?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
ext4: drop ext4dev compat
[linux-2.6/x86.git] / fs / ext4 / super.c
blob312211ee05af39618274736a49096cb88ff067de
1 /*
2 * linux/fs/ext4/super.c
3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19 #include <linux/module.h>
20 #include <linux/string.h>
21 #include <linux/fs.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/smp_lock.h>
30 #include <linux/buffer_head.h>
31 #include <linux/exportfs.h>
32 #include <linux/vfs.h>
33 #include <linux/random.h>
34 #include <linux/mount.h>
35 #include <linux/namei.h>
36 #include <linux/quotaops.h>
37 #include <linux/seq_file.h>
38 #include <linux/proc_fs.h>
39 #include <linux/ctype.h>
40 #include <linux/log2.h>
41 #include <linux/crc16.h>
42 #include <asm/uaccess.h>
43
44 #include "ext4.h"
45 #include "ext4_jbd2.h"
46 #include "xattr.h"
47 #include "acl.h"
48 #include "mballoc.h"
49
50 #define CREATE_TRACE_POINTS
51 #include <trace/events/ext4.h>
52
53 struct proc_dir_entry *ext4_proc_root;
54 static struct kset *ext4_kset;
55
56 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
57 unsigned long journal_devnum);
58 static int ext4_commit_super(struct super_block *sb, int sync);
59 static void ext4_mark_recovery_complete(struct super_block *sb,
60 struct ext4_super_block *es);
61 static void ext4_clear_journal_err(struct super_block *sb,
62 struct ext4_super_block *es);
63 static int ext4_sync_fs(struct super_block *sb, int wait);
64 static const char *ext4_decode_error(struct super_block *sb, int errno,
65 char nbuf[16]);
66 static int ext4_remount(struct super_block *sb, int *flags, char *data);
67 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
68 static int ext4_unfreeze(struct super_block *sb);
69 static void ext4_write_super(struct super_block *sb);
70 static int ext4_freeze(struct super_block *sb);
71
72
73 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
74 struct ext4_group_desc *bg)
75 {
76 return le32_to_cpu(bg->bg_block_bitmap_lo) |
77 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
78 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
79 }
80
81 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
82 struct ext4_group_desc *bg)
83 {
84 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
85 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
86 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
87 }
88
89 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
90 struct ext4_group_desc *bg)
91 {
92 return le32_to_cpu(bg->bg_inode_table_lo) |
93 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
94 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
95 }
96
97 __u32 ext4_free_blks_count(struct super_block *sb,
98 struct ext4_group_desc *bg)
99 {
100 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
101 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
102 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
103 }
104
105 __u32 ext4_free_inodes_count(struct super_block *sb,
106 struct ext4_group_desc *bg)
107 {
108 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
109 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
110 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
111 }
112
113 __u32 ext4_used_dirs_count(struct super_block *sb,
114 struct ext4_group_desc *bg)
115 {
116 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
117 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
118 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
119 }
120
121 __u32 ext4_itable_unused_count(struct super_block *sb,
122 struct ext4_group_desc *bg)
123 {
124 return le16_to_cpu(bg->bg_itable_unused_lo) |
125 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
126 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
127 }
128
129 void ext4_block_bitmap_set(struct super_block *sb,
130 struct ext4_group_desc *bg, ext4_fsblk_t blk)
131 {
132 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
133 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
134 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
135 }
136
137 void ext4_inode_bitmap_set(struct super_block *sb,
138 struct ext4_group_desc *bg, ext4_fsblk_t blk)
139 {
140 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
141 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
142 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
143 }
144
145 void ext4_inode_table_set(struct super_block *sb,
146 struct ext4_group_desc *bg, ext4_fsblk_t blk)
147 {
148 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
149 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
150 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
151 }
152
153 void ext4_free_blks_set(struct super_block *sb,
154 struct ext4_group_desc *bg, __u32 count)
155 {
156 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
157 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
158 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
159 }
160
161 void ext4_free_inodes_set(struct super_block *sb,
162 struct ext4_group_desc *bg, __u32 count)
163 {
164 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
165 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
166 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
167 }
168
169 void ext4_used_dirs_set(struct super_block *sb,
170 struct ext4_group_desc *bg, __u32 count)
171 {
172 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
173 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
174 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
175 }
176
177 void ext4_itable_unused_set(struct super_block *sb,
178 struct ext4_group_desc *bg, __u32 count)
179 {
180 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
181 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
182 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
183 }
184
185
186 /* Just increment the non-pointer handle value */
187 static handle_t *ext4_get_nojournal(void)
188 {
189 handle_t *handle = current->journal_info;
190 unsigned long ref_cnt = (unsigned long)handle;
191
192 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
193
194 ref_cnt++;
195 handle = (handle_t *)ref_cnt;
196
197 current->journal_info = handle;
198 return handle;
199 }
200
201
202 /* Decrement the non-pointer handle value */
203 static void ext4_put_nojournal(handle_t *handle)
204 {
205 unsigned long ref_cnt = (unsigned long)handle;
206
207 BUG_ON(ref_cnt == 0);
208
209 ref_cnt--;
210 handle = (handle_t *)ref_cnt;
211
212 current->journal_info = handle;
213 }
214
215 /*
216 * Wrappers for jbd2_journal_start/end.
217 *
218 * The only special thing we need to do here is to make sure that all
219 * journal_end calls result in the superblock being marked dirty, so
220 * that sync() will call the filesystem's write_super callback if
221 * appropriate.
222 */
223 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
224 {
225 journal_t *journal;
226
227 if (sb->s_flags & MS_RDONLY)
228 return ERR_PTR(-EROFS);
229
230 /* Special case here: if the journal has aborted behind our
231 * backs (eg. EIO in the commit thread), then we still need to
232 * take the FS itself readonly cleanly. */
233 journal = EXT4_SB(sb)->s_journal;
234 if (journal) {
235 if (is_journal_aborted(journal)) {
236 ext4_abort(sb, __func__, "Detected aborted journal");
237 return ERR_PTR(-EROFS);
238 }
239 return jbd2_journal_start(journal, nblocks);
240 }
241 return ext4_get_nojournal();
242 }
243
244 /*
245 * The only special thing we need to do here is to make sure that all
246 * jbd2_journal_stop calls result in the superblock being marked dirty, so
247 * that sync() will call the filesystem's write_super callback if
248 * appropriate.
249 */
250 int __ext4_journal_stop(const char *where, handle_t *handle)
251 {
252 struct super_block *sb;
253 int err;
254 int rc;
255
256 if (!ext4_handle_valid(handle)) {
257 ext4_put_nojournal(handle);
258 return 0;
259 }
260 sb = handle->h_transaction->t_journal->j_private;
261 err = handle->h_err;
262 rc = jbd2_journal_stop(handle);
263
264 if (!err)
265 err = rc;
266 if (err)
267 __ext4_std_error(sb, where, err);
268 return err;
269 }
270
271 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
272 struct buffer_head *bh, handle_t *handle, int err)
273 {
274 char nbuf[16];
275 const char *errstr = ext4_decode_error(NULL, err, nbuf);
276
277 BUG_ON(!ext4_handle_valid(handle));
278
279 if (bh)
280 BUFFER_TRACE(bh, "abort");
281
282 if (!handle->h_err)
283 handle->h_err = err;
284
285 if (is_handle_aborted(handle))
286 return;
287
288 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
289 caller, errstr, err_fn);
290
291 jbd2_journal_abort_handle(handle);
292 }
293
294 /* Deal with the reporting of failure conditions on a filesystem such as
295 * inconsistencies detected or read IO failures.
296 *
297 * On ext2, we can store the error state of the filesystem in the
298 * superblock. That is not possible on ext4, because we may have other
299 * write ordering constraints on the superblock which prevent us from
300 * writing it out straight away; and given that the journal is about to
301 * be aborted, we can't rely on the current, or future, transactions to
302 * write out the superblock safely.
303 *
304 * We'll just use the jbd2_journal_abort() error code to record an error in
305 * the journal instead. On recovery, the journal will compain about
306 * that error until we've noted it down and cleared it.
307 */
308
309 static void ext4_handle_error(struct super_block *sb)
310 {
311 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
312
313 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
314 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
315
316 if (sb->s_flags & MS_RDONLY)
317 return;
318
319 if (!test_opt(sb, ERRORS_CONT)) {
320 journal_t *journal = EXT4_SB(sb)->s_journal;
321
322 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
323 if (journal)
324 jbd2_journal_abort(journal, -EIO);
325 }
326 if (test_opt(sb, ERRORS_RO)) {
327 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
328 sb->s_flags |= MS_RDONLY;
329 }
330 ext4_commit_super(sb, 1);
331 if (test_opt(sb, ERRORS_PANIC))
332 panic("EXT4-fs (device %s): panic forced after error\n",
333 sb->s_id);
334 }
335
336 void ext4_error(struct super_block *sb, const char *function,
337 const char *fmt, ...)
338 {
339 va_list args;
340
341 va_start(args, fmt);
342 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
343 vprintk(fmt, args);
344 printk("\n");
345 va_end(args);
346
347 ext4_handle_error(sb);
348 }
349
350 static const char *ext4_decode_error(struct super_block *sb, int errno,
351 char nbuf[16])
352 {
353 char *errstr = NULL;
354
355 switch (errno) {
356 case -EIO:
357 errstr = "IO failure";
358 break;
359 case -ENOMEM:
360 errstr = "Out of memory";
361 break;
362 case -EROFS:
363 if (!sb || (EXT4_SB(sb)->s_journal &&
364 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
365 errstr = "Journal has aborted";
366 else
367 errstr = "Readonly filesystem";
368 break;
369 default:
370 /* If the caller passed in an extra buffer for unknown
371 * errors, textualise them now. Else we just return
372 * NULL. */
373 if (nbuf) {
374 /* Check for truncated error codes... */
375 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
376 errstr = nbuf;
377 }
378 break;
379 }
380
381 return errstr;
382 }
383
384 /* __ext4_std_error decodes expected errors from journaling functions
385 * automatically and invokes the appropriate error response. */
386
387 void __ext4_std_error(struct super_block *sb, const char *function, int errno)
388 {
389 char nbuf[16];
390 const char *errstr;
391
392 /* Special case: if the error is EROFS, and we're not already
393 * inside a transaction, then there's really no point in logging
394 * an error. */
395 if (errno == -EROFS && journal_current_handle() == NULL &&
396 (sb->s_flags & MS_RDONLY))
397 return;
398
399 errstr = ext4_decode_error(sb, errno, nbuf);
400 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
401 sb->s_id, function, errstr);
402
403 ext4_handle_error(sb);
404 }
405
406 /*
407 * ext4_abort is a much stronger failure handler than ext4_error. The
408 * abort function may be used to deal with unrecoverable failures such
409 * as journal IO errors or ENOMEM at a critical moment in log management.
410 *
411 * We unconditionally force the filesystem into an ABORT|READONLY state,
412 * unless the error response on the fs has been set to panic in which
413 * case we take the easy way out and panic immediately.
414 */
415
416 void ext4_abort(struct super_block *sb, const char *function,
417 const char *fmt, ...)
418 {
419 va_list args;
420
421 va_start(args, fmt);
422 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
423 vprintk(fmt, args);
424 printk("\n");
425 va_end(args);
426
427 if (test_opt(sb, ERRORS_PANIC))
428 panic("EXT4-fs panic from previous error\n");
429
430 if (sb->s_flags & MS_RDONLY)
431 return;
432
433 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
434 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
435 sb->s_flags |= MS_RDONLY;
436 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
437 if (EXT4_SB(sb)->s_journal)
438 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
439 }
440
441 void ext4_msg (struct super_block * sb, const char *prefix,
442 const char *fmt, ...)
443 {
444 va_list args;
445
446 va_start(args, fmt);
447 printk("%sEXT4-fs (%s): ", prefix, sb->s_id);
448 vprintk(fmt, args);
449 printk("\n");
450 va_end(args);
451 }
452
453 void ext4_warning(struct super_block *sb, const char *function,
454 const char *fmt, ...)
455 {
456 va_list args;
457
458 va_start(args, fmt);
459 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
460 sb->s_id, function);
461 vprintk(fmt, args);
462 printk("\n");
463 va_end(args);
464 }
465
466 void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp,
467 const char *function, const char *fmt, ...)
468 __releases(bitlock)
469 __acquires(bitlock)
470 {
471 va_list args;
472 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
473
474 va_start(args, fmt);
475 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
476 vprintk(fmt, args);
477 printk("\n");
478 va_end(args);
479
480 if (test_opt(sb, ERRORS_CONT)) {
481 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
482 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
483 ext4_commit_super(sb, 0);
484 return;
485 }
486 ext4_unlock_group(sb, grp);
487 ext4_handle_error(sb);
488 /*
489 * We only get here in the ERRORS_RO case; relocking the group
490 * may be dangerous, but nothing bad will happen since the
491 * filesystem will have already been marked read/only and the
492 * journal has been aborted. We return 1 as a hint to callers
493 * who might what to use the return value from
494 * ext4_grp_locked_error() to distinguish beween the
495 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
496 * aggressively from the ext4 function in question, with a
497 * more appropriate error code.
498 */
499 ext4_lock_group(sb, grp);
500 return;
501 }
502
503 void ext4_update_dynamic_rev(struct super_block *sb)
504 {
505 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
506
507 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
508 return;
509
510 ext4_warning(sb, __func__,
511 "updating to rev %d because of new feature flag, "
512 "running e2fsck is recommended",
513 EXT4_DYNAMIC_REV);
514
515 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
516 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
517 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
518 /* leave es->s_feature_*compat flags alone */
519 /* es->s_uuid will be set by e2fsck if empty */
520
521 /*
522 * The rest of the superblock fields should be zero, and if not it
523 * means they are likely already in use, so leave them alone. We
524 * can leave it up to e2fsck to clean up any inconsistencies there.
525 */
526 }
527
528 /*
529 * Open the external journal device
530 */
531 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
532 {
533 struct block_device *bdev;
534 char b[BDEVNAME_SIZE];
535
536 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
537 if (IS_ERR(bdev))
538 goto fail;
539 return bdev;
540
541 fail:
542 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
543 __bdevname(dev, b), PTR_ERR(bdev));
544 return NULL;
545 }
546
547 /*
548 * Release the journal device
549 */
550 static int ext4_blkdev_put(struct block_device *bdev)
551 {
552 bd_release(bdev);
553 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
554 }
555
556 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
557 {
558 struct block_device *bdev;
559 int ret = -ENODEV;
560
561 bdev = sbi->journal_bdev;
562 if (bdev) {
563 ret = ext4_blkdev_put(bdev);
564 sbi->journal_bdev = NULL;
565 }
566 return ret;
567 }
568
569 static inline struct inode *orphan_list_entry(struct list_head *l)
570 {
571 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
572 }
573
574 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
575 {
576 struct list_head *l;
577
578 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
579 le32_to_cpu(sbi->s_es->s_last_orphan));
580
581 printk(KERN_ERR "sb_info orphan list:\n");
582 list_for_each(l, &sbi->s_orphan) {
583 struct inode *inode = orphan_list_entry(l);
584 printk(KERN_ERR " "
585 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
586 inode->i_sb->s_id, inode->i_ino, inode,
587 inode->i_mode, inode->i_nlink,
588 NEXT_ORPHAN(inode));
589 }
590 }
591
592 static void ext4_put_super(struct super_block *sb)
593 {
594 struct ext4_sb_info *sbi = EXT4_SB(sb);
595 struct ext4_super_block *es = sbi->s_es;
596 int i, err;
597
598 flush_workqueue(sbi->dio_unwritten_wq);
599 destroy_workqueue(sbi->dio_unwritten_wq);
600
601 lock_super(sb);
602 lock_kernel();
603 if (sb->s_dirt)
604 ext4_commit_super(sb, 1);
605
606 ext4_release_system_zone(sb);
607 ext4_mb_release(sb);
608 ext4_ext_release(sb);
609 ext4_xattr_put_super(sb);
610 if (sbi->s_journal) {
611 err = jbd2_journal_destroy(sbi->s_journal);
612 sbi->s_journal = NULL;
613 if (err < 0)
614 ext4_abort(sb, __func__,
615 "Couldn't clean up the journal");
616 }
617 if (!(sb->s_flags & MS_RDONLY)) {
618 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
619 es->s_state = cpu_to_le16(sbi->s_mount_state);
620 ext4_commit_super(sb, 1);
621 }
622 if (sbi->s_proc) {
623 remove_proc_entry(sb->s_id, ext4_proc_root);
624 }
625 kobject_del(&sbi->s_kobj);
626
627 for (i = 0; i < sbi->s_gdb_count; i++)
628 brelse(sbi->s_group_desc[i]);
629 kfree(sbi->s_group_desc);
630 if (is_vmalloc_addr(sbi->s_flex_groups))
631 vfree(sbi->s_flex_groups);
632 else
633 kfree(sbi->s_flex_groups);
634 percpu_counter_destroy(&sbi->s_freeblocks_counter);
635 percpu_counter_destroy(&sbi->s_freeinodes_counter);
636 percpu_counter_destroy(&sbi->s_dirs_counter);
637 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
638 brelse(sbi->s_sbh);
639 #ifdef CONFIG_QUOTA
640 for (i = 0; i < MAXQUOTAS; i++)
641 kfree(sbi->s_qf_names[i]);
642 #endif
643
644 /* Debugging code just in case the in-memory inode orphan list
645 * isn't empty. The on-disk one can be non-empty if we've
646 * detected an error and taken the fs readonly, but the
647 * in-memory list had better be clean by this point. */
648 if (!list_empty(&sbi->s_orphan))
649 dump_orphan_list(sb, sbi);
650 J_ASSERT(list_empty(&sbi->s_orphan));
651
652 invalidate_bdev(sb->s_bdev);
653 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
654 /*
655 * Invalidate the journal device's buffers. We don't want them
656 * floating about in memory - the physical journal device may
657 * hotswapped, and it breaks the `ro-after' testing code.
658 */
659 sync_blockdev(sbi->journal_bdev);
660 invalidate_bdev(sbi->journal_bdev);
661 ext4_blkdev_remove(sbi);
662 }
663 sb->s_fs_info = NULL;
664 /*
665 * Now that we are completely done shutting down the
666 * superblock, we need to actually destroy the kobject.
667 */
668 unlock_kernel();
669 unlock_super(sb);
670 kobject_put(&sbi->s_kobj);
671 wait_for_completion(&sbi->s_kobj_unregister);
672 kfree(sbi->s_blockgroup_lock);
673 kfree(sbi);
674 }
675
676 static struct kmem_cache *ext4_inode_cachep;
677
678 /*
679 * Called inside transaction, so use GFP_NOFS
680 */
681 static struct inode *ext4_alloc_inode(struct super_block *sb)
682 {
683 struct ext4_inode_info *ei;
684
685 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
686 if (!ei)
687 return NULL;
688
689 ei->vfs_inode.i_version = 1;
690 ei->vfs_inode.i_data.writeback_index = 0;
691 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
692 INIT_LIST_HEAD(&ei->i_prealloc_list);
693 spin_lock_init(&ei->i_prealloc_lock);
694 /*
695 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
696 * therefore it can be null here. Don't check it, just initialize
697 * jinode.
698 */
699 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
700 ei->i_reserved_data_blocks = 0;
701 ei->i_reserved_meta_blocks = 0;
702 ei->i_allocated_meta_blocks = 0;
703 ei->i_delalloc_reserved_flag = 0;
704 spin_lock_init(&(ei->i_block_reservation_lock));
705 INIT_LIST_HEAD(&ei->i_aio_dio_complete_list);
706 ei->cur_aio_dio = NULL;
707
708 return &ei->vfs_inode;
709 }
710
711 static void ext4_destroy_inode(struct inode *inode)
712 {
713 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
714 ext4_msg(inode->i_sb, KERN_ERR,
715 "Inode %lu (%p): orphan list check failed!",
716 inode->i_ino, EXT4_I(inode));
717 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
718 EXT4_I(inode), sizeof(struct ext4_inode_info),
719 true);
720 dump_stack();
721 }
722 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
723 }
724
725 static void init_once(void *foo)
726 {
727 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
728
729 INIT_LIST_HEAD(&ei->i_orphan);
730 #ifdef CONFIG_EXT4_FS_XATTR
731 init_rwsem(&ei->xattr_sem);
732 #endif
733 init_rwsem(&ei->i_data_sem);
734 inode_init_once(&ei->vfs_inode);
735 }
736
737 static int init_inodecache(void)
738 {
739 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
740 sizeof(struct ext4_inode_info),
741 0, (SLAB_RECLAIM_ACCOUNT|
742 SLAB_MEM_SPREAD),
743 init_once);
744 if (ext4_inode_cachep == NULL)
745 return -ENOMEM;
746 return 0;
747 }
748
749 static void destroy_inodecache(void)
750 {
751 kmem_cache_destroy(ext4_inode_cachep);
752 }
753
754 static void ext4_clear_inode(struct inode *inode)
755 {
756 ext4_discard_preallocations(inode);
757 if (EXT4_JOURNAL(inode))
758 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
759 &EXT4_I(inode)->jinode);
760 }
761
762 static inline void ext4_show_quota_options(struct seq_file *seq,
763 struct super_block *sb)
764 {
765 #if defined(CONFIG_QUOTA)
766 struct ext4_sb_info *sbi = EXT4_SB(sb);
767
768 if (sbi->s_jquota_fmt)
769 seq_printf(seq, ",jqfmt=%s",
770 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
771
772 if (sbi->s_qf_names[USRQUOTA])
773 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
774
775 if (sbi->s_qf_names[GRPQUOTA])
776 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
777
778 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
779 seq_puts(seq, ",usrquota");
780
781 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
782 seq_puts(seq, ",grpquota");
783 #endif
784 }
785
786 /*
787 * Show an option if
788 * - it's set to a non-default value OR
789 * - if the per-sb default is different from the global default
790 */
791 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
792 {
793 int def_errors;
794 unsigned long def_mount_opts;
795 struct super_block *sb = vfs->mnt_sb;
796 struct ext4_sb_info *sbi = EXT4_SB(sb);
797 struct ext4_super_block *es = sbi->s_es;
798
799 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
800 def_errors = le16_to_cpu(es->s_errors);
801
802 if (sbi->s_sb_block != 1)
803 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
804 if (test_opt(sb, MINIX_DF))
805 seq_puts(seq, ",minixdf");
806 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
807 seq_puts(seq, ",grpid");
808 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
809 seq_puts(seq, ",nogrpid");
810 if (sbi->s_resuid != EXT4_DEF_RESUID ||
811 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
812 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
813 }
814 if (sbi->s_resgid != EXT4_DEF_RESGID ||
815 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
816 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
817 }
818 if (test_opt(sb, ERRORS_RO)) {
819 if (def_errors == EXT4_ERRORS_PANIC ||
820 def_errors == EXT4_ERRORS_CONTINUE) {
821 seq_puts(seq, ",errors=remount-ro");
822 }
823 }
824 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
825 seq_puts(seq, ",errors=continue");
826 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
827 seq_puts(seq, ",errors=panic");
828 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
829 seq_puts(seq, ",nouid32");
830 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
831 seq_puts(seq, ",debug");
832 if (test_opt(sb, OLDALLOC))
833 seq_puts(seq, ",oldalloc");
834 #ifdef CONFIG_EXT4_FS_XATTR
835 if (test_opt(sb, XATTR_USER) &&
836 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
837 seq_puts(seq, ",user_xattr");
838 if (!test_opt(sb, XATTR_USER) &&
839 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
840 seq_puts(seq, ",nouser_xattr");
841 }
842 #endif
843 #ifdef CONFIG_EXT4_FS_POSIX_ACL
844 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
845 seq_puts(seq, ",acl");
846 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
847 seq_puts(seq, ",noacl");
848 #endif
849 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
850 seq_printf(seq, ",commit=%u",
851 (unsigned) (sbi->s_commit_interval / HZ));
852 }
853 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
854 seq_printf(seq, ",min_batch_time=%u",
855 (unsigned) sbi->s_min_batch_time);
856 }
857 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
858 seq_printf(seq, ",max_batch_time=%u",
859 (unsigned) sbi->s_min_batch_time);
860 }
861
862 /*
863 * We're changing the default of barrier mount option, so
864 * let's always display its mount state so it's clear what its
865 * status is.
866 */
867 seq_puts(seq, ",barrier=");
868 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
869 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
870 seq_puts(seq, ",journal_async_commit");
871 if (test_opt(sb, NOBH))
872 seq_puts(seq, ",nobh");
873 if (test_opt(sb, I_VERSION))
874 seq_puts(seq, ",i_version");
875 if (!test_opt(sb, DELALLOC))
876 seq_puts(seq, ",nodelalloc");
877
878
879 if (sbi->s_stripe)
880 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
881 /*
882 * journal mode get enabled in different ways
883 * So just print the value even if we didn't specify it
884 */
885 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
886 seq_puts(seq, ",data=journal");
887 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
888 seq_puts(seq, ",data=ordered");
889 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
890 seq_puts(seq, ",data=writeback");
891
892 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
893 seq_printf(seq, ",inode_readahead_blks=%u",
894 sbi->s_inode_readahead_blks);
895
896 if (test_opt(sb, DATA_ERR_ABORT))
897 seq_puts(seq, ",data_err=abort");
898
899 if (test_opt(sb, NO_AUTO_DA_ALLOC))
900 seq_puts(seq, ",noauto_da_alloc");
901
902 ext4_show_quota_options(seq, sb);
903
904 return 0;
905 }
906
907 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
908 u64 ino, u32 generation)
909 {
910 struct inode *inode;
911
912 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
913 return ERR_PTR(-ESTALE);
914 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
915 return ERR_PTR(-ESTALE);
916
917 /* iget isn't really right if the inode is currently unallocated!!
918 *
919 * ext4_read_inode will return a bad_inode if the inode had been
920 * deleted, so we should be safe.
921 *
922 * Currently we don't know the generation for parent directory, so
923 * a generation of 0 means "accept any"
924 */
925 inode = ext4_iget(sb, ino);
926 if (IS_ERR(inode))
927 return ERR_CAST(inode);
928 if (generation && inode->i_generation != generation) {
929 iput(inode);
930 return ERR_PTR(-ESTALE);
931 }
932
933 return inode;
934 }
935
936 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
937 int fh_len, int fh_type)
938 {
939 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
940 ext4_nfs_get_inode);
941 }
942
943 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
944 int fh_len, int fh_type)
945 {
946 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
947 ext4_nfs_get_inode);
948 }
949
950 /*
951 * Try to release metadata pages (indirect blocks, directories) which are
952 * mapped via the block device. Since these pages could have journal heads
953 * which would prevent try_to_free_buffers() from freeing them, we must use
954 * jbd2 layer's try_to_free_buffers() function to release them.
955 */
956 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
957 gfp_t wait)
958 {
959 journal_t *journal = EXT4_SB(sb)->s_journal;
960
961 WARN_ON(PageChecked(page));
962 if (!page_has_buffers(page))
963 return 0;
964 if (journal)
965 return jbd2_journal_try_to_free_buffers(journal, page,
966 wait & ~__GFP_WAIT);
967 return try_to_free_buffers(page);
968 }
969
970 #ifdef CONFIG_QUOTA
971 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
972 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
973
974 static int ext4_write_dquot(struct dquot *dquot);
975 static int ext4_acquire_dquot(struct dquot *dquot);
976 static int ext4_release_dquot(struct dquot *dquot);
977 static int ext4_mark_dquot_dirty(struct dquot *dquot);
978 static int ext4_write_info(struct super_block *sb, int type);
979 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
980 char *path, int remount);
981 static int ext4_quota_on_mount(struct super_block *sb, int type);
982 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
983 size_t len, loff_t off);
984 static ssize_t ext4_quota_write(struct super_block *sb, int type,
985 const char *data, size_t len, loff_t off);
986
987 static const struct dquot_operations ext4_quota_operations = {
988 .initialize = dquot_initialize,
989 .drop = dquot_drop,
990 .alloc_space = dquot_alloc_space,
991 .reserve_space = dquot_reserve_space,
992 .claim_space = dquot_claim_space,
993 .release_rsv = dquot_release_reserved_space,
994 .get_reserved_space = ext4_get_reserved_space,
995 .alloc_inode = dquot_alloc_inode,
996 .free_space = dquot_free_space,
997 .free_inode = dquot_free_inode,
998 .transfer = dquot_transfer,
999 .write_dquot = ext4_write_dquot,
1000 .acquire_dquot = ext4_acquire_dquot,
1001 .release_dquot = ext4_release_dquot,
1002 .mark_dirty = ext4_mark_dquot_dirty,
1003 .write_info = ext4_write_info,
1004 .alloc_dquot = dquot_alloc,
1005 .destroy_dquot = dquot_destroy,
1006 };
1007
1008 static const struct quotactl_ops ext4_qctl_operations = {
1009 .quota_on = ext4_quota_on,
1010 .quota_off = vfs_quota_off,
1011 .quota_sync = vfs_quota_sync,
1012 .get_info = vfs_get_dqinfo,
1013 .set_info = vfs_set_dqinfo,
1014 .get_dqblk = vfs_get_dqblk,
1015 .set_dqblk = vfs_set_dqblk
1016 };
1017 #endif
1018
1019 static const struct super_operations ext4_sops = {
1020 .alloc_inode = ext4_alloc_inode,
1021 .destroy_inode = ext4_destroy_inode,
1022 .write_inode = ext4_write_inode,
1023 .dirty_inode = ext4_dirty_inode,
1024 .delete_inode = ext4_delete_inode,
1025 .put_super = ext4_put_super,
1026 .sync_fs = ext4_sync_fs,
1027 .freeze_fs = ext4_freeze,
1028 .unfreeze_fs = ext4_unfreeze,
1029 .statfs = ext4_statfs,
1030 .remount_fs = ext4_remount,
1031 .clear_inode = ext4_clear_inode,
1032 .show_options = ext4_show_options,
1033 #ifdef CONFIG_QUOTA
1034 .quota_read = ext4_quota_read,
1035 .quota_write = ext4_quota_write,
1036 #endif
1037 .bdev_try_to_free_page = bdev_try_to_free_page,
1038 };
1039
1040 static const struct super_operations ext4_nojournal_sops = {
1041 .alloc_inode = ext4_alloc_inode,
1042 .destroy_inode = ext4_destroy_inode,
1043 .write_inode = ext4_write_inode,
1044 .dirty_inode = ext4_dirty_inode,
1045 .delete_inode = ext4_delete_inode,
1046 .write_super = ext4_write_super,
1047 .put_super = ext4_put_super,
1048 .statfs = ext4_statfs,
1049 .remount_fs = ext4_remount,
1050 .clear_inode = ext4_clear_inode,
1051 .show_options = ext4_show_options,
1052 #ifdef CONFIG_QUOTA
1053 .quota_read = ext4_quota_read,
1054 .quota_write = ext4_quota_write,
1055 #endif
1056 .bdev_try_to_free_page = bdev_try_to_free_page,
1057 };
1058
1059 static const struct export_operations ext4_export_ops = {
1060 .fh_to_dentry = ext4_fh_to_dentry,
1061 .fh_to_parent = ext4_fh_to_parent,
1062 .get_parent = ext4_get_parent,
1063 };
1064
1065 enum {
1066 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1067 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1068 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1069 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1070 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1071 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1072 Opt_journal_update, Opt_journal_dev,
1073 Opt_journal_checksum, Opt_journal_async_commit,
1074 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1075 Opt_data_err_abort, Opt_data_err_ignore,
1076 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1077 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
1078 Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err, Opt_resize,
1079 Opt_usrquota, Opt_grpquota, Opt_i_version,
1080 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1081 Opt_block_validity, Opt_noblock_validity,
1082 Opt_inode_readahead_blks, Opt_journal_ioprio
1083 };
1084
1085 static const match_table_t tokens = {
1086 {Opt_bsd_df, "bsddf"},
1087 {Opt_minix_df, "minixdf"},
1088 {Opt_grpid, "grpid"},
1089 {Opt_grpid, "bsdgroups"},
1090 {Opt_nogrpid, "nogrpid"},
1091 {Opt_nogrpid, "sysvgroups"},
1092 {Opt_resgid, "resgid=%u"},
1093 {Opt_resuid, "resuid=%u"},
1094 {Opt_sb, "sb=%u"},
1095 {Opt_err_cont, "errors=continue"},
1096 {Opt_err_panic, "errors=panic"},
1097 {Opt_err_ro, "errors=remount-ro"},
1098 {Opt_nouid32, "nouid32"},
1099 {Opt_debug, "debug"},
1100 {Opt_oldalloc, "oldalloc"},
1101 {Opt_orlov, "orlov"},
1102 {Opt_user_xattr, "user_xattr"},
1103 {Opt_nouser_xattr, "nouser_xattr"},
1104 {Opt_acl, "acl"},
1105 {Opt_noacl, "noacl"},
1106 {Opt_noload, "noload"},
1107 {Opt_nobh, "nobh"},
1108 {Opt_bh, "bh"},
1109 {Opt_commit, "commit=%u"},
1110 {Opt_min_batch_time, "min_batch_time=%u"},
1111 {Opt_max_batch_time, "max_batch_time=%u"},
1112 {Opt_journal_update, "journal=update"},
1113 {Opt_journal_dev, "journal_dev=%u"},
1114 {Opt_journal_checksum, "journal_checksum"},
1115 {Opt_journal_async_commit, "journal_async_commit"},
1116 {Opt_abort, "abort"},
1117 {Opt_data_journal, "data=journal"},
1118 {Opt_data_ordered, "data=ordered"},
1119 {Opt_data_writeback, "data=writeback"},
1120 {Opt_data_err_abort, "data_err=abort"},
1121 {Opt_data_err_ignore, "data_err=ignore"},
1122 {Opt_offusrjquota, "usrjquota="},
1123 {Opt_usrjquota, "usrjquota=%s"},
1124 {Opt_offgrpjquota, "grpjquota="},
1125 {Opt_grpjquota, "grpjquota=%s"},
1126 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1127 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1128 {Opt_grpquota, "grpquota"},
1129 {Opt_noquota, "noquota"},
1130 {Opt_quota, "quota"},
1131 {Opt_usrquota, "usrquota"},
1132 {Opt_barrier, "barrier=%u"},
1133 {Opt_barrier, "barrier"},
1134 {Opt_nobarrier, "nobarrier"},
1135 {Opt_i_version, "i_version"},
1136 {Opt_stripe, "stripe=%u"},
1137 {Opt_resize, "resize"},
1138 {Opt_delalloc, "delalloc"},
1139 {Opt_nodelalloc, "nodelalloc"},
1140 {Opt_block_validity, "block_validity"},
1141 {Opt_noblock_validity, "noblock_validity"},
1142 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1143 {Opt_journal_ioprio, "journal_ioprio=%u"},
1144 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1145 {Opt_auto_da_alloc, "auto_da_alloc"},
1146 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1147 {Opt_err, NULL},
1148 };
1149
1150 static ext4_fsblk_t get_sb_block(void **data)
1151 {
1152 ext4_fsblk_t sb_block;
1153 char *options = (char *) *data;
1154
1155 if (!options || strncmp(options, "sb=", 3) != 0)
1156 return 1; /* Default location */
1157
1158 options += 3;
1159 /* TODO: use simple_strtoll with >32bit ext4 */
1160 sb_block = simple_strtoul(options, &options, 0);
1161 if (*options && *options != ',') {
1162 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1163 (char *) *data);
1164 return 1;
1165 }
1166 if (*options == ',')
1167 options++;
1168 *data = (void *) options;
1169
1170 return sb_block;
1171 }
1172
1173 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1174
1175 static int parse_options(char *options, struct super_block *sb,
1176 unsigned long *journal_devnum,
1177 unsigned int *journal_ioprio,
1178 ext4_fsblk_t *n_blocks_count, int is_remount)
1179 {
1180 struct ext4_sb_info *sbi = EXT4_SB(sb);
1181 char *p;
1182 substring_t args[MAX_OPT_ARGS];
1183 int data_opt = 0;
1184 int option;
1185 #ifdef CONFIG_QUOTA
1186 int qtype, qfmt;
1187 char *qname;
1188 #endif
1189
1190 if (!options)
1191 return 1;
1192
1193 while ((p = strsep(&options, ",")) != NULL) {
1194 int token;
1195 if (!*p)
1196 continue;
1197
1198 token = match_token(p, tokens, args);
1199 switch (token) {
1200 case Opt_bsd_df:
1201 clear_opt(sbi->s_mount_opt, MINIX_DF);
1202 break;
1203 case Opt_minix_df:
1204 set_opt(sbi->s_mount_opt, MINIX_DF);
1205 break;
1206 case Opt_grpid:
1207 set_opt(sbi->s_mount_opt, GRPID);
1208 break;
1209 case Opt_nogrpid:
1210 clear_opt(sbi->s_mount_opt, GRPID);
1211 break;
1212 case Opt_resuid:
1213 if (match_int(&args[0], &option))
1214 return 0;
1215 sbi->s_resuid = option;
1216 break;
1217 case Opt_resgid:
1218 if (match_int(&args[0], &option))
1219 return 0;
1220 sbi->s_resgid = option;
1221 break;
1222 case Opt_sb:
1223 /* handled by get_sb_block() instead of here */
1224 /* *sb_block = match_int(&args[0]); */
1225 break;
1226 case Opt_err_panic:
1227 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1228 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1229 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1230 break;
1231 case Opt_err_ro:
1232 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1233 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1234 set_opt(sbi->s_mount_opt, ERRORS_RO);
1235 break;
1236 case Opt_err_cont:
1237 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1238 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1239 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1240 break;
1241 case Opt_nouid32:
1242 set_opt(sbi->s_mount_opt, NO_UID32);
1243 break;
1244 case Opt_debug:
1245 set_opt(sbi->s_mount_opt, DEBUG);
1246 break;
1247 case Opt_oldalloc:
1248 set_opt(sbi->s_mount_opt, OLDALLOC);
1249 break;
1250 case Opt_orlov:
1251 clear_opt(sbi->s_mount_opt, OLDALLOC);
1252 break;
1253 #ifdef CONFIG_EXT4_FS_XATTR
1254 case Opt_user_xattr:
1255 set_opt(sbi->s_mount_opt, XATTR_USER);
1256 break;
1257 case Opt_nouser_xattr:
1258 clear_opt(sbi->s_mount_opt, XATTR_USER);
1259 break;
1260 #else
1261 case Opt_user_xattr:
1262 case Opt_nouser_xattr:
1263 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1264 break;
1265 #endif
1266 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1267 case Opt_acl:
1268 set_opt(sbi->s_mount_opt, POSIX_ACL);
1269 break;
1270 case Opt_noacl:
1271 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1272 break;
1273 #else
1274 case Opt_acl:
1275 case Opt_noacl:
1276 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1277 break;
1278 #endif
1279 case Opt_journal_update:
1280 /* @@@ FIXME */
1281 /* Eventually we will want to be able to create
1282 a journal file here. For now, only allow the
1283 user to specify an existing inode to be the
1284 journal file. */
1285 if (is_remount) {
1286 ext4_msg(sb, KERN_ERR,
1287 "Cannot specify journal on remount");
1288 return 0;
1289 }
1290 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1291 break;
1292 case Opt_journal_dev:
1293 if (is_remount) {
1294 ext4_msg(sb, KERN_ERR,
1295 "Cannot specify journal on remount");
1296 return 0;
1297 }
1298 if (match_int(&args[0], &option))
1299 return 0;
1300 *journal_devnum = option;
1301 break;
1302 case Opt_journal_checksum:
1303 break; /* Kept for backwards compatibility */
1304 case Opt_journal_async_commit:
1305 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1306 break;
1307 case Opt_noload:
1308 set_opt(sbi->s_mount_opt, NOLOAD);
1309 break;
1310 case Opt_commit:
1311 if (match_int(&args[0], &option))
1312 return 0;
1313 if (option < 0)
1314 return 0;
1315 if (option == 0)
1316 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1317 sbi->s_commit_interval = HZ * option;
1318 break;
1319 case Opt_max_batch_time:
1320 if (match_int(&args[0], &option))
1321 return 0;
1322 if (option < 0)
1323 return 0;
1324 if (option == 0)
1325 option = EXT4_DEF_MAX_BATCH_TIME;
1326 sbi->s_max_batch_time = option;
1327 break;
1328 case Opt_min_batch_time:
1329 if (match_int(&args[0], &option))
1330 return 0;
1331 if (option < 0)
1332 return 0;
1333 sbi->s_min_batch_time = option;
1334 break;
1335 case Opt_data_journal:
1336 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1337 goto datacheck;
1338 case Opt_data_ordered:
1339 data_opt = EXT4_MOUNT_ORDERED_DATA;
1340 goto datacheck;
1341 case Opt_data_writeback:
1342 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1343 datacheck:
1344 if (is_remount) {
1345 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1346 != data_opt) {
1347 ext4_msg(sb, KERN_ERR,
1348 "Cannot change data mode on remount");
1349 return 0;
1350 }
1351 } else {
1352 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1353 sbi->s_mount_opt |= data_opt;
1354 }
1355 break;
1356 case Opt_data_err_abort:
1357 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1358 break;
1359 case Opt_data_err_ignore:
1360 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1361 break;
1362 #ifdef CONFIG_QUOTA
1363 case Opt_usrjquota:
1364 qtype = USRQUOTA;
1365 goto set_qf_name;
1366 case Opt_grpjquota:
1367 qtype = GRPQUOTA;
1368 set_qf_name:
1369 if (sb_any_quota_loaded(sb) &&
1370 !sbi->s_qf_names[qtype]) {
1371 ext4_msg(sb, KERN_ERR,
1372 "Cannot change journaled "
1373 "quota options when quota turned on");
1374 return 0;
1375 }
1376 qname = match_strdup(&args[0]);
1377 if (!qname) {
1378 ext4_msg(sb, KERN_ERR,
1379 "Not enough memory for "
1380 "storing quotafile name");
1381 return 0;
1382 }
1383 if (sbi->s_qf_names[qtype] &&
1384 strcmp(sbi->s_qf_names[qtype], qname)) {
1385 ext4_msg(sb, KERN_ERR,
1386 "%s quota file already "
1387 "specified", QTYPE2NAME(qtype));
1388 kfree(qname);
1389 return 0;
1390 }
1391 sbi->s_qf_names[qtype] = qname;
1392 if (strchr(sbi->s_qf_names[qtype], '/')) {
1393 ext4_msg(sb, KERN_ERR,
1394 "quotafile must be on "
1395 "filesystem root");
1396 kfree(sbi->s_qf_names[qtype]);
1397 sbi->s_qf_names[qtype] = NULL;
1398 return 0;
1399 }
1400 set_opt(sbi->s_mount_opt, QUOTA);
1401 break;
1402 case Opt_offusrjquota:
1403 qtype = USRQUOTA;
1404 goto clear_qf_name;
1405 case Opt_offgrpjquota:
1406 qtype = GRPQUOTA;
1407 clear_qf_name:
1408 if (sb_any_quota_loaded(sb) &&
1409 sbi->s_qf_names[qtype]) {
1410 ext4_msg(sb, KERN_ERR, "Cannot change "
1411 "journaled quota options when "
1412 "quota turned on");
1413 return 0;
1414 }
1415 /*
1416 * The space will be released later when all options
1417 * are confirmed to be correct
1418 */
1419 sbi->s_qf_names[qtype] = NULL;
1420 break;
1421 case Opt_jqfmt_vfsold:
1422 qfmt = QFMT_VFS_OLD;
1423 goto set_qf_format;
1424 case Opt_jqfmt_vfsv0:
1425 qfmt = QFMT_VFS_V0;
1426 set_qf_format:
1427 if (sb_any_quota_loaded(sb) &&
1428 sbi->s_jquota_fmt != qfmt) {
1429 ext4_msg(sb, KERN_ERR, "Cannot change "
1430 "journaled quota options when "
1431 "quota turned on");
1432 return 0;
1433 }
1434 sbi->s_jquota_fmt = qfmt;
1435 break;
1436 case Opt_quota:
1437 case Opt_usrquota:
1438 set_opt(sbi->s_mount_opt, QUOTA);
1439 set_opt(sbi->s_mount_opt, USRQUOTA);
1440 break;
1441 case Opt_grpquota:
1442 set_opt(sbi->s_mount_opt, QUOTA);
1443 set_opt(sbi->s_mount_opt, GRPQUOTA);
1444 break;
1445 case Opt_noquota:
1446 if (sb_any_quota_loaded(sb)) {
1447 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1448 "options when quota turned on");
1449 return 0;
1450 }
1451 clear_opt(sbi->s_mount_opt, QUOTA);
1452 clear_opt(sbi->s_mount_opt, USRQUOTA);
1453 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1454 break;
1455 #else
1456 case Opt_quota:
1457 case Opt_usrquota:
1458 case Opt_grpquota:
1459 ext4_msg(sb, KERN_ERR,
1460 "quota options not supported");
1461 break;
1462 case Opt_usrjquota:
1463 case Opt_grpjquota:
1464 case Opt_offusrjquota:
1465 case Opt_offgrpjquota:
1466 case Opt_jqfmt_vfsold:
1467 case Opt_jqfmt_vfsv0:
1468 ext4_msg(sb, KERN_ERR,
1469 "journaled quota options not supported");
1470 break;
1471 case Opt_noquota:
1472 break;
1473 #endif
1474 case Opt_abort:
1475 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1476 break;
1477 case Opt_nobarrier:
1478 clear_opt(sbi->s_mount_opt, BARRIER);
1479 break;
1480 case Opt_barrier:
1481 if (match_int(&args[0], &option)) {
1482 set_opt(sbi->s_mount_opt, BARRIER);
1483 break;
1484 }
1485 if (option)
1486 set_opt(sbi->s_mount_opt, BARRIER);
1487 else
1488 clear_opt(sbi->s_mount_opt, BARRIER);
1489 break;
1490 case Opt_ignore:
1491 break;
1492 case Opt_resize:
1493 if (!is_remount) {
1494 ext4_msg(sb, KERN_ERR,
1495 "resize option only available "
1496 "for remount");
1497 return 0;
1498 }
1499 if (match_int(&args[0], &option) != 0)
1500 return 0;
1501 *n_blocks_count = option;
1502 break;
1503 case Opt_nobh:
1504 set_opt(sbi->s_mount_opt, NOBH);
1505 break;
1506 case Opt_bh:
1507 clear_opt(sbi->s_mount_opt, NOBH);
1508 break;
1509 case Opt_i_version:
1510 set_opt(sbi->s_mount_opt, I_VERSION);
1511 sb->s_flags |= MS_I_VERSION;
1512 break;
1513 case Opt_nodelalloc:
1514 clear_opt(sbi->s_mount_opt, DELALLOC);
1515 break;
1516 case Opt_stripe:
1517 if (match_int(&args[0], &option))
1518 return 0;
1519 if (option < 0)
1520 return 0;
1521 sbi->s_stripe = option;
1522 break;
1523 case Opt_delalloc:
1524 set_opt(sbi->s_mount_opt, DELALLOC);
1525 break;
1526 case Opt_block_validity:
1527 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1528 break;
1529 case Opt_noblock_validity:
1530 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1531 break;
1532 case Opt_inode_readahead_blks:
1533 if (match_int(&args[0], &option))
1534 return 0;
1535 if (option < 0 || option > (1 << 30))
1536 return 0;
1537 if (!is_power_of_2(option)) {
1538 ext4_msg(sb, KERN_ERR,
1539 "EXT4-fs: inode_readahead_blks"
1540 " must be a power of 2");
1541 return 0;
1542 }
1543 sbi->s_inode_readahead_blks = option;
1544 break;
1545 case Opt_journal_ioprio:
1546 if (match_int(&args[0], &option))
1547 return 0;
1548 if (option < 0 || option > 7)
1549 break;
1550 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1551 option);
1552 break;
1553 case Opt_noauto_da_alloc:
1554 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1555 break;
1556 case Opt_auto_da_alloc:
1557 if (match_int(&args[0], &option)) {
1558 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1559 break;
1560 }
1561 if (option)
1562 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1563 else
1564 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1565 break;
1566 default:
1567 ext4_msg(sb, KERN_ERR,
1568 "Unrecognized mount option \"%s\" "
1569 "or missing value", p);
1570 return 0;
1571 }
1572 }
1573 #ifdef CONFIG_QUOTA
1574 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1575 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1576 sbi->s_qf_names[USRQUOTA])
1577 clear_opt(sbi->s_mount_opt, USRQUOTA);
1578
1579 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1580 sbi->s_qf_names[GRPQUOTA])
1581 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1582
1583 if ((sbi->s_qf_names[USRQUOTA] &&
1584 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1585 (sbi->s_qf_names[GRPQUOTA] &&
1586 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1587 ext4_msg(sb, KERN_ERR, "old and new quota "
1588 "format mixing");
1589 return 0;
1590 }
1591
1592 if (!sbi->s_jquota_fmt) {
1593 ext4_msg(sb, KERN_ERR, "journaled quota format "
1594 "not specified");
1595 return 0;
1596 }
1597 } else {
1598 if (sbi->s_jquota_fmt) {
1599 ext4_msg(sb, KERN_ERR, "journaled quota format "
1600 "specified with no journaling "
1601 "enabled");
1602 return 0;
1603 }
1604 }
1605 #endif
1606 return 1;
1607 }
1608
1609 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1610 int read_only)
1611 {
1612 struct ext4_sb_info *sbi = EXT4_SB(sb);
1613 int res = 0;
1614
1615 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1616 ext4_msg(sb, KERN_ERR, "revision level too high, "
1617 "forcing read-only mode");
1618 res = MS_RDONLY;
1619 }
1620 if (read_only)
1621 return res;
1622 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1623 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1624 "running e2fsck is recommended");
1625 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1626 ext4_msg(sb, KERN_WARNING,
1627 "warning: mounting fs with errors, "
1628 "running e2fsck is recommended");
1629 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1630 le16_to_cpu(es->s_mnt_count) >=
1631 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1632 ext4_msg(sb, KERN_WARNING,
1633 "warning: maximal mount count reached, "
1634 "running e2fsck is recommended");
1635 else if (le32_to_cpu(es->s_checkinterval) &&
1636 (le32_to_cpu(es->s_lastcheck) +
1637 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1638 ext4_msg(sb, KERN_WARNING,
1639 "warning: checktime reached, "
1640 "running e2fsck is recommended");
1641 if (!sbi->s_journal)
1642 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1643 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1644 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1645 le16_add_cpu(&es->s_mnt_count, 1);
1646 es->s_mtime = cpu_to_le32(get_seconds());
1647 ext4_update_dynamic_rev(sb);
1648 if (sbi->s_journal)
1649 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1650
1651 ext4_commit_super(sb, 1);
1652 if (test_opt(sb, DEBUG))
1653 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1654 "bpg=%lu, ipg=%lu, mo=%04x]\n",
1655 sb->s_blocksize,
1656 sbi->s_groups_count,
1657 EXT4_BLOCKS_PER_GROUP(sb),
1658 EXT4_INODES_PER_GROUP(sb),
1659 sbi->s_mount_opt);
1660
1661 return res;
1662 }
1663
1664 static int ext4_fill_flex_info(struct super_block *sb)
1665 {
1666 struct ext4_sb_info *sbi = EXT4_SB(sb);
1667 struct ext4_group_desc *gdp = NULL;
1668 ext4_group_t flex_group_count;
1669 ext4_group_t flex_group;
1670 int groups_per_flex = 0;
1671 size_t size;
1672 int i;
1673
1674 if (!sbi->s_es->s_log_groups_per_flex) {
1675 sbi->s_log_groups_per_flex = 0;
1676 return 1;
1677 }
1678
1679 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1680 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1681
1682 /* We allocate both existing and potentially added groups */
1683 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1684 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1685 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1686 size = flex_group_count * sizeof(struct flex_groups);
1687 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1688 if (sbi->s_flex_groups == NULL) {
1689 sbi->s_flex_groups = vmalloc(size);
1690 if (sbi->s_flex_groups)
1691 memset(sbi->s_flex_groups, 0, size);
1692 }
1693 if (sbi->s_flex_groups == NULL) {
1694 ext4_msg(sb, KERN_ERR, "not enough memory for "
1695 "%u flex groups", flex_group_count);
1696 goto failed;
1697 }
1698
1699 for (i = 0; i < sbi->s_groups_count; i++) {
1700 gdp = ext4_get_group_desc(sb, i, NULL);
1701
1702 flex_group = ext4_flex_group(sbi, i);
1703 atomic_add(ext4_free_inodes_count(sb, gdp),
1704 &sbi->s_flex_groups[flex_group].free_inodes);
1705 atomic_add(ext4_free_blks_count(sb, gdp),
1706 &sbi->s_flex_groups[flex_group].free_blocks);
1707 atomic_add(ext4_used_dirs_count(sb, gdp),
1708 &sbi->s_flex_groups[flex_group].used_dirs);
1709 }
1710
1711 return 1;
1712 failed:
1713 return 0;
1714 }
1715
1716 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1717 struct ext4_group_desc *gdp)
1718 {
1719 __u16 crc = 0;
1720
1721 if (sbi->s_es->s_feature_ro_compat &
1722 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1723 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1724 __le32 le_group = cpu_to_le32(block_group);
1725
1726 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1727 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1728 crc = crc16(crc, (__u8 *)gdp, offset);
1729 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1730 /* for checksum of struct ext4_group_desc do the rest...*/
1731 if ((sbi->s_es->s_feature_incompat &
1732 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1733 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1734 crc = crc16(crc, (__u8 *)gdp + offset,
1735 le16_to_cpu(sbi->s_es->s_desc_size) -
1736 offset);
1737 }
1738
1739 return cpu_to_le16(crc);
1740 }
1741
1742 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1743 struct ext4_group_desc *gdp)
1744 {
1745 if ((sbi->s_es->s_feature_ro_compat &
1746 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1747 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1748 return 0;
1749
1750 return 1;
1751 }
1752
1753 /* Called at mount-time, super-block is locked */
1754 static int ext4_check_descriptors(struct super_block *sb)
1755 {
1756 struct ext4_sb_info *sbi = EXT4_SB(sb);
1757 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1758 ext4_fsblk_t last_block;
1759 ext4_fsblk_t block_bitmap;
1760 ext4_fsblk_t inode_bitmap;
1761 ext4_fsblk_t inode_table;
1762 int flexbg_flag = 0;
1763 ext4_group_t i;
1764
1765 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1766 flexbg_flag = 1;
1767
1768 ext4_debug("Checking group descriptors");
1769
1770 for (i = 0; i < sbi->s_groups_count; i++) {
1771 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1772
1773 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1774 last_block = ext4_blocks_count(sbi->s_es) - 1;
1775 else
1776 last_block = first_block +
1777 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1778
1779 block_bitmap = ext4_block_bitmap(sb, gdp);
1780 if (block_bitmap < first_block || block_bitmap > last_block) {
1781 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1782 "Block bitmap for group %u not in group "
1783 "(block %llu)!", i, block_bitmap);
1784 return 0;
1785 }
1786 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1787 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1788 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1789 "Inode bitmap for group %u not in group "
1790 "(block %llu)!", i, inode_bitmap);
1791 return 0;
1792 }
1793 inode_table = ext4_inode_table(sb, gdp);
1794 if (inode_table < first_block ||
1795 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1796 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1797 "Inode table for group %u not in group "
1798 "(block %llu)!", i, inode_table);
1799 return 0;
1800 }
1801 ext4_lock_group(sb, i);
1802 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1803 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1804 "Checksum for group %u failed (%u!=%u)",
1805 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1806 gdp)), le16_to_cpu(gdp->bg_checksum));
1807 if (!(sb->s_flags & MS_RDONLY)) {
1808 ext4_unlock_group(sb, i);
1809 return 0;
1810 }
1811 }
1812 ext4_unlock_group(sb, i);
1813 if (!flexbg_flag)
1814 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1815 }
1816
1817 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1818 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
1819 return 1;
1820 }
1821
1822 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1823 * the superblock) which were deleted from all directories, but held open by
1824 * a process at the time of a crash. We walk the list and try to delete these
1825 * inodes at recovery time (only with a read-write filesystem).
1826 *
1827 * In order to keep the orphan inode chain consistent during traversal (in
1828 * case of crash during recovery), we link each inode into the superblock
1829 * orphan list_head and handle it the same way as an inode deletion during
1830 * normal operation (which journals the operations for us).
1831 *
1832 * We only do an iget() and an iput() on each inode, which is very safe if we
1833 * accidentally point at an in-use or already deleted inode. The worst that
1834 * can happen in this case is that we get a "bit already cleared" message from
1835 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1836 * e2fsck was run on this filesystem, and it must have already done the orphan
1837 * inode cleanup for us, so we can safely abort without any further action.
1838 */
1839 static void ext4_orphan_cleanup(struct super_block *sb,
1840 struct ext4_super_block *es)
1841 {
1842 unsigned int s_flags = sb->s_flags;
1843 int nr_orphans = 0, nr_truncates = 0;
1844 #ifdef CONFIG_QUOTA
1845 int i;
1846 #endif
1847 if (!es->s_last_orphan) {
1848 jbd_debug(4, "no orphan inodes to clean up\n");
1849 return;
1850 }
1851
1852 if (bdev_read_only(sb->s_bdev)) {
1853 ext4_msg(sb, KERN_ERR, "write access "
1854 "unavailable, skipping orphan cleanup");
1855 return;
1856 }
1857
1858 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1859 if (es->s_last_orphan)
1860 jbd_debug(1, "Errors on filesystem, "
1861 "clearing orphan list.\n");
1862 es->s_last_orphan = 0;
1863 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1864 return;
1865 }
1866
1867 if (s_flags & MS_RDONLY) {
1868 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1869 sb->s_flags &= ~MS_RDONLY;
1870 }
1871 #ifdef CONFIG_QUOTA
1872 /* Needed for iput() to work correctly and not trash data */
1873 sb->s_flags |= MS_ACTIVE;
1874 /* Turn on quotas so that they are updated correctly */
1875 for (i = 0; i < MAXQUOTAS; i++) {
1876 if (EXT4_SB(sb)->s_qf_names[i]) {
1877 int ret = ext4_quota_on_mount(sb, i);
1878 if (ret < 0)
1879 ext4_msg(sb, KERN_ERR,
1880 "Cannot turn on journaled "
1881 "quota: error %d", ret);
1882 }
1883 }
1884 #endif
1885
1886 while (es->s_last_orphan) {
1887 struct inode *inode;
1888
1889 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1890 if (IS_ERR(inode)) {
1891 es->s_last_orphan = 0;
1892 break;
1893 }
1894
1895 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1896 vfs_dq_init(inode);
1897 if (inode->i_nlink) {
1898 ext4_msg(sb, KERN_DEBUG,
1899 "%s: truncating inode %lu to %lld bytes",
1900 __func__, inode->i_ino, inode->i_size);
1901 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1902 inode->i_ino, inode->i_size);
1903 ext4_truncate(inode);
1904 nr_truncates++;
1905 } else {
1906 ext4_msg(sb, KERN_DEBUG,
1907 "%s: deleting unreferenced inode %lu",
1908 __func__, inode->i_ino);
1909 jbd_debug(2, "deleting unreferenced inode %lu\n",
1910 inode->i_ino);
1911 nr_orphans++;
1912 }
1913 iput(inode); /* The delete magic happens here! */
1914 }
1915
1916 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1917
1918 if (nr_orphans)
1919 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1920 PLURAL(nr_orphans));
1921 if (nr_truncates)
1922 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1923 PLURAL(nr_truncates));
1924 #ifdef CONFIG_QUOTA
1925 /* Turn quotas off */
1926 for (i = 0; i < MAXQUOTAS; i++) {
1927 if (sb_dqopt(sb)->files[i])
1928 vfs_quota_off(sb, i, 0);
1929 }
1930 #endif
1931 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1932 }
1933
1934 /*
1935 * Maximal extent format file size.
1936 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1937 * extent format containers, within a sector_t, and within i_blocks
1938 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1939 * so that won't be a limiting factor.
1940 *
1941 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1942 */
1943 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1944 {
1945 loff_t res;
1946 loff_t upper_limit = MAX_LFS_FILESIZE;
1947
1948 /* small i_blocks in vfs inode? */
1949 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1950 /*
1951 * CONFIG_LBDAF is not enabled implies the inode
1952 * i_block represent total blocks in 512 bytes
1953 * 32 == size of vfs inode i_blocks * 8
1954 */
1955 upper_limit = (1LL << 32) - 1;
1956
1957 /* total blocks in file system block size */
1958 upper_limit >>= (blkbits - 9);
1959 upper_limit <<= blkbits;
1960 }
1961
1962 /* 32-bit extent-start container, ee_block */
1963 res = 1LL << 32;
1964 res <<= blkbits;
1965 res -= 1;
1966
1967 /* Sanity check against vm- & vfs- imposed limits */
1968 if (res > upper_limit)
1969 res = upper_limit;
1970
1971 return res;
1972 }
1973
1974 /*
1975 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1976 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1977 * We need to be 1 filesystem block less than the 2^48 sector limit.
1978 */
1979 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
1980 {
1981 loff_t res = EXT4_NDIR_BLOCKS;
1982 int meta_blocks;
1983 loff_t upper_limit;
1984 /* This is calculated to be the largest file size for a dense, block
1985 * mapped file such that the file's total number of 512-byte sectors,
1986 * including data and all indirect blocks, does not exceed (2^48 - 1).
1987 *
1988 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
1989 * number of 512-byte sectors of the file.
1990 */
1991
1992 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1993 /*
1994 * !has_huge_files or CONFIG_LBDAF not enabled implies that
1995 * the inode i_block field represents total file blocks in
1996 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
1997 */
1998 upper_limit = (1LL << 32) - 1;
1999
2000 /* total blocks in file system block size */
2001 upper_limit >>= (bits - 9);
2002
2003 } else {
2004 /*
2005 * We use 48 bit ext4_inode i_blocks
2006 * With EXT4_HUGE_FILE_FL set the i_blocks
2007 * represent total number of blocks in
2008 * file system block size
2009 */
2010 upper_limit = (1LL << 48) - 1;
2011
2012 }
2013
2014 /* indirect blocks */
2015 meta_blocks = 1;
2016 /* double indirect blocks */
2017 meta_blocks += 1 + (1LL << (bits-2));
2018 /* tripple indirect blocks */
2019 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2020
2021 upper_limit -= meta_blocks;
2022 upper_limit <<= bits;
2023
2024 res += 1LL << (bits-2);
2025 res += 1LL << (2*(bits-2));
2026 res += 1LL << (3*(bits-2));
2027 res <<= bits;
2028 if (res > upper_limit)
2029 res = upper_limit;
2030
2031 if (res > MAX_LFS_FILESIZE)
2032 res = MAX_LFS_FILESIZE;
2033
2034 return res;
2035 }
2036
2037 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2038 ext4_fsblk_t logical_sb_block, int nr)
2039 {
2040 struct ext4_sb_info *sbi = EXT4_SB(sb);
2041 ext4_group_t bg, first_meta_bg;
2042 int has_super = 0;
2043
2044 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2045
2046 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2047 nr < first_meta_bg)
2048 return logical_sb_block + nr + 1;
2049 bg = sbi->s_desc_per_block * nr;
2050 if (ext4_bg_has_super(sb, bg))
2051 has_super = 1;
2052
2053 return (has_super + ext4_group_first_block_no(sb, bg));
2054 }
2055
2056 /**
2057 * ext4_get_stripe_size: Get the stripe size.
2058 * @sbi: In memory super block info
2059 *
2060 * If we have specified it via mount option, then
2061 * use the mount option value. If the value specified at mount time is
2062 * greater than the blocks per group use the super block value.
2063 * If the super block value is greater than blocks per group return 0.
2064 * Allocator needs it be less than blocks per group.
2065 *
2066 */
2067 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2068 {
2069 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2070 unsigned long stripe_width =
2071 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2072
2073 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2074 return sbi->s_stripe;
2075
2076 if (stripe_width <= sbi->s_blocks_per_group)
2077 return stripe_width;
2078
2079 if (stride <= sbi->s_blocks_per_group)
2080 return stride;
2081
2082 return 0;
2083 }
2084
2085 /* sysfs supprt */
2086
2087 struct ext4_attr {
2088 struct attribute attr;
2089 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2090 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2091 const char *, size_t);
2092 int offset;
2093 };
2094
2095 static int parse_strtoul(const char *buf,
2096 unsigned long max, unsigned long *value)
2097 {
2098 char *endp;
2099
2100 while (*buf && isspace(*buf))
2101 buf++;
2102 *value = simple_strtoul(buf, &endp, 0);
2103 while (*endp && isspace(*endp))
2104 endp++;
2105 if (*endp || *value > max)
2106 return -EINVAL;
2107
2108 return 0;
2109 }
2110
2111 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2112 struct ext4_sb_info *sbi,
2113 char *buf)
2114 {
2115 return snprintf(buf, PAGE_SIZE, "%llu\n",
2116 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2117 }
2118
2119 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2120 struct ext4_sb_info *sbi, char *buf)
2121 {
2122 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2123
2124 return snprintf(buf, PAGE_SIZE, "%lu\n",
2125 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2126 sbi->s_sectors_written_start) >> 1);
2127 }
2128
2129 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2130 struct ext4_sb_info *sbi, char *buf)
2131 {
2132 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2133
2134 return snprintf(buf, PAGE_SIZE, "%llu\n",
2135 sbi->s_kbytes_written +
2136 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2137 EXT4_SB(sb)->s_sectors_written_start) >> 1));
2138 }
2139
2140 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2141 struct ext4_sb_info *sbi,
2142 const char *buf, size_t count)
2143 {
2144 unsigned long t;
2145
2146 if (parse_strtoul(buf, 0x40000000, &t))
2147 return -EINVAL;
2148
2149 if (!is_power_of_2(t))
2150 return -EINVAL;
2151
2152 sbi->s_inode_readahead_blks = t;
2153 return count;
2154 }
2155
2156 static ssize_t sbi_ui_show(struct ext4_attr *a,
2157 struct ext4_sb_info *sbi, char *buf)
2158 {
2159 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2160
2161 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2162 }
2163
2164 static ssize_t sbi_ui_store(struct ext4_attr *a,
2165 struct ext4_sb_info *sbi,
2166 const char *buf, size_t count)
2167 {
2168 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2169 unsigned long t;
2170
2171 if (parse_strtoul(buf, 0xffffffff, &t))
2172 return -EINVAL;
2173 *ui = t;
2174 return count;
2175 }
2176
2177 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2178 static struct ext4_attr ext4_attr_##_name = { \
2179 .attr = {.name = __stringify(_name), .mode = _mode }, \
2180 .show = _show, \
2181 .store = _store, \
2182 .offset = offsetof(struct ext4_sb_info, _elname), \
2183 }
2184 #define EXT4_ATTR(name, mode, show, store) \
2185 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2186
2187 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2188 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2189 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2190 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2191 #define ATTR_LIST(name) &ext4_attr_##name.attr
2192
2193 EXT4_RO_ATTR(delayed_allocation_blocks);
2194 EXT4_RO_ATTR(session_write_kbytes);
2195 EXT4_RO_ATTR(lifetime_write_kbytes);
2196 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2197 inode_readahead_blks_store, s_inode_readahead_blks);
2198 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2199 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2200 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2201 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2202 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2203 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2204 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2205 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2206
2207 static struct attribute *ext4_attrs[] = {
2208 ATTR_LIST(delayed_allocation_blocks),
2209 ATTR_LIST(session_write_kbytes),
2210 ATTR_LIST(lifetime_write_kbytes),
2211 ATTR_LIST(inode_readahead_blks),
2212 ATTR_LIST(inode_goal),
2213 ATTR_LIST(mb_stats),
2214 ATTR_LIST(mb_max_to_scan),
2215 ATTR_LIST(mb_min_to_scan),
2216 ATTR_LIST(mb_order2_req),
2217 ATTR_LIST(mb_stream_req),
2218 ATTR_LIST(mb_group_prealloc),
2219 ATTR_LIST(max_writeback_mb_bump),
2220 NULL,
2221 };
2222
2223 static ssize_t ext4_attr_show(struct kobject *kobj,
2224 struct attribute *attr, char *buf)
2225 {
2226 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2227 s_kobj);
2228 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2229
2230 return a->show ? a->show(a, sbi, buf) : 0;
2231 }
2232
2233 static ssize_t ext4_attr_store(struct kobject *kobj,
2234 struct attribute *attr,
2235 const char *buf, size_t len)
2236 {
2237 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2238 s_kobj);
2239 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2240
2241 return a->store ? a->store(a, sbi, buf, len) : 0;
2242 }
2243
2244 static void ext4_sb_release(struct kobject *kobj)
2245 {
2246 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2247 s_kobj);
2248 complete(&sbi->s_kobj_unregister);
2249 }
2250
2251
2252 static struct sysfs_ops ext4_attr_ops = {
2253 .show = ext4_attr_show,
2254 .store = ext4_attr_store,
2255 };
2256
2257 static struct kobj_type ext4_ktype = {
2258 .default_attrs = ext4_attrs,
2259 .sysfs_ops = &ext4_attr_ops,
2260 .release = ext4_sb_release,
2261 };
2262
2263 /*
2264 * Check whether this filesystem can be mounted based on
2265 * the features present and the RDONLY/RDWR mount requested.
2266 * Returns 1 if this filesystem can be mounted as requested,
2267 * 0 if it cannot be.
2268 */
2269 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2270 {
2271 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2272 ext4_msg(sb, KERN_ERR,
2273 "Couldn't mount because of "
2274 "unsupported optional features (%x)",
2275 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2276 ~EXT4_FEATURE_INCOMPAT_SUPP));
2277 return 0;
2278 }
2279
2280 if (readonly)
2281 return 1;
2282
2283 /* Check that feature set is OK for a read-write mount */
2284 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2285 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2286 "unsupported optional features (%x)",
2287 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2288 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2289 return 0;
2290 }
2291 /*
2292 * Large file size enabled file system can only be mounted
2293 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2294 */
2295 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2296 if (sizeof(blkcnt_t) < sizeof(u64)) {
2297 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2298 "cannot be mounted RDWR without "
2299 "CONFIG_LBDAF");
2300 return 0;
2301 }
2302 }
2303 return 1;
2304 }
2305
2306 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2307 __releases(kernel_lock)
2308 __acquires(kernel_lock)
2309 {
2310 struct buffer_head *bh;
2311 struct ext4_super_block *es = NULL;
2312 struct ext4_sb_info *sbi;
2313 ext4_fsblk_t block;
2314 ext4_fsblk_t sb_block = get_sb_block(&data);
2315 ext4_fsblk_t logical_sb_block;
2316 unsigned long offset = 0;
2317 unsigned long journal_devnum = 0;
2318 unsigned long def_mount_opts;
2319 struct inode *root;
2320 char *cp;
2321 const char *descr;
2322 int ret = -EINVAL;
2323 int blocksize;
2324 unsigned int db_count;
2325 unsigned int i;
2326 int needs_recovery, has_huge_files;
2327 __u64 blocks_count;
2328 int err;
2329 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2330
2331 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2332 if (!sbi)
2333 return -ENOMEM;
2334
2335 sbi->s_blockgroup_lock =
2336 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2337 if (!sbi->s_blockgroup_lock) {
2338 kfree(sbi);
2339 return -ENOMEM;
2340 }
2341 sb->s_fs_info = sbi;
2342 sbi->s_mount_opt = 0;
2343 sbi->s_resuid = EXT4_DEF_RESUID;
2344 sbi->s_resgid = EXT4_DEF_RESGID;
2345 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2346 sbi->s_sb_block = sb_block;
2347 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2348 sectors[1]);
2349
2350 unlock_kernel();
2351
2352 /* Cleanup superblock name */
2353 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2354 *cp = '!';
2355
2356 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2357 if (!blocksize) {
2358 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
2359 goto out_fail;
2360 }
2361
2362 /*
2363 * The ext4 superblock will not be buffer aligned for other than 1kB
2364 * block sizes. We need to calculate the offset from buffer start.
2365 */
2366 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2367 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2368 offset = do_div(logical_sb_block, blocksize);
2369 } else {
2370 logical_sb_block = sb_block;
2371 }
2372
2373 if (!(bh = sb_bread(sb, logical_sb_block))) {
2374 ext4_msg(sb, KERN_ERR, "unable to read superblock");
2375 goto out_fail;
2376 }
2377 /*
2378 * Note: s_es must be initialized as soon as possible because
2379 * some ext4 macro-instructions depend on its value
2380 */
2381 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2382 sbi->s_es = es;
2383 sb->s_magic = le16_to_cpu(es->s_magic);
2384 if (sb->s_magic != EXT4_SUPER_MAGIC)
2385 goto cantfind_ext4;
2386 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2387
2388 /* Set defaults before we parse the mount options */
2389 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2390 if (def_mount_opts & EXT4_DEFM_DEBUG)
2391 set_opt(sbi->s_mount_opt, DEBUG);
2392 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
2393 set_opt(sbi->s_mount_opt, GRPID);
2394 if (def_mount_opts & EXT4_DEFM_UID16)
2395 set_opt(sbi->s_mount_opt, NO_UID32);
2396 #ifdef CONFIG_EXT4_FS_XATTR
2397 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2398 set_opt(sbi->s_mount_opt, XATTR_USER);
2399 #endif
2400 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2401 if (def_mount_opts & EXT4_DEFM_ACL)
2402 set_opt(sbi->s_mount_opt, POSIX_ACL);
2403 #endif
2404 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2405 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2406 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2407 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2408 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2409 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2410
2411 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2412 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2413 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2414 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2415 else
2416 set_opt(sbi->s_mount_opt, ERRORS_RO);
2417
2418 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2419 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2420 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2421 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2422 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2423
2424 set_opt(sbi->s_mount_opt, BARRIER);
2425
2426 /*
2427 * enable delayed allocation by default
2428 * Use -o nodelalloc to turn it off
2429 */
2430 set_opt(sbi->s_mount_opt, DELALLOC);
2431
2432 if (!parse_options((char *) data, sb, &journal_devnum,
2433 &journal_ioprio, NULL, 0))
2434 goto failed_mount;
2435
2436 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2437 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2438
2439 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2440 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2441 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2442 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2443 ext4_msg(sb, KERN_WARNING,
2444 "feature flags set on rev 0 fs, "
2445 "running e2fsck is recommended");
2446
2447 /*
2448 * Check feature flags regardless of the revision level, since we
2449 * previously didn't change the revision level when setting the flags,
2450 * so there is a chance incompat flags are set on a rev 0 filesystem.
2451 */
2452 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
2453 goto failed_mount;
2454
2455 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2456
2457 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2458 blocksize > EXT4_MAX_BLOCK_SIZE) {
2459 ext4_msg(sb, KERN_ERR,
2460 "Unsupported filesystem blocksize %d", blocksize);
2461 goto failed_mount;
2462 }
2463
2464 if (sb->s_blocksize != blocksize) {
2465 /* Validate the filesystem blocksize */
2466 if (!sb_set_blocksize(sb, blocksize)) {
2467 ext4_msg(sb, KERN_ERR, "bad block size %d",
2468 blocksize);
2469 goto failed_mount;
2470 }
2471
2472 brelse(bh);
2473 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2474 offset = do_div(logical_sb_block, blocksize);
2475 bh = sb_bread(sb, logical_sb_block);
2476 if (!bh) {
2477 ext4_msg(sb, KERN_ERR,
2478 "Can't read superblock on 2nd try");
2479 goto failed_mount;
2480 }
2481 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2482 sbi->s_es = es;
2483 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2484 ext4_msg(sb, KERN_ERR,
2485 "Magic mismatch, very weird!");
2486 goto failed_mount;
2487 }
2488 }
2489
2490 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2491 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2492 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2493 has_huge_files);
2494 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2495
2496 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2497 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2498 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2499 } else {
2500 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2501 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2502 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2503 (!is_power_of_2(sbi->s_inode_size)) ||
2504 (sbi->s_inode_size > blocksize)) {
2505 ext4_msg(sb, KERN_ERR,
2506 "unsupported inode size: %d",
2507 sbi->s_inode_size);
2508 goto failed_mount;
2509 }
2510 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2511 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2512 }
2513
2514 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2515 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2516 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2517 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2518 !is_power_of_2(sbi->s_desc_size)) {
2519 ext4_msg(sb, KERN_ERR,
2520 "unsupported descriptor size %lu",
2521 sbi->s_desc_size);
2522 goto failed_mount;
2523 }
2524 } else
2525 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2526
2527 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2528 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2529 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2530 goto cantfind_ext4;
2531
2532 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2533 if (sbi->s_inodes_per_block == 0)
2534 goto cantfind_ext4;
2535 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2536 sbi->s_inodes_per_block;
2537 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2538 sbi->s_sbh = bh;
2539 sbi->s_mount_state = le16_to_cpu(es->s_state);
2540 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2541 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2542
2543 for (i = 0; i < 4; i++)
2544 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2545 sbi->s_def_hash_version = es->s_def_hash_version;
2546 i = le32_to_cpu(es->s_flags);
2547 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2548 sbi->s_hash_unsigned = 3;
2549 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2550 #ifdef __CHAR_UNSIGNED__
2551 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2552 sbi->s_hash_unsigned = 3;
2553 #else
2554 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2555 #endif
2556 sb->s_dirt = 1;
2557 }
2558
2559 if (sbi->s_blocks_per_group > blocksize * 8) {
2560 ext4_msg(sb, KERN_ERR,
2561 "#blocks per group too big: %lu",
2562 sbi->s_blocks_per_group);
2563 goto failed_mount;
2564 }
2565 if (sbi->s_inodes_per_group > blocksize * 8) {
2566 ext4_msg(sb, KERN_ERR,
2567 "#inodes per group too big: %lu",
2568 sbi->s_inodes_per_group);
2569 goto failed_mount;
2570 }
2571
2572 /*
2573 * Test whether we have more sectors than will fit in sector_t,
2574 * and whether the max offset is addressable by the page cache.
2575 */
2576 if ((ext4_blocks_count(es) >
2577 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2578 (ext4_blocks_count(es) >
2579 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
2580 ext4_msg(sb, KERN_ERR, "filesystem"
2581 " too large to mount safely on this system");
2582 if (sizeof(sector_t) < 8)
2583 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
2584 ret = -EFBIG;
2585 goto failed_mount;
2586 }
2587
2588 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2589 goto cantfind_ext4;
2590
2591 /* check blocks count against device size */
2592 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2593 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2594 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2595 "exceeds size of device (%llu blocks)",
2596 ext4_blocks_count(es), blocks_count);
2597 goto failed_mount;
2598 }
2599
2600 /*
2601 * It makes no sense for the first data block to be beyond the end
2602 * of the filesystem.
2603 */
2604 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2605 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2606 "block %u is beyond end of filesystem (%llu)",
2607 le32_to_cpu(es->s_first_data_block),
2608 ext4_blocks_count(es));
2609 goto failed_mount;
2610 }
2611 blocks_count = (ext4_blocks_count(es) -
2612 le32_to_cpu(es->s_first_data_block) +
2613 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2614 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2615 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2616 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
2617 "(block count %llu, first data block %u, "
2618 "blocks per group %lu)", sbi->s_groups_count,
2619 ext4_blocks_count(es),
2620 le32_to_cpu(es->s_first_data_block),
2621 EXT4_BLOCKS_PER_GROUP(sb));
2622 goto failed_mount;
2623 }
2624 sbi->s_groups_count = blocks_count;
2625 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
2626 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
2627 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2628 EXT4_DESC_PER_BLOCK(sb);
2629 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2630 GFP_KERNEL);
2631 if (sbi->s_group_desc == NULL) {
2632 ext4_msg(sb, KERN_ERR, "not enough memory");
2633 goto failed_mount;
2634 }
2635
2636 #ifdef CONFIG_PROC_FS
2637 if (ext4_proc_root)
2638 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2639 #endif
2640
2641 bgl_lock_init(sbi->s_blockgroup_lock);
2642
2643 for (i = 0; i < db_count; i++) {
2644 block = descriptor_loc(sb, logical_sb_block, i);
2645 sbi->s_group_desc[i] = sb_bread(sb, block);
2646 if (!sbi->s_group_desc[i]) {
2647 ext4_msg(sb, KERN_ERR,
2648 "can't read group descriptor %d", i);
2649 db_count = i;
2650 goto failed_mount2;
2651 }
2652 }
2653 if (!ext4_check_descriptors(sb)) {
2654 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
2655 goto failed_mount2;
2656 }
2657 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2658 if (!ext4_fill_flex_info(sb)) {
2659 ext4_msg(sb, KERN_ERR,
2660 "unable to initialize "
2661 "flex_bg meta info!");
2662 goto failed_mount2;
2663 }
2664
2665 sbi->s_gdb_count = db_count;
2666 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2667 spin_lock_init(&sbi->s_next_gen_lock);
2668
2669 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2670 ext4_count_free_blocks(sb));
2671 if (!err) {
2672 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2673 ext4_count_free_inodes(sb));
2674 }
2675 if (!err) {
2676 err = percpu_counter_init(&sbi->s_dirs_counter,
2677 ext4_count_dirs(sb));
2678 }
2679 if (!err) {
2680 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2681 }
2682 if (err) {
2683 ext4_msg(sb, KERN_ERR, "insufficient memory");
2684 goto failed_mount3;
2685 }
2686
2687 sbi->s_stripe = ext4_get_stripe_size(sbi);
2688 sbi->s_max_writeback_mb_bump = 128;
2689
2690 /*
2691 * set up enough so that it can read an inode
2692 */
2693 if (!test_opt(sb, NOLOAD) &&
2694 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2695 sb->s_op = &ext4_sops;
2696 else
2697 sb->s_op = &ext4_nojournal_sops;
2698 sb->s_export_op = &ext4_export_ops;
2699 sb->s_xattr = ext4_xattr_handlers;
2700 #ifdef CONFIG_QUOTA
2701 sb->s_qcop = &ext4_qctl_operations;
2702 sb->dq_op = &ext4_quota_operations;
2703 #endif
2704 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2705 mutex_init(&sbi->s_orphan_lock);
2706 mutex_init(&sbi->s_resize_lock);
2707
2708 sb->s_root = NULL;
2709
2710 needs_recovery = (es->s_last_orphan != 0 ||
2711 EXT4_HAS_INCOMPAT_FEATURE(sb,
2712 EXT4_FEATURE_INCOMPAT_RECOVER));
2713
2714 /*
2715 * The first inode we look at is the journal inode. Don't try
2716 * root first: it may be modified in the journal!
2717 */
2718 if (!test_opt(sb, NOLOAD) &&
2719 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2720 if (ext4_load_journal(sb, es, journal_devnum))
2721 goto failed_mount3;
2722 if (!(sb->s_flags & MS_RDONLY) &&
2723 EXT4_SB(sb)->s_journal->j_failed_commit) {
2724 ext4_msg(sb, KERN_CRIT, "error: "
2725 "ext4_fill_super: Journal transaction "
2726 "%u is corrupt",
2727 EXT4_SB(sb)->s_journal->j_failed_commit);
2728 if (test_opt(sb, ERRORS_RO)) {
2729 ext4_msg(sb, KERN_CRIT,
2730 "Mounting filesystem read-only");
2731 sb->s_flags |= MS_RDONLY;
2732 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2733 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2734 }
2735 if (test_opt(sb, ERRORS_PANIC)) {
2736 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2737 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2738 ext4_commit_super(sb, 1);
2739 goto failed_mount4;
2740 }
2741 }
2742 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2743 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2744 ext4_msg(sb, KERN_ERR, "required journal recovery "
2745 "suppressed and not mounted read-only");
2746 goto failed_mount4;
2747 } else {
2748 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2749 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2750 sbi->s_journal = NULL;
2751 needs_recovery = 0;
2752 goto no_journal;
2753 }
2754
2755 if (ext4_blocks_count(es) > 0xffffffffULL &&
2756 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2757 JBD2_FEATURE_INCOMPAT_64BIT)) {
2758 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
2759 goto failed_mount4;
2760 }
2761
2762 jbd2_journal_set_features(sbi->s_journal,
2763 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2764 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
2765 jbd2_journal_set_features(sbi->s_journal, 0, 0,
2766 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2767 else
2768 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2769 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2770
2771 /* We have now updated the journal if required, so we can
2772 * validate the data journaling mode. */
2773 switch (test_opt(sb, DATA_FLAGS)) {
2774 case 0:
2775 /* No mode set, assume a default based on the journal
2776 * capabilities: ORDERED_DATA if the journal can
2777 * cope, else JOURNAL_DATA
2778 */
2779 if (jbd2_journal_check_available_features
2780 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2781 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2782 else
2783 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2784 break;
2785
2786 case EXT4_MOUNT_ORDERED_DATA:
2787 case EXT4_MOUNT_WRITEBACK_DATA:
2788 if (!jbd2_journal_check_available_features
2789 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2790 ext4_msg(sb, KERN_ERR, "Journal does not support "
2791 "requested data journaling mode");
2792 goto failed_mount4;
2793 }
2794 default:
2795 break;
2796 }
2797 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2798
2799 no_journal:
2800
2801 if (test_opt(sb, NOBH)) {
2802 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2803 ext4_msg(sb, KERN_WARNING, "Ignoring nobh option - "
2804 "its supported only with writeback mode");
2805 clear_opt(sbi->s_mount_opt, NOBH);
2806 }
2807 }
2808 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
2809 if (!EXT4_SB(sb)->dio_unwritten_wq) {
2810 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
2811 goto failed_mount_wq;
2812 }
2813
2814 /*
2815 * The jbd2_journal_load will have done any necessary log recovery,
2816 * so we can safely mount the rest of the filesystem now.
2817 */
2818
2819 root = ext4_iget(sb, EXT4_ROOT_INO);
2820 if (IS_ERR(root)) {
2821 ext4_msg(sb, KERN_ERR, "get root inode failed");
2822 ret = PTR_ERR(root);
2823 goto failed_mount4;
2824 }
2825 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2826 iput(root);
2827 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
2828 goto failed_mount4;
2829 }
2830 sb->s_root = d_alloc_root(root);
2831 if (!sb->s_root) {
2832 ext4_msg(sb, KERN_ERR, "get root dentry failed");
2833 iput(root);
2834 ret = -ENOMEM;
2835 goto failed_mount4;
2836 }
2837
2838 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2839
2840 /* determine the minimum size of new large inodes, if present */
2841 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2842 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2843 EXT4_GOOD_OLD_INODE_SIZE;
2844 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2845 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2846 if (sbi->s_want_extra_isize <
2847 le16_to_cpu(es->s_want_extra_isize))
2848 sbi->s_want_extra_isize =
2849 le16_to_cpu(es->s_want_extra_isize);
2850 if (sbi->s_want_extra_isize <
2851 le16_to_cpu(es->s_min_extra_isize))
2852 sbi->s_want_extra_isize =
2853 le16_to_cpu(es->s_min_extra_isize);
2854 }
2855 }
2856 /* Check if enough inode space is available */
2857 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2858 sbi->s_inode_size) {
2859 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2860 EXT4_GOOD_OLD_INODE_SIZE;
2861 ext4_msg(sb, KERN_INFO, "required extra inode space not"
2862 "available");
2863 }
2864
2865 if (test_opt(sb, DELALLOC) &&
2866 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
2867 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
2868 "requested data journaling mode");
2869 clear_opt(sbi->s_mount_opt, DELALLOC);
2870 }
2871
2872 err = ext4_setup_system_zone(sb);
2873 if (err) {
2874 ext4_msg(sb, KERN_ERR, "failed to initialize system "
2875 "zone (%d)\n", err);
2876 goto failed_mount4;
2877 }
2878
2879 ext4_ext_init(sb);
2880 err = ext4_mb_init(sb, needs_recovery);
2881 if (err) {
2882 ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
2883 err);
2884 goto failed_mount4;
2885 }
2886
2887 sbi->s_kobj.kset = ext4_kset;
2888 init_completion(&sbi->s_kobj_unregister);
2889 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2890 "%s", sb->s_id);
2891 if (err) {
2892 ext4_mb_release(sb);
2893 ext4_ext_release(sb);
2894 goto failed_mount4;
2895 };
2896
2897 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2898 ext4_orphan_cleanup(sb, es);
2899 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2900 if (needs_recovery) {
2901 ext4_msg(sb, KERN_INFO, "recovery complete");
2902 ext4_mark_recovery_complete(sb, es);
2903 }
2904 if (EXT4_SB(sb)->s_journal) {
2905 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2906 descr = " journalled data mode";
2907 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2908 descr = " ordered data mode";
2909 else
2910 descr = " writeback data mode";
2911 } else
2912 descr = "out journal";
2913
2914 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s", descr);
2915
2916 lock_kernel();
2917 return 0;
2918
2919 cantfind_ext4:
2920 if (!silent)
2921 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
2922 goto failed_mount;
2923
2924 failed_mount4:
2925 ext4_msg(sb, KERN_ERR, "mount failed");
2926 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
2927 failed_mount_wq:
2928 ext4_release_system_zone(sb);
2929 if (sbi->s_journal) {
2930 jbd2_journal_destroy(sbi->s_journal);
2931 sbi->s_journal = NULL;
2932 }
2933 failed_mount3:
2934 if (sbi->s_flex_groups) {
2935 if (is_vmalloc_addr(sbi->s_flex_groups))
2936 vfree(sbi->s_flex_groups);
2937 else
2938 kfree(sbi->s_flex_groups);
2939 }
2940 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2941 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2942 percpu_counter_destroy(&sbi->s_dirs_counter);
2943 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2944 failed_mount2:
2945 for (i = 0; i < db_count; i++)
2946 brelse(sbi->s_group_desc[i]);
2947 kfree(sbi->s_group_desc);
2948 failed_mount:
2949 if (sbi->s_proc) {
2950 remove_proc_entry(sb->s_id, ext4_proc_root);
2951 }
2952 #ifdef CONFIG_QUOTA
2953 for (i = 0; i < MAXQUOTAS; i++)
2954 kfree(sbi->s_qf_names[i]);
2955 #endif
2956 ext4_blkdev_remove(sbi);
2957 brelse(bh);
2958 out_fail:
2959 sb->s_fs_info = NULL;
2960 kfree(sbi->s_blockgroup_lock);
2961 kfree(sbi);
2962 lock_kernel();
2963 return ret;
2964 }
2965
2966 /*
2967 * Setup any per-fs journal parameters now. We'll do this both on
2968 * initial mount, once the journal has been initialised but before we've
2969 * done any recovery; and again on any subsequent remount.
2970 */
2971 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2972 {
2973 struct ext4_sb_info *sbi = EXT4_SB(sb);
2974
2975 journal->j_commit_interval = sbi->s_commit_interval;
2976 journal->j_min_batch_time = sbi->s_min_batch_time;
2977 journal->j_max_batch_time = sbi->s_max_batch_time;
2978
2979 spin_lock(&journal->j_state_lock);
2980 if (test_opt(sb, BARRIER))
2981 journal->j_flags |= JBD2_BARRIER;
2982 else
2983 journal->j_flags &= ~JBD2_BARRIER;
2984 if (test_opt(sb, DATA_ERR_ABORT))
2985 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2986 else
2987 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
2988 spin_unlock(&journal->j_state_lock);
2989 }
2990
2991 static journal_t *ext4_get_journal(struct super_block *sb,
2992 unsigned int journal_inum)
2993 {
2994 struct inode *journal_inode;
2995 journal_t *journal;
2996
2997 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2998
2999 /* First, test for the existence of a valid inode on disk. Bad
3000 * things happen if we iget() an unused inode, as the subsequent
3001 * iput() will try to delete it. */
3002
3003 journal_inode = ext4_iget(sb, journal_inum);
3004 if (IS_ERR(journal_inode)) {
3005 ext4_msg(sb, KERN_ERR, "no journal found");
3006 return NULL;
3007 }
3008 if (!journal_inode->i_nlink) {
3009 make_bad_inode(journal_inode);
3010 iput(journal_inode);
3011 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3012 return NULL;
3013 }
3014
3015 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3016 journal_inode, journal_inode->i_size);
3017 if (!S_ISREG(journal_inode->i_mode)) {
3018 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3019 iput(journal_inode);
3020 return NULL;
3021 }
3022
3023 journal = jbd2_journal_init_inode(journal_inode);
3024 if (!journal) {
3025 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3026 iput(journal_inode);
3027 return NULL;
3028 }
3029 journal->j_private = sb;
3030 ext4_init_journal_params(sb, journal);
3031 return journal;
3032 }
3033
3034 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3035 dev_t j_dev)
3036 {
3037 struct buffer_head *bh;
3038 journal_t *journal;
3039 ext4_fsblk_t start;
3040 ext4_fsblk_t len;
3041 int hblock, blocksize;
3042 ext4_fsblk_t sb_block;
3043 unsigned long offset;
3044 struct ext4_super_block *es;
3045 struct block_device *bdev;
3046
3047 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3048
3049 bdev = ext4_blkdev_get(j_dev, sb);
3050 if (bdev == NULL)
3051 return NULL;
3052
3053 if (bd_claim(bdev, sb)) {
3054 ext4_msg(sb, KERN_ERR,
3055 "failed to claim external journal device");
3056 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3057 return NULL;
3058 }
3059
3060 blocksize = sb->s_blocksize;
3061 hblock = bdev_logical_block_size(bdev);
3062 if (blocksize < hblock) {
3063 ext4_msg(sb, KERN_ERR,
3064 "blocksize too small for journal device");
3065 goto out_bdev;
3066 }
3067
3068 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3069 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3070 set_blocksize(bdev, blocksize);
3071 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3072 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3073 "external journal");
3074 goto out_bdev;
3075 }
3076
3077 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3078 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3079 !(le32_to_cpu(es->s_feature_incompat) &
3080 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3081 ext4_msg(sb, KERN_ERR, "external journal has "
3082 "bad superblock");
3083 brelse(bh);
3084 goto out_bdev;
3085 }
3086
3087 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3088 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3089 brelse(bh);
3090 goto out_bdev;
3091 }
3092
3093 len = ext4_blocks_count(es);
3094 start = sb_block + 1;
3095 brelse(bh); /* we're done with the superblock */
3096
3097 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3098 start, len, blocksize);
3099 if (!journal) {
3100 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3101 goto out_bdev;
3102 }
3103 journal->j_private = sb;
3104 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3105 wait_on_buffer(journal->j_sb_buffer);
3106 if (!buffer_uptodate(journal->j_sb_buffer)) {
3107 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3108 goto out_journal;
3109 }
3110 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3111 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3112 "user (unsupported) - %d",
3113 be32_to_cpu(journal->j_superblock->s_nr_users));
3114 goto out_journal;
3115 }
3116 EXT4_SB(sb)->journal_bdev = bdev;
3117 ext4_init_journal_params(sb, journal);
3118 return journal;
3119
3120 out_journal:
3121 jbd2_journal_destroy(journal);
3122 out_bdev:
3123 ext4_blkdev_put(bdev);
3124 return NULL;
3125 }
3126
3127 static int ext4_load_journal(struct super_block *sb,
3128 struct ext4_super_block *es,
3129 unsigned long journal_devnum)
3130 {
3131 journal_t *journal;
3132 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3133 dev_t journal_dev;
3134 int err = 0;
3135 int really_read_only;
3136
3137 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3138
3139 if (journal_devnum &&
3140 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3141 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3142 "numbers have changed");
3143 journal_dev = new_decode_dev(journal_devnum);
3144 } else
3145 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3146
3147 really_read_only = bdev_read_only(sb->s_bdev);
3148
3149 /*
3150 * Are we loading a blank journal or performing recovery after a
3151 * crash? For recovery, we need to check in advance whether we
3152 * can get read-write access to the device.
3153 */
3154 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3155 if (sb->s_flags & MS_RDONLY) {
3156 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3157 "required on readonly filesystem");
3158 if (really_read_only) {
3159 ext4_msg(sb, KERN_ERR, "write access "
3160 "unavailable, cannot proceed");
3161 return -EROFS;
3162 }
3163 ext4_msg(sb, KERN_INFO, "write access will "
3164 "be enabled during recovery");
3165 }
3166 }
3167
3168 if (journal_inum && journal_dev) {
3169 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3170 "and inode journals!");
3171 return -EINVAL;
3172 }
3173
3174 if (journal_inum) {
3175 if (!(journal = ext4_get_journal(sb, journal_inum)))
3176 return -EINVAL;
3177 } else {
3178 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3179 return -EINVAL;
3180 }
3181
3182 if (!(journal->j_flags & JBD2_BARRIER))
3183 ext4_msg(sb, KERN_INFO, "barriers disabled");
3184
3185 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3186 err = jbd2_journal_update_format(journal);
3187 if (err) {
3188 ext4_msg(sb, KERN_ERR, "error updating journal");
3189 jbd2_journal_destroy(journal);
3190 return err;
3191 }
3192 }
3193
3194 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3195 err = jbd2_journal_wipe(journal, !really_read_only);
3196 if (!err)
3197 err = jbd2_journal_load(journal);
3198
3199 if (err) {
3200 ext4_msg(sb, KERN_ERR, "error loading journal");
3201 jbd2_journal_destroy(journal);
3202 return err;
3203 }
3204
3205 EXT4_SB(sb)->s_journal = journal;
3206 ext4_clear_journal_err(sb, es);
3207
3208 if (journal_devnum &&
3209 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3210 es->s_journal_dev = cpu_to_le32(journal_devnum);
3211
3212 /* Make sure we flush the recovery flag to disk. */
3213 ext4_commit_super(sb, 1);
3214 }
3215
3216 return 0;
3217 }
3218
3219 static int ext4_commit_super(struct super_block *sb, int sync)
3220 {
3221 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3222 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3223 int error = 0;
3224
3225 if (!sbh)
3226 return error;
3227 if (buffer_write_io_error(sbh)) {
3228 /*
3229 * Oh, dear. A previous attempt to write the
3230 * superblock failed. This could happen because the
3231 * USB device was yanked out. Or it could happen to
3232 * be a transient write error and maybe the block will
3233 * be remapped. Nothing we can do but to retry the
3234 * write and hope for the best.
3235 */
3236 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3237 "superblock detected");
3238 clear_buffer_write_io_error(sbh);
3239 set_buffer_uptodate(sbh);
3240 }
3241 /*
3242 * If the file system is mounted read-only, don't update the
3243 * superblock write time. This avoids updating the superblock
3244 * write time when we are mounting the root file system
3245 * read/only but we need to replay the journal; at that point,
3246 * for people who are east of GMT and who make their clock
3247 * tick in localtime for Windows bug-for-bug compatibility,
3248 * the clock is set in the future, and this will cause e2fsck
3249 * to complain and force a full file system check.
3250 */
3251 if (!(sb->s_flags & MS_RDONLY))
3252 es->s_wtime = cpu_to_le32(get_seconds());
3253 es->s_kbytes_written =
3254 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3255 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3256 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3257 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3258 &EXT4_SB(sb)->s_freeblocks_counter));
3259 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3260 &EXT4_SB(sb)->s_freeinodes_counter));
3261 sb->s_dirt = 0;
3262 BUFFER_TRACE(sbh, "marking dirty");
3263 mark_buffer_dirty(sbh);
3264 if (sync) {
3265 error = sync_dirty_buffer(sbh);
3266 if (error)
3267 return error;
3268
3269 error = buffer_write_io_error(sbh);
3270 if (error) {
3271 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3272 "superblock");
3273 clear_buffer_write_io_error(sbh);
3274 set_buffer_uptodate(sbh);
3275 }
3276 }
3277 return error;
3278 }
3279
3280 /*
3281 * Have we just finished recovery? If so, and if we are mounting (or
3282 * remounting) the filesystem readonly, then we will end up with a
3283 * consistent fs on disk. Record that fact.
3284 */
3285 static void ext4_mark_recovery_complete(struct super_block *sb,
3286 struct ext4_super_block *es)
3287 {
3288 journal_t *journal = EXT4_SB(sb)->s_journal;
3289
3290 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3291 BUG_ON(journal != NULL);
3292 return;
3293 }
3294 jbd2_journal_lock_updates(journal);
3295 if (jbd2_journal_flush(journal) < 0)
3296 goto out;
3297
3298 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3299 sb->s_flags & MS_RDONLY) {
3300 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3301 ext4_commit_super(sb, 1);
3302 }
3303
3304 out:
3305 jbd2_journal_unlock_updates(journal);
3306 }
3307
3308 /*
3309 * If we are mounting (or read-write remounting) a filesystem whose journal
3310 * has recorded an error from a previous lifetime, move that error to the
3311 * main filesystem now.
3312 */
3313 static void ext4_clear_journal_err(struct super_block *sb,
3314 struct ext4_super_block *es)
3315 {
3316 journal_t *journal;
3317 int j_errno;
3318 const char *errstr;
3319
3320 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3321
3322 journal = EXT4_SB(sb)->s_journal;
3323
3324 /*
3325 * Now check for any error status which may have been recorded in the
3326 * journal by a prior ext4_error() or ext4_abort()
3327 */
3328
3329 j_errno = jbd2_journal_errno(journal);
3330 if (j_errno) {
3331 char nbuf[16];
3332
3333 errstr = ext4_decode_error(sb, j_errno, nbuf);
3334 ext4_warning(sb, __func__, "Filesystem error recorded "
3335 "from previous mount: %s", errstr);
3336 ext4_warning(sb, __func__, "Marking fs in need of "
3337 "filesystem check.");
3338
3339 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3340 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3341 ext4_commit_super(sb, 1);
3342
3343 jbd2_journal_clear_err(journal);
3344 }
3345 }
3346
3347 /*
3348 * Force the running and committing transactions to commit,
3349 * and wait on the commit.
3350 */
3351 int ext4_force_commit(struct super_block *sb)
3352 {
3353 journal_t *journal;
3354 int ret = 0;
3355
3356 if (sb->s_flags & MS_RDONLY)
3357 return 0;
3358
3359 journal = EXT4_SB(sb)->s_journal;
3360 if (journal)
3361 ret = ext4_journal_force_commit(journal);
3362
3363 return ret;
3364 }
3365
3366 static void ext4_write_super(struct super_block *sb)
3367 {
3368 lock_super(sb);
3369 ext4_commit_super(sb, 1);
3370 unlock_super(sb);
3371 }
3372
3373 static int ext4_sync_fs(struct super_block *sb, int wait)
3374 {
3375 int ret = 0;
3376 tid_t target;
3377 struct ext4_sb_info *sbi = EXT4_SB(sb);
3378
3379 trace_ext4_sync_fs(sb, wait);
3380 flush_workqueue(sbi->dio_unwritten_wq);
3381 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
3382 if (wait)
3383 jbd2_log_wait_commit(sbi->s_journal, target);
3384 }
3385 return ret;
3386 }
3387
3388 /*
3389 * LVM calls this function before a (read-only) snapshot is created. This
3390 * gives us a chance to flush the journal completely and mark the fs clean.
3391 */
3392 static int ext4_freeze(struct super_block *sb)
3393 {
3394 int error = 0;
3395 journal_t *journal;
3396
3397 if (sb->s_flags & MS_RDONLY)
3398 return 0;
3399
3400 journal = EXT4_SB(sb)->s_journal;
3401
3402 /* Now we set up the journal barrier. */
3403 jbd2_journal_lock_updates(journal);
3404
3405 /*
3406 * Don't clear the needs_recovery flag if we failed to flush
3407 * the journal.
3408 */
3409 error = jbd2_journal_flush(journal);
3410 if (error < 0) {
3411 out:
3412 jbd2_journal_unlock_updates(journal);
3413 return error;
3414 }
3415
3416 /* Journal blocked and flushed, clear needs_recovery flag. */
3417 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3418 error = ext4_commit_super(sb, 1);
3419 if (error)
3420 goto out;
3421 return 0;
3422 }
3423
3424 /*
3425 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3426 * flag here, even though the filesystem is not technically dirty yet.
3427 */
3428 static int ext4_unfreeze(struct super_block *sb)
3429 {
3430 if (sb->s_flags & MS_RDONLY)
3431 return 0;
3432
3433 lock_super(sb);
3434 /* Reset the needs_recovery flag before the fs is unlocked. */
3435 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3436 ext4_commit_super(sb, 1);
3437 unlock_super(sb);
3438 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3439 return 0;
3440 }
3441
3442 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3443 {
3444 struct ext4_super_block *es;
3445 struct ext4_sb_info *sbi = EXT4_SB(sb);
3446 ext4_fsblk_t n_blocks_count = 0;
3447 unsigned long old_sb_flags;
3448 struct ext4_mount_options old_opts;
3449 ext4_group_t g;
3450 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3451 int err;
3452 #ifdef CONFIG_QUOTA
3453 int i;
3454 #endif
3455
3456 lock_kernel();
3457
3458 /* Store the original options */
3459 lock_super(sb);
3460 old_sb_flags = sb->s_flags;
3461 old_opts.s_mount_opt = sbi->s_mount_opt;
3462 old_opts.s_resuid = sbi->s_resuid;
3463 old_opts.s_resgid = sbi->s_resgid;
3464 old_opts.s_commit_interval = sbi->s_commit_interval;
3465 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3466 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3467 #ifdef CONFIG_QUOTA
3468 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3469 for (i = 0; i < MAXQUOTAS; i++)
3470 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3471 #endif
3472 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3473 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3474
3475 /*
3476 * Allow the "check" option to be passed as a remount option.
3477 */
3478 if (!parse_options(data, sb, NULL, &journal_ioprio,
3479 &n_blocks_count, 1)) {
3480 err = -EINVAL;
3481 goto restore_opts;
3482 }
3483
3484 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
3485 ext4_abort(sb, __func__, "Abort forced by user");
3486
3487 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3488 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3489
3490 es = sbi->s_es;
3491
3492 if (sbi->s_journal) {
3493 ext4_init_journal_params(sb, sbi->s_journal);
3494 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3495 }
3496
3497 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3498 n_blocks_count > ext4_blocks_count(es)) {
3499 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
3500 err = -EROFS;
3501 goto restore_opts;
3502 }
3503
3504 if (*flags & MS_RDONLY) {
3505 /*
3506 * First of all, the unconditional stuff we have to do
3507 * to disable replay of the journal when we next remount
3508 */
3509 sb->s_flags |= MS_RDONLY;
3510
3511 /*
3512 * OK, test if we are remounting a valid rw partition
3513 * readonly, and if so set the rdonly flag and then
3514 * mark the partition as valid again.
3515 */
3516 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3517 (sbi->s_mount_state & EXT4_VALID_FS))
3518 es->s_state = cpu_to_le16(sbi->s_mount_state);
3519
3520 if (sbi->s_journal)
3521 ext4_mark_recovery_complete(sb, es);
3522 } else {
3523 /* Make sure we can mount this feature set readwrite */
3524 if (!ext4_feature_set_ok(sb, 0)) {
3525 err = -EROFS;
3526 goto restore_opts;
3527 }
3528 /*
3529 * Make sure the group descriptor checksums
3530 * are sane. If they aren't, refuse to remount r/w.
3531 */
3532 for (g = 0; g < sbi->s_groups_count; g++) {
3533 struct ext4_group_desc *gdp =
3534 ext4_get_group_desc(sb, g, NULL);
3535
3536 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3537 ext4_msg(sb, KERN_ERR,
3538 "ext4_remount: Checksum for group %u failed (%u!=%u)",
3539 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3540 le16_to_cpu(gdp->bg_checksum));
3541 err = -EINVAL;
3542 goto restore_opts;
3543 }
3544 }
3545
3546 /*
3547 * If we have an unprocessed orphan list hanging
3548 * around from a previously readonly bdev mount,
3549 * require a full umount/remount for now.
3550 */
3551 if (es->s_last_orphan) {
3552 ext4_msg(sb, KERN_WARNING, "Couldn't "
3553 "remount RDWR because of unprocessed "
3554 "orphan inode list. Please "
3555 "umount/remount instead");
3556 err = -EINVAL;
3557 goto restore_opts;
3558 }
3559
3560 /*
3561 * Mounting a RDONLY partition read-write, so reread
3562 * and store the current valid flag. (It may have
3563 * been changed by e2fsck since we originally mounted
3564 * the partition.)
3565 */
3566 if (sbi->s_journal)
3567 ext4_clear_journal_err(sb, es);
3568 sbi->s_mount_state = le16_to_cpu(es->s_state);
3569 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3570 goto restore_opts;
3571 if (!ext4_setup_super(sb, es, 0))
3572 sb->s_flags &= ~MS_RDONLY;
3573 }
3574 }
3575 ext4_setup_system_zone(sb);
3576 if (sbi->s_journal == NULL)
3577 ext4_commit_super(sb, 1);
3578
3579 #ifdef CONFIG_QUOTA
3580 /* Release old quota file names */
3581 for (i = 0; i < MAXQUOTAS; i++)
3582 if (old_opts.s_qf_names[i] &&
3583 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3584 kfree(old_opts.s_qf_names[i]);
3585 #endif
3586 unlock_super(sb);
3587 unlock_kernel();
3588 return 0;
3589
3590 restore_opts:
3591 sb->s_flags = old_sb_flags;
3592 sbi->s_mount_opt = old_opts.s_mount_opt;
3593 sbi->s_resuid = old_opts.s_resuid;
3594 sbi->s_resgid = old_opts.s_resgid;
3595 sbi->s_commit_interval = old_opts.s_commit_interval;
3596 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3597 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3598 #ifdef CONFIG_QUOTA
3599 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3600 for (i = 0; i < MAXQUOTAS; i++) {
3601 if (sbi->s_qf_names[i] &&
3602 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3603 kfree(sbi->s_qf_names[i]);
3604 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3605 }
3606 #endif
3607 unlock_super(sb);
3608 unlock_kernel();
3609 return err;
3610 }
3611
3612 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3613 {
3614 struct super_block *sb = dentry->d_sb;
3615 struct ext4_sb_info *sbi = EXT4_SB(sb);
3616 struct ext4_super_block *es = sbi->s_es;
3617 u64 fsid;
3618
3619 if (test_opt(sb, MINIX_DF)) {
3620 sbi->s_overhead_last = 0;
3621 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3622 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3623 ext4_fsblk_t overhead = 0;
3624
3625 /*
3626 * Compute the overhead (FS structures). This is constant
3627 * for a given filesystem unless the number of block groups
3628 * changes so we cache the previous value until it does.
3629 */
3630
3631 /*
3632 * All of the blocks before first_data_block are
3633 * overhead
3634 */
3635 overhead = le32_to_cpu(es->s_first_data_block);
3636
3637 /*
3638 * Add the overhead attributed to the superblock and
3639 * block group descriptors. If the sparse superblocks
3640 * feature is turned on, then not all groups have this.
3641 */
3642 for (i = 0; i < ngroups; i++) {
3643 overhead += ext4_bg_has_super(sb, i) +
3644 ext4_bg_num_gdb(sb, i);
3645 cond_resched();
3646 }
3647
3648 /*
3649 * Every block group has an inode bitmap, a block
3650 * bitmap, and an inode table.
3651 */
3652 overhead += ngroups * (2 + sbi->s_itb_per_group);
3653 sbi->s_overhead_last = overhead;
3654 smp_wmb();
3655 sbi->s_blocks_last = ext4_blocks_count(es);
3656 }
3657
3658 buf->f_type = EXT4_SUPER_MAGIC;
3659 buf->f_bsize = sb->s_blocksize;
3660 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3661 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3662 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3663 ext4_free_blocks_count_set(es, buf->f_bfree);
3664 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3665 if (buf->f_bfree < ext4_r_blocks_count(es))
3666 buf->f_bavail = 0;
3667 buf->f_files = le32_to_cpu(es->s_inodes_count);
3668 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3669 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
3670 buf->f_namelen = EXT4_NAME_LEN;
3671 fsid = le64_to_cpup((void *)es->s_uuid) ^
3672 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3673 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3674 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3675
3676 return 0;
3677 }
3678
3679 /* Helper function for writing quotas on sync - we need to start transaction
3680 * before quota file is locked for write. Otherwise the are possible deadlocks:
3681 * Process 1 Process 2
3682 * ext4_create() quota_sync()
3683 * jbd2_journal_start() write_dquot()
3684 * vfs_dq_init() down(dqio_mutex)
3685 * down(dqio_mutex) jbd2_journal_start()
3686 *
3687 */
3688
3689 #ifdef CONFIG_QUOTA
3690
3691 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3692 {
3693 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3694 }
3695
3696 static int ext4_write_dquot(struct dquot *dquot)
3697 {
3698 int ret, err;
3699 handle_t *handle;
3700 struct inode *inode;
3701
3702 inode = dquot_to_inode(dquot);
3703 handle = ext4_journal_start(inode,
3704 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3705 if (IS_ERR(handle))
3706 return PTR_ERR(handle);
3707 ret = dquot_commit(dquot);
3708 err = ext4_journal_stop(handle);
3709 if (!ret)
3710 ret = err;
3711 return ret;
3712 }
3713
3714 static int ext4_acquire_dquot(struct dquot *dquot)
3715 {
3716 int ret, err;
3717 handle_t *handle;
3718
3719 handle = ext4_journal_start(dquot_to_inode(dquot),
3720 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3721 if (IS_ERR(handle))
3722 return PTR_ERR(handle);
3723 ret = dquot_acquire(dquot);
3724 err = ext4_journal_stop(handle);
3725 if (!ret)
3726 ret = err;
3727 return ret;
3728 }
3729
3730 static int ext4_release_dquot(struct dquot *dquot)
3731 {
3732 int ret, err;
3733 handle_t *handle;
3734
3735 handle = ext4_journal_start(dquot_to_inode(dquot),
3736 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3737 if (IS_ERR(handle)) {
3738 /* Release dquot anyway to avoid endless cycle in dqput() */
3739 dquot_release(dquot);
3740 return PTR_ERR(handle);
3741 }
3742 ret = dquot_release(dquot);
3743 err = ext4_journal_stop(handle);
3744 if (!ret)
3745 ret = err;
3746 return ret;
3747 }
3748
3749 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3750 {
3751 /* Are we journaling quotas? */
3752 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3753 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3754 dquot_mark_dquot_dirty(dquot);
3755 return ext4_write_dquot(dquot);
3756 } else {
3757 return dquot_mark_dquot_dirty(dquot);
3758 }
3759 }
3760
3761 static int ext4_write_info(struct super_block *sb, int type)
3762 {
3763 int ret, err;
3764 handle_t *handle;
3765
3766 /* Data block + inode block */
3767 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3768 if (IS_ERR(handle))
3769 return PTR_ERR(handle);
3770 ret = dquot_commit_info(sb, type);
3771 err = ext4_journal_stop(handle);
3772 if (!ret)
3773 ret = err;
3774 return ret;
3775 }
3776
3777 /*
3778 * Turn on quotas during mount time - we need to find
3779 * the quota file and such...
3780 */
3781 static int ext4_quota_on_mount(struct super_block *sb, int type)
3782 {
3783 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3784 EXT4_SB(sb)->s_jquota_fmt, type);
3785 }
3786
3787 /*
3788 * Standard function to be called on quota_on
3789 */
3790 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3791 char *name, int remount)
3792 {
3793 int err;
3794 struct path path;
3795
3796 if (!test_opt(sb, QUOTA))
3797 return -EINVAL;
3798 /* When remounting, no checks are needed and in fact, name is NULL */
3799 if (remount)
3800 return vfs_quota_on(sb, type, format_id, name, remount);
3801
3802 err = kern_path(name, LOOKUP_FOLLOW, &path);
3803 if (err)
3804 return err;
3805
3806 /* Quotafile not on the same filesystem? */
3807 if (path.mnt->mnt_sb != sb) {
3808 path_put(&path);
3809 return -EXDEV;
3810 }
3811 /* Journaling quota? */
3812 if (EXT4_SB(sb)->s_qf_names[type]) {
3813 /* Quotafile not in fs root? */
3814 if (path.dentry->d_parent != sb->s_root)
3815 ext4_msg(sb, KERN_WARNING,
3816 "Quota file not on filesystem root. "
3817 "Journaled quota will not work");
3818 }
3819
3820 /*
3821 * When we journal data on quota file, we have to flush journal to see
3822 * all updates to the file when we bypass pagecache...
3823 */
3824 if (EXT4_SB(sb)->s_journal &&
3825 ext4_should_journal_data(path.dentry->d_inode)) {
3826 /*
3827 * We don't need to lock updates but journal_flush() could
3828 * otherwise be livelocked...
3829 */
3830 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3831 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3832 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3833 if (err) {
3834 path_put(&path);
3835 return err;
3836 }
3837 }
3838
3839 err = vfs_quota_on_path(sb, type, format_id, &path);
3840 path_put(&path);
3841 return err;
3842 }
3843
3844 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3845 * acquiring the locks... As quota files are never truncated and quota code
3846 * itself serializes the operations (and noone else should touch the files)
3847 * we don't have to be afraid of races */
3848 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3849 size_t len, loff_t off)
3850 {
3851 struct inode *inode = sb_dqopt(sb)->files[type];
3852 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3853 int err = 0;
3854 int offset = off & (sb->s_blocksize - 1);
3855 int tocopy;
3856 size_t toread;
3857 struct buffer_head *bh;
3858 loff_t i_size = i_size_read(inode);
3859
3860 if (off > i_size)
3861 return 0;
3862 if (off+len > i_size)
3863 len = i_size-off;
3864 toread = len;
3865 while (toread > 0) {
3866 tocopy = sb->s_blocksize - offset < toread ?
3867 sb->s_blocksize - offset : toread;
3868 bh = ext4_bread(NULL, inode, blk, 0, &err);
3869 if (err)
3870 return err;
3871 if (!bh) /* A hole? */
3872 memset(data, 0, tocopy);
3873 else
3874 memcpy(data, bh->b_data+offset, tocopy);
3875 brelse(bh);
3876 offset = 0;
3877 toread -= tocopy;
3878 data += tocopy;
3879 blk++;
3880 }
3881 return len;
3882 }
3883
3884 /* Write to quotafile (we know the transaction is already started and has
3885 * enough credits) */
3886 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3887 const char *data, size_t len, loff_t off)
3888 {
3889 struct inode *inode = sb_dqopt(sb)->files[type];
3890 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3891 int err = 0;
3892 int offset = off & (sb->s_blocksize - 1);
3893 int tocopy;
3894 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3895 size_t towrite = len;
3896 struct buffer_head *bh;
3897 handle_t *handle = journal_current_handle();
3898
3899 if (EXT4_SB(sb)->s_journal && !handle) {
3900 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
3901 " cancelled because transaction is not started",
3902 (unsigned long long)off, (unsigned long long)len);
3903 return -EIO;
3904 }
3905 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3906 while (towrite > 0) {
3907 tocopy = sb->s_blocksize - offset < towrite ?
3908 sb->s_blocksize - offset : towrite;
3909 bh = ext4_bread(handle, inode, blk, 1, &err);
3910 if (!bh)
3911 goto out;
3912 if (journal_quota) {
3913 err = ext4_journal_get_write_access(handle, bh);
3914 if (err) {
3915 brelse(bh);
3916 goto out;
3917 }
3918 }
3919 lock_buffer(bh);
3920 memcpy(bh->b_data+offset, data, tocopy);
3921 flush_dcache_page(bh->b_page);
3922 unlock_buffer(bh);
3923 if (journal_quota)
3924 err = ext4_handle_dirty_metadata(handle, NULL, bh);
3925 else {
3926 /* Always do at least ordered writes for quotas */
3927 err = ext4_jbd2_file_inode(handle, inode);
3928 mark_buffer_dirty(bh);
3929 }
3930 brelse(bh);
3931 if (err)
3932 goto out;
3933 offset = 0;
3934 towrite -= tocopy;
3935 data += tocopy;
3936 blk++;
3937 }
3938 out:
3939 if (len == towrite) {
3940 mutex_unlock(&inode->i_mutex);
3941 return err;
3942 }
3943 if (inode->i_size < off+len-towrite) {
3944 i_size_write(inode, off+len-towrite);
3945 EXT4_I(inode)->i_disksize = inode->i_size;
3946 }
3947 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3948 ext4_mark_inode_dirty(handle, inode);
3949 mutex_unlock(&inode->i_mutex);
3950 return len - towrite;
3951 }
3952
3953 #endif
3954
3955 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
3956 const char *dev_name, void *data, struct vfsmount *mnt)
3957 {
3958 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
3959 }
3960
3961 static struct file_system_type ext4_fs_type = {
3962 .owner = THIS_MODULE,
3963 .name = "ext4",
3964 .get_sb = ext4_get_sb,
3965 .kill_sb = kill_block_super,
3966 .fs_flags = FS_REQUIRES_DEV,
3967 };
3968
3969 static int __init init_ext4_fs(void)
3970 {
3971 int err;
3972
3973 err = init_ext4_system_zone();
3974 if (err)
3975 return err;
3976 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
3977 if (!ext4_kset)
3978 goto out4;
3979 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
3980 err = init_ext4_mballoc();
3981 if (err)
3982 goto out3;
3983
3984 err = init_ext4_xattr();
3985 if (err)
3986 goto out2;
3987 err = init_inodecache();
3988 if (err)
3989 goto out1;
3990 err = register_filesystem(&ext4_fs_type);
3991 if (err)
3992 goto out;
3993 return 0;
3994 out:
3995 destroy_inodecache();
3996 out1:
3997 exit_ext4_xattr();
3998 out2:
3999 exit_ext4_mballoc();
4000 out3:
4001 remove_proc_entry("fs/ext4", NULL);
4002 kset_unregister(ext4_kset);
4003 out4:
4004 exit_ext4_system_zone();
4005 return err;
4006 }
4007
4008 static void __exit exit_ext4_fs(void)
4009 {
4010 unregister_filesystem(&ext4_fs_type);
4011 destroy_inodecache();
4012 exit_ext4_xattr();
4013 exit_ext4_mballoc();
4014 remove_proc_entry("fs/ext4", NULL);
4015 kset_unregister(ext4_kset);
4016 exit_ext4_system_zone();
4017 }
4018
4019 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4020 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4021 MODULE_LICENSE("GPL");
4022 module_init(init_ext4_fs)
4023 module_exit(exit_ext4_fs)
x86 "subsystem" repository