?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
drm/radeon/kms: rs6xx/rs740: clamp vram to aperture size
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / fs / ext4 / super.c
blob9ae521749b90cce82d84b46abbe027199b25c473
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 if (sbi->s_journal) {
607 err = jbd2_journal_destroy(sbi->s_journal);
608 sbi->s_journal = NULL;
609 if (err < 0)
610 ext4_abort(sb, __func__,
611 "Couldn't clean up the journal");
612 }
613
614 ext4_release_system_zone(sb);
615 ext4_mb_release(sb);
616 ext4_ext_release(sb);
617 ext4_xattr_put_super(sb);
618
619 if (!(sb->s_flags & MS_RDONLY)) {
620 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
621 es->s_state = cpu_to_le16(sbi->s_mount_state);
622 ext4_commit_super(sb, 1);
623 }
624 if (sbi->s_proc) {
625 remove_proc_entry(sb->s_id, ext4_proc_root);
626 }
627 kobject_del(&sbi->s_kobj);
628
629 for (i = 0; i < sbi->s_gdb_count; i++)
630 brelse(sbi->s_group_desc[i]);
631 kfree(sbi->s_group_desc);
632 if (is_vmalloc_addr(sbi->s_flex_groups))
633 vfree(sbi->s_flex_groups);
634 else
635 kfree(sbi->s_flex_groups);
636 percpu_counter_destroy(&sbi->s_freeblocks_counter);
637 percpu_counter_destroy(&sbi->s_freeinodes_counter);
638 percpu_counter_destroy(&sbi->s_dirs_counter);
639 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
640 brelse(sbi->s_sbh);
641 #ifdef CONFIG_QUOTA
642 for (i = 0; i < MAXQUOTAS; i++)
643 kfree(sbi->s_qf_names[i]);
644 #endif
645
646 /* Debugging code just in case the in-memory inode orphan list
647 * isn't empty. The on-disk one can be non-empty if we've
648 * detected an error and taken the fs readonly, but the
649 * in-memory list had better be clean by this point. */
650 if (!list_empty(&sbi->s_orphan))
651 dump_orphan_list(sb, sbi);
652 J_ASSERT(list_empty(&sbi->s_orphan));
653
654 invalidate_bdev(sb->s_bdev);
655 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
656 /*
657 * Invalidate the journal device's buffers. We don't want them
658 * floating about in memory - the physical journal device may
659 * hotswapped, and it breaks the `ro-after' testing code.
660 */
661 sync_blockdev(sbi->journal_bdev);
662 invalidate_bdev(sbi->journal_bdev);
663 ext4_blkdev_remove(sbi);
664 }
665 sb->s_fs_info = NULL;
666 /*
667 * Now that we are completely done shutting down the
668 * superblock, we need to actually destroy the kobject.
669 */
670 unlock_kernel();
671 unlock_super(sb);
672 kobject_put(&sbi->s_kobj);
673 wait_for_completion(&sbi->s_kobj_unregister);
674 kfree(sbi->s_blockgroup_lock);
675 kfree(sbi);
676 }
677
678 static struct kmem_cache *ext4_inode_cachep;
679
680 /*
681 * Called inside transaction, so use GFP_NOFS
682 */
683 static struct inode *ext4_alloc_inode(struct super_block *sb)
684 {
685 struct ext4_inode_info *ei;
686
687 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
688 if (!ei)
689 return NULL;
690
691 ei->vfs_inode.i_version = 1;
692 ei->vfs_inode.i_data.writeback_index = 0;
693 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
694 INIT_LIST_HEAD(&ei->i_prealloc_list);
695 spin_lock_init(&ei->i_prealloc_lock);
696 /*
697 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
698 * therefore it can be null here. Don't check it, just initialize
699 * jinode.
700 */
701 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
702 ei->i_reserved_data_blocks = 0;
703 ei->i_reserved_meta_blocks = 0;
704 ei->i_allocated_meta_blocks = 0;
705 ei->i_delalloc_reserved_flag = 0;
706 spin_lock_init(&(ei->i_block_reservation_lock));
707 INIT_LIST_HEAD(&ei->i_aio_dio_complete_list);
708 ei->cur_aio_dio = NULL;
709 ei->i_sync_tid = 0;
710 ei->i_datasync_tid = 0;
711
712 return &ei->vfs_inode;
713 }
714
715 static void ext4_destroy_inode(struct inode *inode)
716 {
717 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
718 ext4_msg(inode->i_sb, KERN_ERR,
719 "Inode %lu (%p): orphan list check failed!",
720 inode->i_ino, EXT4_I(inode));
721 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
722 EXT4_I(inode), sizeof(struct ext4_inode_info),
723 true);
724 dump_stack();
725 }
726 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
727 }
728
729 static void init_once(void *foo)
730 {
731 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
732
733 INIT_LIST_HEAD(&ei->i_orphan);
734 #ifdef CONFIG_EXT4_FS_XATTR
735 init_rwsem(&ei->xattr_sem);
736 #endif
737 init_rwsem(&ei->i_data_sem);
738 inode_init_once(&ei->vfs_inode);
739 }
740
741 static int init_inodecache(void)
742 {
743 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
744 sizeof(struct ext4_inode_info),
745 0, (SLAB_RECLAIM_ACCOUNT|
746 SLAB_MEM_SPREAD),
747 init_once);
748 if (ext4_inode_cachep == NULL)
749 return -ENOMEM;
750 return 0;
751 }
752
753 static void destroy_inodecache(void)
754 {
755 kmem_cache_destroy(ext4_inode_cachep);
756 }
757
758 static void ext4_clear_inode(struct inode *inode)
759 {
760 ext4_discard_preallocations(inode);
761 if (EXT4_JOURNAL(inode))
762 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
763 &EXT4_I(inode)->jinode);
764 }
765
766 static inline void ext4_show_quota_options(struct seq_file *seq,
767 struct super_block *sb)
768 {
769 #if defined(CONFIG_QUOTA)
770 struct ext4_sb_info *sbi = EXT4_SB(sb);
771
772 if (sbi->s_jquota_fmt)
773 seq_printf(seq, ",jqfmt=%s",
774 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
775
776 if (sbi->s_qf_names[USRQUOTA])
777 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
778
779 if (sbi->s_qf_names[GRPQUOTA])
780 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
781
782 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
783 seq_puts(seq, ",usrquota");
784
785 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
786 seq_puts(seq, ",grpquota");
787 #endif
788 }
789
790 /*
791 * Show an option if
792 * - it's set to a non-default value OR
793 * - if the per-sb default is different from the global default
794 */
795 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
796 {
797 int def_errors;
798 unsigned long def_mount_opts;
799 struct super_block *sb = vfs->mnt_sb;
800 struct ext4_sb_info *sbi = EXT4_SB(sb);
801 struct ext4_super_block *es = sbi->s_es;
802
803 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
804 def_errors = le16_to_cpu(es->s_errors);
805
806 if (sbi->s_sb_block != 1)
807 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
808 if (test_opt(sb, MINIX_DF))
809 seq_puts(seq, ",minixdf");
810 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
811 seq_puts(seq, ",grpid");
812 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
813 seq_puts(seq, ",nogrpid");
814 if (sbi->s_resuid != EXT4_DEF_RESUID ||
815 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
816 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
817 }
818 if (sbi->s_resgid != EXT4_DEF_RESGID ||
819 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
820 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
821 }
822 if (test_opt(sb, ERRORS_RO)) {
823 if (def_errors == EXT4_ERRORS_PANIC ||
824 def_errors == EXT4_ERRORS_CONTINUE) {
825 seq_puts(seq, ",errors=remount-ro");
826 }
827 }
828 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
829 seq_puts(seq, ",errors=continue");
830 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
831 seq_puts(seq, ",errors=panic");
832 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
833 seq_puts(seq, ",nouid32");
834 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
835 seq_puts(seq, ",debug");
836 if (test_opt(sb, OLDALLOC))
837 seq_puts(seq, ",oldalloc");
838 #ifdef CONFIG_EXT4_FS_XATTR
839 if (test_opt(sb, XATTR_USER) &&
840 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
841 seq_puts(seq, ",user_xattr");
842 if (!test_opt(sb, XATTR_USER) &&
843 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
844 seq_puts(seq, ",nouser_xattr");
845 }
846 #endif
847 #ifdef CONFIG_EXT4_FS_POSIX_ACL
848 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
849 seq_puts(seq, ",acl");
850 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
851 seq_puts(seq, ",noacl");
852 #endif
853 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
854 seq_printf(seq, ",commit=%u",
855 (unsigned) (sbi->s_commit_interval / HZ));
856 }
857 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
858 seq_printf(seq, ",min_batch_time=%u",
859 (unsigned) sbi->s_min_batch_time);
860 }
861 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
862 seq_printf(seq, ",max_batch_time=%u",
863 (unsigned) sbi->s_min_batch_time);
864 }
865
866 /*
867 * We're changing the default of barrier mount option, so
868 * let's always display its mount state so it's clear what its
869 * status is.
870 */
871 seq_puts(seq, ",barrier=");
872 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
873 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
874 seq_puts(seq, ",journal_async_commit");
875 if (test_opt(sb, NOBH))
876 seq_puts(seq, ",nobh");
877 if (test_opt(sb, I_VERSION))
878 seq_puts(seq, ",i_version");
879 if (!test_opt(sb, DELALLOC))
880 seq_puts(seq, ",nodelalloc");
881
882
883 if (sbi->s_stripe)
884 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
885 /*
886 * journal mode get enabled in different ways
887 * So just print the value even if we didn't specify it
888 */
889 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
890 seq_puts(seq, ",data=journal");
891 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
892 seq_puts(seq, ",data=ordered");
893 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
894 seq_puts(seq, ",data=writeback");
895
896 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
897 seq_printf(seq, ",inode_readahead_blks=%u",
898 sbi->s_inode_readahead_blks);
899
900 if (test_opt(sb, DATA_ERR_ABORT))
901 seq_puts(seq, ",data_err=abort");
902
903 if (test_opt(sb, NO_AUTO_DA_ALLOC))
904 seq_puts(seq, ",noauto_da_alloc");
905
906 if (test_opt(sb, DISCARD))
907 seq_puts(seq, ",discard");
908
909 if (test_opt(sb, NOLOAD))
910 seq_puts(seq, ",norecovery");
911
912 ext4_show_quota_options(seq, sb);
913
914 return 0;
915 }
916
917 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
918 u64 ino, u32 generation)
919 {
920 struct inode *inode;
921
922 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
923 return ERR_PTR(-ESTALE);
924 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
925 return ERR_PTR(-ESTALE);
926
927 /* iget isn't really right if the inode is currently unallocated!!
928 *
929 * ext4_read_inode will return a bad_inode if the inode had been
930 * deleted, so we should be safe.
931 *
932 * Currently we don't know the generation for parent directory, so
933 * a generation of 0 means "accept any"
934 */
935 inode = ext4_iget(sb, ino);
936 if (IS_ERR(inode))
937 return ERR_CAST(inode);
938 if (generation && inode->i_generation != generation) {
939 iput(inode);
940 return ERR_PTR(-ESTALE);
941 }
942
943 return inode;
944 }
945
946 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
947 int fh_len, int fh_type)
948 {
949 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
950 ext4_nfs_get_inode);
951 }
952
953 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
954 int fh_len, int fh_type)
955 {
956 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
957 ext4_nfs_get_inode);
958 }
959
960 /*
961 * Try to release metadata pages (indirect blocks, directories) which are
962 * mapped via the block device. Since these pages could have journal heads
963 * which would prevent try_to_free_buffers() from freeing them, we must use
964 * jbd2 layer's try_to_free_buffers() function to release them.
965 */
966 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
967 gfp_t wait)
968 {
969 journal_t *journal = EXT4_SB(sb)->s_journal;
970
971 WARN_ON(PageChecked(page));
972 if (!page_has_buffers(page))
973 return 0;
974 if (journal)
975 return jbd2_journal_try_to_free_buffers(journal, page,
976 wait & ~__GFP_WAIT);
977 return try_to_free_buffers(page);
978 }
979
980 #ifdef CONFIG_QUOTA
981 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
982 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
983
984 static int ext4_write_dquot(struct dquot *dquot);
985 static int ext4_acquire_dquot(struct dquot *dquot);
986 static int ext4_release_dquot(struct dquot *dquot);
987 static int ext4_mark_dquot_dirty(struct dquot *dquot);
988 static int ext4_write_info(struct super_block *sb, int type);
989 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
990 char *path, int remount);
991 static int ext4_quota_on_mount(struct super_block *sb, int type);
992 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
993 size_t len, loff_t off);
994 static ssize_t ext4_quota_write(struct super_block *sb, int type,
995 const char *data, size_t len, loff_t off);
996
997 static const struct dquot_operations ext4_quota_operations = {
998 .initialize = dquot_initialize,
999 .drop = dquot_drop,
1000 .alloc_space = dquot_alloc_space,
1001 .reserve_space = dquot_reserve_space,
1002 .claim_space = dquot_claim_space,
1003 .release_rsv = dquot_release_reserved_space,
1004 .get_reserved_space = ext4_get_reserved_space,
1005 .alloc_inode = dquot_alloc_inode,
1006 .free_space = dquot_free_space,
1007 .free_inode = dquot_free_inode,
1008 .transfer = dquot_transfer,
1009 .write_dquot = ext4_write_dquot,
1010 .acquire_dquot = ext4_acquire_dquot,
1011 .release_dquot = ext4_release_dquot,
1012 .mark_dirty = ext4_mark_dquot_dirty,
1013 .write_info = ext4_write_info,
1014 .alloc_dquot = dquot_alloc,
1015 .destroy_dquot = dquot_destroy,
1016 };
1017
1018 static const struct quotactl_ops ext4_qctl_operations = {
1019 .quota_on = ext4_quota_on,
1020 .quota_off = vfs_quota_off,
1021 .quota_sync = vfs_quota_sync,
1022 .get_info = vfs_get_dqinfo,
1023 .set_info = vfs_set_dqinfo,
1024 .get_dqblk = vfs_get_dqblk,
1025 .set_dqblk = vfs_set_dqblk
1026 };
1027 #endif
1028
1029 static const struct super_operations ext4_sops = {
1030 .alloc_inode = ext4_alloc_inode,
1031 .destroy_inode = ext4_destroy_inode,
1032 .write_inode = ext4_write_inode,
1033 .dirty_inode = ext4_dirty_inode,
1034 .delete_inode = ext4_delete_inode,
1035 .put_super = ext4_put_super,
1036 .sync_fs = ext4_sync_fs,
1037 .freeze_fs = ext4_freeze,
1038 .unfreeze_fs = ext4_unfreeze,
1039 .statfs = ext4_statfs,
1040 .remount_fs = ext4_remount,
1041 .clear_inode = ext4_clear_inode,
1042 .show_options = ext4_show_options,
1043 #ifdef CONFIG_QUOTA
1044 .quota_read = ext4_quota_read,
1045 .quota_write = ext4_quota_write,
1046 #endif
1047 .bdev_try_to_free_page = bdev_try_to_free_page,
1048 };
1049
1050 static const struct super_operations ext4_nojournal_sops = {
1051 .alloc_inode = ext4_alloc_inode,
1052 .destroy_inode = ext4_destroy_inode,
1053 .write_inode = ext4_write_inode,
1054 .dirty_inode = ext4_dirty_inode,
1055 .delete_inode = ext4_delete_inode,
1056 .write_super = ext4_write_super,
1057 .put_super = ext4_put_super,
1058 .statfs = ext4_statfs,
1059 .remount_fs = ext4_remount,
1060 .clear_inode = ext4_clear_inode,
1061 .show_options = ext4_show_options,
1062 #ifdef CONFIG_QUOTA
1063 .quota_read = ext4_quota_read,
1064 .quota_write = ext4_quota_write,
1065 #endif
1066 .bdev_try_to_free_page = bdev_try_to_free_page,
1067 };
1068
1069 static const struct export_operations ext4_export_ops = {
1070 .fh_to_dentry = ext4_fh_to_dentry,
1071 .fh_to_parent = ext4_fh_to_parent,
1072 .get_parent = ext4_get_parent,
1073 };
1074
1075 enum {
1076 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1077 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1078 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1079 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1080 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1081 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1082 Opt_journal_update, Opt_journal_dev,
1083 Opt_journal_checksum, Opt_journal_async_commit,
1084 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1085 Opt_data_err_abort, Opt_data_err_ignore,
1086 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1087 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
1088 Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err, Opt_resize,
1089 Opt_usrquota, Opt_grpquota, Opt_i_version,
1090 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1091 Opt_block_validity, Opt_noblock_validity,
1092 Opt_inode_readahead_blks, Opt_journal_ioprio,
1093 Opt_discard, Opt_nodiscard,
1094 };
1095
1096 static const match_table_t tokens = {
1097 {Opt_bsd_df, "bsddf"},
1098 {Opt_minix_df, "minixdf"},
1099 {Opt_grpid, "grpid"},
1100 {Opt_grpid, "bsdgroups"},
1101 {Opt_nogrpid, "nogrpid"},
1102 {Opt_nogrpid, "sysvgroups"},
1103 {Opt_resgid, "resgid=%u"},
1104 {Opt_resuid, "resuid=%u"},
1105 {Opt_sb, "sb=%u"},
1106 {Opt_err_cont, "errors=continue"},
1107 {Opt_err_panic, "errors=panic"},
1108 {Opt_err_ro, "errors=remount-ro"},
1109 {Opt_nouid32, "nouid32"},
1110 {Opt_debug, "debug"},
1111 {Opt_oldalloc, "oldalloc"},
1112 {Opt_orlov, "orlov"},
1113 {Opt_user_xattr, "user_xattr"},
1114 {Opt_nouser_xattr, "nouser_xattr"},
1115 {Opt_acl, "acl"},
1116 {Opt_noacl, "noacl"},
1117 {Opt_noload, "noload"},
1118 {Opt_noload, "norecovery"},
1119 {Opt_nobh, "nobh"},
1120 {Opt_bh, "bh"},
1121 {Opt_commit, "commit=%u"},
1122 {Opt_min_batch_time, "min_batch_time=%u"},
1123 {Opt_max_batch_time, "max_batch_time=%u"},
1124 {Opt_journal_update, "journal=update"},
1125 {Opt_journal_dev, "journal_dev=%u"},
1126 {Opt_journal_checksum, "journal_checksum"},
1127 {Opt_journal_async_commit, "journal_async_commit"},
1128 {Opt_abort, "abort"},
1129 {Opt_data_journal, "data=journal"},
1130 {Opt_data_ordered, "data=ordered"},
1131 {Opt_data_writeback, "data=writeback"},
1132 {Opt_data_err_abort, "data_err=abort"},
1133 {Opt_data_err_ignore, "data_err=ignore"},
1134 {Opt_offusrjquota, "usrjquota="},
1135 {Opt_usrjquota, "usrjquota=%s"},
1136 {Opt_offgrpjquota, "grpjquota="},
1137 {Opt_grpjquota, "grpjquota=%s"},
1138 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1139 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1140 {Opt_grpquota, "grpquota"},
1141 {Opt_noquota, "noquota"},
1142 {Opt_quota, "quota"},
1143 {Opt_usrquota, "usrquota"},
1144 {Opt_barrier, "barrier=%u"},
1145 {Opt_barrier, "barrier"},
1146 {Opt_nobarrier, "nobarrier"},
1147 {Opt_i_version, "i_version"},
1148 {Opt_stripe, "stripe=%u"},
1149 {Opt_resize, "resize"},
1150 {Opt_delalloc, "delalloc"},
1151 {Opt_nodelalloc, "nodelalloc"},
1152 {Opt_block_validity, "block_validity"},
1153 {Opt_noblock_validity, "noblock_validity"},
1154 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1155 {Opt_journal_ioprio, "journal_ioprio=%u"},
1156 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1157 {Opt_auto_da_alloc, "auto_da_alloc"},
1158 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1159 {Opt_discard, "discard"},
1160 {Opt_nodiscard, "nodiscard"},
1161 {Opt_err, NULL},
1162 };
1163
1164 static ext4_fsblk_t get_sb_block(void **data)
1165 {
1166 ext4_fsblk_t sb_block;
1167 char *options = (char *) *data;
1168
1169 if (!options || strncmp(options, "sb=", 3) != 0)
1170 return 1; /* Default location */
1171
1172 options += 3;
1173 /* TODO: use simple_strtoll with >32bit ext4 */
1174 sb_block = simple_strtoul(options, &options, 0);
1175 if (*options && *options != ',') {
1176 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1177 (char *) *data);
1178 return 1;
1179 }
1180 if (*options == ',')
1181 options++;
1182 *data = (void *) options;
1183
1184 return sb_block;
1185 }
1186
1187 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1188
1189 static int parse_options(char *options, struct super_block *sb,
1190 unsigned long *journal_devnum,
1191 unsigned int *journal_ioprio,
1192 ext4_fsblk_t *n_blocks_count, int is_remount)
1193 {
1194 struct ext4_sb_info *sbi = EXT4_SB(sb);
1195 char *p;
1196 substring_t args[MAX_OPT_ARGS];
1197 int data_opt = 0;
1198 int option;
1199 #ifdef CONFIG_QUOTA
1200 int qtype, qfmt;
1201 char *qname;
1202 #endif
1203
1204 if (!options)
1205 return 1;
1206
1207 while ((p = strsep(&options, ",")) != NULL) {
1208 int token;
1209 if (!*p)
1210 continue;
1211
1212 token = match_token(p, tokens, args);
1213 switch (token) {
1214 case Opt_bsd_df:
1215 clear_opt(sbi->s_mount_opt, MINIX_DF);
1216 break;
1217 case Opt_minix_df:
1218 set_opt(sbi->s_mount_opt, MINIX_DF);
1219 break;
1220 case Opt_grpid:
1221 set_opt(sbi->s_mount_opt, GRPID);
1222 break;
1223 case Opt_nogrpid:
1224 clear_opt(sbi->s_mount_opt, GRPID);
1225 break;
1226 case Opt_resuid:
1227 if (match_int(&args[0], &option))
1228 return 0;
1229 sbi->s_resuid = option;
1230 break;
1231 case Opt_resgid:
1232 if (match_int(&args[0], &option))
1233 return 0;
1234 sbi->s_resgid = option;
1235 break;
1236 case Opt_sb:
1237 /* handled by get_sb_block() instead of here */
1238 /* *sb_block = match_int(&args[0]); */
1239 break;
1240 case Opt_err_panic:
1241 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1242 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1243 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1244 break;
1245 case Opt_err_ro:
1246 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1247 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1248 set_opt(sbi->s_mount_opt, ERRORS_RO);
1249 break;
1250 case Opt_err_cont:
1251 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1252 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1253 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1254 break;
1255 case Opt_nouid32:
1256 set_opt(sbi->s_mount_opt, NO_UID32);
1257 break;
1258 case Opt_debug:
1259 set_opt(sbi->s_mount_opt, DEBUG);
1260 break;
1261 case Opt_oldalloc:
1262 set_opt(sbi->s_mount_opt, OLDALLOC);
1263 break;
1264 case Opt_orlov:
1265 clear_opt(sbi->s_mount_opt, OLDALLOC);
1266 break;
1267 #ifdef CONFIG_EXT4_FS_XATTR
1268 case Opt_user_xattr:
1269 set_opt(sbi->s_mount_opt, XATTR_USER);
1270 break;
1271 case Opt_nouser_xattr:
1272 clear_opt(sbi->s_mount_opt, XATTR_USER);
1273 break;
1274 #else
1275 case Opt_user_xattr:
1276 case Opt_nouser_xattr:
1277 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1278 break;
1279 #endif
1280 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1281 case Opt_acl:
1282 set_opt(sbi->s_mount_opt, POSIX_ACL);
1283 break;
1284 case Opt_noacl:
1285 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1286 break;
1287 #else
1288 case Opt_acl:
1289 case Opt_noacl:
1290 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1291 break;
1292 #endif
1293 case Opt_journal_update:
1294 /* @@@ FIXME */
1295 /* Eventually we will want to be able to create
1296 a journal file here. For now, only allow the
1297 user to specify an existing inode to be the
1298 journal file. */
1299 if (is_remount) {
1300 ext4_msg(sb, KERN_ERR,
1301 "Cannot specify journal on remount");
1302 return 0;
1303 }
1304 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1305 break;
1306 case Opt_journal_dev:
1307 if (is_remount) {
1308 ext4_msg(sb, KERN_ERR,
1309 "Cannot specify journal on remount");
1310 return 0;
1311 }
1312 if (match_int(&args[0], &option))
1313 return 0;
1314 *journal_devnum = option;
1315 break;
1316 case Opt_journal_checksum:
1317 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1318 break;
1319 case Opt_journal_async_commit:
1320 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1321 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1322 break;
1323 case Opt_noload:
1324 set_opt(sbi->s_mount_opt, NOLOAD);
1325 break;
1326 case Opt_commit:
1327 if (match_int(&args[0], &option))
1328 return 0;
1329 if (option < 0)
1330 return 0;
1331 if (option == 0)
1332 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1333 sbi->s_commit_interval = HZ * option;
1334 break;
1335 case Opt_max_batch_time:
1336 if (match_int(&args[0], &option))
1337 return 0;
1338 if (option < 0)
1339 return 0;
1340 if (option == 0)
1341 option = EXT4_DEF_MAX_BATCH_TIME;
1342 sbi->s_max_batch_time = option;
1343 break;
1344 case Opt_min_batch_time:
1345 if (match_int(&args[0], &option))
1346 return 0;
1347 if (option < 0)
1348 return 0;
1349 sbi->s_min_batch_time = option;
1350 break;
1351 case Opt_data_journal:
1352 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1353 goto datacheck;
1354 case Opt_data_ordered:
1355 data_opt = EXT4_MOUNT_ORDERED_DATA;
1356 goto datacheck;
1357 case Opt_data_writeback:
1358 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1359 datacheck:
1360 if (is_remount) {
1361 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1362 != data_opt) {
1363 ext4_msg(sb, KERN_ERR,
1364 "Cannot change data mode on remount");
1365 return 0;
1366 }
1367 } else {
1368 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1369 sbi->s_mount_opt |= data_opt;
1370 }
1371 break;
1372 case Opt_data_err_abort:
1373 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1374 break;
1375 case Opt_data_err_ignore:
1376 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1377 break;
1378 #ifdef CONFIG_QUOTA
1379 case Opt_usrjquota:
1380 qtype = USRQUOTA;
1381 goto set_qf_name;
1382 case Opt_grpjquota:
1383 qtype = GRPQUOTA;
1384 set_qf_name:
1385 if (sb_any_quota_loaded(sb) &&
1386 !sbi->s_qf_names[qtype]) {
1387 ext4_msg(sb, KERN_ERR,
1388 "Cannot change journaled "
1389 "quota options when quota turned on");
1390 return 0;
1391 }
1392 qname = match_strdup(&args[0]);
1393 if (!qname) {
1394 ext4_msg(sb, KERN_ERR,
1395 "Not enough memory for "
1396 "storing quotafile name");
1397 return 0;
1398 }
1399 if (sbi->s_qf_names[qtype] &&
1400 strcmp(sbi->s_qf_names[qtype], qname)) {
1401 ext4_msg(sb, KERN_ERR,
1402 "%s quota file already "
1403 "specified", QTYPE2NAME(qtype));
1404 kfree(qname);
1405 return 0;
1406 }
1407 sbi->s_qf_names[qtype] = qname;
1408 if (strchr(sbi->s_qf_names[qtype], '/')) {
1409 ext4_msg(sb, KERN_ERR,
1410 "quotafile must be on "
1411 "filesystem root");
1412 kfree(sbi->s_qf_names[qtype]);
1413 sbi->s_qf_names[qtype] = NULL;
1414 return 0;
1415 }
1416 set_opt(sbi->s_mount_opt, QUOTA);
1417 break;
1418 case Opt_offusrjquota:
1419 qtype = USRQUOTA;
1420 goto clear_qf_name;
1421 case Opt_offgrpjquota:
1422 qtype = GRPQUOTA;
1423 clear_qf_name:
1424 if (sb_any_quota_loaded(sb) &&
1425 sbi->s_qf_names[qtype]) {
1426 ext4_msg(sb, KERN_ERR, "Cannot change "
1427 "journaled quota options when "
1428 "quota turned on");
1429 return 0;
1430 }
1431 /*
1432 * The space will be released later when all options
1433 * are confirmed to be correct
1434 */
1435 sbi->s_qf_names[qtype] = NULL;
1436 break;
1437 case Opt_jqfmt_vfsold:
1438 qfmt = QFMT_VFS_OLD;
1439 goto set_qf_format;
1440 case Opt_jqfmt_vfsv0:
1441 qfmt = QFMT_VFS_V0;
1442 set_qf_format:
1443 if (sb_any_quota_loaded(sb) &&
1444 sbi->s_jquota_fmt != qfmt) {
1445 ext4_msg(sb, KERN_ERR, "Cannot change "
1446 "journaled quota options when "
1447 "quota turned on");
1448 return 0;
1449 }
1450 sbi->s_jquota_fmt = qfmt;
1451 break;
1452 case Opt_quota:
1453 case Opt_usrquota:
1454 set_opt(sbi->s_mount_opt, QUOTA);
1455 set_opt(sbi->s_mount_opt, USRQUOTA);
1456 break;
1457 case Opt_grpquota:
1458 set_opt(sbi->s_mount_opt, QUOTA);
1459 set_opt(sbi->s_mount_opt, GRPQUOTA);
1460 break;
1461 case Opt_noquota:
1462 if (sb_any_quota_loaded(sb)) {
1463 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1464 "options when quota turned on");
1465 return 0;
1466 }
1467 clear_opt(sbi->s_mount_opt, QUOTA);
1468 clear_opt(sbi->s_mount_opt, USRQUOTA);
1469 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1470 break;
1471 #else
1472 case Opt_quota:
1473 case Opt_usrquota:
1474 case Opt_grpquota:
1475 ext4_msg(sb, KERN_ERR,
1476 "quota options not supported");
1477 break;
1478 case Opt_usrjquota:
1479 case Opt_grpjquota:
1480 case Opt_offusrjquota:
1481 case Opt_offgrpjquota:
1482 case Opt_jqfmt_vfsold:
1483 case Opt_jqfmt_vfsv0:
1484 ext4_msg(sb, KERN_ERR,
1485 "journaled quota options not supported");
1486 break;
1487 case Opt_noquota:
1488 break;
1489 #endif
1490 case Opt_abort:
1491 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1492 break;
1493 case Opt_nobarrier:
1494 clear_opt(sbi->s_mount_opt, BARRIER);
1495 break;
1496 case Opt_barrier:
1497 if (match_int(&args[0], &option)) {
1498 set_opt(sbi->s_mount_opt, BARRIER);
1499 break;
1500 }
1501 if (option)
1502 set_opt(sbi->s_mount_opt, BARRIER);
1503 else
1504 clear_opt(sbi->s_mount_opt, BARRIER);
1505 break;
1506 case Opt_ignore:
1507 break;
1508 case Opt_resize:
1509 if (!is_remount) {
1510 ext4_msg(sb, KERN_ERR,
1511 "resize option only available "
1512 "for remount");
1513 return 0;
1514 }
1515 if (match_int(&args[0], &option) != 0)
1516 return 0;
1517 *n_blocks_count = option;
1518 break;
1519 case Opt_nobh:
1520 set_opt(sbi->s_mount_opt, NOBH);
1521 break;
1522 case Opt_bh:
1523 clear_opt(sbi->s_mount_opt, NOBH);
1524 break;
1525 case Opt_i_version:
1526 set_opt(sbi->s_mount_opt, I_VERSION);
1527 sb->s_flags |= MS_I_VERSION;
1528 break;
1529 case Opt_nodelalloc:
1530 clear_opt(sbi->s_mount_opt, DELALLOC);
1531 break;
1532 case Opt_stripe:
1533 if (match_int(&args[0], &option))
1534 return 0;
1535 if (option < 0)
1536 return 0;
1537 sbi->s_stripe = option;
1538 break;
1539 case Opt_delalloc:
1540 set_opt(sbi->s_mount_opt, DELALLOC);
1541 break;
1542 case Opt_block_validity:
1543 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1544 break;
1545 case Opt_noblock_validity:
1546 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1547 break;
1548 case Opt_inode_readahead_blks:
1549 if (match_int(&args[0], &option))
1550 return 0;
1551 if (option < 0 || option > (1 << 30))
1552 return 0;
1553 if (!is_power_of_2(option)) {
1554 ext4_msg(sb, KERN_ERR,
1555 "EXT4-fs: inode_readahead_blks"
1556 " must be a power of 2");
1557 return 0;
1558 }
1559 sbi->s_inode_readahead_blks = option;
1560 break;
1561 case Opt_journal_ioprio:
1562 if (match_int(&args[0], &option))
1563 return 0;
1564 if (option < 0 || option > 7)
1565 break;
1566 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1567 option);
1568 break;
1569 case Opt_noauto_da_alloc:
1570 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1571 break;
1572 case Opt_auto_da_alloc:
1573 if (match_int(&args[0], &option)) {
1574 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1575 break;
1576 }
1577 if (option)
1578 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1579 else
1580 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1581 break;
1582 case Opt_discard:
1583 set_opt(sbi->s_mount_opt, DISCARD);
1584 break;
1585 case Opt_nodiscard:
1586 clear_opt(sbi->s_mount_opt, DISCARD);
1587 break;
1588 default:
1589 ext4_msg(sb, KERN_ERR,
1590 "Unrecognized mount option \"%s\" "
1591 "or missing value", p);
1592 return 0;
1593 }
1594 }
1595 #ifdef CONFIG_QUOTA
1596 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1597 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1598 sbi->s_qf_names[USRQUOTA])
1599 clear_opt(sbi->s_mount_opt, USRQUOTA);
1600
1601 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1602 sbi->s_qf_names[GRPQUOTA])
1603 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1604
1605 if ((sbi->s_qf_names[USRQUOTA] &&
1606 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1607 (sbi->s_qf_names[GRPQUOTA] &&
1608 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1609 ext4_msg(sb, KERN_ERR, "old and new quota "
1610 "format mixing");
1611 return 0;
1612 }
1613
1614 if (!sbi->s_jquota_fmt) {
1615 ext4_msg(sb, KERN_ERR, "journaled quota format "
1616 "not specified");
1617 return 0;
1618 }
1619 } else {
1620 if (sbi->s_jquota_fmt) {
1621 ext4_msg(sb, KERN_ERR, "journaled quota format "
1622 "specified with no journaling "
1623 "enabled");
1624 return 0;
1625 }
1626 }
1627 #endif
1628 return 1;
1629 }
1630
1631 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1632 int read_only)
1633 {
1634 struct ext4_sb_info *sbi = EXT4_SB(sb);
1635 int res = 0;
1636
1637 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1638 ext4_msg(sb, KERN_ERR, "revision level too high, "
1639 "forcing read-only mode");
1640 res = MS_RDONLY;
1641 }
1642 if (read_only)
1643 return res;
1644 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1645 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1646 "running e2fsck is recommended");
1647 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1648 ext4_msg(sb, KERN_WARNING,
1649 "warning: mounting fs with errors, "
1650 "running e2fsck is recommended");
1651 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1652 le16_to_cpu(es->s_mnt_count) >=
1653 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1654 ext4_msg(sb, KERN_WARNING,
1655 "warning: maximal mount count reached, "
1656 "running e2fsck is recommended");
1657 else if (le32_to_cpu(es->s_checkinterval) &&
1658 (le32_to_cpu(es->s_lastcheck) +
1659 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1660 ext4_msg(sb, KERN_WARNING,
1661 "warning: checktime reached, "
1662 "running e2fsck is recommended");
1663 if (!sbi->s_journal)
1664 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1665 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1666 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1667 le16_add_cpu(&es->s_mnt_count, 1);
1668 es->s_mtime = cpu_to_le32(get_seconds());
1669 ext4_update_dynamic_rev(sb);
1670 if (sbi->s_journal)
1671 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1672
1673 ext4_commit_super(sb, 1);
1674 if (test_opt(sb, DEBUG))
1675 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1676 "bpg=%lu, ipg=%lu, mo=%04x]\n",
1677 sb->s_blocksize,
1678 sbi->s_groups_count,
1679 EXT4_BLOCKS_PER_GROUP(sb),
1680 EXT4_INODES_PER_GROUP(sb),
1681 sbi->s_mount_opt);
1682
1683 return res;
1684 }
1685
1686 static int ext4_fill_flex_info(struct super_block *sb)
1687 {
1688 struct ext4_sb_info *sbi = EXT4_SB(sb);
1689 struct ext4_group_desc *gdp = NULL;
1690 ext4_group_t flex_group_count;
1691 ext4_group_t flex_group;
1692 int groups_per_flex = 0;
1693 size_t size;
1694 int i;
1695
1696 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1697 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1698
1699 if (groups_per_flex < 2) {
1700 sbi->s_log_groups_per_flex = 0;
1701 return 1;
1702 }
1703
1704 /* We allocate both existing and potentially added groups */
1705 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1706 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1707 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1708 size = flex_group_count * sizeof(struct flex_groups);
1709 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1710 if (sbi->s_flex_groups == NULL) {
1711 sbi->s_flex_groups = vmalloc(size);
1712 if (sbi->s_flex_groups)
1713 memset(sbi->s_flex_groups, 0, size);
1714 }
1715 if (sbi->s_flex_groups == NULL) {
1716 ext4_msg(sb, KERN_ERR, "not enough memory for "
1717 "%u flex groups", flex_group_count);
1718 goto failed;
1719 }
1720
1721 for (i = 0; i < sbi->s_groups_count; i++) {
1722 gdp = ext4_get_group_desc(sb, i, NULL);
1723
1724 flex_group = ext4_flex_group(sbi, i);
1725 atomic_add(ext4_free_inodes_count(sb, gdp),
1726 &sbi->s_flex_groups[flex_group].free_inodes);
1727 atomic_add(ext4_free_blks_count(sb, gdp),
1728 &sbi->s_flex_groups[flex_group].free_blocks);
1729 atomic_add(ext4_used_dirs_count(sb, gdp),
1730 &sbi->s_flex_groups[flex_group].used_dirs);
1731 }
1732
1733 return 1;
1734 failed:
1735 return 0;
1736 }
1737
1738 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1739 struct ext4_group_desc *gdp)
1740 {
1741 __u16 crc = 0;
1742
1743 if (sbi->s_es->s_feature_ro_compat &
1744 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1745 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1746 __le32 le_group = cpu_to_le32(block_group);
1747
1748 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1749 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1750 crc = crc16(crc, (__u8 *)gdp, offset);
1751 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1752 /* for checksum of struct ext4_group_desc do the rest...*/
1753 if ((sbi->s_es->s_feature_incompat &
1754 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1755 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1756 crc = crc16(crc, (__u8 *)gdp + offset,
1757 le16_to_cpu(sbi->s_es->s_desc_size) -
1758 offset);
1759 }
1760
1761 return cpu_to_le16(crc);
1762 }
1763
1764 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1765 struct ext4_group_desc *gdp)
1766 {
1767 if ((sbi->s_es->s_feature_ro_compat &
1768 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1769 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1770 return 0;
1771
1772 return 1;
1773 }
1774
1775 /* Called at mount-time, super-block is locked */
1776 static int ext4_check_descriptors(struct super_block *sb)
1777 {
1778 struct ext4_sb_info *sbi = EXT4_SB(sb);
1779 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1780 ext4_fsblk_t last_block;
1781 ext4_fsblk_t block_bitmap;
1782 ext4_fsblk_t inode_bitmap;
1783 ext4_fsblk_t inode_table;
1784 int flexbg_flag = 0;
1785 ext4_group_t i;
1786
1787 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1788 flexbg_flag = 1;
1789
1790 ext4_debug("Checking group descriptors");
1791
1792 for (i = 0; i < sbi->s_groups_count; i++) {
1793 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1794
1795 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1796 last_block = ext4_blocks_count(sbi->s_es) - 1;
1797 else
1798 last_block = first_block +
1799 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1800
1801 block_bitmap = ext4_block_bitmap(sb, gdp);
1802 if (block_bitmap < first_block || block_bitmap > last_block) {
1803 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1804 "Block bitmap for group %u not in group "
1805 "(block %llu)!", i, block_bitmap);
1806 return 0;
1807 }
1808 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1809 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1810 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1811 "Inode bitmap for group %u not in group "
1812 "(block %llu)!", i, inode_bitmap);
1813 return 0;
1814 }
1815 inode_table = ext4_inode_table(sb, gdp);
1816 if (inode_table < first_block ||
1817 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1818 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1819 "Inode table for group %u not in group "
1820 "(block %llu)!", i, inode_table);
1821 return 0;
1822 }
1823 ext4_lock_group(sb, i);
1824 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1825 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1826 "Checksum for group %u failed (%u!=%u)",
1827 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1828 gdp)), le16_to_cpu(gdp->bg_checksum));
1829 if (!(sb->s_flags & MS_RDONLY)) {
1830 ext4_unlock_group(sb, i);
1831 return 0;
1832 }
1833 }
1834 ext4_unlock_group(sb, i);
1835 if (!flexbg_flag)
1836 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1837 }
1838
1839 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1840 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
1841 return 1;
1842 }
1843
1844 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1845 * the superblock) which were deleted from all directories, but held open by
1846 * a process at the time of a crash. We walk the list and try to delete these
1847 * inodes at recovery time (only with a read-write filesystem).
1848 *
1849 * In order to keep the orphan inode chain consistent during traversal (in
1850 * case of crash during recovery), we link each inode into the superblock
1851 * orphan list_head and handle it the same way as an inode deletion during
1852 * normal operation (which journals the operations for us).
1853 *
1854 * We only do an iget() and an iput() on each inode, which is very safe if we
1855 * accidentally point at an in-use or already deleted inode. The worst that
1856 * can happen in this case is that we get a "bit already cleared" message from
1857 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1858 * e2fsck was run on this filesystem, and it must have already done the orphan
1859 * inode cleanup for us, so we can safely abort without any further action.
1860 */
1861 static void ext4_orphan_cleanup(struct super_block *sb,
1862 struct ext4_super_block *es)
1863 {
1864 unsigned int s_flags = sb->s_flags;
1865 int nr_orphans = 0, nr_truncates = 0;
1866 #ifdef CONFIG_QUOTA
1867 int i;
1868 #endif
1869 if (!es->s_last_orphan) {
1870 jbd_debug(4, "no orphan inodes to clean up\n");
1871 return;
1872 }
1873
1874 if (bdev_read_only(sb->s_bdev)) {
1875 ext4_msg(sb, KERN_ERR, "write access "
1876 "unavailable, skipping orphan cleanup");
1877 return;
1878 }
1879
1880 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1881 if (es->s_last_orphan)
1882 jbd_debug(1, "Errors on filesystem, "
1883 "clearing orphan list.\n");
1884 es->s_last_orphan = 0;
1885 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1886 return;
1887 }
1888
1889 if (s_flags & MS_RDONLY) {
1890 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1891 sb->s_flags &= ~MS_RDONLY;
1892 }
1893 #ifdef CONFIG_QUOTA
1894 /* Needed for iput() to work correctly and not trash data */
1895 sb->s_flags |= MS_ACTIVE;
1896 /* Turn on quotas so that they are updated correctly */
1897 for (i = 0; i < MAXQUOTAS; i++) {
1898 if (EXT4_SB(sb)->s_qf_names[i]) {
1899 int ret = ext4_quota_on_mount(sb, i);
1900 if (ret < 0)
1901 ext4_msg(sb, KERN_ERR,
1902 "Cannot turn on journaled "
1903 "quota: error %d", ret);
1904 }
1905 }
1906 #endif
1907
1908 while (es->s_last_orphan) {
1909 struct inode *inode;
1910
1911 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1912 if (IS_ERR(inode)) {
1913 es->s_last_orphan = 0;
1914 break;
1915 }
1916
1917 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1918 vfs_dq_init(inode);
1919 if (inode->i_nlink) {
1920 ext4_msg(sb, KERN_DEBUG,
1921 "%s: truncating inode %lu to %lld bytes",
1922 __func__, inode->i_ino, inode->i_size);
1923 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1924 inode->i_ino, inode->i_size);
1925 ext4_truncate(inode);
1926 nr_truncates++;
1927 } else {
1928 ext4_msg(sb, KERN_DEBUG,
1929 "%s: deleting unreferenced inode %lu",
1930 __func__, inode->i_ino);
1931 jbd_debug(2, "deleting unreferenced inode %lu\n",
1932 inode->i_ino);
1933 nr_orphans++;
1934 }
1935 iput(inode); /* The delete magic happens here! */
1936 }
1937
1938 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1939
1940 if (nr_orphans)
1941 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1942 PLURAL(nr_orphans));
1943 if (nr_truncates)
1944 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1945 PLURAL(nr_truncates));
1946 #ifdef CONFIG_QUOTA
1947 /* Turn quotas off */
1948 for (i = 0; i < MAXQUOTAS; i++) {
1949 if (sb_dqopt(sb)->files[i])
1950 vfs_quota_off(sb, i, 0);
1951 }
1952 #endif
1953 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1954 }
1955
1956 /*
1957 * Maximal extent format file size.
1958 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1959 * extent format containers, within a sector_t, and within i_blocks
1960 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1961 * so that won't be a limiting factor.
1962 *
1963 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1964 */
1965 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1966 {
1967 loff_t res;
1968 loff_t upper_limit = MAX_LFS_FILESIZE;
1969
1970 /* small i_blocks in vfs inode? */
1971 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1972 /*
1973 * CONFIG_LBDAF is not enabled implies the inode
1974 * i_block represent total blocks in 512 bytes
1975 * 32 == size of vfs inode i_blocks * 8
1976 */
1977 upper_limit = (1LL << 32) - 1;
1978
1979 /* total blocks in file system block size */
1980 upper_limit >>= (blkbits - 9);
1981 upper_limit <<= blkbits;
1982 }
1983
1984 /* 32-bit extent-start container, ee_block */
1985 res = 1LL << 32;
1986 res <<= blkbits;
1987 res -= 1;
1988
1989 /* Sanity check against vm- & vfs- imposed limits */
1990 if (res > upper_limit)
1991 res = upper_limit;
1992
1993 return res;
1994 }
1995
1996 /*
1997 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1998 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1999 * We need to be 1 filesystem block less than the 2^48 sector limit.
2000 */
2001 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2002 {
2003 loff_t res = EXT4_NDIR_BLOCKS;
2004 int meta_blocks;
2005 loff_t upper_limit;
2006 /* This is calculated to be the largest file size for a dense, block
2007 * mapped file such that the file's total number of 512-byte sectors,
2008 * including data and all indirect blocks, does not exceed (2^48 - 1).
2009 *
2010 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2011 * number of 512-byte sectors of the file.
2012 */
2013
2014 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2015 /*
2016 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2017 * the inode i_block field represents total file blocks in
2018 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2019 */
2020 upper_limit = (1LL << 32) - 1;
2021
2022 /* total blocks in file system block size */
2023 upper_limit >>= (bits - 9);
2024
2025 } else {
2026 /*
2027 * We use 48 bit ext4_inode i_blocks
2028 * With EXT4_HUGE_FILE_FL set the i_blocks
2029 * represent total number of blocks in
2030 * file system block size
2031 */
2032 upper_limit = (1LL << 48) - 1;
2033
2034 }
2035
2036 /* indirect blocks */
2037 meta_blocks = 1;
2038 /* double indirect blocks */
2039 meta_blocks += 1 + (1LL << (bits-2));
2040 /* tripple indirect blocks */
2041 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2042
2043 upper_limit -= meta_blocks;
2044 upper_limit <<= bits;
2045
2046 res += 1LL << (bits-2);
2047 res += 1LL << (2*(bits-2));
2048 res += 1LL << (3*(bits-2));
2049 res <<= bits;
2050 if (res > upper_limit)
2051 res = upper_limit;
2052
2053 if (res > MAX_LFS_FILESIZE)
2054 res = MAX_LFS_FILESIZE;
2055
2056 return res;
2057 }
2058
2059 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2060 ext4_fsblk_t logical_sb_block, int nr)
2061 {
2062 struct ext4_sb_info *sbi = EXT4_SB(sb);
2063 ext4_group_t bg, first_meta_bg;
2064 int has_super = 0;
2065
2066 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2067
2068 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2069 nr < first_meta_bg)
2070 return logical_sb_block + nr + 1;
2071 bg = sbi->s_desc_per_block * nr;
2072 if (ext4_bg_has_super(sb, bg))
2073 has_super = 1;
2074
2075 return (has_super + ext4_group_first_block_no(sb, bg));
2076 }
2077
2078 /**
2079 * ext4_get_stripe_size: Get the stripe size.
2080 * @sbi: In memory super block info
2081 *
2082 * If we have specified it via mount option, then
2083 * use the mount option value. If the value specified at mount time is
2084 * greater than the blocks per group use the super block value.
2085 * If the super block value is greater than blocks per group return 0.
2086 * Allocator needs it be less than blocks per group.
2087 *
2088 */
2089 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2090 {
2091 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2092 unsigned long stripe_width =
2093 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2094
2095 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2096 return sbi->s_stripe;
2097
2098 if (stripe_width <= sbi->s_blocks_per_group)
2099 return stripe_width;
2100
2101 if (stride <= sbi->s_blocks_per_group)
2102 return stride;
2103
2104 return 0;
2105 }
2106
2107 /* sysfs supprt */
2108
2109 struct ext4_attr {
2110 struct attribute attr;
2111 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2112 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2113 const char *, size_t);
2114 int offset;
2115 };
2116
2117 static int parse_strtoul(const char *buf,
2118 unsigned long max, unsigned long *value)
2119 {
2120 char *endp;
2121
2122 while (*buf && isspace(*buf))
2123 buf++;
2124 *value = simple_strtoul(buf, &endp, 0);
2125 while (*endp && isspace(*endp))
2126 endp++;
2127 if (*endp || *value > max)
2128 return -EINVAL;
2129
2130 return 0;
2131 }
2132
2133 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2134 struct ext4_sb_info *sbi,
2135 char *buf)
2136 {
2137 return snprintf(buf, PAGE_SIZE, "%llu\n",
2138 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2139 }
2140
2141 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2142 struct ext4_sb_info *sbi, char *buf)
2143 {
2144 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2145
2146 return snprintf(buf, PAGE_SIZE, "%lu\n",
2147 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2148 sbi->s_sectors_written_start) >> 1);
2149 }
2150
2151 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2152 struct ext4_sb_info *sbi, char *buf)
2153 {
2154 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2155
2156 return snprintf(buf, PAGE_SIZE, "%llu\n",
2157 sbi->s_kbytes_written +
2158 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2159 EXT4_SB(sb)->s_sectors_written_start) >> 1));
2160 }
2161
2162 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2163 struct ext4_sb_info *sbi,
2164 const char *buf, size_t count)
2165 {
2166 unsigned long t;
2167
2168 if (parse_strtoul(buf, 0x40000000, &t))
2169 return -EINVAL;
2170
2171 if (!is_power_of_2(t))
2172 return -EINVAL;
2173
2174 sbi->s_inode_readahead_blks = t;
2175 return count;
2176 }
2177
2178 static ssize_t sbi_ui_show(struct ext4_attr *a,
2179 struct ext4_sb_info *sbi, char *buf)
2180 {
2181 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2182
2183 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2184 }
2185
2186 static ssize_t sbi_ui_store(struct ext4_attr *a,
2187 struct ext4_sb_info *sbi,
2188 const char *buf, size_t count)
2189 {
2190 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2191 unsigned long t;
2192
2193 if (parse_strtoul(buf, 0xffffffff, &t))
2194 return -EINVAL;
2195 *ui = t;
2196 return count;
2197 }
2198
2199 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2200 static struct ext4_attr ext4_attr_##_name = { \
2201 .attr = {.name = __stringify(_name), .mode = _mode }, \
2202 .show = _show, \
2203 .store = _store, \
2204 .offset = offsetof(struct ext4_sb_info, _elname), \
2205 }
2206 #define EXT4_ATTR(name, mode, show, store) \
2207 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2208
2209 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2210 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2211 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2212 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2213 #define ATTR_LIST(name) &ext4_attr_##name.attr
2214
2215 EXT4_RO_ATTR(delayed_allocation_blocks);
2216 EXT4_RO_ATTR(session_write_kbytes);
2217 EXT4_RO_ATTR(lifetime_write_kbytes);
2218 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2219 inode_readahead_blks_store, s_inode_readahead_blks);
2220 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2221 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2222 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2223 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2224 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2225 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2226 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2227 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2228
2229 static struct attribute *ext4_attrs[] = {
2230 ATTR_LIST(delayed_allocation_blocks),
2231 ATTR_LIST(session_write_kbytes),
2232 ATTR_LIST(lifetime_write_kbytes),
2233 ATTR_LIST(inode_readahead_blks),
2234 ATTR_LIST(inode_goal),
2235 ATTR_LIST(mb_stats),
2236 ATTR_LIST(mb_max_to_scan),
2237 ATTR_LIST(mb_min_to_scan),
2238 ATTR_LIST(mb_order2_req),
2239 ATTR_LIST(mb_stream_req),
2240 ATTR_LIST(mb_group_prealloc),
2241 ATTR_LIST(max_writeback_mb_bump),
2242 NULL,
2243 };
2244
2245 static ssize_t ext4_attr_show(struct kobject *kobj,
2246 struct attribute *attr, char *buf)
2247 {
2248 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2249 s_kobj);
2250 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2251
2252 return a->show ? a->show(a, sbi, buf) : 0;
2253 }
2254
2255 static ssize_t ext4_attr_store(struct kobject *kobj,
2256 struct attribute *attr,
2257 const char *buf, size_t len)
2258 {
2259 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2260 s_kobj);
2261 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2262
2263 return a->store ? a->store(a, sbi, buf, len) : 0;
2264 }
2265
2266 static void ext4_sb_release(struct kobject *kobj)
2267 {
2268 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2269 s_kobj);
2270 complete(&sbi->s_kobj_unregister);
2271 }
2272
2273
2274 static struct sysfs_ops ext4_attr_ops = {
2275 .show = ext4_attr_show,
2276 .store = ext4_attr_store,
2277 };
2278
2279 static struct kobj_type ext4_ktype = {
2280 .default_attrs = ext4_attrs,
2281 .sysfs_ops = &ext4_attr_ops,
2282 .release = ext4_sb_release,
2283 };
2284
2285 /*
2286 * Check whether this filesystem can be mounted based on
2287 * the features present and the RDONLY/RDWR mount requested.
2288 * Returns 1 if this filesystem can be mounted as requested,
2289 * 0 if it cannot be.
2290 */
2291 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2292 {
2293 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2294 ext4_msg(sb, KERN_ERR,
2295 "Couldn't mount because of "
2296 "unsupported optional features (%x)",
2297 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2298 ~EXT4_FEATURE_INCOMPAT_SUPP));
2299 return 0;
2300 }
2301
2302 if (readonly)
2303 return 1;
2304
2305 /* Check that feature set is OK for a read-write mount */
2306 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2307 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2308 "unsupported optional features (%x)",
2309 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2310 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2311 return 0;
2312 }
2313 /*
2314 * Large file size enabled file system can only be mounted
2315 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2316 */
2317 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2318 if (sizeof(blkcnt_t) < sizeof(u64)) {
2319 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2320 "cannot be mounted RDWR without "
2321 "CONFIG_LBDAF");
2322 return 0;
2323 }
2324 }
2325 return 1;
2326 }
2327
2328 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2329 __releases(kernel_lock)
2330 __acquires(kernel_lock)
2331 {
2332 struct buffer_head *bh;
2333 struct ext4_super_block *es = NULL;
2334 struct ext4_sb_info *sbi;
2335 ext4_fsblk_t block;
2336 ext4_fsblk_t sb_block = get_sb_block(&data);
2337 ext4_fsblk_t logical_sb_block;
2338 unsigned long offset = 0;
2339 unsigned long journal_devnum = 0;
2340 unsigned long def_mount_opts;
2341 struct inode *root;
2342 char *cp;
2343 const char *descr;
2344 int ret = -EINVAL;
2345 int blocksize;
2346 unsigned int db_count;
2347 unsigned int i;
2348 int needs_recovery, has_huge_files;
2349 __u64 blocks_count;
2350 int err;
2351 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2352
2353 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2354 if (!sbi)
2355 return -ENOMEM;
2356
2357 sbi->s_blockgroup_lock =
2358 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2359 if (!sbi->s_blockgroup_lock) {
2360 kfree(sbi);
2361 return -ENOMEM;
2362 }
2363 sb->s_fs_info = sbi;
2364 sbi->s_mount_opt = 0;
2365 sbi->s_resuid = EXT4_DEF_RESUID;
2366 sbi->s_resgid = EXT4_DEF_RESGID;
2367 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2368 sbi->s_sb_block = sb_block;
2369 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2370 sectors[1]);
2371
2372 unlock_kernel();
2373
2374 /* Cleanup superblock name */
2375 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2376 *cp = '!';
2377
2378 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2379 if (!blocksize) {
2380 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
2381 goto out_fail;
2382 }
2383
2384 /*
2385 * The ext4 superblock will not be buffer aligned for other than 1kB
2386 * block sizes. We need to calculate the offset from buffer start.
2387 */
2388 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2389 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2390 offset = do_div(logical_sb_block, blocksize);
2391 } else {
2392 logical_sb_block = sb_block;
2393 }
2394
2395 if (!(bh = sb_bread(sb, logical_sb_block))) {
2396 ext4_msg(sb, KERN_ERR, "unable to read superblock");
2397 goto out_fail;
2398 }
2399 /*
2400 * Note: s_es must be initialized as soon as possible because
2401 * some ext4 macro-instructions depend on its value
2402 */
2403 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2404 sbi->s_es = es;
2405 sb->s_magic = le16_to_cpu(es->s_magic);
2406 if (sb->s_magic != EXT4_SUPER_MAGIC)
2407 goto cantfind_ext4;
2408 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2409
2410 /* Set defaults before we parse the mount options */
2411 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2412 if (def_mount_opts & EXT4_DEFM_DEBUG)
2413 set_opt(sbi->s_mount_opt, DEBUG);
2414 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
2415 set_opt(sbi->s_mount_opt, GRPID);
2416 if (def_mount_opts & EXT4_DEFM_UID16)
2417 set_opt(sbi->s_mount_opt, NO_UID32);
2418 #ifdef CONFIG_EXT4_FS_XATTR
2419 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2420 set_opt(sbi->s_mount_opt, XATTR_USER);
2421 #endif
2422 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2423 if (def_mount_opts & EXT4_DEFM_ACL)
2424 set_opt(sbi->s_mount_opt, POSIX_ACL);
2425 #endif
2426 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2427 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2428 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2429 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2430 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2431 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2432
2433 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2434 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2435 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2436 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2437 else
2438 set_opt(sbi->s_mount_opt, ERRORS_RO);
2439
2440 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2441 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2442 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2443 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2444 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2445
2446 set_opt(sbi->s_mount_opt, BARRIER);
2447
2448 /*
2449 * enable delayed allocation by default
2450 * Use -o nodelalloc to turn it off
2451 */
2452 set_opt(sbi->s_mount_opt, DELALLOC);
2453
2454 if (!parse_options((char *) data, sb, &journal_devnum,
2455 &journal_ioprio, NULL, 0))
2456 goto failed_mount;
2457
2458 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2459 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2460
2461 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2462 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2463 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2464 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2465 ext4_msg(sb, KERN_WARNING,
2466 "feature flags set on rev 0 fs, "
2467 "running e2fsck is recommended");
2468
2469 /*
2470 * Check feature flags regardless of the revision level, since we
2471 * previously didn't change the revision level when setting the flags,
2472 * so there is a chance incompat flags are set on a rev 0 filesystem.
2473 */
2474 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
2475 goto failed_mount;
2476
2477 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2478
2479 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2480 blocksize > EXT4_MAX_BLOCK_SIZE) {
2481 ext4_msg(sb, KERN_ERR,
2482 "Unsupported filesystem blocksize %d", blocksize);
2483 goto failed_mount;
2484 }
2485
2486 if (sb->s_blocksize != blocksize) {
2487 /* Validate the filesystem blocksize */
2488 if (!sb_set_blocksize(sb, blocksize)) {
2489 ext4_msg(sb, KERN_ERR, "bad block size %d",
2490 blocksize);
2491 goto failed_mount;
2492 }
2493
2494 brelse(bh);
2495 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2496 offset = do_div(logical_sb_block, blocksize);
2497 bh = sb_bread(sb, logical_sb_block);
2498 if (!bh) {
2499 ext4_msg(sb, KERN_ERR,
2500 "Can't read superblock on 2nd try");
2501 goto failed_mount;
2502 }
2503 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2504 sbi->s_es = es;
2505 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2506 ext4_msg(sb, KERN_ERR,
2507 "Magic mismatch, very weird!");
2508 goto failed_mount;
2509 }
2510 }
2511
2512 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2513 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2514 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2515 has_huge_files);
2516 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2517
2518 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2519 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2520 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2521 } else {
2522 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2523 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2524 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2525 (!is_power_of_2(sbi->s_inode_size)) ||
2526 (sbi->s_inode_size > blocksize)) {
2527 ext4_msg(sb, KERN_ERR,
2528 "unsupported inode size: %d",
2529 sbi->s_inode_size);
2530 goto failed_mount;
2531 }
2532 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2533 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2534 }
2535
2536 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2537 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2538 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2539 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2540 !is_power_of_2(sbi->s_desc_size)) {
2541 ext4_msg(sb, KERN_ERR,
2542 "unsupported descriptor size %lu",
2543 sbi->s_desc_size);
2544 goto failed_mount;
2545 }
2546 } else
2547 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2548
2549 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2550 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2551 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2552 goto cantfind_ext4;
2553
2554 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2555 if (sbi->s_inodes_per_block == 0)
2556 goto cantfind_ext4;
2557 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2558 sbi->s_inodes_per_block;
2559 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2560 sbi->s_sbh = bh;
2561 sbi->s_mount_state = le16_to_cpu(es->s_state);
2562 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2563 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2564
2565 for (i = 0; i < 4; i++)
2566 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2567 sbi->s_def_hash_version = es->s_def_hash_version;
2568 i = le32_to_cpu(es->s_flags);
2569 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2570 sbi->s_hash_unsigned = 3;
2571 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2572 #ifdef __CHAR_UNSIGNED__
2573 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2574 sbi->s_hash_unsigned = 3;
2575 #else
2576 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2577 #endif
2578 sb->s_dirt = 1;
2579 }
2580
2581 if (sbi->s_blocks_per_group > blocksize * 8) {
2582 ext4_msg(sb, KERN_ERR,
2583 "#blocks per group too big: %lu",
2584 sbi->s_blocks_per_group);
2585 goto failed_mount;
2586 }
2587 if (sbi->s_inodes_per_group > blocksize * 8) {
2588 ext4_msg(sb, KERN_ERR,
2589 "#inodes per group too big: %lu",
2590 sbi->s_inodes_per_group);
2591 goto failed_mount;
2592 }
2593
2594 /*
2595 * Test whether we have more sectors than will fit in sector_t,
2596 * and whether the max offset is addressable by the page cache.
2597 */
2598 if ((ext4_blocks_count(es) >
2599 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2600 (ext4_blocks_count(es) >
2601 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
2602 ext4_msg(sb, KERN_ERR, "filesystem"
2603 " too large to mount safely on this system");
2604 if (sizeof(sector_t) < 8)
2605 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
2606 ret = -EFBIG;
2607 goto failed_mount;
2608 }
2609
2610 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2611 goto cantfind_ext4;
2612
2613 /* check blocks count against device size */
2614 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2615 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2616 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2617 "exceeds size of device (%llu blocks)",
2618 ext4_blocks_count(es), blocks_count);
2619 goto failed_mount;
2620 }
2621
2622 /*
2623 * It makes no sense for the first data block to be beyond the end
2624 * of the filesystem.
2625 */
2626 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2627 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2628 "block %u is beyond end of filesystem (%llu)",
2629 le32_to_cpu(es->s_first_data_block),
2630 ext4_blocks_count(es));
2631 goto failed_mount;
2632 }
2633 blocks_count = (ext4_blocks_count(es) -
2634 le32_to_cpu(es->s_first_data_block) +
2635 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2636 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2637 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2638 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
2639 "(block count %llu, first data block %u, "
2640 "blocks per group %lu)", sbi->s_groups_count,
2641 ext4_blocks_count(es),
2642 le32_to_cpu(es->s_first_data_block),
2643 EXT4_BLOCKS_PER_GROUP(sb));
2644 goto failed_mount;
2645 }
2646 sbi->s_groups_count = blocks_count;
2647 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
2648 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
2649 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2650 EXT4_DESC_PER_BLOCK(sb);
2651 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2652 GFP_KERNEL);
2653 if (sbi->s_group_desc == NULL) {
2654 ext4_msg(sb, KERN_ERR, "not enough memory");
2655 goto failed_mount;
2656 }
2657
2658 #ifdef CONFIG_PROC_FS
2659 if (ext4_proc_root)
2660 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2661 #endif
2662
2663 bgl_lock_init(sbi->s_blockgroup_lock);
2664
2665 for (i = 0; i < db_count; i++) {
2666 block = descriptor_loc(sb, logical_sb_block, i);
2667 sbi->s_group_desc[i] = sb_bread(sb, block);
2668 if (!sbi->s_group_desc[i]) {
2669 ext4_msg(sb, KERN_ERR,
2670 "can't read group descriptor %d", i);
2671 db_count = i;
2672 goto failed_mount2;
2673 }
2674 }
2675 if (!ext4_check_descriptors(sb)) {
2676 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
2677 goto failed_mount2;
2678 }
2679 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2680 if (!ext4_fill_flex_info(sb)) {
2681 ext4_msg(sb, KERN_ERR,
2682 "unable to initialize "
2683 "flex_bg meta info!");
2684 goto failed_mount2;
2685 }
2686
2687 sbi->s_gdb_count = db_count;
2688 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2689 spin_lock_init(&sbi->s_next_gen_lock);
2690
2691 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2692 ext4_count_free_blocks(sb));
2693 if (!err) {
2694 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2695 ext4_count_free_inodes(sb));
2696 }
2697 if (!err) {
2698 err = percpu_counter_init(&sbi->s_dirs_counter,
2699 ext4_count_dirs(sb));
2700 }
2701 if (!err) {
2702 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2703 }
2704 if (err) {
2705 ext4_msg(sb, KERN_ERR, "insufficient memory");
2706 goto failed_mount3;
2707 }
2708
2709 sbi->s_stripe = ext4_get_stripe_size(sbi);
2710 sbi->s_max_writeback_mb_bump = 128;
2711
2712 /*
2713 * set up enough so that it can read an inode
2714 */
2715 if (!test_opt(sb, NOLOAD) &&
2716 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2717 sb->s_op = &ext4_sops;
2718 else
2719 sb->s_op = &ext4_nojournal_sops;
2720 sb->s_export_op = &ext4_export_ops;
2721 sb->s_xattr = ext4_xattr_handlers;
2722 #ifdef CONFIG_QUOTA
2723 sb->s_qcop = &ext4_qctl_operations;
2724 sb->dq_op = &ext4_quota_operations;
2725 #endif
2726 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2727 mutex_init(&sbi->s_orphan_lock);
2728 mutex_init(&sbi->s_resize_lock);
2729
2730 sb->s_root = NULL;
2731
2732 needs_recovery = (es->s_last_orphan != 0 ||
2733 EXT4_HAS_INCOMPAT_FEATURE(sb,
2734 EXT4_FEATURE_INCOMPAT_RECOVER));
2735
2736 /*
2737 * The first inode we look at is the journal inode. Don't try
2738 * root first: it may be modified in the journal!
2739 */
2740 if (!test_opt(sb, NOLOAD) &&
2741 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2742 if (ext4_load_journal(sb, es, journal_devnum))
2743 goto failed_mount3;
2744 if (!(sb->s_flags & MS_RDONLY) &&
2745 EXT4_SB(sb)->s_journal->j_failed_commit) {
2746 ext4_msg(sb, KERN_CRIT, "error: "
2747 "ext4_fill_super: Journal transaction "
2748 "%u is corrupt",
2749 EXT4_SB(sb)->s_journal->j_failed_commit);
2750 if (test_opt(sb, ERRORS_RO)) {
2751 ext4_msg(sb, KERN_CRIT,
2752 "Mounting filesystem read-only");
2753 sb->s_flags |= MS_RDONLY;
2754 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2755 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2756 }
2757 if (test_opt(sb, ERRORS_PANIC)) {
2758 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2759 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2760 ext4_commit_super(sb, 1);
2761 goto failed_mount4;
2762 }
2763 }
2764 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2765 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2766 ext4_msg(sb, KERN_ERR, "required journal recovery "
2767 "suppressed and not mounted read-only");
2768 goto failed_mount4;
2769 } else {
2770 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2771 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2772 sbi->s_journal = NULL;
2773 needs_recovery = 0;
2774 goto no_journal;
2775 }
2776
2777 if (ext4_blocks_count(es) > 0xffffffffULL &&
2778 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2779 JBD2_FEATURE_INCOMPAT_64BIT)) {
2780 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
2781 goto failed_mount4;
2782 }
2783
2784 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2785 jbd2_journal_set_features(sbi->s_journal,
2786 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2787 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2788 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2789 jbd2_journal_set_features(sbi->s_journal,
2790 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2791 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2792 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2793 } else {
2794 jbd2_journal_clear_features(sbi->s_journal,
2795 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2796 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2797 }
2798
2799 /* We have now updated the journal if required, so we can
2800 * validate the data journaling mode. */
2801 switch (test_opt(sb, DATA_FLAGS)) {
2802 case 0:
2803 /* No mode set, assume a default based on the journal
2804 * capabilities: ORDERED_DATA if the journal can
2805 * cope, else JOURNAL_DATA
2806 */
2807 if (jbd2_journal_check_available_features
2808 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2809 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2810 else
2811 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2812 break;
2813
2814 case EXT4_MOUNT_ORDERED_DATA:
2815 case EXT4_MOUNT_WRITEBACK_DATA:
2816 if (!jbd2_journal_check_available_features
2817 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2818 ext4_msg(sb, KERN_ERR, "Journal does not support "
2819 "requested data journaling mode");
2820 goto failed_mount4;
2821 }
2822 default:
2823 break;
2824 }
2825 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2826
2827 no_journal:
2828
2829 if (test_opt(sb, NOBH)) {
2830 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2831 ext4_msg(sb, KERN_WARNING, "Ignoring nobh option - "
2832 "its supported only with writeback mode");
2833 clear_opt(sbi->s_mount_opt, NOBH);
2834 }
2835 }
2836 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
2837 if (!EXT4_SB(sb)->dio_unwritten_wq) {
2838 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
2839 goto failed_mount_wq;
2840 }
2841
2842 /*
2843 * The jbd2_journal_load will have done any necessary log recovery,
2844 * so we can safely mount the rest of the filesystem now.
2845 */
2846
2847 root = ext4_iget(sb, EXT4_ROOT_INO);
2848 if (IS_ERR(root)) {
2849 ext4_msg(sb, KERN_ERR, "get root inode failed");
2850 ret = PTR_ERR(root);
2851 goto failed_mount4;
2852 }
2853 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2854 iput(root);
2855 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
2856 goto failed_mount4;
2857 }
2858 sb->s_root = d_alloc_root(root);
2859 if (!sb->s_root) {
2860 ext4_msg(sb, KERN_ERR, "get root dentry failed");
2861 iput(root);
2862 ret = -ENOMEM;
2863 goto failed_mount4;
2864 }
2865
2866 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2867
2868 /* determine the minimum size of new large inodes, if present */
2869 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2870 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2871 EXT4_GOOD_OLD_INODE_SIZE;
2872 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2873 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2874 if (sbi->s_want_extra_isize <
2875 le16_to_cpu(es->s_want_extra_isize))
2876 sbi->s_want_extra_isize =
2877 le16_to_cpu(es->s_want_extra_isize);
2878 if (sbi->s_want_extra_isize <
2879 le16_to_cpu(es->s_min_extra_isize))
2880 sbi->s_want_extra_isize =
2881 le16_to_cpu(es->s_min_extra_isize);
2882 }
2883 }
2884 /* Check if enough inode space is available */
2885 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2886 sbi->s_inode_size) {
2887 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2888 EXT4_GOOD_OLD_INODE_SIZE;
2889 ext4_msg(sb, KERN_INFO, "required extra inode space not"
2890 "available");
2891 }
2892
2893 if (test_opt(sb, DELALLOC) &&
2894 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
2895 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
2896 "requested data journaling mode");
2897 clear_opt(sbi->s_mount_opt, DELALLOC);
2898 }
2899
2900 err = ext4_setup_system_zone(sb);
2901 if (err) {
2902 ext4_msg(sb, KERN_ERR, "failed to initialize system "
2903 "zone (%d)\n", err);
2904 goto failed_mount4;
2905 }
2906
2907 ext4_ext_init(sb);
2908 err = ext4_mb_init(sb, needs_recovery);
2909 if (err) {
2910 ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
2911 err);
2912 goto failed_mount4;
2913 }
2914
2915 sbi->s_kobj.kset = ext4_kset;
2916 init_completion(&sbi->s_kobj_unregister);
2917 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2918 "%s", sb->s_id);
2919 if (err) {
2920 ext4_mb_release(sb);
2921 ext4_ext_release(sb);
2922 goto failed_mount4;
2923 };
2924
2925 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2926 ext4_orphan_cleanup(sb, es);
2927 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2928 if (needs_recovery) {
2929 ext4_msg(sb, KERN_INFO, "recovery complete");
2930 ext4_mark_recovery_complete(sb, es);
2931 }
2932 if (EXT4_SB(sb)->s_journal) {
2933 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2934 descr = " journalled data mode";
2935 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2936 descr = " ordered data mode";
2937 else
2938 descr = " writeback data mode";
2939 } else
2940 descr = "out journal";
2941
2942 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s", descr);
2943
2944 lock_kernel();
2945 return 0;
2946
2947 cantfind_ext4:
2948 if (!silent)
2949 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
2950 goto failed_mount;
2951
2952 failed_mount4:
2953 ext4_msg(sb, KERN_ERR, "mount failed");
2954 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
2955 failed_mount_wq:
2956 ext4_release_system_zone(sb);
2957 if (sbi->s_journal) {
2958 jbd2_journal_destroy(sbi->s_journal);
2959 sbi->s_journal = NULL;
2960 }
2961 failed_mount3:
2962 if (sbi->s_flex_groups) {
2963 if (is_vmalloc_addr(sbi->s_flex_groups))
2964 vfree(sbi->s_flex_groups);
2965 else
2966 kfree(sbi->s_flex_groups);
2967 }
2968 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2969 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2970 percpu_counter_destroy(&sbi->s_dirs_counter);
2971 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2972 failed_mount2:
2973 for (i = 0; i < db_count; i++)
2974 brelse(sbi->s_group_desc[i]);
2975 kfree(sbi->s_group_desc);
2976 failed_mount:
2977 if (sbi->s_proc) {
2978 remove_proc_entry(sb->s_id, ext4_proc_root);
2979 }
2980 #ifdef CONFIG_QUOTA
2981 for (i = 0; i < MAXQUOTAS; i++)
2982 kfree(sbi->s_qf_names[i]);
2983 #endif
2984 ext4_blkdev_remove(sbi);
2985 brelse(bh);
2986 out_fail:
2987 sb->s_fs_info = NULL;
2988 kfree(sbi->s_blockgroup_lock);
2989 kfree(sbi);
2990 lock_kernel();
2991 return ret;
2992 }
2993
2994 /*
2995 * Setup any per-fs journal parameters now. We'll do this both on
2996 * initial mount, once the journal has been initialised but before we've
2997 * done any recovery; and again on any subsequent remount.
2998 */
2999 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3000 {
3001 struct ext4_sb_info *sbi = EXT4_SB(sb);
3002
3003 journal->j_commit_interval = sbi->s_commit_interval;
3004 journal->j_min_batch_time = sbi->s_min_batch_time;
3005 journal->j_max_batch_time = sbi->s_max_batch_time;
3006
3007 spin_lock(&journal->j_state_lock);
3008 if (test_opt(sb, BARRIER))
3009 journal->j_flags |= JBD2_BARRIER;
3010 else
3011 journal->j_flags &= ~JBD2_BARRIER;
3012 if (test_opt(sb, DATA_ERR_ABORT))
3013 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3014 else
3015 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3016 spin_unlock(&journal->j_state_lock);
3017 }
3018
3019 static journal_t *ext4_get_journal(struct super_block *sb,
3020 unsigned int journal_inum)
3021 {
3022 struct inode *journal_inode;
3023 journal_t *journal;
3024
3025 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3026
3027 /* First, test for the existence of a valid inode on disk. Bad
3028 * things happen if we iget() an unused inode, as the subsequent
3029 * iput() will try to delete it. */
3030
3031 journal_inode = ext4_iget(sb, journal_inum);
3032 if (IS_ERR(journal_inode)) {
3033 ext4_msg(sb, KERN_ERR, "no journal found");
3034 return NULL;
3035 }
3036 if (!journal_inode->i_nlink) {
3037 make_bad_inode(journal_inode);
3038 iput(journal_inode);
3039 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3040 return NULL;
3041 }
3042
3043 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3044 journal_inode, journal_inode->i_size);
3045 if (!S_ISREG(journal_inode->i_mode)) {
3046 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3047 iput(journal_inode);
3048 return NULL;
3049 }
3050
3051 journal = jbd2_journal_init_inode(journal_inode);
3052 if (!journal) {
3053 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3054 iput(journal_inode);
3055 return NULL;
3056 }
3057 journal->j_private = sb;
3058 ext4_init_journal_params(sb, journal);
3059 return journal;
3060 }
3061
3062 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3063 dev_t j_dev)
3064 {
3065 struct buffer_head *bh;
3066 journal_t *journal;
3067 ext4_fsblk_t start;
3068 ext4_fsblk_t len;
3069 int hblock, blocksize;
3070 ext4_fsblk_t sb_block;
3071 unsigned long offset;
3072 struct ext4_super_block *es;
3073 struct block_device *bdev;
3074
3075 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3076
3077 bdev = ext4_blkdev_get(j_dev, sb);
3078 if (bdev == NULL)
3079 return NULL;
3080
3081 if (bd_claim(bdev, sb)) {
3082 ext4_msg(sb, KERN_ERR,
3083 "failed to claim external journal device");
3084 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3085 return NULL;
3086 }
3087
3088 blocksize = sb->s_blocksize;
3089 hblock = bdev_logical_block_size(bdev);
3090 if (blocksize < hblock) {
3091 ext4_msg(sb, KERN_ERR,
3092 "blocksize too small for journal device");
3093 goto out_bdev;
3094 }
3095
3096 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3097 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3098 set_blocksize(bdev, blocksize);
3099 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3100 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3101 "external journal");
3102 goto out_bdev;
3103 }
3104
3105 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3106 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3107 !(le32_to_cpu(es->s_feature_incompat) &
3108 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3109 ext4_msg(sb, KERN_ERR, "external journal has "
3110 "bad superblock");
3111 brelse(bh);
3112 goto out_bdev;
3113 }
3114
3115 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3116 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3117 brelse(bh);
3118 goto out_bdev;
3119 }
3120
3121 len = ext4_blocks_count(es);
3122 start = sb_block + 1;
3123 brelse(bh); /* we're done with the superblock */
3124
3125 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3126 start, len, blocksize);
3127 if (!journal) {
3128 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3129 goto out_bdev;
3130 }
3131 journal->j_private = sb;
3132 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3133 wait_on_buffer(journal->j_sb_buffer);
3134 if (!buffer_uptodate(journal->j_sb_buffer)) {
3135 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3136 goto out_journal;
3137 }
3138 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3139 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3140 "user (unsupported) - %d",
3141 be32_to_cpu(journal->j_superblock->s_nr_users));
3142 goto out_journal;
3143 }
3144 EXT4_SB(sb)->journal_bdev = bdev;
3145 ext4_init_journal_params(sb, journal);
3146 return journal;
3147
3148 out_journal:
3149 jbd2_journal_destroy(journal);
3150 out_bdev:
3151 ext4_blkdev_put(bdev);
3152 return NULL;
3153 }
3154
3155 static int ext4_load_journal(struct super_block *sb,
3156 struct ext4_super_block *es,
3157 unsigned long journal_devnum)
3158 {
3159 journal_t *journal;
3160 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3161 dev_t journal_dev;
3162 int err = 0;
3163 int really_read_only;
3164
3165 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3166
3167 if (journal_devnum &&
3168 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3169 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3170 "numbers have changed");
3171 journal_dev = new_decode_dev(journal_devnum);
3172 } else
3173 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3174
3175 really_read_only = bdev_read_only(sb->s_bdev);
3176
3177 /*
3178 * Are we loading a blank journal or performing recovery after a
3179 * crash? For recovery, we need to check in advance whether we
3180 * can get read-write access to the device.
3181 */
3182 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3183 if (sb->s_flags & MS_RDONLY) {
3184 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3185 "required on readonly filesystem");
3186 if (really_read_only) {
3187 ext4_msg(sb, KERN_ERR, "write access "
3188 "unavailable, cannot proceed");
3189 return -EROFS;
3190 }
3191 ext4_msg(sb, KERN_INFO, "write access will "
3192 "be enabled during recovery");
3193 }
3194 }
3195
3196 if (journal_inum && journal_dev) {
3197 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3198 "and inode journals!");
3199 return -EINVAL;
3200 }
3201
3202 if (journal_inum) {
3203 if (!(journal = ext4_get_journal(sb, journal_inum)))
3204 return -EINVAL;
3205 } else {
3206 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3207 return -EINVAL;
3208 }
3209
3210 if (!(journal->j_flags & JBD2_BARRIER))
3211 ext4_msg(sb, KERN_INFO, "barriers disabled");
3212
3213 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3214 err = jbd2_journal_update_format(journal);
3215 if (err) {
3216 ext4_msg(sb, KERN_ERR, "error updating journal");
3217 jbd2_journal_destroy(journal);
3218 return err;
3219 }
3220 }
3221
3222 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3223 err = jbd2_journal_wipe(journal, !really_read_only);
3224 if (!err)
3225 err = jbd2_journal_load(journal);
3226
3227 if (err) {
3228 ext4_msg(sb, KERN_ERR, "error loading journal");
3229 jbd2_journal_destroy(journal);
3230 return err;
3231 }
3232
3233 EXT4_SB(sb)->s_journal = journal;
3234 ext4_clear_journal_err(sb, es);
3235
3236 if (journal_devnum &&
3237 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3238 es->s_journal_dev = cpu_to_le32(journal_devnum);
3239
3240 /* Make sure we flush the recovery flag to disk. */
3241 ext4_commit_super(sb, 1);
3242 }
3243
3244 return 0;
3245 }
3246
3247 static int ext4_commit_super(struct super_block *sb, int sync)
3248 {
3249 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3250 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3251 int error = 0;
3252
3253 if (!sbh)
3254 return error;
3255 if (buffer_write_io_error(sbh)) {
3256 /*
3257 * Oh, dear. A previous attempt to write the
3258 * superblock failed. This could happen because the
3259 * USB device was yanked out. Or it could happen to
3260 * be a transient write error and maybe the block will
3261 * be remapped. Nothing we can do but to retry the
3262 * write and hope for the best.
3263 */
3264 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3265 "superblock detected");
3266 clear_buffer_write_io_error(sbh);
3267 set_buffer_uptodate(sbh);
3268 }
3269 /*
3270 * If the file system is mounted read-only, don't update the
3271 * superblock write time. This avoids updating the superblock
3272 * write time when we are mounting the root file system
3273 * read/only but we need to replay the journal; at that point,
3274 * for people who are east of GMT and who make their clock
3275 * tick in localtime for Windows bug-for-bug compatibility,
3276 * the clock is set in the future, and this will cause e2fsck
3277 * to complain and force a full file system check.
3278 */
3279 if (!(sb->s_flags & MS_RDONLY))
3280 es->s_wtime = cpu_to_le32(get_seconds());
3281 es->s_kbytes_written =
3282 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3283 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3284 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3285 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3286 &EXT4_SB(sb)->s_freeblocks_counter));
3287 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3288 &EXT4_SB(sb)->s_freeinodes_counter));
3289 sb->s_dirt = 0;
3290 BUFFER_TRACE(sbh, "marking dirty");
3291 mark_buffer_dirty(sbh);
3292 if (sync) {
3293 error = sync_dirty_buffer(sbh);
3294 if (error)
3295 return error;
3296
3297 error = buffer_write_io_error(sbh);
3298 if (error) {
3299 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3300 "superblock");
3301 clear_buffer_write_io_error(sbh);
3302 set_buffer_uptodate(sbh);
3303 }
3304 }
3305 return error;
3306 }
3307
3308 /*
3309 * Have we just finished recovery? If so, and if we are mounting (or
3310 * remounting) the filesystem readonly, then we will end up with a
3311 * consistent fs on disk. Record that fact.
3312 */
3313 static void ext4_mark_recovery_complete(struct super_block *sb,
3314 struct ext4_super_block *es)
3315 {
3316 journal_t *journal = EXT4_SB(sb)->s_journal;
3317
3318 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3319 BUG_ON(journal != NULL);
3320 return;
3321 }
3322 jbd2_journal_lock_updates(journal);
3323 if (jbd2_journal_flush(journal) < 0)
3324 goto out;
3325
3326 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3327 sb->s_flags & MS_RDONLY) {
3328 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3329 ext4_commit_super(sb, 1);
3330 }
3331
3332 out:
3333 jbd2_journal_unlock_updates(journal);
3334 }
3335
3336 /*
3337 * If we are mounting (or read-write remounting) a filesystem whose journal
3338 * has recorded an error from a previous lifetime, move that error to the
3339 * main filesystem now.
3340 */
3341 static void ext4_clear_journal_err(struct super_block *sb,
3342 struct ext4_super_block *es)
3343 {
3344 journal_t *journal;
3345 int j_errno;
3346 const char *errstr;
3347
3348 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3349
3350 journal = EXT4_SB(sb)->s_journal;
3351
3352 /*
3353 * Now check for any error status which may have been recorded in the
3354 * journal by a prior ext4_error() or ext4_abort()
3355 */
3356
3357 j_errno = jbd2_journal_errno(journal);
3358 if (j_errno) {
3359 char nbuf[16];
3360
3361 errstr = ext4_decode_error(sb, j_errno, nbuf);
3362 ext4_warning(sb, __func__, "Filesystem error recorded "
3363 "from previous mount: %s", errstr);
3364 ext4_warning(sb, __func__, "Marking fs in need of "
3365 "filesystem check.");
3366
3367 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3368 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3369 ext4_commit_super(sb, 1);
3370
3371 jbd2_journal_clear_err(journal);
3372 }
3373 }
3374
3375 /*
3376 * Force the running and committing transactions to commit,
3377 * and wait on the commit.
3378 */
3379 int ext4_force_commit(struct super_block *sb)
3380 {
3381 journal_t *journal;
3382 int ret = 0;
3383
3384 if (sb->s_flags & MS_RDONLY)
3385 return 0;
3386
3387 journal = EXT4_SB(sb)->s_journal;
3388 if (journal)
3389 ret = ext4_journal_force_commit(journal);
3390
3391 return ret;
3392 }
3393
3394 static void ext4_write_super(struct super_block *sb)
3395 {
3396 lock_super(sb);
3397 ext4_commit_super(sb, 1);
3398 unlock_super(sb);
3399 }
3400
3401 static int ext4_sync_fs(struct super_block *sb, int wait)
3402 {
3403 int ret = 0;
3404 tid_t target;
3405 struct ext4_sb_info *sbi = EXT4_SB(sb);
3406
3407 trace_ext4_sync_fs(sb, wait);
3408 flush_workqueue(sbi->dio_unwritten_wq);
3409 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
3410 if (wait)
3411 jbd2_log_wait_commit(sbi->s_journal, target);
3412 }
3413 return ret;
3414 }
3415
3416 /*
3417 * LVM calls this function before a (read-only) snapshot is created. This
3418 * gives us a chance to flush the journal completely and mark the fs clean.
3419 */
3420 static int ext4_freeze(struct super_block *sb)
3421 {
3422 int error = 0;
3423 journal_t *journal;
3424
3425 if (sb->s_flags & MS_RDONLY)
3426 return 0;
3427
3428 journal = EXT4_SB(sb)->s_journal;
3429
3430 /* Now we set up the journal barrier. */
3431 jbd2_journal_lock_updates(journal);
3432
3433 /*
3434 * Don't clear the needs_recovery flag if we failed to flush
3435 * the journal.
3436 */
3437 error = jbd2_journal_flush(journal);
3438 if (error < 0) {
3439 out:
3440 jbd2_journal_unlock_updates(journal);
3441 return error;
3442 }
3443
3444 /* Journal blocked and flushed, clear needs_recovery flag. */
3445 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3446 error = ext4_commit_super(sb, 1);
3447 if (error)
3448 goto out;
3449 return 0;
3450 }
3451
3452 /*
3453 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3454 * flag here, even though the filesystem is not technically dirty yet.
3455 */
3456 static int ext4_unfreeze(struct super_block *sb)
3457 {
3458 if (sb->s_flags & MS_RDONLY)
3459 return 0;
3460
3461 lock_super(sb);
3462 /* Reset the needs_recovery flag before the fs is unlocked. */
3463 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3464 ext4_commit_super(sb, 1);
3465 unlock_super(sb);
3466 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3467 return 0;
3468 }
3469
3470 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3471 {
3472 struct ext4_super_block *es;
3473 struct ext4_sb_info *sbi = EXT4_SB(sb);
3474 ext4_fsblk_t n_blocks_count = 0;
3475 unsigned long old_sb_flags;
3476 struct ext4_mount_options old_opts;
3477 ext4_group_t g;
3478 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3479 int err;
3480 #ifdef CONFIG_QUOTA
3481 int i;
3482 #endif
3483
3484 lock_kernel();
3485
3486 /* Store the original options */
3487 lock_super(sb);
3488 old_sb_flags = sb->s_flags;
3489 old_opts.s_mount_opt = sbi->s_mount_opt;
3490 old_opts.s_resuid = sbi->s_resuid;
3491 old_opts.s_resgid = sbi->s_resgid;
3492 old_opts.s_commit_interval = sbi->s_commit_interval;
3493 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3494 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3495 #ifdef CONFIG_QUOTA
3496 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3497 for (i = 0; i < MAXQUOTAS; i++)
3498 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3499 #endif
3500 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3501 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3502
3503 /*
3504 * Allow the "check" option to be passed as a remount option.
3505 */
3506 if (!parse_options(data, sb, NULL, &journal_ioprio,
3507 &n_blocks_count, 1)) {
3508 err = -EINVAL;
3509 goto restore_opts;
3510 }
3511
3512 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
3513 ext4_abort(sb, __func__, "Abort forced by user");
3514
3515 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3516 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3517
3518 es = sbi->s_es;
3519
3520 if (sbi->s_journal) {
3521 ext4_init_journal_params(sb, sbi->s_journal);
3522 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3523 }
3524
3525 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3526 n_blocks_count > ext4_blocks_count(es)) {
3527 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
3528 err = -EROFS;
3529 goto restore_opts;
3530 }
3531
3532 if (*flags & MS_RDONLY) {
3533 /*
3534 * First of all, the unconditional stuff we have to do
3535 * to disable replay of the journal when we next remount
3536 */
3537 sb->s_flags |= MS_RDONLY;
3538
3539 /*
3540 * OK, test if we are remounting a valid rw partition
3541 * readonly, and if so set the rdonly flag and then
3542 * mark the partition as valid again.
3543 */
3544 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3545 (sbi->s_mount_state & EXT4_VALID_FS))
3546 es->s_state = cpu_to_le16(sbi->s_mount_state);
3547
3548 if (sbi->s_journal)
3549 ext4_mark_recovery_complete(sb, es);
3550 } else {
3551 /* Make sure we can mount this feature set readwrite */
3552 if (!ext4_feature_set_ok(sb, 0)) {
3553 err = -EROFS;
3554 goto restore_opts;
3555 }
3556 /*
3557 * Make sure the group descriptor checksums
3558 * are sane. If they aren't, refuse to remount r/w.
3559 */
3560 for (g = 0; g < sbi->s_groups_count; g++) {
3561 struct ext4_group_desc *gdp =
3562 ext4_get_group_desc(sb, g, NULL);
3563
3564 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3565 ext4_msg(sb, KERN_ERR,
3566 "ext4_remount: Checksum for group %u failed (%u!=%u)",
3567 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3568 le16_to_cpu(gdp->bg_checksum));
3569 err = -EINVAL;
3570 goto restore_opts;
3571 }
3572 }
3573
3574 /*
3575 * If we have an unprocessed orphan list hanging
3576 * around from a previously readonly bdev mount,
3577 * require a full umount/remount for now.
3578 */
3579 if (es->s_last_orphan) {
3580 ext4_msg(sb, KERN_WARNING, "Couldn't "
3581 "remount RDWR because of unprocessed "
3582 "orphan inode list. Please "
3583 "umount/remount instead");
3584 err = -EINVAL;
3585 goto restore_opts;
3586 }
3587
3588 /*
3589 * Mounting a RDONLY partition read-write, so reread
3590 * and store the current valid flag. (It may have
3591 * been changed by e2fsck since we originally mounted
3592 * the partition.)
3593 */
3594 if (sbi->s_journal)
3595 ext4_clear_journal_err(sb, es);
3596 sbi->s_mount_state = le16_to_cpu(es->s_state);
3597 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3598 goto restore_opts;
3599 if (!ext4_setup_super(sb, es, 0))
3600 sb->s_flags &= ~MS_RDONLY;
3601 }
3602 }
3603 ext4_setup_system_zone(sb);
3604 if (sbi->s_journal == NULL)
3605 ext4_commit_super(sb, 1);
3606
3607 #ifdef CONFIG_QUOTA
3608 /* Release old quota file names */
3609 for (i = 0; i < MAXQUOTAS; i++)
3610 if (old_opts.s_qf_names[i] &&
3611 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3612 kfree(old_opts.s_qf_names[i]);
3613 #endif
3614 unlock_super(sb);
3615 unlock_kernel();
3616 return 0;
3617
3618 restore_opts:
3619 sb->s_flags = old_sb_flags;
3620 sbi->s_mount_opt = old_opts.s_mount_opt;
3621 sbi->s_resuid = old_opts.s_resuid;
3622 sbi->s_resgid = old_opts.s_resgid;
3623 sbi->s_commit_interval = old_opts.s_commit_interval;
3624 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3625 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3626 #ifdef CONFIG_QUOTA
3627 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3628 for (i = 0; i < MAXQUOTAS; i++) {
3629 if (sbi->s_qf_names[i] &&
3630 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3631 kfree(sbi->s_qf_names[i]);
3632 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3633 }
3634 #endif
3635 unlock_super(sb);
3636 unlock_kernel();
3637 return err;
3638 }
3639
3640 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3641 {
3642 struct super_block *sb = dentry->d_sb;
3643 struct ext4_sb_info *sbi = EXT4_SB(sb);
3644 struct ext4_super_block *es = sbi->s_es;
3645 u64 fsid;
3646
3647 if (test_opt(sb, MINIX_DF)) {
3648 sbi->s_overhead_last = 0;
3649 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3650 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3651 ext4_fsblk_t overhead = 0;
3652
3653 /*
3654 * Compute the overhead (FS structures). This is constant
3655 * for a given filesystem unless the number of block groups
3656 * changes so we cache the previous value until it does.
3657 */
3658
3659 /*
3660 * All of the blocks before first_data_block are
3661 * overhead
3662 */
3663 overhead = le32_to_cpu(es->s_first_data_block);
3664
3665 /*
3666 * Add the overhead attributed to the superblock and
3667 * block group descriptors. If the sparse superblocks
3668 * feature is turned on, then not all groups have this.
3669 */
3670 for (i = 0; i < ngroups; i++) {
3671 overhead += ext4_bg_has_super(sb, i) +
3672 ext4_bg_num_gdb(sb, i);
3673 cond_resched();
3674 }
3675
3676 /*
3677 * Every block group has an inode bitmap, a block
3678 * bitmap, and an inode table.
3679 */
3680 overhead += ngroups * (2 + sbi->s_itb_per_group);
3681 sbi->s_overhead_last = overhead;
3682 smp_wmb();
3683 sbi->s_blocks_last = ext4_blocks_count(es);
3684 }
3685
3686 buf->f_type = EXT4_SUPER_MAGIC;
3687 buf->f_bsize = sb->s_blocksize;
3688 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3689 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3690 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3691 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3692 if (buf->f_bfree < ext4_r_blocks_count(es))
3693 buf->f_bavail = 0;
3694 buf->f_files = le32_to_cpu(es->s_inodes_count);
3695 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3696 buf->f_namelen = EXT4_NAME_LEN;
3697 fsid = le64_to_cpup((void *)es->s_uuid) ^
3698 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3699 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3700 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3701
3702 return 0;
3703 }
3704
3705 /* Helper function for writing quotas on sync - we need to start transaction
3706 * before quota file is locked for write. Otherwise the are possible deadlocks:
3707 * Process 1 Process 2
3708 * ext4_create() quota_sync()
3709 * jbd2_journal_start() write_dquot()
3710 * vfs_dq_init() down(dqio_mutex)
3711 * down(dqio_mutex) jbd2_journal_start()
3712 *
3713 */
3714
3715 #ifdef CONFIG_QUOTA
3716
3717 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3718 {
3719 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3720 }
3721
3722 static int ext4_write_dquot(struct dquot *dquot)
3723 {
3724 int ret, err;
3725 handle_t *handle;
3726 struct inode *inode;
3727
3728 inode = dquot_to_inode(dquot);
3729 handle = ext4_journal_start(inode,
3730 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3731 if (IS_ERR(handle))
3732 return PTR_ERR(handle);
3733 ret = dquot_commit(dquot);
3734 err = ext4_journal_stop(handle);
3735 if (!ret)
3736 ret = err;
3737 return ret;
3738 }
3739
3740 static int ext4_acquire_dquot(struct dquot *dquot)
3741 {
3742 int ret, err;
3743 handle_t *handle;
3744
3745 handle = ext4_journal_start(dquot_to_inode(dquot),
3746 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3747 if (IS_ERR(handle))
3748 return PTR_ERR(handle);
3749 ret = dquot_acquire(dquot);
3750 err = ext4_journal_stop(handle);
3751 if (!ret)
3752 ret = err;
3753 return ret;
3754 }
3755
3756 static int ext4_release_dquot(struct dquot *dquot)
3757 {
3758 int ret, err;
3759 handle_t *handle;
3760
3761 handle = ext4_journal_start(dquot_to_inode(dquot),
3762 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3763 if (IS_ERR(handle)) {
3764 /* Release dquot anyway to avoid endless cycle in dqput() */
3765 dquot_release(dquot);
3766 return PTR_ERR(handle);
3767 }
3768 ret = dquot_release(dquot);
3769 err = ext4_journal_stop(handle);
3770 if (!ret)
3771 ret = err;
3772 return ret;
3773 }
3774
3775 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3776 {
3777 /* Are we journaling quotas? */
3778 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3779 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3780 dquot_mark_dquot_dirty(dquot);
3781 return ext4_write_dquot(dquot);
3782 } else {
3783 return dquot_mark_dquot_dirty(dquot);
3784 }
3785 }
3786
3787 static int ext4_write_info(struct super_block *sb, int type)
3788 {
3789 int ret, err;
3790 handle_t *handle;
3791
3792 /* Data block + inode block */
3793 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3794 if (IS_ERR(handle))
3795 return PTR_ERR(handle);
3796 ret = dquot_commit_info(sb, type);
3797 err = ext4_journal_stop(handle);
3798 if (!ret)
3799 ret = err;
3800 return ret;
3801 }
3802
3803 /*
3804 * Turn on quotas during mount time - we need to find
3805 * the quota file and such...
3806 */
3807 static int ext4_quota_on_mount(struct super_block *sb, int type)
3808 {
3809 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3810 EXT4_SB(sb)->s_jquota_fmt, type);
3811 }
3812
3813 /*
3814 * Standard function to be called on quota_on
3815 */
3816 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3817 char *name, int remount)
3818 {
3819 int err;
3820 struct path path;
3821
3822 if (!test_opt(sb, QUOTA))
3823 return -EINVAL;
3824 /* When remounting, no checks are needed and in fact, name is NULL */
3825 if (remount)
3826 return vfs_quota_on(sb, type, format_id, name, remount);
3827
3828 err = kern_path(name, LOOKUP_FOLLOW, &path);
3829 if (err)
3830 return err;
3831
3832 /* Quotafile not on the same filesystem? */
3833 if (path.mnt->mnt_sb != sb) {
3834 path_put(&path);
3835 return -EXDEV;
3836 }
3837 /* Journaling quota? */
3838 if (EXT4_SB(sb)->s_qf_names[type]) {
3839 /* Quotafile not in fs root? */
3840 if (path.dentry->d_parent != sb->s_root)
3841 ext4_msg(sb, KERN_WARNING,
3842 "Quota file not on filesystem root. "
3843 "Journaled quota will not work");
3844 }
3845
3846 /*
3847 * When we journal data on quota file, we have to flush journal to see
3848 * all updates to the file when we bypass pagecache...
3849 */
3850 if (EXT4_SB(sb)->s_journal &&
3851 ext4_should_journal_data(path.dentry->d_inode)) {
3852 /*
3853 * We don't need to lock updates but journal_flush() could
3854 * otherwise be livelocked...
3855 */
3856 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3857 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3858 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3859 if (err) {
3860 path_put(&path);
3861 return err;
3862 }
3863 }
3864
3865 err = vfs_quota_on_path(sb, type, format_id, &path);
3866 path_put(&path);
3867 return err;
3868 }
3869
3870 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3871 * acquiring the locks... As quota files are never truncated and quota code
3872 * itself serializes the operations (and noone else should touch the files)
3873 * we don't have to be afraid of races */
3874 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3875 size_t len, loff_t off)
3876 {
3877 struct inode *inode = sb_dqopt(sb)->files[type];
3878 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3879 int err = 0;
3880 int offset = off & (sb->s_blocksize - 1);
3881 int tocopy;
3882 size_t toread;
3883 struct buffer_head *bh;
3884 loff_t i_size = i_size_read(inode);
3885
3886 if (off > i_size)
3887 return 0;
3888 if (off+len > i_size)
3889 len = i_size-off;
3890 toread = len;
3891 while (toread > 0) {
3892 tocopy = sb->s_blocksize - offset < toread ?
3893 sb->s_blocksize - offset : toread;
3894 bh = ext4_bread(NULL, inode, blk, 0, &err);
3895 if (err)
3896 return err;
3897 if (!bh) /* A hole? */
3898 memset(data, 0, tocopy);
3899 else
3900 memcpy(data, bh->b_data+offset, tocopy);
3901 brelse(bh);
3902 offset = 0;
3903 toread -= tocopy;
3904 data += tocopy;
3905 blk++;
3906 }
3907 return len;
3908 }
3909
3910 /* Write to quotafile (we know the transaction is already started and has
3911 * enough credits) */
3912 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3913 const char *data, size_t len, loff_t off)
3914 {
3915 struct inode *inode = sb_dqopt(sb)->files[type];
3916 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3917 int err = 0;
3918 int offset = off & (sb->s_blocksize - 1);
3919 int tocopy;
3920 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3921 size_t towrite = len;
3922 struct buffer_head *bh;
3923 handle_t *handle = journal_current_handle();
3924
3925 if (EXT4_SB(sb)->s_journal && !handle) {
3926 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
3927 " cancelled because transaction is not started",
3928 (unsigned long long)off, (unsigned long long)len);
3929 return -EIO;
3930 }
3931 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3932 while (towrite > 0) {
3933 tocopy = sb->s_blocksize - offset < towrite ?
3934 sb->s_blocksize - offset : towrite;
3935 bh = ext4_bread(handle, inode, blk, 1, &err);
3936 if (!bh)
3937 goto out;
3938 if (journal_quota) {
3939 err = ext4_journal_get_write_access(handle, bh);
3940 if (err) {
3941 brelse(bh);
3942 goto out;
3943 }
3944 }
3945 lock_buffer(bh);
3946 memcpy(bh->b_data+offset, data, tocopy);
3947 flush_dcache_page(bh->b_page);
3948 unlock_buffer(bh);
3949 if (journal_quota)
3950 err = ext4_handle_dirty_metadata(handle, NULL, bh);
3951 else {
3952 /* Always do at least ordered writes for quotas */
3953 err = ext4_jbd2_file_inode(handle, inode);
3954 mark_buffer_dirty(bh);
3955 }
3956 brelse(bh);
3957 if (err)
3958 goto out;
3959 offset = 0;
3960 towrite -= tocopy;
3961 data += tocopy;
3962 blk++;
3963 }
3964 out:
3965 if (len == towrite) {
3966 mutex_unlock(&inode->i_mutex);
3967 return err;
3968 }
3969 if (inode->i_size < off+len-towrite) {
3970 i_size_write(inode, off+len-towrite);
3971 EXT4_I(inode)->i_disksize = inode->i_size;
3972 }
3973 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3974 ext4_mark_inode_dirty(handle, inode);
3975 mutex_unlock(&inode->i_mutex);
3976 return len - towrite;
3977 }
3978
3979 #endif
3980
3981 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
3982 const char *dev_name, void *data, struct vfsmount *mnt)
3983 {
3984 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
3985 }
3986
3987 static struct file_system_type ext4_fs_type = {
3988 .owner = THIS_MODULE,
3989 .name = "ext4",
3990 .get_sb = ext4_get_sb,
3991 .kill_sb = kill_block_super,
3992 .fs_flags = FS_REQUIRES_DEV,
3993 };
3994
3995 static int __init init_ext4_fs(void)
3996 {
3997 int err;
3998
3999 err = init_ext4_system_zone();
4000 if (err)
4001 return err;
4002 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4003 if (!ext4_kset)
4004 goto out4;
4005 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
4006 err = init_ext4_mballoc();
4007 if (err)
4008 goto out3;
4009
4010 err = init_ext4_xattr();
4011 if (err)
4012 goto out2;
4013 err = init_inodecache();
4014 if (err)
4015 goto out1;
4016 err = register_filesystem(&ext4_fs_type);
4017 if (err)
4018 goto out;
4019 return 0;
4020 out:
4021 destroy_inodecache();
4022 out1:
4023 exit_ext4_xattr();
4024 out2:
4025 exit_ext4_mballoc();
4026 out3:
4027 remove_proc_entry("fs/ext4", NULL);
4028 kset_unregister(ext4_kset);
4029 out4:
4030 exit_ext4_system_zone();
4031 return err;
4032 }
4033
4034 static void __exit exit_ext4_fs(void)
4035 {
4036 unregister_filesystem(&ext4_fs_type);
4037 destroy_inodecache();
4038 exit_ext4_xattr();
4039 exit_ext4_mballoc();
4040 remove_proc_entry("fs/ext4", NULL);
4041 kset_unregister(ext4_kset);
4042 exit_ext4_system_zone();
4043 }
4044
4045 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4046 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4047 MODULE_LICENSE("GPL");
4048 module_init(init_ext4_fs)
4049 module_exit(exit_ext4_fs)
Linux 2.6 ibm-acpi/thinkpad-acpi driver git tree