (%u!=%u)",
1842 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1843 gdp)), le16_to_cpu(gdp->bg_checksum));
1844 if (!(sb->s_flags & MS_RDONLY)) {
1845 ext4_unlock_group(sb, i);
1846 return 0;
1847 }
1848 }
1849 ext4_unlock_group(sb, i);
1850 if (!flexbg_flag)
1851 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1852 }
1853
1854 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1855 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
1856 return 1;
1857 }
1858
1859 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1860 * the superblock) which were deleted from all directories, but held open by
1861 * a process at the time of a crash. We walk the list and try to delete these
1862 * inodes at recovery time (only with a read-write filesystem).
1863 *
1864 * In order to keep the orphan inode chain consistent during traversal (in
1865 * case of crash during recovery), we link each inode into the superblock
1866 * orphan list_head and handle it the same way as an inode deletion during
1867 * normal operation (which journals the operations for us).
1868 *
1869 * We only do an iget() and an iput() on each inode, which is very safe if we
1870 * accidentally point at an in-use or already deleted inode. The worst that
1871 * can happen in this case is that we get a "bit already cleared" message from
1872 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1873 * e2fsck was run on this filesystem, and it must have already done the orphan
1874 * inode cleanup for us, so we can safely abort without any further action.
1875 */
1876 static void ext4_orphan_cleanup(struct super_block *sb,
1877 struct ext4_super_block *es)
1878 {
1879 unsigned int s_flags = sb->s_flags;
1880 int nr_orphans = 0, nr_truncates = 0;
1881 #ifdef CONFIG_QUOTA
1882 int i;
1883 #endif
1884 if (!es->s_last_orphan) {
1885 jbd_debug(4, "no orphan inodes to clean up\n");
1886 return;
1887 }
1888
1889 if (bdev_read_only(sb->s_bdev)) {
1890 ext4_msg(sb, KERN_ERR, "write access "
1891 "unavailable, skipping orphan cleanup");
1892 return;
1893 }
1894
1895 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1896 if (es->s_last_orphan)
1897 jbd_debug(1, "Errors on filesystem, "
1898 "clearing orphan list.\n");
1899 es->s_last_orphan = 0;
1900 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1901 return;
1902 }
1903
1904 if (s_flags & MS_RDONLY) {
1905 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1906 sb->s_flags &= ~MS_RDONLY;
1907 }
1908 #ifdef CONFIG_QUOTA
1909 /* Needed for iput() to work correctly and not trash data */
1910 sb->s_flags |= MS_ACTIVE;
1911 /* Turn on quotas so that they are updated correctly */
1912 for (i = 0; i < MAXQUOTAS; i++) {
1913 if (EXT4_SB(sb)->s_qf_names[i]) {
1914 int ret = ext4_quota_on_mount(sb, i);
1915 if (ret < 0)
1916 ext4_msg(sb, KERN_ERR,
1917 "Cannot turn on journaled "
1918 "quota: error %d", ret);
1919 }
1920 }
1921 #endif
1922
1923 while (es->s_last_orphan) {
1924 struct inode *inode;
1925
1926 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1927 if (IS_ERR(inode)) {
1928 es->s_last_orphan = 0;
1929 break;
1930 }
1931
1932 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1933 vfs_dq_init(inode);
1934 if (inode->i_nlink) {
1935 ext4_msg(sb, KERN_DEBUG,
1936 "%s: truncating inode %lu to %lld bytes",
1937 __func__, inode->i_ino, inode->i_size);
1938 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1939 inode->i_ino, inode->i_size);
1940 ext4_truncate(inode);
1941 nr_truncates++;
1942 } else {
1943 ext4_msg(sb, KERN_DEBUG,
1944 "%s: deleting unreferenced inode %lu",
1945 __func__, inode->i_ino);
1946 jbd_debug(2, "deleting unreferenced inode %lu\n",
1947 inode->i_ino);
1948 nr_orphans++;
1949 }
1950 iput(inode); /* The delete magic happens here! */
1951 }
1952
1953 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1954
1955 if (nr_orphans)
1956 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1957 PLURAL(nr_orphans));
1958 if (nr_truncates)
1959 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1960 PLURAL(nr_truncates));
1961 #ifdef CONFIG_QUOTA
1962 /* Turn quotas off */
1963 for (i = 0; i < MAXQUOTAS; i++) {
1964 if (sb_dqopt(sb)->files[i])
1965 vfs_quota_off(sb, i, 0);
1966 }
1967 #endif
1968 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1969 }
1970
1971 /*
1972 * Maximal extent format file size.
1973 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1974 * extent format containers, within a sector_t, and within i_blocks
1975 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1976 * so that won't be a limiting factor.
1977 *
1978 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1979 */
1980 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1981 {
1982 loff_t res;
1983 loff_t upper_limit = MAX_LFS_FILESIZE;
1984
1985 /* small i_blocks in vfs inode? */
1986 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1987 /*
1988 * CONFIG_LBDAF is not enabled implies the inode
1989 * i_block represent total blocks in 512 bytes
1990 * 32 == size of vfs inode i_blocks * 8
1991 */
1992 upper_limit = (1LL << 32) - 1;
1993
1994 /* total blocks in file system block size */
1995 upper_limit >>= (blkbits - 9);
1996 upper_limit <<= blkbits;
1997 }
1998
1999 /* 32-bit extent-start container, ee_block */
2000 res = 1LL << 32;
2001 res <<= blkbits;
2002 res -= 1;
2003
2004 /* Sanity check against vm- & vfs- imposed limits */
2005 if (res > upper_limit)
2006 res = upper_limit;
2007
2008 return res;
2009 }
2010
2011 /*
2012 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2013 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2014 * We need to be 1 filesystem block less than the 2^48 sector limit.
2015 */
2016 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2017 {
2018 loff_t res = EXT4_NDIR_BLOCKS;
2019 int meta_blocks;
2020 loff_t upper_limit;
2021 /* This is calculated to be the largest file size for a dense, block
2022 * mapped file such that the file's total number of 512-byte sectors,
2023 * including data and all indirect blocks, does not exceed (2^48 - 1).
2024 *
2025 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2026 * number of 512-byte sectors of the file.
2027 */
2028
2029 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2030 /*
2031 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2032 * the inode i_block field represents total file blocks in
2033 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2034 */
2035 upper_limit = (1LL << 32) - 1;
2036
2037 /* total blocks in file system block size */
2038 upper_limit >>= (bits - 9);
2039
2040 } else {
2041 /*
2042 * We use 48 bit ext4_inode i_blocks
2043 * With EXT4_HUGE_FILE_FL set the i_blocks
2044 * represent total number of blocks in
2045 * file system block size
2046 */
2047 upper_limit = (1LL << 48) - 1;
2048
2049 }
2050
2051 /* indirect blocks */
2052 meta_blocks = 1;
2053 /* double indirect blocks */
2054 meta_blocks += 1 + (1LL << (bits-2));
2055 /* tripple indirect blocks */
2056 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2057
2058 upper_limit -= meta_blocks;
2059 upper_limit <<= bits;
2060
2061 res += 1LL << (bits-2);
2062 res += 1LL << (2*(bits-2));
2063 res += 1LL << (3*(bits-2));
2064 res <<= bits;
2065 if (res > upper_limit)
2066 res = upper_limit;
2067
2068 if (res > MAX_LFS_FILESIZE)
2069 res = MAX_LFS_FILESIZE;
2070
2071 return res;
2072 }
2073
2074 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2075 ext4_fsblk_t logical_sb_block, int nr)
2076 {
2077 struct ext4_sb_info *sbi = EXT4_SB(sb);
2078 ext4_group_t bg, first_meta_bg;
2079 int has_super = 0;
2080
2081 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2082
2083 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2084 nr < first_meta_bg)
2085 return logical_sb_block + nr + 1;
2086 bg = sbi->s_desc_per_block * nr;
2087 if (ext4_bg_has_super(sb, bg))
2088 has_super = 1;
2089
2090 return (has_super + ext4_group_first_block_no(sb, bg));
2091 }
2092
2093 /**
2094 * ext4_get_stripe_size: Get the stripe size.
2095 * @sbi: In memory super block info
2096 *
2097 * If we have specified it via mount option, then
2098 * use the mount option value. If the value specified at mount time is
2099 * greater than the blocks per group use the super block value.
2100 * If the super block value is greater than blocks per group return 0.
2101 * Allocator needs it be less than blocks per group.
2102 *
2103 */
2104 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2105 {
2106 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2107 unsigned long stripe_width =
2108 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2109
2110 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2111 return sbi->s_stripe;
2112
2113 if (stripe_width <= sbi->s_blocks_per_group)
2114 return stripe_width;
2115
2116 if (stride <= sbi->s_blocks_per_group)
2117 return stride;
2118
2119 return 0;
2120 }
2121
2122 /* sysfs supprt */
2123
2124 struct ext4_attr {
2125 struct attribute attr;
2126 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2127 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2128 const char *, size_t);
2129 int offset;
2130 };
2131
2132 static int parse_strtoul(const char *buf,
2133 unsigned long max, unsigned long *value)
2134 {
2135 char *endp;
2136
2137 while (*buf && isspace(*buf))
2138 buf++;
2139 *value = simple_strtoul(buf, &endp, 0);
2140 while (*endp && isspace(*endp))
2141 endp++;
2142 if (*endp || *value > max)
2143 return -EINVAL;
2144
2145 return 0;
2146 }
2147
2148 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2149 struct ext4_sb_info *sbi,
2150 char *buf)
2151 {
2152 return snprintf(buf, PAGE_SIZE, "%llu\n",
2153 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2154 }
2155
2156 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2157 struct ext4_sb_info *sbi, char *buf)
2158 {
2159 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2160
2161 return snprintf(buf, PAGE_SIZE, "%lu\n",
2162 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2163 sbi->s_sectors_written_start) >> 1);
2164 }
2165
2166 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2167 struct ext4_sb_info *sbi, char *buf)
2168 {
2169 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2170
2171 return snprintf(buf, PAGE_SIZE, "%llu\n",
2172 sbi->s_kbytes_written +
2173 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2174 EXT4_SB(sb)->s_sectors_written_start) >> 1));
2175 }
2176
2177 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2178 struct ext4_sb_info *sbi,
2179 const char *buf, size_t count)
2180 {
2181 unsigned long t;
2182
2183 if (parse_strtoul(buf, 0x40000000, &t))
2184 return -EINVAL;
2185
2186 if (!is_power_of_2(t))
2187 return -EINVAL;
2188
2189 sbi->s_inode_readahead_blks = t;
2190 return count;
2191 }
2192
2193 static ssize_t sbi_ui_show(struct ext4_attr *a,
2194 struct ext4_sb_info *sbi, char *buf)
2195 {
2196 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2197
2198 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2199 }
2200
2201 static ssize_t sbi_ui_store(struct ext4_attr *a,
2202 struct ext4_sb_info *sbi,
2203 const char *buf, size_t count)
2204 {
2205 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2206 unsigned long t;
2207
2208 if (parse_strtoul(buf, 0xffffffff, &t))
2209 return -EINVAL;
2210 *ui = t;
2211 return count;
2212 }
2213
2214 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2215 static struct ext4_attr ext4_attr_##_name = { \
2216 .attr = {.name = __stringify(_name), .mode = _mode }, \
2217 .show = _show, \
2218 .store = _store, \
2219 .offset = offsetof(struct ext4_sb_info, _elname), \
2220 }
2221 #define EXT4_ATTR(name, mode, show, store) \
2222 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2223
2224 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2225 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2226 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2227 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2228 #define ATTR_LIST(name) &ext4_attr_##name.attr
2229
2230 EXT4_RO_ATTR(delayed_allocation_blocks);
2231 EXT4_RO_ATTR(session_write_kbytes);
2232 EXT4_RO_ATTR(lifetime_write_kbytes);
2233 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2234 inode_readahead_blks_store, s_inode_readahead_blks);
2235 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2236 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2237 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2238 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2239 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2240 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2241 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2242 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2243
2244 static struct attribute *ext4_attrs[] = {
2245 ATTR_LIST(delayed_allocation_blocks),
2246 ATTR_LIST(session_write_kbytes),
2247 ATTR_LIST(lifetime_write_kbytes),
2248 ATTR_LIST(inode_readahead_blks),
2249 ATTR_LIST(inode_goal),
2250 ATTR_LIST(mb_stats),
2251 ATTR_LIST(mb_max_to_scan),
2252 ATTR_LIST(mb_min_to_scan),
2253 ATTR_LIST(mb_order2_req),
2254 ATTR_LIST(mb_stream_req),
2255 ATTR_LIST(mb_group_prealloc),
2256 ATTR_LIST(max_writeback_mb_bump),
2257 NULL,
2258 };
2259
2260 static ssize_t ext4_attr_show(struct kobject *kobj,
2261 struct attribute *attr, char *buf)
2262 {
2263 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2264 s_kobj);
2265 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2266
2267 return a->show ? a->show(a, sbi, buf) : 0;
2268 }
2269
2270 static ssize_t ext4_attr_store(struct kobject *kobj,
2271 struct attribute *attr,
2272 const char *buf, size_t len)
2273 {
2274 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2275 s_kobj);
2276 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2277
2278 return a->store ? a->store(a, sbi, buf, len) : 0;
2279 }
2280
2281 static void ext4_sb_release(struct kobject *kobj)
2282 {
2283 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2284 s_kobj);
2285 complete(&sbi->s_kobj_unregister);
2286 }
2287
2288
2289 static struct sysfs_ops ext4_attr_ops = {
2290 .show = ext4_attr_show,
2291 .store = ext4_attr_store,
2292 };
2293
2294 static struct kobj_type ext4_ktype = {
2295 .default_attrs = ext4_attrs,
2296 .sysfs_ops = &ext4_attr_ops,
2297 .release = ext4_sb_release,
2298 };
2299
2300 /*
2301 * Check whether this filesystem can be mounted based on
2302 * the features present and the RDONLY/RDWR mount requested.
2303 * Returns 1 if this filesystem can be mounted as requested,
2304 * 0 if it cannot be.
2305 */
2306 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2307 {
2308 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2309 ext4_msg(sb, KERN_ERR,
2310 "Couldn't mount because of "
2311 "unsupported optional features (%x)",
2312 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2313 ~EXT4_FEATURE_INCOMPAT_SUPP));
2314 return 0;
2315 }
2316
2317 if (readonly)
2318 return 1;
2319
2320 /* Check that feature set is OK for a read-write mount */
2321 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2322 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2323 "unsupported optional features (%x)",
2324 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2325 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2326 return 0;
2327 }
2328 /*
2329 * Large file size enabled file system can only be mounted
2330 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2331 */
2332 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2333 if (sizeof(blkcnt_t) < sizeof(u64)) {
2334 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2335 "cannot be mounted RDWR without "
2336 "CONFIG_LBDAF");
2337 return 0;
2338 }
2339 }
2340 return 1;
2341 }
2342
2343 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2344 __releases(kernel_lock)
2345 __acquires(kernel_lock)
2346 {
2347 struct buffer_head *bh;
2348 struct ext4_super_block *es = NULL;
2349 struct ext4_sb_info *sbi;
2350 ext4_fsblk_t block;
2351 ext4_fsblk_t sb_block = get_sb_block(&data);
2352 ext4_fsblk_t logical_sb_block;
2353 unsigned long offset = 0;
2354 unsigned long journal_devnum = 0;
2355 unsigned long def_mount_opts;
2356 struct inode *root;
2357 char *cp;
2358 const char *descr;
2359 int ret = -EINVAL;
2360 int blocksize;
2361 unsigned int db_count;
2362 unsigned int i;
2363 int needs_recovery, has_huge_files;
2364 __u64 blocks_count;
2365 int err;
2366 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2367
2368 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2369 if (!sbi)
2370 return -ENOMEM;
2371
2372 sbi->s_blockgroup_lock =
2373 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2374 if (!sbi->s_blockgroup_lock) {
2375 kfree(sbi);
2376 return -ENOMEM;
2377 }
2378 sb->s_fs_info = sbi;
2379 sbi->s_mount_opt = 0;
2380 sbi->s_resuid = EXT4_DEF_RESUID;
2381 sbi->s_resgid = EXT4_DEF_RESGID;
2382 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2383 sbi->s_sb_block = sb_block;
2384 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2385 sectors[1]);
2386
2387 unlock_kernel();
2388
2389 /* Cleanup superblock name */
2390 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2391 *cp = '!';
2392
2393 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2394 if (!blocksize) {
2395 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
2396 goto out_fail;
2397 }
2398
2399 /*
2400 * The ext4 superblock will not be buffer aligned for other than 1kB
2401 * block sizes. We need to calculate the offset from buffer start.
2402 */
2403 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2404 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2405 offset = do_div(logical_sb_block, blocksize);
2406 } else {
2407 logical_sb_block = sb_block;
2408 }
2409
2410 if (!(bh = sb_bread(sb, logical_sb_block))) {
2411 ext4_msg(sb, KERN_ERR, "unable to read superblock");
2412 goto out_fail;
2413 }
2414 /*
2415 * Note: s_es must be initialized as soon as possible because
2416 * some ext4 macro-instructions depend on its value
2417 */
2418 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2419 sbi->s_es = es;
2420 sb->s_magic = le16_to_cpu(es->s_magic);
2421 if (sb->s_magic != EXT4_SUPER_MAGIC)
2422 goto cantfind_ext4;
2423 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2424
2425 /* Set defaults before we parse the mount options */
2426 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2427 if (def_mount_opts & EXT4_DEFM_DEBUG)
2428 set_opt(sbi->s_mount_opt, DEBUG);
2429 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
2430 set_opt(sbi->s_mount_opt, GRPID);
2431 if (def_mount_opts & EXT4_DEFM_UID16)
2432 set_opt(sbi->s_mount_opt, NO_UID32);
2433 #ifdef CONFIG_EXT4_FS_XATTR
2434 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2435 set_opt(sbi->s_mount_opt, XATTR_USER);
2436 #endif
2437 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2438 if (def_mount_opts & EXT4_DEFM_ACL)
2439 set_opt(sbi->s_mount_opt, POSIX_ACL);
2440 #endif
2441 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2442 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2443 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2444 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2445 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2446 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2447
2448 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2449 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2450 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2451 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2452 else
2453 set_opt(sbi->s_mount_opt, ERRORS_RO);
2454
2455 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2456 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2457 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2458 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2459 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2460
2461 set_opt(sbi->s_mount_opt, BARRIER);
2462
2463 /*
2464 * enable delayed allocation by default
2465 * Use -o nodelalloc to turn it off
2466 */
2467 set_opt(sbi->s_mount_opt, DELALLOC);
2468
2469 if (!parse_options((char *) data, sb, &journal_devnum,
2470 &journal_ioprio, NULL, 0))
2471 goto failed_mount;
2472
2473 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2474 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2475
2476 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2477 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2478 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2479 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2480 ext4_msg(sb, KERN_WARNING,
2481 "feature flags set on rev 0 fs, "
2482 "running e2fsck is recommended");
2483
2484 /*
2485 * Check feature flags regardless of the revision level, since we
2486 * previously didn't change the revision level when setting the flags,
2487 * so there is a chance incompat flags are set on a rev 0 filesystem.
2488 */
2489 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
2490 goto failed_mount;
2491
2492 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2493
2494 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2495 blocksize > EXT4_MAX_BLOCK_SIZE) {
2496 ext4_msg(sb, KERN_ERR,
2497 "Unsupported filesystem blocksize %d", blocksize);
2498 goto failed_mount;
2499 }
2500
2501 if (sb->s_blocksize != blocksize) {
2502 /* Validate the filesystem blocksize */
2503 if (!sb_set_blocksize(sb, blocksize)) {
2504 ext4_msg(sb, KERN_ERR, "bad block size %d",
2505 blocksize);
2506 goto failed_mount;
2507 }
2508
2509 brelse(bh);
2510 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2511 offset = do_div(logical_sb_block, blocksize);
2512 bh = sb_bread(sb, logical_sb_block);
2513 if (!bh) {
2514 ext4_msg(sb, KERN_ERR,
2515 "Can't read superblock on 2nd try");
2516 goto failed_mount;
2517 }
2518 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2519 sbi->s_es = es;
2520 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2521 ext4_msg(sb, KERN_ERR,
2522 "Magic mismatch, very weird!");
2523 goto failed_mount;
2524 }
2525 }
2526
2527 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2528 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2529 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2530 has_huge_files);
2531 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2532
2533 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2534 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2535 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2536 } else {
2537 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2538 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2539 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2540 (!is_power_of_2(sbi->s_inode_size)) ||
2541 (sbi->s_inode_size > blocksize)) {
2542 ext4_msg(sb, KERN_ERR,
2543 "unsupported inode size: %d",
2544 sbi->s_inode_size);
2545 goto failed_mount;
2546 }
2547 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2548 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2549 }
2550
2551 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2552 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2553 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2554 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2555 !is_power_of_2(sbi->s_desc_size)) {
2556 ext4_msg(sb, KERN_ERR,
2557 "unsupported descriptor size %lu",
2558 sbi->s_desc_size);
2559 goto failed_mount;
2560 }
2561 } else
2562 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2563
2564 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2565 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2566 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2567 goto cantfind_ext4;
2568
2569 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2570 if (sbi->s_inodes_per_block == 0)
2571 goto cantfind_ext4;
2572 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2573 sbi->s_inodes_per_block;
2574 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2575 sbi->s_sbh = bh;
2576 sbi->s_mount_state = le16_to_cpu(es->s_state);
2577 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2578 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2579
2580 for (i = 0; i < 4; i++)
2581 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2582 sbi->s_def_hash_version = es->s_def_hash_version;
2583 i = le32_to_cpu(es->s_flags);
2584 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2585 sbi->s_hash_unsigned = 3;
2586 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2587 #ifdef __CHAR_UNSIGNED__
2588 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2589 sbi->s_hash_unsigned = 3;
2590 #else
2591 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2592 #endif
2593 sb->s_dirt = 1;
2594 }
2595
2596 if (sbi->s_blocks_per_group > blocksize * 8) {
2597 ext4_msg(sb, KERN_ERR,
2598 "#blocks per group too big: %lu",
2599 sbi->s_blocks_per_group);
2600 goto failed_mount;
2601 }
2602 if (sbi->s_inodes_per_group > blocksize * 8) {
2603 ext4_msg(sb, KERN_ERR,
2604 "#inodes per group too big: %lu",
2605 sbi->s_inodes_per_group);
2606 goto failed_mount;
2607 }
2608
2609 /*
2610 * Test whether we have more sectors than will fit in sector_t,
2611 * and whether the max offset is addressable by the page cache.
2612 */
2613 if ((ext4_blocks_count(es) >
2614 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2615 (ext4_blocks_count(es) >
2616 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
2617 ext4_msg(sb, KERN_ERR, "filesystem"
2618 " too large to mount safely on this system");
2619 if (sizeof(sector_t) < 8)
2620 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
2621 ret = -EFBIG;
2622 goto failed_mount;
2623 }
2624
2625 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2626 goto cantfind_ext4;
2627
2628 /* check blocks count against device size */
2629 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2630 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2631 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2632 "exceeds size of device (%llu blocks)",
2633 ext4_blocks_count(es), blocks_count);
2634 goto failed_mount;
2635 }
2636
2637 /*
2638 * It makes no sense for the first data block to be beyond the end
2639 * of the filesystem.
2640 */
2641 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2642 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2643 "block %u is beyond end of filesystem (%llu)",
2644 le32_to_cpu(es->s_first_data_block),
2645 ext4_blocks_count(es));
2646 goto failed_mount;
2647 }
2648 blocks_count = (ext4_blocks_count(es) -
2649 le32_to_cpu(es->s_first_data_block) +
2650 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2651 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2652 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2653 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
2654 "(block count %llu, first data block %u, "
2655 "blocks per group %lu)", sbi->s_groups_count,
2656 ext4_blocks_count(es),
2657 le32_to_cpu(es->s_first_data_block),
2658 EXT4_BLOCKS_PER_GROUP(sb));
2659 goto failed_mount;
2660 }
2661 sbi->s_groups_count = blocks_count;
2662 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
2663 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
2664 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2665 EXT4_DESC_PER_BLOCK(sb);
2666 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2667 GFP_KERNEL);
2668 if (sbi->s_group_desc == NULL) {
2669 ext4_msg(sb, KERN_ERR, "not enough memory");
2670 goto failed_mount;
2671 }
2672
2673 #ifdef CONFIG_PROC_FS
2674 if (ext4_proc_root)
2675 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2676 #endif
2677
2678 bgl_lock_init(sbi->s_blockgroup_lock);
2679
2680 for (i = 0; i < db_count; i++) {
2681 block = descriptor_loc(sb, logical_sb_block, i);
2682 sbi->s_group_desc[i] = sb_bread(sb, block);
2683 if (!sbi->s_group_desc[i]) {
2684 ext4_msg(sb, KERN_ERR,
2685 "can't read group descriptor %d", i);
2686 db_count = i;
2687 goto failed_mount2;
2688 }
2689 }
2690 if (!ext4_check_descriptors(sb)) {
2691 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
2692 goto failed_mount2;
2693 }
2694 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2695 if (!ext4_fill_flex_info(sb)) {
2696 ext4_msg(sb, KERN_ERR,
2697 "unable to initialize "
2698 "flex_bg meta info!");
2699 goto failed_mount2;
2700 }
2701
2702 sbi->s_gdb_count = db_count;
2703 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2704 spin_lock_init(&sbi->s_next_gen_lock);
2705
2706 sbi->s_stripe = ext4_get_stripe_size(sbi);
2707 sbi->s_max_writeback_mb_bump = 128;
2708
2709 /*
2710 * set up enough so that it can read an inode
2711 */
2712 if (!test_opt(sb, NOLOAD) &&
2713 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2714 sb->s_op = &ext4_sops;
2715 else
2716 sb->s_op = &ext4_nojournal_sops;
2717 sb->s_export_op = &ext4_export_ops;
2718 sb->s_xattr = ext4_xattr_handlers;
2719 #ifdef CONFIG_QUOTA
2720 sb->s_qcop = &ext4_qctl_operations;
2721 sb->dq_op = &ext4_quota_operations;
2722 #endif
2723 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2724 mutex_init(&sbi->s_orphan_lock);
2725 mutex_init(&sbi->s_resize_lock);
2726
2727 sb->s_root = NULL;
2728
2729 needs_recovery = (es->s_last_orphan != 0 ||
2730 EXT4_HAS_INCOMPAT_FEATURE(sb,
2731 EXT4_FEATURE_INCOMPAT_RECOVER));
2732
2733 /*
2734 * The first inode we look at is the journal inode. Don't try
2735 * root first: it may be modified in the journal!
2736 */
2737 if (!test_opt(sb, NOLOAD) &&
2738 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2739 if (ext4_load_journal(sb, es, journal_devnum))
2740 goto failed_mount3;
2741 if (!(sb->s_flags & MS_RDONLY) &&
2742 EXT4_SB(sb)->s_journal->j_failed_commit) {
2743 ext4_msg(sb, KERN_CRIT, "error: "
2744 "ext4_fill_super: Journal transaction "
2745 "%u is corrupt",
2746 EXT4_SB(sb)->s_journal->j_failed_commit);
2747 if (test_opt(sb, ERRORS_RO)) {
2748 ext4_msg(sb, KERN_CRIT,
2749 "Mounting filesystem read-only");
2750 sb->s_flags |= MS_RDONLY;
2751 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2752 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2753 }
2754 if (test_opt(sb, ERRORS_PANIC)) {
2755 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2756 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2757 ext4_commit_super(sb, 1);
2758 goto failed_mount4;
2759 }
2760 }
2761 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2762 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2763 ext4_msg(sb, KERN_ERR, "required journal recovery "
2764 "suppressed and not mounted read-only");
2765 goto failed_mount4;
2766 } else {
2767 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2768 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2769 sbi->s_journal = NULL;
2770 needs_recovery = 0;
2771 goto no_journal;
2772 }
2773
2774 if (ext4_blocks_count(es) > 0xffffffffULL &&
2775 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2776 JBD2_FEATURE_INCOMPAT_64BIT)) {
2777 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
2778 goto failed_mount4;
2779 }
2780
2781 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2782 jbd2_journal_set_features(sbi->s_journal,
2783 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2784 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2785 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2786 jbd2_journal_set_features(sbi->s_journal,
2787 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2788 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2789 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2790 } else {
2791 jbd2_journal_clear_features(sbi->s_journal,
2792 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2793 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2794 }
2795
2796 /* We have now updated the journal if required, so we can
2797 * validate the data journaling mode. */
2798 switch (test_opt(sb, DATA_FLAGS)) {
2799 case 0:
2800 /* No mode set, assume a default based on the journal
2801 * capabilities: ORDERED_DATA if the journal can
2802 * cope, else JOURNAL_DATA
2803 */
2804 if (jbd2_journal_check_available_features
2805 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2806 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2807 else
2808 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2809 break;
2810
2811 case EXT4_MOUNT_ORDERED_DATA:
2812 case EXT4_MOUNT_WRITEBACK_DATA:
2813 if (!jbd2_journal_check_available_features
2814 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2815 ext4_msg(sb, KERN_ERR, "Journal does not support "
2816 "requested data journaling mode");
2817 goto failed_mount4;
2818 }
2819 default:
2820 break;
2821 }
2822 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2823
2824 no_journal:
2825 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2826 ext4_count_free_blocks(sb));
2827 if (!err)
2828 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2829 ext4_count_free_inodes(sb));
2830 if (!err)
2831 err = percpu_counter_init(&sbi->s_dirs_counter,
2832 ext4_count_dirs(sb));
2833 if (!err)
2834 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2835 if (err) {
2836 ext4_msg(sb, KERN_ERR, "insufficient memory");
2837 goto failed_mount_wq;
2838 }
2839 if (test_opt(sb, NOBH)) {
2840 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2841 ext4_msg(sb, KERN_WARNING, "Ignoring nobh option - "
2842 "its supported only with writeback mode");
2843 clear_opt(sbi->s_mount_opt, NOBH);
2844 }
2845 }
2846 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
2847 if (!EXT4_SB(sb)->dio_unwritten_wq) {
2848 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
2849 goto failed_mount_wq;
2850 }
2851
2852 /*
2853 * The jbd2_journal_load will have done any necessary log recovery,
2854 * so we can safely mount the rest of the filesystem now.
2855 */
2856
2857 root = ext4_iget(sb, EXT4_ROOT_INO);
2858 if (IS_ERR(root)) {
2859 ext4_msg(sb, KERN_ERR, "get root inode failed");
2860 ret = PTR_ERR(root);
2861 goto failed_mount4;
2862 }
2863 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2864 iput(root);
2865 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
2866 goto failed_mount4;
2867 }
2868 sb->s_root = d_alloc_root(root);
2869 if (!sb->s_root) {
2870 ext4_msg(sb, KERN_ERR, "get root dentry failed");
2871 iput(root);
2872 ret = -ENOMEM;
2873 goto failed_mount4;
2874 }
2875
2876 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2877
2878 /* determine the minimum size of new large inodes, if present */
2879 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2880 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2881 EXT4_GOOD_OLD_INODE_SIZE;
2882 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2883 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2884 if (sbi->s_want_extra_isize <
2885 le16_to_cpu(es->s_want_extra_isize))
2886 sbi->s_want_extra_isize =
2887 le16_to_cpu(es->s_want_extra_isize);
2888 if (sbi->s_want_extra_isize <
2889 le16_to_cpu(es->s_min_extra_isize))
2890 sbi->s_want_extra_isize =
2891 le16_to_cpu(es->s_min_extra_isize);
2892 }
2893 }
2894 /* Check if enough inode space is available */
2895 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2896 sbi->s_inode_size) {
2897 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2898 EXT4_GOOD_OLD_INODE_SIZE;
2899 ext4_msg(sb, KERN_INFO, "required extra inode space not"
2900 "available");
2901 }
2902
2903 if (test_opt(sb, DELALLOC) &&
2904 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
2905 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
2906 "requested data journaling mode");
2907 clear_opt(sbi->s_mount_opt, DELALLOC);
2908 }
2909
2910 err = ext4_setup_system_zone(sb);
2911 if (err) {
2912 ext4_msg(sb, KERN_ERR, "failed to initialize system "
2913 "zone (%d)", err);
2914 goto failed_mount4;
2915 }
2916
2917 ext4_ext_init(sb);
2918 err = ext4_mb_init(sb, needs_recovery);
2919 if (err) {
2920 ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
2921 err);
2922 goto failed_mount4;
2923 }
2924
2925 sbi->s_kobj.kset = ext4_kset;
2926 init_completion(&sbi->s_kobj_unregister);
2927 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2928 "%s", sb->s_id);
2929 if (err) {
2930 ext4_mb_release(sb);
2931 ext4_ext_release(sb);
2932 goto failed_mount4;
2933 };
2934
2935 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2936 ext4_orphan_cleanup(sb, es);
2937 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2938 if (needs_recovery) {
2939 ext4_msg(sb, KERN_INFO, "recovery complete");
2940 ext4_mark_recovery_complete(sb, es);
2941 }
2942 if (EXT4_SB(sb)->s_journal) {
2943 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2944 descr = " journalled data mode";
2945 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2946 descr = " ordered data mode";
2947 else
2948 descr = " writeback data mode";
2949 } else
2950 descr = "out journal";
2951
2952 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s", descr);
2953
2954 lock_kernel();
2955 return 0;
2956
2957 cantfind_ext4:
2958 if (!silent)
2959 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
2960 goto failed_mount;
2961
2962 failed_mount4:
2963 ext4_msg(sb, KERN_ERR, "mount failed");
2964 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
2965 failed_mount_wq:
2966 ext4_release_system_zone(sb);
2967 if (sbi->s_journal) {
2968 jbd2_journal_destroy(sbi->s_journal);
2969 sbi->s_journal = NULL;
2970 }
2971 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2972 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2973 percpu_counter_destroy(&sbi->s_dirs_counter);
2974 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2975 failed_mount3:
2976 if (sbi->s_flex_groups) {
2977 if (is_vmalloc_addr(sbi->s_flex_groups))
2978 vfree(sbi->s_flex_groups);
2979 else
2980 kfree(sbi->s_flex_groups);
2981 }
2982 failed_mount2:
2983 for (i = 0; i < db_count; i++)
2984 brelse(sbi->s_group_desc[i]);
2985 kfree(sbi->s_group_desc);
2986 failed_mount:
2987 if (sbi->s_proc) {
2988 remove_proc_entry(sb->s_id, ext4_proc_root);
2989 }
2990 #ifdef CONFIG_QUOTA
2991 for (i = 0; i < MAXQUOTAS; i++)
2992 kfree(sbi->s_qf_names[i]);
2993 #endif
2994 ext4_blkdev_remove(sbi);
2995 brelse(bh);
2996 out_fail:
2997 sb->s_fs_info = NULL;
2998 kfree(sbi->s_blockgroup_lock);
2999 kfree(sbi);
3000 lock_kernel();
3001 return ret;
3002 }
3003
3004 /*
3005 * Setup any per-fs journal parameters now. We'll do this both on
3006 * initial mount, once the journal has been initialised but before we've
3007 * done any recovery; and again on any subsequent remount.
3008 */
3009 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3010 {
3011 struct ext4_sb_info *sbi = EXT4_SB(sb);
3012
3013 journal->j_commit_interval = sbi->s_commit_interval;
3014 journal->j_min_batch_time = sbi->s_min_batch_time;
3015 journal->j_max_batch_time = sbi->s_max_batch_time;
3016
3017 spin_lock(&journal->j_state_lock);
3018 if (test_opt(sb, BARRIER))
3019 journal->j_flags |= JBD2_BARRIER;
3020 else
3021 journal->j_flags &= ~JBD2_BARRIER;
3022 if (test_opt(sb, DATA_ERR_ABORT))
3023 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3024 else
3025 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3026 spin_unlock(&journal->j_state_lock);
3027 }
3028
3029 static journal_t *ext4_get_journal(struct super_block *sb,
3030 unsigned int journal_inum)
3031 {
3032 struct inode *journal_inode;
3033 journal_t *journal;
3034
3035 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3036
3037 /* First, test for the existence of a valid inode on disk. Bad
3038 * things happen if we iget() an unused inode, as the subsequent
3039 * iput() will try to delete it. */
3040
3041 journal_inode = ext4_iget(sb, journal_inum);
3042 if (IS_ERR(journal_inode)) {
3043 ext4_msg(sb, KERN_ERR, "no journal found");
3044 return NULL;
3045 }
3046 if (!journal_inode->i_nlink) {
3047 make_bad_inode(journal_inode);
3048 iput(journal_inode);
3049 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3050 return NULL;
3051 }
3052
3053 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3054 journal_inode, journal_inode->i_size);
3055 if (!S_ISREG(journal_inode->i_mode)) {
3056 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3057 iput(journal_inode);
3058 return NULL;
3059 }
3060
3061 journal = jbd2_journal_init_inode(journal_inode);
3062 if (!journal) {
3063 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3064 iput(journal_inode);
3065 return NULL;
3066 }
3067 journal->j_private = sb;
3068 ext4_init_journal_params(sb, journal);
3069 return journal;
3070 }
3071
3072 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3073 dev_t j_dev)
3074 {
3075 struct buffer_head *bh;
3076 journal_t *journal;
3077 ext4_fsblk_t start;
3078 ext4_fsblk_t len;
3079 int hblock, blocksize;
3080 ext4_fsblk_t sb_block;
3081 unsigned long offset;
3082 struct ext4_super_block *es;
3083 struct block_device *bdev;
3084
3085 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3086
3087 bdev = ext4_blkdev_get(j_dev, sb);
3088 if (bdev == NULL)
3089 return NULL;
3090
3091 if (bd_claim(bdev, sb)) {
3092 ext4_msg(sb, KERN_ERR,
3093 "failed to claim external journal device");
3094 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3095 return NULL;
3096 }
3097
3098 blocksize = sb->s_blocksize;
3099 hblock = bdev_logical_block_size(bdev);
3100 if (blocksize < hblock) {
3101 ext4_msg(sb, KERN_ERR,
3102 "blocksize too small for journal device");
3103 goto out_bdev;
3104 }
3105
3106 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3107 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3108 set_blocksize(bdev, blocksize);
3109 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3110 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3111 "external journal");
3112 goto out_bdev;
3113 }
3114
3115 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3116 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3117 !(le32_to_cpu(es->s_feature_incompat) &
3118 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3119 ext4_msg(sb, KERN_ERR, "external journal has "
3120 "bad superblock");
3121 brelse(bh);
3122 goto out_bdev;
3123 }
3124
3125 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3126 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3127 brelse(bh);
3128 goto out_bdev;
3129 }
3130
3131 len = ext4_blocks_count(es);
3132 start = sb_block + 1;
3133 brelse(bh); /* we're done with the superblock */
3134
3135 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3136 start, len, blocksize);
3137 if (!journal) {
3138 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3139 goto out_bdev;
3140 }
3141 journal->j_private = sb;
3142 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3143 wait_on_buffer(journal->j_sb_buffer);
3144 if (!buffer_uptodate(journal->j_sb_buffer)) {
3145 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3146 goto out_journal;
3147 }
3148 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3149 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3150 "user (unsupported) - %d",
3151 be32_to_cpu(journal->j_superblock->s_nr_users));
3152 goto out_journal;
3153 }
3154 EXT4_SB(sb)->journal_bdev = bdev;
3155 ext4_init_journal_params(sb, journal);
3156 return journal;
3157
3158 out_journal:
3159 jbd2_journal_destroy(journal);
3160 out_bdev:
3161 ext4_blkdev_put(bdev);
3162 return NULL;
3163 }
3164
3165 static int ext4_load_journal(struct super_block *sb,
3166 struct ext4_super_block *es,
3167 unsigned long journal_devnum)
3168 {
3169 journal_t *journal;
3170 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3171 dev_t journal_dev;
3172 int err = 0;
3173 int really_read_only;
3174
3175 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3176
3177 if (journal_devnum &&
3178 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3179 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3180 "numbers have changed");
3181 journal_dev = new_decode_dev(journal_devnum);
3182 } else
3183 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3184
3185 really_read_only = bdev_read_only(sb->s_bdev);
3186
3187 /*
3188 * Are we loading a blank journal or performing recovery after a
3189 * crash? For recovery, we need to check in advance whether we
3190 * can get read-write access to the device.
3191 */
3192 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3193 if (sb->s_flags & MS_RDONLY) {
3194 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3195 "required on readonly filesystem");
3196 if (really_read_only) {
3197 ext4_msg(sb, KERN_ERR, "write access "
3198 "unavailable, cannot proceed");
3199 return -EROFS;
3200 }
3201 ext4_msg(sb, KERN_INFO, "write access will "
3202 "be enabled during recovery");
3203 }
3204 }
3205
3206 if (journal_inum && journal_dev) {
3207 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3208 "and inode journals!");
3209 return -EINVAL;
3210 }
3211
3212 if (journal_inum) {
3213 if (!(journal = ext4_get_journal(sb, journal_inum)))
3214 return -EINVAL;
3215 } else {
3216 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3217 return -EINVAL;
3218 }
3219
3220 if (!(journal->j_flags & JBD2_BARRIER))
3221 ext4_msg(sb, KERN_INFO, "barriers disabled");
3222
3223 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3224 err = jbd2_journal_update_format(journal);
3225 if (err) {
3226 ext4_msg(sb, KERN_ERR, "error updating journal");
3227 jbd2_journal_destroy(journal);
3228 return err;
3229 }
3230 }
3231
3232 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3233 err = jbd2_journal_wipe(journal, !really_read_only);
3234 if (!err)
3235 err = jbd2_journal_load(journal);
3236
3237 if (err) {
3238 ext4_msg(sb, KERN_ERR, "error loading journal");
3239 jbd2_journal_destroy(journal);
3240 return err;
3241 }
3242
3243 EXT4_SB(sb)->s_journal = journal;
3244 ext4_clear_journal_err(sb, es);
3245
3246 if (journal_devnum &&
3247 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3248 es->s_journal_dev = cpu_to_le32(journal_devnum);
3249
3250 /* Make sure we flush the recovery flag to disk. */
3251 ext4_commit_super(sb, 1);
3252 }
3253
3254 return 0;
3255 }
3256
3257 static int ext4_commit_super(struct super_block *sb, int sync)
3258 {
3259 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3260 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3261 int error = 0;
3262
3263 if (!sbh)
3264 return error;
3265 if (buffer_write_io_error(sbh)) {
3266 /*
3267 * Oh, dear. A previous attempt to write the
3268 * superblock failed. This could happen because the
3269 * USB device was yanked out. Or it could happen to
3270 * be a transient write error and maybe the block will
3271 * be remapped. Nothing we can do but to retry the
3272 * write and hope for the best.
3273 */
3274 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3275 "superblock detected");
3276 clear_buffer_write_io_error(sbh);
3277 set_buffer_uptodate(sbh);
3278 }
3279 /*
3280 * If the file system is mounted read-only, don't update the
3281 * superblock write time. This avoids updating the superblock
3282 * write time when we are mounting the root file system
3283 * read/only but we need to replay the journal; at that point,
3284 * for people who are east of GMT and who make their clock
3285 * tick in localtime for Windows bug-for-bug compatibility,
3286 * the clock is set in the future, and this will cause e2fsck
3287 * to complain and force a full file system check.
3288 */
3289 if (!(sb->s_flags & MS_RDONLY))
3290 es->s_wtime = cpu_to_le32(get_seconds());
3291 es->s_kbytes_written =
3292 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3293 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3294 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3295 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3296 &EXT4_SB(sb)->s_freeblocks_counter));
3297 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3298 &EXT4_SB(sb)->s_freeinodes_counter));
3299 sb->s_dirt = 0;
3300 BUFFER_TRACE(sbh, "marking dirty");
3301 mark_buffer_dirty(sbh);
3302 if (sync) {
3303 error = sync_dirty_buffer(sbh);
3304 if (error)
3305 return error;
3306
3307 error = buffer_write_io_error(sbh);
3308 if (error) {
3309 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3310 "superblock");
3311 clear_buffer_write_io_error(sbh);
3312 set_buffer_uptodate(sbh);
3313 }
3314 }
3315 return error;
3316 }
3317
3318 /*
3319 * Have we just finished recovery? If so, and if we are mounting (or
3320 * remounting) the filesystem readonly, then we will end up with a
3321 * consistent fs on disk. Record that fact.
3322 */
3323 static void ext4_mark_recovery_complete(struct super_block *sb,
3324 struct ext4_super_block *es)
3325 {
3326 journal_t *journal = EXT4_SB(sb)->s_journal;
3327
3328 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3329 BUG_ON(journal != NULL);
3330 return;
3331 }
3332 jbd2_journal_lock_updates(journal);
3333 if (jbd2_journal_flush(journal) < 0)
3334 goto out;
3335
3336 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3337 sb->s_flags & MS_RDONLY) {
3338 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3339 ext4_commit_super(sb, 1);
3340 }
3341
3342 out:
3343 jbd2_journal_unlock_updates(journal);
3344 }
3345
3346 /*
3347 * If we are mounting (or read-write remounting) a filesystem whose journal
3348 * has recorded an error from a previous lifetime, move that error to the
3349 * main filesystem now.
3350 */
3351 static void ext4_clear_journal_err(struct super_block *sb,
3352 struct ext4_super_block *es)
3353 {
3354 journal_t *journal;
3355 int j_errno;
3356 const char *errstr;
3357
3358 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3359
3360 journal = EXT4_SB(sb)->s_journal;
3361
3362 /*
3363 * Now check for any error status which may have been recorded in the
3364 * journal by a prior ext4_error() or ext4_abort()
3365 */
3366
3367 j_errno = jbd2_journal_errno(journal);
3368 if (j_errno) {
3369 char nbuf[16];
3370
3371 errstr = ext4_decode_error(sb, j_errno, nbuf);
3372 ext4_warning(sb, __func__, "Filesystem error recorded "
3373 "from previous mount: %s", errstr);
3374 ext4_warning(sb, __func__, "Marking fs in need of "
3375 "filesystem check.");
3376
3377 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3378 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3379 ext4_commit_super(sb, 1);
3380
3381 jbd2_journal_clear_err(journal);
3382 }
3383 }
3384
3385 /*
3386 * Force the running and committing transactions to commit,
3387 * and wait on the commit.
3388 */
3389 int ext4_force_commit(struct super_block *sb)
3390 {
3391 journal_t *journal;
3392 int ret = 0;
3393
3394 if (sb->s_flags & MS_RDONLY)
3395 return 0;
3396
3397 journal = EXT4_SB(sb)->s_journal;
3398 if (journal) {
3399 vfs_check_frozen(sb, SB_FREEZE_TRANS);
3400 ret = ext4_journal_force_commit(journal);
3401 }
3402
3403 return ret;
3404 }
3405
3406 static void ext4_write_super(struct super_block *sb)
3407 {
3408 lock_super(sb);
3409 ext4_commit_super(sb, 1);
3410 unlock_super(sb);
3411 }
3412
3413 static int ext4_sync_fs(struct super_block *sb, int wait)
3414 {
3415 int ret = 0;
3416 tid_t target;
3417 struct ext4_sb_info *sbi = EXT4_SB(sb);
3418
3419 trace_ext4_sync_fs(sb, wait);
3420 flush_workqueue(sbi->dio_unwritten_wq);
3421 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
3422 if (wait)
3423 jbd2_log_wait_commit(sbi->s_journal, target);
3424 }
3425 return ret;
3426 }
3427
3428 /*
3429 * LVM calls this function before a (read-only) snapshot is created. This
3430 * gives us a chance to flush the journal completely and mark the fs clean.
3431 */
3432 static int ext4_freeze(struct super_block *sb)
3433 {
3434 int error = 0;
3435 journal_t *journal;
3436
3437 if (sb->s_flags & MS_RDONLY)
3438 return 0;
3439
3440 journal = EXT4_SB(sb)->s_journal;
3441
3442 /* Now we set up the journal barrier. */
3443 jbd2_journal_lock_updates(journal);
3444
3445 /*
3446 * Don't clear the needs_recovery flag if we failed to flush
3447 * the journal.
3448 */
3449 error = jbd2_journal_flush(journal);
3450 if (error < 0)
3451 goto out;
3452
3453 /* Journal blocked and flushed, clear needs_recovery flag. */
3454 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3455 error = ext4_commit_super(sb, 1);
3456 out:
3457 /* we rely on s_frozen to stop further updates */
3458 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3459 return error;
3460 }
3461
3462 /*
3463 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3464 * flag here, even though the filesystem is not technically dirty yet.
3465 */
3466 static int ext4_unfreeze(struct super_block *sb)
3467 {
3468 if (sb->s_flags & MS_RDONLY)
3469 return 0;
3470
3471 lock_super(sb);
3472 /* Reset the needs_recovery flag before the fs is unlocked. */
3473 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3474 ext4_commit_super(sb, 1);
3475 unlock_super(sb);
3476 return 0;
3477 }
3478
3479 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3480 {
3481 struct ext4_super_block *es;
3482 struct ext4_sb_info *sbi = EXT4_SB(sb);
3483 ext4_fsblk_t n_blocks_count = 0;
3484 unsigned long old_sb_flags;
3485 struct ext4_mount_options old_opts;
3486 ext4_group_t g;
3487 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3488 int err;
3489 #ifdef CONFIG_QUOTA
3490 int i;
3491 #endif
3492
3493 lock_kernel();
3494
3495 /* Store the original options */
3496 lock_super(sb);
3497 old_sb_flags = sb->s_flags;
3498 old_opts.s_mount_opt = sbi->s_mount_opt;
3499 old_opts.s_resuid = sbi->s_resuid;
3500 old_opts.s_resgid = sbi->s_resgid;
3501 old_opts.s_commit_interval = sbi->s_commit_interval;
3502 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3503 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3504 #ifdef CONFIG_QUOTA
3505 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3506 for (i = 0; i < MAXQUOTAS; i++)
3507 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3508 #endif
3509 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3510 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3511
3512 /*
3513 * Allow the "check" option to be passed as a remount option.
3514 */
3515 if (!parse_options(data, sb, NULL, &journal_ioprio,
3516 &n_blocks_count, 1)) {
3517 err = -EINVAL;
3518 goto restore_opts;
3519 }
3520
3521 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
3522 ext4_abort(sb, __func__, "Abort forced by user");
3523
3524 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3525 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3526
3527 es = sbi->s_es;
3528
3529 if (sbi->s_journal) {
3530 ext4_init_journal_params(sb, sbi->s_journal);
3531 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3532 }
3533
3534 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3535 n_blocks_count > ext4_blocks_count(es)) {
3536 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
3537 err = -EROFS;
3538 goto restore_opts;
3539 }
3540
3541 if (*flags & MS_RDONLY) {
3542 /*
3543 * First of all, the unconditional stuff we have to do
3544 * to disable replay of the journal when we next remount
3545 */
3546 sb->s_flags |= MS_RDONLY;
3547
3548 /*
3549 * OK, test if we are remounting a valid rw partition
3550 * readonly, and if so set the rdonly flag and then
3551 * mark the partition as valid again.
3552 */
3553 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3554 (sbi->s_mount_state & EXT4_VALID_FS))
3555 es->s_state = cpu_to_le16(sbi->s_mount_state);
3556
3557 if (sbi->s_journal)
3558 ext4_mark_recovery_complete(sb, es);
3559 } else {
3560 /* Make sure we can mount this feature set readwrite */
3561 if (!ext4_feature_set_ok(sb, 0)) {
3562 err = -EROFS;
3563 goto restore_opts;
3564 }
3565 /*
3566 * Make sure the group descriptor checksums
3567 * are sane. If they aren't, refuse to remount r/w.
3568 */
3569 for (g = 0; g < sbi->s_groups_count; g++) {
3570 struct ext4_group_desc *gdp =
3571 ext4_get_group_desc(sb, g, NULL);
3572
3573 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3574 ext4_msg(sb, KERN_ERR,
3575 "ext4_remount: Checksum for group %u failed (%u!=%u)",
3576 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3577 le16_to_cpu(gdp->bg_checksum));
3578 err = -EINVAL;
3579 goto restore_opts;
3580 }
3581 }
3582
3583 /*
3584 * If we have an unprocessed orphan list hanging
3585 * around from a previously readonly bdev mount,
3586 * require a full umount/remount for now.
3587 */
3588 if (es->s_last_orphan) {
3589 ext4_msg(sb, KERN_WARNING, "Couldn't "
3590 "remount RDWR because of unprocessed "
3591 "orphan inode list. Please "
3592 "umount/remount instead");
3593 err = -EINVAL;
3594 goto restore_opts;
3595 }
3596
3597 /*
3598 * Mounting a RDONLY partition read-write, so reread
3599 * and store the current valid flag. (It may have
3600 * been changed by e2fsck since we originally mounted
3601 * the partition.)
3602 */
3603 if (sbi->s_journal)
3604 ext4_clear_journal_err(sb, es);
3605 sbi->s_mount_state = le16_to_cpu(es->s_state);
3606 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3607 goto restore_opts;
3608 if (!ext4_setup_super(sb, es, 0))
3609 sb->s_flags &= ~MS_RDONLY;
3610 }
3611 }
3612 ext4_setup_system_zone(sb);
3613 if (sbi->s_journal == NULL)
3614 ext4_commit_super(sb, 1);
3615
3616 #ifdef CONFIG_QUOTA
3617 /* Release old quota file names */
3618 for (i = 0; i < MAXQUOTAS; i++)
3619 if (old_opts.s_qf_names[i] &&
3620 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3621 kfree(old_opts.s_qf_names[i]);
3622 #endif
3623 unlock_super(sb);
3624 unlock_kernel();
3625 return 0;
3626
3627 restore_opts:
3628 sb->s_flags = old_sb_flags;
3629 sbi->s_mount_opt = old_opts.s_mount_opt;
3630 sbi->s_resuid = old_opts.s_resuid;
3631 sbi->s_resgid = old_opts.s_resgid;
3632 sbi->s_commit_interval = old_opts.s_commit_interval;
3633 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3634 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3635 #ifdef CONFIG_QUOTA
3636 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3637 for (i = 0; i < MAXQUOTAS; i++) {
3638 if (sbi->s_qf_names[i] &&
3639 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3640 kfree(sbi->s_qf_names[i]);
3641 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3642 }
3643 #endif
3644 unlock_super(sb);
3645 unlock_kernel();
3646 return err;
3647 }
3648
3649 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3650 {
3651 struct super_block *sb = dentry->d_sb;
3652 struct ext4_sb_info *sbi = EXT4_SB(sb);
3653 struct ext4_super_block *es = sbi->s_es;
3654 u64 fsid;
3655
3656 if (test_opt(sb, MINIX_DF)) {
3657 sbi->s_overhead_last = 0;
3658 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3659 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3660 ext4_fsblk_t overhead = 0;
3661
3662 /*
3663 * Compute the overhead (FS structures). This is constant
3664 * for a given filesystem unless the number of block groups
3665 * changes so we cache the previous value until it does.
3666 */
3667
3668 /*
3669 * All of the blocks before first_data_block are
3670 * overhead
3671 */
3672 overhead = le32_to_cpu(es->s_first_data_block);
3673
3674 /*
3675 * Add the overhead attributed to the superblock and
3676 * block group descriptors. If the sparse superblocks
3677 * feature is turned on, then not all groups have this.
3678 */
3679 for (i = 0; i < ngroups; i++) {
3680 overhead += ext4_bg_has_super(sb, i) +
3681 ext4_bg_num_gdb(sb, i);
3682 cond_resched();
3683 }
3684
3685 /*
3686 * Every block group has an inode bitmap, a block
3687 * bitmap, and an inode table.
3688 */
3689 overhead += ngroups * (2 + sbi->s_itb_per_group);
3690 sbi->s_overhead_last = overhead;
3691 smp_wmb();
3692 sbi->s_blocks_last = ext4_blocks_count(es);
3693 }
3694
3695 buf->f_type = EXT4_SUPER_MAGIC;
3696 buf->f_bsize = sb->s_blocksize;
3697 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3698 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3699 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3700 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3701 if (buf->f_bfree < ext4_r_blocks_count(es))
3702 buf->f_bavail = 0;
3703 buf->f_files = le32_to_cpu(es->s_inodes_count);
3704 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3705 buf->f_namelen = EXT4_NAME_LEN;
3706 fsid = le64_to_cpup((void *)es->s_uuid) ^
3707 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3708 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3709 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3710
3711 return 0;
3712 }
3713
3714 /* Helper function for writing quotas on sync - we need to start transaction
3715 * before quota file is locked for write. Otherwise the are possible deadlocks:
3716 * Process 1 Process 2
3717 * ext4_create() quota_sync()
3718 * jbd2_journal_start() write_dquot()
3719 * vfs_dq_init() down(dqio_mutex)
3720 * down(dqio_mutex) jbd2_journal_start()
3721 *
3722 */
3723
3724 #ifdef CONFIG_QUOTA
3725
3726 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3727 {
3728 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3729 }
3730
3731 static int ext4_write_dquot(struct dquot *dquot)
3732 {
3733 int ret, err;
3734 handle_t *handle;
3735 struct inode *inode;
3736
3737 inode = dquot_to_inode(dquot);
3738 handle = ext4_journal_start(inode,
3739 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3740 if (IS_ERR(handle))
3741 return PTR_ERR(handle);
3742 ret = dquot_commit(dquot);
3743 err = ext4_journal_stop(handle);
3744 if (!ret)
3745 ret = err;
3746 return ret;
3747 }
3748
3749 static int ext4_acquire_dquot(struct dquot *dquot)
3750 {
3751 int ret, err;
3752 handle_t *handle;
3753
3754 handle = ext4_journal_start(dquot_to_inode(dquot),
3755 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3756 if (IS_ERR(handle))
3757 return PTR_ERR(handle);
3758 ret = dquot_acquire(dquot);
3759 err = ext4_journal_stop(handle);
3760 if (!ret)
3761 ret = err;
3762 return ret;
3763 }
3764
3765 static int ext4_release_dquot(struct dquot *dquot)
3766 {
3767 int ret, err;
3768 handle_t *handle;
3769
3770 handle = ext4_journal_start(dquot_to_inode(dquot),
3771 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3772 if (IS_ERR(handle)) {
3773 /* Release dquot anyway to avoid endless cycle in dqput() */
3774 dquot_release(dquot);
3775 return PTR_ERR(handle);
3776 }
3777 ret = dquot_release(dquot);
3778 err = ext4_journal_stop(handle);
3779 if (!ret)
3780 ret = err;
3781 return ret;
3782 }
3783
3784 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3785 {
3786 /* Are we journaling quotas? */
3787 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3788 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3789 dquot_mark_dquot_dirty(dquot);
3790 return ext4_write_dquot(dquot);
3791 } else {
3792 return dquot_mark_dquot_dirty(dquot);
3793 }
3794 }
3795
3796 static int ext4_write_info(struct super_block *sb, int type)
3797 {
3798 int ret, err;
3799 handle_t *handle;
3800
3801 /* Data block + inode block */
3802 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3803 if (IS_ERR(handle))
3804 return PTR_ERR(handle);
3805 ret = dquot_commit_info(sb, type);
3806 err = ext4_journal_stop(handle);
3807 if (!ret)
3808 ret = err;
3809 return ret;
3810 }
3811
3812 /*
3813 * Turn on quotas during mount time - we need to find
3814 * the quota file and such...
3815 */
3816 static int ext4_quota_on_mount(struct super_block *sb, int type)
3817 {
3818 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3819 EXT4_SB(sb)->s_jquota_fmt, type);
3820 }
3821
3822 /*
3823 * Standard function to be called on quota_on
3824 */
3825 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3826 char *name, int remount)
3827 {
3828 int err;
3829 struct path path;
3830
3831 if (!test_opt(sb, QUOTA))
3832 return -EINVAL;
3833 /* When remounting, no checks are needed and in fact, name is NULL */
3834 if (remount)
3835 return vfs_quota_on(sb, type, format_id, name, remount);
3836
3837 err = kern_path(name, LOOKUP_FOLLOW, &path);
3838 if (err)
3839 return err;
3840
3841 /* Quotafile not on the same filesystem? */
3842 if (path.mnt->mnt_sb != sb) {
3843 path_put(&path);
3844 return -EXDEV;
3845 }
3846 /* Journaling quota? */
3847 if (EXT4_SB(sb)->s_qf_names[type]) {
3848 /* Quotafile not in fs root? */
3849 if (path.dentry->d_parent != sb->s_root)
3850 ext4_msg(sb, KERN_WARNING,
3851 "Quota file not on filesystem root. "
3852 "Journaled quota will not work");
3853 }
3854
3855 /*
3856 * When we journal data on quota file, we have to flush journal to see
3857 * all updates to the file when we bypass pagecache...
3858 */
3859 if (EXT4_SB(sb)->s_journal &&
3860 ext4_should_journal_data(path.dentry->d_inode)) {
3861 /*
3862 * We don't need to lock updates but journal_flush() could
3863 * otherwise be livelocked...
3864 */
3865 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3866 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3867 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3868 if (err) {
3869 path_put(&path);
3870 return err;
3871 }
3872 }
3873
3874 err = vfs_quota_on_path(sb, type, format_id, &path);
3875 path_put(&path);
3876 return err;
3877 }
3878
3879 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3880 * acquiring the locks... As quota files are never truncated and quota code
3881 * itself serializes the operations (and noone else should touch the files)
3882 * we don't have to be afraid of races */
3883 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3884 size_t len, loff_t off)
3885 {
3886 struct inode *inode = sb_dqopt(sb)->files[type];
3887 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3888 int err = 0;
3889 int offset = off & (sb->s_blocksize - 1);
3890 int tocopy;
3891 size_t toread;
3892 struct buffer_head *bh;
3893 loff_t i_size = i_size_read(inode);
3894
3895 if (off > i_size)
3896 return 0;
3897 if (off+len > i_size)
3898 len = i_size-off;
3899 toread = len;
3900 while (toread > 0) {
3901 tocopy = sb->s_blocksize - offset < toread ?
3902 sb->s_blocksize - offset : toread;
3903 bh = ext4_bread(NULL, inode, blk, 0, &err);
3904 if (err)
3905 return err;
3906 if (!bh) /* A hole? */
3907 memset(data, 0, tocopy);
3908 else
3909 memcpy(data, bh->b_data+offset, tocopy);
3910 brelse(bh);
3911 offset = 0;
3912 toread -= tocopy;
3913 data += tocopy;
3914 blk++;
3915 }
3916 return len;
3917 }
3918
3919 /* Write to quotafile (we know the transaction is already started and has
3920 * enough credits) */
3921 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3922 const char *data, size_t len, loff_t off)
3923 {
3924 struct inode *inode = sb_dqopt(sb)->files[type];
3925 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3926 int err = 0;
3927 int offset = off & (sb->s_blocksize - 1);
3928 int tocopy;
3929 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3930 size_t towrite = len;
3931 struct buffer_head *bh;
3932 handle_t *handle = journal_current_handle();
3933
3934 if (EXT4_SB(sb)->s_journal && !handle) {
3935 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
3936 " cancelled because transaction is not started",
3937 (unsigned long long)off, (unsigned long long)len);
3938 return -EIO;
3939 }
3940 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3941 while (towrite > 0) {
3942 tocopy = sb->s_blocksize - offset < towrite ?
3943 sb->s_blocksize - offset : towrite;
3944 bh = ext4_bread(handle, inode, blk, 1, &err);
3945 if (!bh)
3946 goto out;
3947 if (journal_quota) {
3948 err = ext4_journal_get_write_access(handle, bh);
3949 if (err) {
3950 brelse(bh);
3951 goto out;
3952 }
3953 }
3954 lock_buffer(bh);
3955 memcpy(bh->b_data+offset, data, tocopy);
3956 flush_dcache_page(bh->b_page);
3957 unlock_buffer(bh);
3958 if (journal_quota)
3959 err = ext4_handle_dirty_metadata(handle, NULL, bh);
3960 else {
3961 /* Always do at least ordered writes for quotas */
3962 err = ext4_jbd2_file_inode(handle, inode);
3963 mark_buffer_dirty(bh);
3964 }
3965 brelse(bh);
3966 if (err)
3967 goto out;
3968 offset = 0;
3969 towrite -= tocopy;
3970 data += tocopy;
3971 blk++;
3972 }
3973 out:
3974 if (len == towrite) {
3975 mutex_unlock(&inode->i_mutex);
3976 return err;
3977 }
3978 if (inode->i_size < off+len-towrite) {
3979 i_size_write(inode, off+len-towrite);
3980 EXT4_I(inode)->i_disksize = inode->i_size;
3981 }
3982 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3983 ext4_mark_inode_dirty(handle, inode);
3984 mutex_unlock(&inode->i_mutex);
3985 return len - towrite;
3986 }
3987
3988 #endif
3989
3990 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
3991 const char *dev_name, void *data, struct vfsmount *mnt)
3992 {
3993 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
3994 }
3995
3996 static struct file_system_type ext4_fs_type = {
3997 .owner = THIS_MODULE,
3998 .name = "ext4",
3999 .get_sb = ext4_get_sb,
4000 .kill_sb = kill_block_super,
4001 .fs_flags = FS_REQUIRES_DEV,
4002 };
4003
4004 static int __init init_ext4_fs(void)
4005 {
4006 int err;
4007
4008 ext4_check_flag_values();
4009 err = init_ext4_system_zone();
4010 if (err)
4011 return err;
4012 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4013 if (!ext4_kset)
4014 goto out4;
4015 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
4016 err = init_ext4_mballoc();
4017 if (err)
4018 goto out3;
4019
4020 err = init_ext4_xattr();
4021 if (err)
4022 goto out2;
4023 err = init_inodecache();
4024 if (err)
4025 goto out1;
4026 err = register_filesystem(&ext4_fs_type);
4027 if (err)
4028 goto out;
4029 return 0;
4030 out:
4031 destroy_inodecache();
4032 out1:
4033 exit_ext4_xattr();
4034 out2:
4035 exit_ext4_mballoc();
4036 out3:
4037 remove_proc_entry("fs/ext4", NULL);
4038 kset_unregister(ext4_kset);
4039 out4:
4040 exit_ext4_system_zone();
4041 return err;
4042 }
4043
4044 static void __exit exit_ext4_fs(void)
4045 {
4046 unregister_filesystem(&ext4_fs_type);
4047 destroy_inodecache();
4048 exit_ext4_xattr();
4049 exit_ext4_mballoc();
4050 remove_proc_entry("fs/ext4", NULL);
4051 kset_unregister(ext4_kset);
4052 exit_ext4_system_zone();
4053 }
4054
4055 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4056 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4057 MODULE_LICENSE("GPL");
4058 module_init(init_ext4_fs)
4059 module_exit(exit_ext4_fs)
Linux 2.6 ibm-acpi/thinkpad-acpi driver git tree