?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
thinkpad-acpi: drop HKEY event 0x5010
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / fs / ext4 / super.c
blob3278acf504230d3919335045d51a4e5ea8eb589d
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 static int default_mb_history_length = 1000;
54
55 module_param_named(default_mb_history_length, default_mb_history_length,
56 int, 0644);
57 MODULE_PARM_DESC(default_mb_history_length,
58 "Default number of entries saved for mb_history");
59
60 struct proc_dir_entry *ext4_proc_root;
61 static struct kset *ext4_kset;
62
63 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
64 unsigned long journal_devnum);
65 static int ext4_commit_super(struct super_block *sb, int sync);
66 static void ext4_mark_recovery_complete(struct super_block *sb,
67 struct ext4_super_block *es);
68 static void ext4_clear_journal_err(struct super_block *sb,
69 struct ext4_super_block *es);
70 static int ext4_sync_fs(struct super_block *sb, int wait);
71 static const char *ext4_decode_error(struct super_block *sb, int errno,
72 char nbuf[16]);
73 static int ext4_remount(struct super_block *sb, int *flags, char *data);
74 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
75 static int ext4_unfreeze(struct super_block *sb);
76 static void ext4_write_super(struct super_block *sb);
77 static int ext4_freeze(struct super_block *sb);
78
79
80 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
81 struct ext4_group_desc *bg)
82 {
83 return le32_to_cpu(bg->bg_block_bitmap_lo) |
84 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
85 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
86 }
87
88 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
89 struct ext4_group_desc *bg)
90 {
91 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
92 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
93 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
94 }
95
96 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
97 struct ext4_group_desc *bg)
98 {
99 return le32_to_cpu(bg->bg_inode_table_lo) |
100 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
101 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
102 }
103
104 __u32 ext4_free_blks_count(struct super_block *sb,
105 struct ext4_group_desc *bg)
106 {
107 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
108 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
109 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
110 }
111
112 __u32 ext4_free_inodes_count(struct super_block *sb,
113 struct ext4_group_desc *bg)
114 {
115 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
116 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
117 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
118 }
119
120 __u32 ext4_used_dirs_count(struct super_block *sb,
121 struct ext4_group_desc *bg)
122 {
123 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
124 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
125 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
126 }
127
128 __u32 ext4_itable_unused_count(struct super_block *sb,
129 struct ext4_group_desc *bg)
130 {
131 return le16_to_cpu(bg->bg_itable_unused_lo) |
132 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
133 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
134 }
135
136 void ext4_block_bitmap_set(struct super_block *sb,
137 struct ext4_group_desc *bg, ext4_fsblk_t blk)
138 {
139 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
140 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
141 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
142 }
143
144 void ext4_inode_bitmap_set(struct super_block *sb,
145 struct ext4_group_desc *bg, ext4_fsblk_t blk)
146 {
147 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
148 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
149 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
150 }
151
152 void ext4_inode_table_set(struct super_block *sb,
153 struct ext4_group_desc *bg, ext4_fsblk_t blk)
154 {
155 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
156 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
157 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
158 }
159
160 void ext4_free_blks_set(struct super_block *sb,
161 struct ext4_group_desc *bg, __u32 count)
162 {
163 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
164 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
165 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
166 }
167
168 void ext4_free_inodes_set(struct super_block *sb,
169 struct ext4_group_desc *bg, __u32 count)
170 {
171 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
172 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
173 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
174 }
175
176 void ext4_used_dirs_set(struct super_block *sb,
177 struct ext4_group_desc *bg, __u32 count)
178 {
179 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
180 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
181 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
182 }
183
184 void ext4_itable_unused_set(struct super_block *sb,
185 struct ext4_group_desc *bg, __u32 count)
186 {
187 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
188 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
189 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
190 }
191
192
193 /* Just increment the non-pointer handle value */
194 static handle_t *ext4_get_nojournal(void)
195 {
196 handle_t *handle = current->journal_info;
197 unsigned long ref_cnt = (unsigned long)handle;
198
199 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
200
201 ref_cnt++;
202 handle = (handle_t *)ref_cnt;
203
204 current->journal_info = handle;
205 return handle;
206 }
207
208
209 /* Decrement the non-pointer handle value */
210 static void ext4_put_nojournal(handle_t *handle)
211 {
212 unsigned long ref_cnt = (unsigned long)handle;
213
214 BUG_ON(ref_cnt == 0);
215
216 ref_cnt--;
217 handle = (handle_t *)ref_cnt;
218
219 current->journal_info = handle;
220 }
221
222 /*
223 * Wrappers for jbd2_journal_start/end.
224 *
225 * The only special thing we need to do here is to make sure that all
226 * journal_end calls result in the superblock being marked dirty, so
227 * that sync() will call the filesystem's write_super callback if
228 * appropriate.
229 */
230 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
231 {
232 journal_t *journal;
233
234 if (sb->s_flags & MS_RDONLY)
235 return ERR_PTR(-EROFS);
236
237 /* Special case here: if the journal has aborted behind our
238 * backs (eg. EIO in the commit thread), then we still need to
239 * take the FS itself readonly cleanly. */
240 journal = EXT4_SB(sb)->s_journal;
241 if (journal) {
242 if (is_journal_aborted(journal)) {
243 ext4_abort(sb, __func__, "Detected aborted journal");
244 return ERR_PTR(-EROFS);
245 }
246 return jbd2_journal_start(journal, nblocks);
247 }
248 return ext4_get_nojournal();
249 }
250
251 /*
252 * The only special thing we need to do here is to make sure that all
253 * jbd2_journal_stop calls result in the superblock being marked dirty, so
254 * that sync() will call the filesystem's write_super callback if
255 * appropriate.
256 */
257 int __ext4_journal_stop(const char *where, handle_t *handle)
258 {
259 struct super_block *sb;
260 int err;
261 int rc;
262
263 if (!ext4_handle_valid(handle)) {
264 ext4_put_nojournal(handle);
265 return 0;
266 }
267 sb = handle->h_transaction->t_journal->j_private;
268 err = handle->h_err;
269 rc = jbd2_journal_stop(handle);
270
271 if (!err)
272 err = rc;
273 if (err)
274 __ext4_std_error(sb, where, err);
275 return err;
276 }
277
278 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
279 struct buffer_head *bh, handle_t *handle, int err)
280 {
281 char nbuf[16];
282 const char *errstr = ext4_decode_error(NULL, err, nbuf);
283
284 BUG_ON(!ext4_handle_valid(handle));
285
286 if (bh)
287 BUFFER_TRACE(bh, "abort");
288
289 if (!handle->h_err)
290 handle->h_err = err;
291
292 if (is_handle_aborted(handle))
293 return;
294
295 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
296 caller, errstr, err_fn);
297
298 jbd2_journal_abort_handle(handle);
299 }
300
301 /* Deal with the reporting of failure conditions on a filesystem such as
302 * inconsistencies detected or read IO failures.
303 *
304 * On ext2, we can store the error state of the filesystem in the
305 * superblock. That is not possible on ext4, because we may have other
306 * write ordering constraints on the superblock which prevent us from
307 * writing it out straight away; and given that the journal is about to
308 * be aborted, we can't rely on the current, or future, transactions to
309 * write out the superblock safely.
310 *
311 * We'll just use the jbd2_journal_abort() error code to record an error in
312 * the journal instead. On recovery, the journal will compain about
313 * that error until we've noted it down and cleared it.
314 */
315
316 static void ext4_handle_error(struct super_block *sb)
317 {
318 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
319
320 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
321 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
322
323 if (sb->s_flags & MS_RDONLY)
324 return;
325
326 if (!test_opt(sb, ERRORS_CONT)) {
327 journal_t *journal = EXT4_SB(sb)->s_journal;
328
329 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
330 if (journal)
331 jbd2_journal_abort(journal, -EIO);
332 }
333 if (test_opt(sb, ERRORS_RO)) {
334 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
335 sb->s_flags |= MS_RDONLY;
336 }
337 ext4_commit_super(sb, 1);
338 if (test_opt(sb, ERRORS_PANIC))
339 panic("EXT4-fs (device %s): panic forced after error\n",
340 sb->s_id);
341 }
342
343 void ext4_error(struct super_block *sb, const char *function,
344 const char *fmt, ...)
345 {
346 va_list args;
347
348 va_start(args, fmt);
349 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
350 vprintk(fmt, args);
351 printk("\n");
352 va_end(args);
353
354 ext4_handle_error(sb);
355 }
356
357 static const char *ext4_decode_error(struct super_block *sb, int errno,
358 char nbuf[16])
359 {
360 char *errstr = NULL;
361
362 switch (errno) {
363 case -EIO:
364 errstr = "IO failure";
365 break;
366 case -ENOMEM:
367 errstr = "Out of memory";
368 break;
369 case -EROFS:
370 if (!sb || (EXT4_SB(sb)->s_journal &&
371 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
372 errstr = "Journal has aborted";
373 else
374 errstr = "Readonly filesystem";
375 break;
376 default:
377 /* If the caller passed in an extra buffer for unknown
378 * errors, textualise them now. Else we just return
379 * NULL. */
380 if (nbuf) {
381 /* Check for truncated error codes... */
382 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
383 errstr = nbuf;
384 }
385 break;
386 }
387
388 return errstr;
389 }
390
391 /* __ext4_std_error decodes expected errors from journaling functions
392 * automatically and invokes the appropriate error response. */
393
394 void __ext4_std_error(struct super_block *sb, const char *function, int errno)
395 {
396 char nbuf[16];
397 const char *errstr;
398
399 /* Special case: if the error is EROFS, and we're not already
400 * inside a transaction, then there's really no point in logging
401 * an error. */
402 if (errno == -EROFS && journal_current_handle() == NULL &&
403 (sb->s_flags & MS_RDONLY))
404 return;
405
406 errstr = ext4_decode_error(sb, errno, nbuf);
407 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
408 sb->s_id, function, errstr);
409
410 ext4_handle_error(sb);
411 }
412
413 /*
414 * ext4_abort is a much stronger failure handler than ext4_error. The
415 * abort function may be used to deal with unrecoverable failures such
416 * as journal IO errors or ENOMEM at a critical moment in log management.
417 *
418 * We unconditionally force the filesystem into an ABORT|READONLY state,
419 * unless the error response on the fs has been set to panic in which
420 * case we take the easy way out and panic immediately.
421 */
422
423 void ext4_abort(struct super_block *sb, const char *function,
424 const char *fmt, ...)
425 {
426 va_list args;
427
428 va_start(args, fmt);
429 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
430 vprintk(fmt, args);
431 printk("\n");
432 va_end(args);
433
434 if (test_opt(sb, ERRORS_PANIC))
435 panic("EXT4-fs panic from previous error\n");
436
437 if (sb->s_flags & MS_RDONLY)
438 return;
439
440 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
441 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
442 sb->s_flags |= MS_RDONLY;
443 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
444 if (EXT4_SB(sb)->s_journal)
445 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
446 }
447
448 void ext4_msg (struct super_block * sb, const char *prefix,
449 const char *fmt, ...)
450 {
451 va_list args;
452
453 va_start(args, fmt);
454 printk("%sEXT4-fs (%s): ", prefix, sb->s_id);
455 vprintk(fmt, args);
456 printk("\n");
457 va_end(args);
458 }
459
460 void ext4_warning(struct super_block *sb, const char *function,
461 const char *fmt, ...)
462 {
463 va_list args;
464
465 va_start(args, fmt);
466 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
467 sb->s_id, function);
468 vprintk(fmt, args);
469 printk("\n");
470 va_end(args);
471 }
472
473 void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp,
474 const char *function, const char *fmt, ...)
475 __releases(bitlock)
476 __acquires(bitlock)
477 {
478 va_list args;
479 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
480
481 va_start(args, fmt);
482 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
483 vprintk(fmt, args);
484 printk("\n");
485 va_end(args);
486
487 if (test_opt(sb, ERRORS_CONT)) {
488 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
489 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
490 ext4_commit_super(sb, 0);
491 return;
492 }
493 ext4_unlock_group(sb, grp);
494 ext4_handle_error(sb);
495 /*
496 * We only get here in the ERRORS_RO case; relocking the group
497 * may be dangerous, but nothing bad will happen since the
498 * filesystem will have already been marked read/only and the
499 * journal has been aborted. We return 1 as a hint to callers
500 * who might what to use the return value from
501 * ext4_grp_locked_error() to distinguish beween the
502 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
503 * aggressively from the ext4 function in question, with a
504 * more appropriate error code.
505 */
506 ext4_lock_group(sb, grp);
507 return;
508 }
509
510 void ext4_update_dynamic_rev(struct super_block *sb)
511 {
512 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
513
514 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
515 return;
516
517 ext4_warning(sb, __func__,
518 "updating to rev %d because of new feature flag, "
519 "running e2fsck is recommended",
520 EXT4_DYNAMIC_REV);
521
522 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
523 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
524 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
525 /* leave es->s_feature_*compat flags alone */
526 /* es->s_uuid will be set by e2fsck if empty */
527
528 /*
529 * The rest of the superblock fields should be zero, and if not it
530 * means they are likely already in use, so leave them alone. We
531 * can leave it up to e2fsck to clean up any inconsistencies there.
532 */
533 }
534
535 /*
536 * Open the external journal device
537 */
538 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
539 {
540 struct block_device *bdev;
541 char b[BDEVNAME_SIZE];
542
543 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
544 if (IS_ERR(bdev))
545 goto fail;
546 return bdev;
547
548 fail:
549 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
550 __bdevname(dev, b), PTR_ERR(bdev));
551 return NULL;
552 }
553
554 /*
555 * Release the journal device
556 */
557 static int ext4_blkdev_put(struct block_device *bdev)
558 {
559 bd_release(bdev);
560 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
561 }
562
563 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
564 {
565 struct block_device *bdev;
566 int ret = -ENODEV;
567
568 bdev = sbi->journal_bdev;
569 if (bdev) {
570 ret = ext4_blkdev_put(bdev);
571 sbi->journal_bdev = NULL;
572 }
573 return ret;
574 }
575
576 static inline struct inode *orphan_list_entry(struct list_head *l)
577 {
578 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
579 }
580
581 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
582 {
583 struct list_head *l;
584
585 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
586 le32_to_cpu(sbi->s_es->s_last_orphan));
587
588 printk(KERN_ERR "sb_info orphan list:\n");
589 list_for_each(l, &sbi->s_orphan) {
590 struct inode *inode = orphan_list_entry(l);
591 printk(KERN_ERR " "
592 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
593 inode->i_sb->s_id, inode->i_ino, inode,
594 inode->i_mode, inode->i_nlink,
595 NEXT_ORPHAN(inode));
596 }
597 }
598
599 static void ext4_put_super(struct super_block *sb)
600 {
601 struct ext4_sb_info *sbi = EXT4_SB(sb);
602 struct ext4_super_block *es = sbi->s_es;
603 int i, err;
604
605 flush_workqueue(sbi->dio_unwritten_wq);
606 destroy_workqueue(sbi->dio_unwritten_wq);
607
608 lock_super(sb);
609 lock_kernel();
610 if (sb->s_dirt)
611 ext4_commit_super(sb, 1);
612
613 if (sbi->s_journal) {
614 err = jbd2_journal_destroy(sbi->s_journal);
615 sbi->s_journal = NULL;
616 if (err < 0)
617 ext4_abort(sb, __func__,
618 "Couldn't clean up the journal");
619 }
620
621 ext4_release_system_zone(sb);
622 ext4_mb_release(sb);
623 ext4_ext_release(sb);
624 ext4_xattr_put_super(sb);
625
626 if (!(sb->s_flags & MS_RDONLY)) {
627 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
628 es->s_state = cpu_to_le16(sbi->s_mount_state);
629 ext4_commit_super(sb, 1);
630 }
631 if (sbi->s_proc) {
632 remove_proc_entry(sb->s_id, ext4_proc_root);
633 }
634 kobject_del(&sbi->s_kobj);
635
636 for (i = 0; i < sbi->s_gdb_count; i++)
637 brelse(sbi->s_group_desc[i]);
638 kfree(sbi->s_group_desc);
639 if (is_vmalloc_addr(sbi->s_flex_groups))
640 vfree(sbi->s_flex_groups);
641 else
642 kfree(sbi->s_flex_groups);
643 percpu_counter_destroy(&sbi->s_freeblocks_counter);
644 percpu_counter_destroy(&sbi->s_freeinodes_counter);
645 percpu_counter_destroy(&sbi->s_dirs_counter);
646 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
647 brelse(sbi->s_sbh);
648 #ifdef CONFIG_QUOTA
649 for (i = 0; i < MAXQUOTAS; i++)
650 kfree(sbi->s_qf_names[i]);
651 #endif
652
653 /* Debugging code just in case the in-memory inode orphan list
654 * isn't empty. The on-disk one can be non-empty if we've
655 * detected an error and taken the fs readonly, but the
656 * in-memory list had better be clean by this point. */
657 if (!list_empty(&sbi->s_orphan))
658 dump_orphan_list(sb, sbi);
659 J_ASSERT(list_empty(&sbi->s_orphan));
660
661 invalidate_bdev(sb->s_bdev);
662 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
663 /*
664 * Invalidate the journal device's buffers. We don't want them
665 * floating about in memory - the physical journal device may
666 * hotswapped, and it breaks the `ro-after' testing code.
667 */
668 sync_blockdev(sbi->journal_bdev);
669 invalidate_bdev(sbi->journal_bdev);
670 ext4_blkdev_remove(sbi);
671 }
672 sb->s_fs_info = NULL;
673 /*
674 * Now that we are completely done shutting down the
675 * superblock, we need to actually destroy the kobject.
676 */
677 unlock_kernel();
678 unlock_super(sb);
679 kobject_put(&sbi->s_kobj);
680 wait_for_completion(&sbi->s_kobj_unregister);
681 kfree(sbi->s_blockgroup_lock);
682 kfree(sbi);
683 }
684
685 static struct kmem_cache *ext4_inode_cachep;
686
687 /*
688 * Called inside transaction, so use GFP_NOFS
689 */
690 static struct inode *ext4_alloc_inode(struct super_block *sb)
691 {
692 struct ext4_inode_info *ei;
693
694 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
695 if (!ei)
696 return NULL;
697
698 ei->vfs_inode.i_version = 1;
699 ei->vfs_inode.i_data.writeback_index = 0;
700 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
701 INIT_LIST_HEAD(&ei->i_prealloc_list);
702 spin_lock_init(&ei->i_prealloc_lock);
703 /*
704 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
705 * therefore it can be null here. Don't check it, just initialize
706 * jinode.
707 */
708 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
709 ei->i_reserved_data_blocks = 0;
710 ei->i_reserved_meta_blocks = 0;
711 ei->i_allocated_meta_blocks = 0;
712 ei->i_delalloc_reserved_flag = 0;
713 spin_lock_init(&(ei->i_block_reservation_lock));
714 #ifdef CONFIG_QUOTA
715 ei->i_reserved_quota = 0;
716 #endif
717 INIT_LIST_HEAD(&ei->i_aio_dio_complete_list);
718 ei->cur_aio_dio = NULL;
719 ei->i_sync_tid = 0;
720 ei->i_datasync_tid = 0;
721
722 return &ei->vfs_inode;
723 }
724
725 static void ext4_destroy_inode(struct inode *inode)
726 {
727 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
728 ext4_msg(inode->i_sb, KERN_ERR,
729 "Inode %lu (%p): orphan list check failed!",
730 inode->i_ino, EXT4_I(inode));
731 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
732 EXT4_I(inode), sizeof(struct ext4_inode_info),
733 true);
734 dump_stack();
735 }
736 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
737 }
738
739 static void init_once(void *foo)
740 {
741 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
742
743 INIT_LIST_HEAD(&ei->i_orphan);
744 #ifdef CONFIG_EXT4_FS_XATTR
745 init_rwsem(&ei->xattr_sem);
746 #endif
747 init_rwsem(&ei->i_data_sem);
748 inode_init_once(&ei->vfs_inode);
749 }
750
751 static int init_inodecache(void)
752 {
753 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
754 sizeof(struct ext4_inode_info),
755 0, (SLAB_RECLAIM_ACCOUNT|
756 SLAB_MEM_SPREAD),
757 init_once);
758 if (ext4_inode_cachep == NULL)
759 return -ENOMEM;
760 return 0;
761 }
762
763 static void destroy_inodecache(void)
764 {
765 kmem_cache_destroy(ext4_inode_cachep);
766 }
767
768 static void ext4_clear_inode(struct inode *inode)
769 {
770 ext4_discard_preallocations(inode);
771 if (EXT4_JOURNAL(inode))
772 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
773 &EXT4_I(inode)->jinode);
774 }
775
776 static inline void ext4_show_quota_options(struct seq_file *seq,
777 struct super_block *sb)
778 {
779 #if defined(CONFIG_QUOTA)
780 struct ext4_sb_info *sbi = EXT4_SB(sb);
781
782 if (sbi->s_jquota_fmt)
783 seq_printf(seq, ",jqfmt=%s",
784 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
785
786 if (sbi->s_qf_names[USRQUOTA])
787 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
788
789 if (sbi->s_qf_names[GRPQUOTA])
790 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
791
792 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
793 seq_puts(seq, ",usrquota");
794
795 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
796 seq_puts(seq, ",grpquota");
797 #endif
798 }
799
800 /*
801 * Show an option if
802 * - it's set to a non-default value OR
803 * - if the per-sb default is different from the global default
804 */
805 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
806 {
807 int def_errors;
808 unsigned long def_mount_opts;
809 struct super_block *sb = vfs->mnt_sb;
810 struct ext4_sb_info *sbi = EXT4_SB(sb);
811 struct ext4_super_block *es = sbi->s_es;
812
813 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
814 def_errors = le16_to_cpu(es->s_errors);
815
816 if (sbi->s_sb_block != 1)
817 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
818 if (test_opt(sb, MINIX_DF))
819 seq_puts(seq, ",minixdf");
820 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
821 seq_puts(seq, ",grpid");
822 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
823 seq_puts(seq, ",nogrpid");
824 if (sbi->s_resuid != EXT4_DEF_RESUID ||
825 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
826 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
827 }
828 if (sbi->s_resgid != EXT4_DEF_RESGID ||
829 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
830 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
831 }
832 if (test_opt(sb, ERRORS_RO)) {
833 if (def_errors == EXT4_ERRORS_PANIC ||
834 def_errors == EXT4_ERRORS_CONTINUE) {
835 seq_puts(seq, ",errors=remount-ro");
836 }
837 }
838 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
839 seq_puts(seq, ",errors=continue");
840 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
841 seq_puts(seq, ",errors=panic");
842 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
843 seq_puts(seq, ",nouid32");
844 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
845 seq_puts(seq, ",debug");
846 if (test_opt(sb, OLDALLOC))
847 seq_puts(seq, ",oldalloc");
848 #ifdef CONFIG_EXT4_FS_XATTR
849 if (test_opt(sb, XATTR_USER) &&
850 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
851 seq_puts(seq, ",user_xattr");
852 if (!test_opt(sb, XATTR_USER) &&
853 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
854 seq_puts(seq, ",nouser_xattr");
855 }
856 #endif
857 #ifdef CONFIG_EXT4_FS_POSIX_ACL
858 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
859 seq_puts(seq, ",acl");
860 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
861 seq_puts(seq, ",noacl");
862 #endif
863 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
864 seq_printf(seq, ",commit=%u",
865 (unsigned) (sbi->s_commit_interval / HZ));
866 }
867 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
868 seq_printf(seq, ",min_batch_time=%u",
869 (unsigned) sbi->s_min_batch_time);
870 }
871 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
872 seq_printf(seq, ",max_batch_time=%u",
873 (unsigned) sbi->s_min_batch_time);
874 }
875
876 /*
877 * We're changing the default of barrier mount option, so
878 * let's always display its mount state so it's clear what its
879 * status is.
880 */
881 seq_puts(seq, ",barrier=");
882 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
883 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
884 seq_puts(seq, ",journal_async_commit");
885 if (test_opt(sb, NOBH))
886 seq_puts(seq, ",nobh");
887 if (test_opt(sb, I_VERSION))
888 seq_puts(seq, ",i_version");
889 if (!test_opt(sb, DELALLOC))
890 seq_puts(seq, ",nodelalloc");
891
892
893 if (sbi->s_stripe)
894 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
895 /*
896 * journal mode get enabled in different ways
897 * So just print the value even if we didn't specify it
898 */
899 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
900 seq_puts(seq, ",data=journal");
901 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
902 seq_puts(seq, ",data=ordered");
903 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
904 seq_puts(seq, ",data=writeback");
905
906 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
907 seq_printf(seq, ",inode_readahead_blks=%u",
908 sbi->s_inode_readahead_blks);
909
910 if (test_opt(sb, DATA_ERR_ABORT))
911 seq_puts(seq, ",data_err=abort");
912
913 if (test_opt(sb, NO_AUTO_DA_ALLOC))
914 seq_puts(seq, ",noauto_da_alloc");
915
916 if (test_opt(sb, DISCARD))
917 seq_puts(seq, ",discard");
918
919 if (test_opt(sb, NOLOAD))
920 seq_puts(seq, ",norecovery");
921
922 ext4_show_quota_options(seq, sb);
923
924 return 0;
925 }
926
927 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
928 u64 ino, u32 generation)
929 {
930 struct inode *inode;
931
932 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
933 return ERR_PTR(-ESTALE);
934 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
935 return ERR_PTR(-ESTALE);
936
937 /* iget isn't really right if the inode is currently unallocated!!
938 *
939 * ext4_read_inode will return a bad_inode if the inode had been
940 * deleted, so we should be safe.
941 *
942 * Currently we don't know the generation for parent directory, so
943 * a generation of 0 means "accept any"
944 */
945 inode = ext4_iget(sb, ino);
946 if (IS_ERR(inode))
947 return ERR_CAST(inode);
948 if (generation && inode->i_generation != generation) {
949 iput(inode);
950 return ERR_PTR(-ESTALE);
951 }
952
953 return inode;
954 }
955
956 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
957 int fh_len, int fh_type)
958 {
959 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
960 ext4_nfs_get_inode);
961 }
962
963 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
964 int fh_len, int fh_type)
965 {
966 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
967 ext4_nfs_get_inode);
968 }
969
970 /*
971 * Try to release metadata pages (indirect blocks, directories) which are
972 * mapped via the block device. Since these pages could have journal heads
973 * which would prevent try_to_free_buffers() from freeing them, we must use
974 * jbd2 layer's try_to_free_buffers() function to release them.
975 */
976 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
977 gfp_t wait)
978 {
979 journal_t *journal = EXT4_SB(sb)->s_journal;
980
981 WARN_ON(PageChecked(page));
982 if (!page_has_buffers(page))
983 return 0;
984 if (journal)
985 return jbd2_journal_try_to_free_buffers(journal, page,
986 wait & ~__GFP_WAIT);
987 return try_to_free_buffers(page);
988 }
989
990 #ifdef CONFIG_QUOTA
991 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
992 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
993
994 static int ext4_write_dquot(struct dquot *dquot);
995 static int ext4_acquire_dquot(struct dquot *dquot);
996 static int ext4_release_dquot(struct dquot *dquot);
997 static int ext4_mark_dquot_dirty(struct dquot *dquot);
998 static int ext4_write_info(struct super_block *sb, int type);
999 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1000 char *path, int remount);
1001 static int ext4_quota_on_mount(struct super_block *sb, int type);
1002 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1003 size_t len, loff_t off);
1004 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1005 const char *data, size_t len, loff_t off);
1006
1007 static struct dquot_operations ext4_quota_operations = {
1008 .initialize = dquot_initialize,
1009 .drop = dquot_drop,
1010 .alloc_space = dquot_alloc_space,
1011 .reserve_space = dquot_reserve_space,
1012 .claim_space = dquot_claim_space,
1013 .release_rsv = dquot_release_reserved_space,
1014 #ifdef CONFIG_QUOTA
1015 .get_reserved_space = ext4_get_reserved_space,
1016 #endif
1017 .alloc_inode = dquot_alloc_inode,
1018 .free_space = dquot_free_space,
1019 .free_inode = dquot_free_inode,
1020 .transfer = dquot_transfer,
1021 .write_dquot = ext4_write_dquot,
1022 .acquire_dquot = ext4_acquire_dquot,
1023 .release_dquot = ext4_release_dquot,
1024 .mark_dirty = ext4_mark_dquot_dirty,
1025 .write_info = ext4_write_info,
1026 .alloc_dquot = dquot_alloc,
1027 .destroy_dquot = dquot_destroy,
1028 };
1029
1030 static struct quotactl_ops ext4_qctl_operations = {
1031 .quota_on = ext4_quota_on,
1032 .quota_off = vfs_quota_off,
1033 .quota_sync = vfs_quota_sync,
1034 .get_info = vfs_get_dqinfo,
1035 .set_info = vfs_set_dqinfo,
1036 .get_dqblk = vfs_get_dqblk,
1037 .set_dqblk = vfs_set_dqblk
1038 };
1039 #endif
1040
1041 static const struct super_operations ext4_sops = {
1042 .alloc_inode = ext4_alloc_inode,
1043 .destroy_inode = ext4_destroy_inode,
1044 .write_inode = ext4_write_inode,
1045 .dirty_inode = ext4_dirty_inode,
1046 .delete_inode = ext4_delete_inode,
1047 .put_super = ext4_put_super,
1048 .sync_fs = ext4_sync_fs,
1049 .freeze_fs = ext4_freeze,
1050 .unfreeze_fs = ext4_unfreeze,
1051 .statfs = ext4_statfs,
1052 .remount_fs = ext4_remount,
1053 .clear_inode = ext4_clear_inode,
1054 .show_options = ext4_show_options,
1055 #ifdef CONFIG_QUOTA
1056 .quota_read = ext4_quota_read,
1057 .quota_write = ext4_quota_write,
1058 #endif
1059 .bdev_try_to_free_page = bdev_try_to_free_page,
1060 };
1061
1062 static const struct super_operations ext4_nojournal_sops = {
1063 .alloc_inode = ext4_alloc_inode,
1064 .destroy_inode = ext4_destroy_inode,
1065 .write_inode = ext4_write_inode,
1066 .dirty_inode = ext4_dirty_inode,
1067 .delete_inode = ext4_delete_inode,
1068 .write_super = ext4_write_super,
1069 .put_super = ext4_put_super,
1070 .statfs = ext4_statfs,
1071 .remount_fs = ext4_remount,
1072 .clear_inode = ext4_clear_inode,
1073 .show_options = ext4_show_options,
1074 #ifdef CONFIG_QUOTA
1075 .quota_read = ext4_quota_read,
1076 .quota_write = ext4_quota_write,
1077 #endif
1078 .bdev_try_to_free_page = bdev_try_to_free_page,
1079 };
1080
1081 static const struct export_operations ext4_export_ops = {
1082 .fh_to_dentry = ext4_fh_to_dentry,
1083 .fh_to_parent = ext4_fh_to_parent,
1084 .get_parent = ext4_get_parent,
1085 };
1086
1087 enum {
1088 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1089 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1090 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1091 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1092 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1093 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1094 Opt_journal_update, Opt_journal_dev,
1095 Opt_journal_checksum, Opt_journal_async_commit,
1096 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1097 Opt_data_err_abort, Opt_data_err_ignore, Opt_mb_history_length,
1098 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1099 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
1100 Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err, Opt_resize,
1101 Opt_usrquota, Opt_grpquota, Opt_i_version,
1102 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1103 Opt_block_validity, Opt_noblock_validity,
1104 Opt_inode_readahead_blks, Opt_journal_ioprio,
1105 Opt_discard, Opt_nodiscard,
1106 };
1107
1108 static const match_table_t tokens = {
1109 {Opt_bsd_df, "bsddf"},
1110 {Opt_minix_df, "minixdf"},
1111 {Opt_grpid, "grpid"},
1112 {Opt_grpid, "bsdgroups"},
1113 {Opt_nogrpid, "nogrpid"},
1114 {Opt_nogrpid, "sysvgroups"},
1115 {Opt_resgid, "resgid=%u"},
1116 {Opt_resuid, "resuid=%u"},
1117 {Opt_sb, "sb=%u"},
1118 {Opt_err_cont, "errors=continue"},
1119 {Opt_err_panic, "errors=panic"},
1120 {Opt_err_ro, "errors=remount-ro"},
1121 {Opt_nouid32, "nouid32"},
1122 {Opt_debug, "debug"},
1123 {Opt_oldalloc, "oldalloc"},
1124 {Opt_orlov, "orlov"},
1125 {Opt_user_xattr, "user_xattr"},
1126 {Opt_nouser_xattr, "nouser_xattr"},
1127 {Opt_acl, "acl"},
1128 {Opt_noacl, "noacl"},
1129 {Opt_noload, "noload"},
1130 {Opt_noload, "norecovery"},
1131 {Opt_nobh, "nobh"},
1132 {Opt_bh, "bh"},
1133 {Opt_commit, "commit=%u"},
1134 {Opt_min_batch_time, "min_batch_time=%u"},
1135 {Opt_max_batch_time, "max_batch_time=%u"},
1136 {Opt_journal_update, "journal=update"},
1137 {Opt_journal_dev, "journal_dev=%u"},
1138 {Opt_journal_checksum, "journal_checksum"},
1139 {Opt_journal_async_commit, "journal_async_commit"},
1140 {Opt_abort, "abort"},
1141 {Opt_data_journal, "data=journal"},
1142 {Opt_data_ordered, "data=ordered"},
1143 {Opt_data_writeback, "data=writeback"},
1144 {Opt_data_err_abort, "data_err=abort"},
1145 {Opt_data_err_ignore, "data_err=ignore"},
1146 {Opt_mb_history_length, "mb_history_length=%u"},
1147 {Opt_offusrjquota, "usrjquota="},
1148 {Opt_usrjquota, "usrjquota=%s"},
1149 {Opt_offgrpjquota, "grpjquota="},
1150 {Opt_grpjquota, "grpjquota=%s"},
1151 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1152 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1153 {Opt_grpquota, "grpquota"},
1154 {Opt_noquota, "noquota"},
1155 {Opt_quota, "quota"},
1156 {Opt_usrquota, "usrquota"},
1157 {Opt_barrier, "barrier=%u"},
1158 {Opt_barrier, "barrier"},
1159 {Opt_nobarrier, "nobarrier"},
1160 {Opt_i_version, "i_version"},
1161 {Opt_stripe, "stripe=%u"},
1162 {Opt_resize, "resize"},
1163 {Opt_delalloc, "delalloc"},
1164 {Opt_nodelalloc, "nodelalloc"},
1165 {Opt_block_validity, "block_validity"},
1166 {Opt_noblock_validity, "noblock_validity"},
1167 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1168 {Opt_journal_ioprio, "journal_ioprio=%u"},
1169 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1170 {Opt_auto_da_alloc, "auto_da_alloc"},
1171 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1172 {Opt_discard, "discard"},
1173 {Opt_nodiscard, "nodiscard"},
1174 {Opt_err, NULL},
1175 };
1176
1177 static ext4_fsblk_t get_sb_block(void **data)
1178 {
1179 ext4_fsblk_t sb_block;
1180 char *options = (char *) *data;
1181
1182 if (!options || strncmp(options, "sb=", 3) != 0)
1183 return 1; /* Default location */
1184
1185 options += 3;
1186 /* TODO: use simple_strtoll with >32bit ext4 */
1187 sb_block = simple_strtoul(options, &options, 0);
1188 if (*options && *options != ',') {
1189 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1190 (char *) *data);
1191 return 1;
1192 }
1193 if (*options == ',')
1194 options++;
1195 *data = (void *) options;
1196
1197 return sb_block;
1198 }
1199
1200 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1201
1202 static int parse_options(char *options, struct super_block *sb,
1203 unsigned long *journal_devnum,
1204 unsigned int *journal_ioprio,
1205 ext4_fsblk_t *n_blocks_count, int is_remount)
1206 {
1207 struct ext4_sb_info *sbi = EXT4_SB(sb);
1208 char *p;
1209 substring_t args[MAX_OPT_ARGS];
1210 int data_opt = 0;
1211 int option;
1212 #ifdef CONFIG_QUOTA
1213 int qtype, qfmt;
1214 char *qname;
1215 #endif
1216
1217 if (!options)
1218 return 1;
1219
1220 while ((p = strsep(&options, ",")) != NULL) {
1221 int token;
1222 if (!*p)
1223 continue;
1224
1225 token = match_token(p, tokens, args);
1226 switch (token) {
1227 case Opt_bsd_df:
1228 clear_opt(sbi->s_mount_opt, MINIX_DF);
1229 break;
1230 case Opt_minix_df:
1231 set_opt(sbi->s_mount_opt, MINIX_DF);
1232 break;
1233 case Opt_grpid:
1234 set_opt(sbi->s_mount_opt, GRPID);
1235 break;
1236 case Opt_nogrpid:
1237 clear_opt(sbi->s_mount_opt, GRPID);
1238 break;
1239 case Opt_resuid:
1240 if (match_int(&args[0], &option))
1241 return 0;
1242 sbi->s_resuid = option;
1243 break;
1244 case Opt_resgid:
1245 if (match_int(&args[0], &option))
1246 return 0;
1247 sbi->s_resgid = option;
1248 break;
1249 case Opt_sb:
1250 /* handled by get_sb_block() instead of here */
1251 /* *sb_block = match_int(&args[0]); */
1252 break;
1253 case Opt_err_panic:
1254 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1255 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1256 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1257 break;
1258 case Opt_err_ro:
1259 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1260 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1261 set_opt(sbi->s_mount_opt, ERRORS_RO);
1262 break;
1263 case Opt_err_cont:
1264 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1265 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1266 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1267 break;
1268 case Opt_nouid32:
1269 set_opt(sbi->s_mount_opt, NO_UID32);
1270 break;
1271 case Opt_debug:
1272 set_opt(sbi->s_mount_opt, DEBUG);
1273 break;
1274 case Opt_oldalloc:
1275 set_opt(sbi->s_mount_opt, OLDALLOC);
1276 break;
1277 case Opt_orlov:
1278 clear_opt(sbi->s_mount_opt, OLDALLOC);
1279 break;
1280 #ifdef CONFIG_EXT4_FS_XATTR
1281 case Opt_user_xattr:
1282 set_opt(sbi->s_mount_opt, XATTR_USER);
1283 break;
1284 case Opt_nouser_xattr:
1285 clear_opt(sbi->s_mount_opt, XATTR_USER);
1286 break;
1287 #else
1288 case Opt_user_xattr:
1289 case Opt_nouser_xattr:
1290 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1291 break;
1292 #endif
1293 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1294 case Opt_acl:
1295 set_opt(sbi->s_mount_opt, POSIX_ACL);
1296 break;
1297 case Opt_noacl:
1298 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1299 break;
1300 #else
1301 case Opt_acl:
1302 case Opt_noacl:
1303 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1304 break;
1305 #endif
1306 case Opt_journal_update:
1307 /* @@@ FIXME */
1308 /* Eventually we will want to be able to create
1309 a journal file here. For now, only allow the
1310 user to specify an existing inode to be the
1311 journal file. */
1312 if (is_remount) {
1313 ext4_msg(sb, KERN_ERR,
1314 "Cannot specify journal on remount");
1315 return 0;
1316 }
1317 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1318 break;
1319 case Opt_journal_dev:
1320 if (is_remount) {
1321 ext4_msg(sb, KERN_ERR,
1322 "Cannot specify journal on remount");
1323 return 0;
1324 }
1325 if (match_int(&args[0], &option))
1326 return 0;
1327 *journal_devnum = option;
1328 break;
1329 case Opt_journal_checksum:
1330 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1331 break;
1332 case Opt_journal_async_commit:
1333 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1334 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1335 break;
1336 case Opt_noload:
1337 set_opt(sbi->s_mount_opt, NOLOAD);
1338 break;
1339 case Opt_commit:
1340 if (match_int(&args[0], &option))
1341 return 0;
1342 if (option < 0)
1343 return 0;
1344 if (option == 0)
1345 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1346 sbi->s_commit_interval = HZ * option;
1347 break;
1348 case Opt_max_batch_time:
1349 if (match_int(&args[0], &option))
1350 return 0;
1351 if (option < 0)
1352 return 0;
1353 if (option == 0)
1354 option = EXT4_DEF_MAX_BATCH_TIME;
1355 sbi->s_max_batch_time = option;
1356 break;
1357 case Opt_min_batch_time:
1358 if (match_int(&args[0], &option))
1359 return 0;
1360 if (option < 0)
1361 return 0;
1362 sbi->s_min_batch_time = option;
1363 break;
1364 case Opt_data_journal:
1365 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1366 goto datacheck;
1367 case Opt_data_ordered:
1368 data_opt = EXT4_MOUNT_ORDERED_DATA;
1369 goto datacheck;
1370 case Opt_data_writeback:
1371 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1372 datacheck:
1373 if (is_remount) {
1374 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1375 != data_opt) {
1376 ext4_msg(sb, KERN_ERR,
1377 "Cannot change data mode on remount");
1378 return 0;
1379 }
1380 } else {
1381 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1382 sbi->s_mount_opt |= data_opt;
1383 }
1384 break;
1385 case Opt_data_err_abort:
1386 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1387 break;
1388 case Opt_data_err_ignore:
1389 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1390 break;
1391 case Opt_mb_history_length:
1392 if (match_int(&args[0], &option))
1393 return 0;
1394 if (option < 0)
1395 return 0;
1396 sbi->s_mb_history_max = option;
1397 break;
1398 #ifdef CONFIG_QUOTA
1399 case Opt_usrjquota:
1400 qtype = USRQUOTA;
1401 goto set_qf_name;
1402 case Opt_grpjquota:
1403 qtype = GRPQUOTA;
1404 set_qf_name:
1405 if (sb_any_quota_loaded(sb) &&
1406 !sbi->s_qf_names[qtype]) {
1407 ext4_msg(sb, KERN_ERR,
1408 "Cannot change journaled "
1409 "quota options when quota turned on");
1410 return 0;
1411 }
1412 qname = match_strdup(&args[0]);
1413 if (!qname) {
1414 ext4_msg(sb, KERN_ERR,
1415 "Not enough memory for "
1416 "storing quotafile name");
1417 return 0;
1418 }
1419 if (sbi->s_qf_names[qtype] &&
1420 strcmp(sbi->s_qf_names[qtype], qname)) {
1421 ext4_msg(sb, KERN_ERR,
1422 "%s quota file already "
1423 "specified", QTYPE2NAME(qtype));
1424 kfree(qname);
1425 return 0;
1426 }
1427 sbi->s_qf_names[qtype] = qname;
1428 if (strchr(sbi->s_qf_names[qtype], '/')) {
1429 ext4_msg(sb, KERN_ERR,
1430 "quotafile must be on "
1431 "filesystem root");
1432 kfree(sbi->s_qf_names[qtype]);
1433 sbi->s_qf_names[qtype] = NULL;
1434 return 0;
1435 }
1436 set_opt(sbi->s_mount_opt, QUOTA);
1437 break;
1438 case Opt_offusrjquota:
1439 qtype = USRQUOTA;
1440 goto clear_qf_name;
1441 case Opt_offgrpjquota:
1442 qtype = GRPQUOTA;
1443 clear_qf_name:
1444 if (sb_any_quota_loaded(sb) &&
1445 sbi->s_qf_names[qtype]) {
1446 ext4_msg(sb, KERN_ERR, "Cannot change "
1447 "journaled quota options when "
1448 "quota turned on");
1449 return 0;
1450 }
1451 /*
1452 * The space will be released later when all options
1453 * are confirmed to be correct
1454 */
1455 sbi->s_qf_names[qtype] = NULL;
1456 break;
1457 case Opt_jqfmt_vfsold:
1458 qfmt = QFMT_VFS_OLD;
1459 goto set_qf_format;
1460 case Opt_jqfmt_vfsv0:
1461 qfmt = QFMT_VFS_V0;
1462 set_qf_format:
1463 if (sb_any_quota_loaded(sb) &&
1464 sbi->s_jquota_fmt != qfmt) {
1465 ext4_msg(sb, KERN_ERR, "Cannot change "
1466 "journaled quota options when "
1467 "quota turned on");
1468 return 0;
1469 }
1470 sbi->s_jquota_fmt = qfmt;
1471 break;
1472 case Opt_quota:
1473 case Opt_usrquota:
1474 set_opt(sbi->s_mount_opt, QUOTA);
1475 set_opt(sbi->s_mount_opt, USRQUOTA);
1476 break;
1477 case Opt_grpquota:
1478 set_opt(sbi->s_mount_opt, QUOTA);
1479 set_opt(sbi->s_mount_opt, GRPQUOTA);
1480 break;
1481 case Opt_noquota:
1482 if (sb_any_quota_loaded(sb)) {
1483 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1484 "options when quota turned on");
1485 return 0;
1486 }
1487 clear_opt(sbi->s_mount_opt, QUOTA);
1488 clear_opt(sbi->s_mount_opt, USRQUOTA);
1489 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1490 break;
1491 #else
1492 case Opt_quota:
1493 case Opt_usrquota:
1494 case Opt_grpquota:
1495 ext4_msg(sb, KERN_ERR,
1496 "quota options not supported");
1497 break;
1498 case Opt_usrjquota:
1499 case Opt_grpjquota:
1500 case Opt_offusrjquota:
1501 case Opt_offgrpjquota:
1502 case Opt_jqfmt_vfsold:
1503 case Opt_jqfmt_vfsv0:
1504 ext4_msg(sb, KERN_ERR,
1505 "journaled quota options not supported");
1506 break;
1507 case Opt_noquota:
1508 break;
1509 #endif
1510 case Opt_abort:
1511 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1512 break;
1513 case Opt_nobarrier:
1514 clear_opt(sbi->s_mount_opt, BARRIER);
1515 break;
1516 case Opt_barrier:
1517 if (match_int(&args[0], &option)) {
1518 set_opt(sbi->s_mount_opt, BARRIER);
1519 break;
1520 }
1521 if (option)
1522 set_opt(sbi->s_mount_opt, BARRIER);
1523 else
1524 clear_opt(sbi->s_mount_opt, BARRIER);
1525 break;
1526 case Opt_ignore:
1527 break;
1528 case Opt_resize:
1529 if (!is_remount) {
1530 ext4_msg(sb, KERN_ERR,
1531 "resize option only available "
1532 "for remount");
1533 return 0;
1534 }
1535 if (match_int(&args[0], &option) != 0)
1536 return 0;
1537 *n_blocks_count = option;
1538 break;
1539 case Opt_nobh:
1540 set_opt(sbi->s_mount_opt, NOBH);
1541 break;
1542 case Opt_bh:
1543 clear_opt(sbi->s_mount_opt, NOBH);
1544 break;
1545 case Opt_i_version:
1546 set_opt(sbi->s_mount_opt, I_VERSION);
1547 sb->s_flags |= MS_I_VERSION;
1548 break;
1549 case Opt_nodelalloc:
1550 clear_opt(sbi->s_mount_opt, DELALLOC);
1551 break;
1552 case Opt_stripe:
1553 if (match_int(&args[0], &option))
1554 return 0;
1555 if (option < 0)
1556 return 0;
1557 sbi->s_stripe = option;
1558 break;
1559 case Opt_delalloc:
1560 set_opt(sbi->s_mount_opt, DELALLOC);
1561 break;
1562 case Opt_block_validity:
1563 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1564 break;
1565 case Opt_noblock_validity:
1566 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1567 break;
1568 case Opt_inode_readahead_blks:
1569 if (match_int(&args[0], &option))
1570 return 0;
1571 if (option < 0 || option > (1 << 30))
1572 return 0;
1573 if (!is_power_of_2(option)) {
1574 ext4_msg(sb, KERN_ERR,
1575 "EXT4-fs: inode_readahead_blks"
1576 " must be a power of 2");
1577 return 0;
1578 }
1579 sbi->s_inode_readahead_blks = option;
1580 break;
1581 case Opt_journal_ioprio:
1582 if (match_int(&args[0], &option))
1583 return 0;
1584 if (option < 0 || option > 7)
1585 break;
1586 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1587 option);
1588 break;
1589 case Opt_noauto_da_alloc:
1590 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1591 break;
1592 case Opt_auto_da_alloc:
1593 if (match_int(&args[0], &option)) {
1594 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1595 break;
1596 }
1597 if (option)
1598 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1599 else
1600 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1601 break;
1602 case Opt_discard:
1603 set_opt(sbi->s_mount_opt, DISCARD);
1604 break;
1605 case Opt_nodiscard:
1606 clear_opt(sbi->s_mount_opt, DISCARD);
1607 break;
1608 default:
1609 ext4_msg(sb, KERN_ERR,
1610 "Unrecognized mount option \"%s\" "
1611 "or missing value", p);
1612 return 0;
1613 }
1614 }
1615 #ifdef CONFIG_QUOTA
1616 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1617 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1618 sbi->s_qf_names[USRQUOTA])
1619 clear_opt(sbi->s_mount_opt, USRQUOTA);
1620
1621 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1622 sbi->s_qf_names[GRPQUOTA])
1623 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1624
1625 if ((sbi->s_qf_names[USRQUOTA] &&
1626 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1627 (sbi->s_qf_names[GRPQUOTA] &&
1628 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1629 ext4_msg(sb, KERN_ERR, "old and new quota "
1630 "format mixing");
1631 return 0;
1632 }
1633
1634 if (!sbi->s_jquota_fmt) {
1635 ext4_msg(sb, KERN_ERR, "journaled quota format "
1636 "not specified");
1637 return 0;
1638 }
1639 } else {
1640 if (sbi->s_jquota_fmt) {
1641 ext4_msg(sb, KERN_ERR, "journaled quota format "
1642 "specified with no journaling "
1643 "enabled");
1644 return 0;
1645 }
1646 }
1647 #endif
1648 return 1;
1649 }
1650
1651 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1652 int read_only)
1653 {
1654 struct ext4_sb_info *sbi = EXT4_SB(sb);
1655 int res = 0;
1656
1657 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1658 ext4_msg(sb, KERN_ERR, "revision level too high, "
1659 "forcing read-only mode");
1660 res = MS_RDONLY;
1661 }
1662 if (read_only)
1663 return res;
1664 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1665 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1666 "running e2fsck is recommended");
1667 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1668 ext4_msg(sb, KERN_WARNING,
1669 "warning: mounting fs with errors, "
1670 "running e2fsck is recommended");
1671 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1672 le16_to_cpu(es->s_mnt_count) >=
1673 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1674 ext4_msg(sb, KERN_WARNING,
1675 "warning: maximal mount count reached, "
1676 "running e2fsck is recommended");
1677 else if (le32_to_cpu(es->s_checkinterval) &&
1678 (le32_to_cpu(es->s_lastcheck) +
1679 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1680 ext4_msg(sb, KERN_WARNING,
1681 "warning: checktime reached, "
1682 "running e2fsck is recommended");
1683 if (!sbi->s_journal)
1684 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1685 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1686 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1687 le16_add_cpu(&es->s_mnt_count, 1);
1688 es->s_mtime = cpu_to_le32(get_seconds());
1689 ext4_update_dynamic_rev(sb);
1690 if (sbi->s_journal)
1691 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1692
1693 ext4_commit_super(sb, 1);
1694 if (test_opt(sb, DEBUG))
1695 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1696 "bpg=%lu, ipg=%lu, mo=%04x]\n",
1697 sb->s_blocksize,
1698 sbi->s_groups_count,
1699 EXT4_BLOCKS_PER_GROUP(sb),
1700 EXT4_INODES_PER_GROUP(sb),
1701 sbi->s_mount_opt);
1702
1703 if (EXT4_SB(sb)->s_journal) {
1704 ext4_msg(sb, KERN_INFO, "%s journal on %s",
1705 EXT4_SB(sb)->s_journal->j_inode ? "internal" :
1706 "external", EXT4_SB(sb)->s_journal->j_devname);
1707 } else {
1708 ext4_msg(sb, KERN_INFO, "no journal");
1709 }
1710 return res;
1711 }
1712
1713 static int ext4_fill_flex_info(struct super_block *sb)
1714 {
1715 struct ext4_sb_info *sbi = EXT4_SB(sb);
1716 struct ext4_group_desc *gdp = NULL;
1717 ext4_group_t flex_group_count;
1718 ext4_group_t flex_group;
1719 int groups_per_flex = 0;
1720 size_t size;
1721 int i;
1722
1723 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1724 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1725
1726 if (groups_per_flex < 2) {
1727 sbi->s_log_groups_per_flex = 0;
1728 return 1;
1729 }
1730
1731 /* We allocate both existing and potentially added groups */
1732 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1733 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1734 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1735 size = flex_group_count * sizeof(struct flex_groups);
1736 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1737 if (sbi->s_flex_groups == NULL) {
1738 sbi->s_flex_groups = vmalloc(size);
1739 if (sbi->s_flex_groups)
1740 memset(sbi->s_flex_groups, 0, size);
1741 }
1742 if (sbi->s_flex_groups == NULL) {
1743 ext4_msg(sb, KERN_ERR, "not enough memory for "
1744 "%u flex groups", flex_group_count);
1745 goto failed;
1746 }
1747
1748 for (i = 0; i < sbi->s_groups_count; i++) {
1749 gdp = ext4_get_group_desc(sb, i, NULL);
1750
1751 flex_group = ext4_flex_group(sbi, i);
1752 atomic_add(ext4_free_inodes_count(sb, gdp),
1753 &sbi->s_flex_groups[flex_group].free_inodes);
1754 atomic_add(ext4_free_blks_count(sb, gdp),
1755 &sbi->s_flex_groups[flex_group].free_blocks);
1756 atomic_add(ext4_used_dirs_count(sb, gdp),
1757 &sbi->s_flex_groups[flex_group].used_dirs);
1758 }
1759
1760 return 1;
1761 failed:
1762 return 0;
1763 }
1764
1765 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1766 struct ext4_group_desc *gdp)
1767 {
1768 __u16 crc = 0;
1769
1770 if (sbi->s_es->s_feature_ro_compat &
1771 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1772 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1773 __le32 le_group = cpu_to_le32(block_group);
1774
1775 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1776 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1777 crc = crc16(crc, (__u8 *)gdp, offset);
1778 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1779 /* for checksum of struct ext4_group_desc do the rest...*/
1780 if ((sbi->s_es->s_feature_incompat &
1781 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1782 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1783 crc = crc16(crc, (__u8 *)gdp + offset,
1784 le16_to_cpu(sbi->s_es->s_desc_size) -
1785 offset);
1786 }
1787
1788 return cpu_to_le16(crc);
1789 }
1790
1791 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1792 struct ext4_group_desc *gdp)
1793 {
1794 if ((sbi->s_es->s_feature_ro_compat &
1795 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1796 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1797 return 0;
1798
1799 return 1;
1800 }
1801
1802 /* Called at mount-time, super-block is locked */
1803 static int ext4_check_descriptors(struct super_block *sb)
1804 {
1805 struct ext4_sb_info *sbi = EXT4_SB(sb);
1806 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1807 ext4_fsblk_t last_block;
1808 ext4_fsblk_t block_bitmap;
1809 ext4_fsblk_t inode_bitmap;
1810 ext4_fsblk_t inode_table;
1811 int flexbg_flag = 0;
1812 ext4_group_t i;
1813
1814 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1815 flexbg_flag = 1;
1816
1817 ext4_debug("Checking group descriptors");
1818
1819 for (i = 0; i < sbi->s_groups_count; i++) {
1820 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1821
1822 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1823 last_block = ext4_blocks_count(sbi->s_es) - 1;
1824 else
1825 last_block = first_block +
1826 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1827
1828 block_bitmap = ext4_block_bitmap(sb, gdp);
1829 if (block_bitmap < first_block || block_bitmap > last_block) {
1830 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1831 "Block bitmap for group %u not in group "
1832 "(block %llu)!", i, block_bitmap);
1833 return 0;
1834 }
1835 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1836 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1837 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1838 "Inode bitmap for group %u not in group "
1839 "(block %llu)!", i, inode_bitmap);
1840 return 0;
1841 }
1842 inode_table = ext4_inode_table(sb, gdp);
1843 if (inode_table < first_block ||
1844 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1845 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1846 "Inode table for group %u not in group "
1847 "(block %llu)!", i, inode_table);
1848 return 0;
1849 }
1850 ext4_lock_group(sb, i);
1851 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1852 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1853 "Checksum for group %u failed (%u!=%u)",
1854 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1855 gdp)), le16_to_cpu(gdp->bg_checksum));
1856 if (!(sb->s_flags & MS_RDONLY)) {
1857 ext4_unlock_group(sb, i);
1858 return 0;
1859 }
1860 }
1861 ext4_unlock_group(sb, i);
1862 if (!flexbg_flag)
1863 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1864 }
1865
1866 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1867 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
1868 return 1;
1869 }
1870
1871 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1872 * the superblock) which were deleted from all directories, but held open by
1873 * a process at the time of a crash. We walk the list and try to delete these
1874 * inodes at recovery time (only with a read-write filesystem).
1875 *
1876 * In order to keep the orphan inode chain consistent during traversal (in
1877 * case of crash during recovery), we link each inode into the superblock
1878 * orphan list_head and handle it the same way as an inode deletion during
1879 * normal operation (which journals the operations for us).
1880 *
1881 * We only do an iget() and an iput() on each inode, which is very safe if we
1882 * accidentally point at an in-use or already deleted inode. The worst that
1883 * can happen in this case is that we get a "bit already cleared" message from
1884 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1885 * e2fsck was run on this filesystem, and it must have already done the orphan
1886 * inode cleanup for us, so we can safely abort without any further action.
1887 */
1888 static void ext4_orphan_cleanup(struct super_block *sb,
1889 struct ext4_super_block *es)
1890 {
1891 unsigned int s_flags = sb->s_flags;
1892 int nr_orphans = 0, nr_truncates = 0;
1893 #ifdef CONFIG_QUOTA
1894 int i;
1895 #endif
1896 if (!es->s_last_orphan) {
1897 jbd_debug(4, "no orphan inodes to clean up\n");
1898 return;
1899 }
1900
1901 if (bdev_read_only(sb->s_bdev)) {
1902 ext4_msg(sb, KERN_ERR, "write access "
1903 "unavailable, skipping orphan cleanup");
1904 return;
1905 }
1906
1907 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1908 if (es->s_last_orphan)
1909 jbd_debug(1, "Errors on filesystem, "
1910 "clearing orphan list.\n");
1911 es->s_last_orphan = 0;
1912 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1913 return;
1914 }
1915
1916 if (s_flags & MS_RDONLY) {
1917 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1918 sb->s_flags &= ~MS_RDONLY;
1919 }
1920 #ifdef CONFIG_QUOTA
1921 /* Needed for iput() to work correctly and not trash data */
1922 sb->s_flags |= MS_ACTIVE;
1923 /* Turn on quotas so that they are updated correctly */
1924 for (i = 0; i < MAXQUOTAS; i++) {
1925 if (EXT4_SB(sb)->s_qf_names[i]) {
1926 int ret = ext4_quota_on_mount(sb, i);
1927 if (ret < 0)
1928 ext4_msg(sb, KERN_ERR,
1929 "Cannot turn on journaled "
1930 "quota: error %d", ret);
1931 }
1932 }
1933 #endif
1934
1935 while (es->s_last_orphan) {
1936 struct inode *inode;
1937
1938 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1939 if (IS_ERR(inode)) {
1940 es->s_last_orphan = 0;
1941 break;
1942 }
1943
1944 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1945 vfs_dq_init(inode);
1946 if (inode->i_nlink) {
1947 ext4_msg(sb, KERN_DEBUG,
1948 "%s: truncating inode %lu to %lld bytes",
1949 __func__, inode->i_ino, inode->i_size);
1950 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1951 inode->i_ino, inode->i_size);
1952 ext4_truncate(inode);
1953 nr_truncates++;
1954 } else {
1955 ext4_msg(sb, KERN_DEBUG,
1956 "%s: deleting unreferenced inode %lu",
1957 __func__, inode->i_ino);
1958 jbd_debug(2, "deleting unreferenced inode %lu\n",
1959 inode->i_ino);
1960 nr_orphans++;
1961 }
1962 iput(inode); /* The delete magic happens here! */
1963 }
1964
1965 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1966
1967 if (nr_orphans)
1968 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1969 PLURAL(nr_orphans));
1970 if (nr_truncates)
1971 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1972 PLURAL(nr_truncates));
1973 #ifdef CONFIG_QUOTA
1974 /* Turn quotas off */
1975 for (i = 0; i < MAXQUOTAS; i++) {
1976 if (sb_dqopt(sb)->files[i])
1977 vfs_quota_off(sb, i, 0);
1978 }
1979 #endif
1980 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1981 }
1982
1983 /*
1984 * Maximal extent format file size.
1985 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1986 * extent format containers, within a sector_t, and within i_blocks
1987 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1988 * so that won't be a limiting factor.
1989 *
1990 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1991 */
1992 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1993 {
1994 loff_t res;
1995 loff_t upper_limit = MAX_LFS_FILESIZE;
1996
1997 /* small i_blocks in vfs inode? */
1998 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1999 /*
2000 * CONFIG_LBDAF is not enabled implies the inode
2001 * i_block represent total blocks in 512 bytes
2002 * 32 == size of vfs inode i_blocks * 8
2003 */
2004 upper_limit = (1LL << 32) - 1;
2005
2006 /* total blocks in file system block size */
2007 upper_limit >>= (blkbits - 9);
2008 upper_limit <<= blkbits;
2009 }
2010
2011 /* 32-bit extent-start container, ee_block */
2012 res = 1LL << 32;
2013 res <<= blkbits;
2014 res -= 1;
2015
2016 /* Sanity check against vm- & vfs- imposed limits */
2017 if (res > upper_limit)
2018 res = upper_limit;
2019
2020 return res;
2021 }
2022
2023 /*
2024 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2025 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2026 * We need to be 1 filesystem block less than the 2^48 sector limit.
2027 */
2028 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2029 {
2030 loff_t res = EXT4_NDIR_BLOCKS;
2031 int meta_blocks;
2032 loff_t upper_limit;
2033 /* This is calculated to be the largest file size for a dense, block
2034 * mapped file such that the file's total number of 512-byte sectors,
2035 * including data and all indirect blocks, does not exceed (2^48 - 1).
2036 *
2037 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2038 * number of 512-byte sectors of the file.
2039 */
2040
2041 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2042 /*
2043 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2044 * the inode i_block field represents total file blocks in
2045 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2046 */
2047 upper_limit = (1LL << 32) - 1;
2048
2049 /* total blocks in file system block size */
2050 upper_limit >>= (bits - 9);
2051
2052 } else {
2053 /*
2054 * We use 48 bit ext4_inode i_blocks
2055 * With EXT4_HUGE_FILE_FL set the i_blocks
2056 * represent total number of blocks in
2057 * file system block size
2058 */
2059 upper_limit = (1LL << 48) - 1;
2060
2061 }
2062
2063 /* indirect blocks */
2064 meta_blocks = 1;
2065 /* double indirect blocks */
2066 meta_blocks += 1 + (1LL << (bits-2));
2067 /* tripple indirect blocks */
2068 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2069
2070 upper_limit -= meta_blocks;
2071 upper_limit <<= bits;
2072
2073 res += 1LL << (bits-2);
2074 res += 1LL << (2*(bits-2));
2075 res += 1LL << (3*(bits-2));
2076 res <<= bits;
2077 if (res > upper_limit)
2078 res = upper_limit;
2079
2080 if (res > MAX_LFS_FILESIZE)
2081 res = MAX_LFS_FILESIZE;
2082
2083 return res;
2084 }
2085
2086 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2087 ext4_fsblk_t logical_sb_block, int nr)
2088 {
2089 struct ext4_sb_info *sbi = EXT4_SB(sb);
2090 ext4_group_t bg, first_meta_bg;
2091 int has_super = 0;
2092
2093 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2094
2095 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2096 nr < first_meta_bg)
2097 return logical_sb_block + nr + 1;
2098 bg = sbi->s_desc_per_block * nr;
2099 if (ext4_bg_has_super(sb, bg))
2100 has_super = 1;
2101
2102 return (has_super + ext4_group_first_block_no(sb, bg));
2103 }
2104
2105 /**
2106 * ext4_get_stripe_size: Get the stripe size.
2107 * @sbi: In memory super block info
2108 *
2109 * If we have specified it via mount option, then
2110 * use the mount option value. If the value specified at mount time is
2111 * greater than the blocks per group use the super block value.
2112 * If the super block value is greater than blocks per group return 0.
2113 * Allocator needs it be less than blocks per group.
2114 *
2115 */
2116 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2117 {
2118 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2119 unsigned long stripe_width =
2120 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2121
2122 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2123 return sbi->s_stripe;
2124
2125 if (stripe_width <= sbi->s_blocks_per_group)
2126 return stripe_width;
2127
2128 if (stride <= sbi->s_blocks_per_group)
2129 return stride;
2130
2131 return 0;
2132 }
2133
2134 /* sysfs supprt */
2135
2136 struct ext4_attr {
2137 struct attribute attr;
2138 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2139 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2140 const char *, size_t);
2141 int offset;
2142 };
2143
2144 static int parse_strtoul(const char *buf,
2145 unsigned long max, unsigned long *value)
2146 {
2147 char *endp;
2148
2149 while (*buf && isspace(*buf))
2150 buf++;
2151 *value = simple_strtoul(buf, &endp, 0);
2152 while (*endp && isspace(*endp))
2153 endp++;
2154 if (*endp || *value > max)
2155 return -EINVAL;
2156
2157 return 0;
2158 }
2159
2160 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2161 struct ext4_sb_info *sbi,
2162 char *buf)
2163 {
2164 return snprintf(buf, PAGE_SIZE, "%llu\n",
2165 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2166 }
2167
2168 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2169 struct ext4_sb_info *sbi, char *buf)
2170 {
2171 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2172
2173 return snprintf(buf, PAGE_SIZE, "%lu\n",
2174 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2175 sbi->s_sectors_written_start) >> 1);
2176 }
2177
2178 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2179 struct ext4_sb_info *sbi, char *buf)
2180 {
2181 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2182
2183 return snprintf(buf, PAGE_SIZE, "%llu\n",
2184 sbi->s_kbytes_written +
2185 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2186 EXT4_SB(sb)->s_sectors_written_start) >> 1));
2187 }
2188
2189 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2190 struct ext4_sb_info *sbi,
2191 const char *buf, size_t count)
2192 {
2193 unsigned long t;
2194
2195 if (parse_strtoul(buf, 0x40000000, &t))
2196 return -EINVAL;
2197
2198 if (!is_power_of_2(t))
2199 return -EINVAL;
2200
2201 sbi->s_inode_readahead_blks = t;
2202 return count;
2203 }
2204
2205 static ssize_t sbi_ui_show(struct ext4_attr *a,
2206 struct ext4_sb_info *sbi, char *buf)
2207 {
2208 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2209
2210 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2211 }
2212
2213 static ssize_t sbi_ui_store(struct ext4_attr *a,
2214 struct ext4_sb_info *sbi,
2215 const char *buf, size_t count)
2216 {
2217 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2218 unsigned long t;
2219
2220 if (parse_strtoul(buf, 0xffffffff, &t))
2221 return -EINVAL;
2222 *ui = t;
2223 return count;
2224 }
2225
2226 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2227 static struct ext4_attr ext4_attr_##_name = { \
2228 .attr = {.name = __stringify(_name), .mode = _mode }, \
2229 .show = _show, \
2230 .store = _store, \
2231 .offset = offsetof(struct ext4_sb_info, _elname), \
2232 }
2233 #define EXT4_ATTR(name, mode, show, store) \
2234 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2235
2236 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2237 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2238 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2239 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2240 #define ATTR_LIST(name) &ext4_attr_##name.attr
2241
2242 EXT4_RO_ATTR(delayed_allocation_blocks);
2243 EXT4_RO_ATTR(session_write_kbytes);
2244 EXT4_RO_ATTR(lifetime_write_kbytes);
2245 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2246 inode_readahead_blks_store, s_inode_readahead_blks);
2247 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2248 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2249 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2250 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2251 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2252 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2253 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2254 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2255
2256 static struct attribute *ext4_attrs[] = {
2257 ATTR_LIST(delayed_allocation_blocks),
2258 ATTR_LIST(session_write_kbytes),
2259 ATTR_LIST(lifetime_write_kbytes),
2260 ATTR_LIST(inode_readahead_blks),
2261 ATTR_LIST(inode_goal),
2262 ATTR_LIST(mb_stats),
2263 ATTR_LIST(mb_max_to_scan),
2264 ATTR_LIST(mb_min_to_scan),
2265 ATTR_LIST(mb_order2_req),
2266 ATTR_LIST(mb_stream_req),
2267 ATTR_LIST(mb_group_prealloc),
2268 ATTR_LIST(max_writeback_mb_bump),
2269 NULL,
2270 };
2271
2272 static ssize_t ext4_attr_show(struct kobject *kobj,
2273 struct attribute *attr, char *buf)
2274 {
2275 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2276 s_kobj);
2277 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2278
2279 return a->show ? a->show(a, sbi, buf) : 0;
2280 }
2281
2282 static ssize_t ext4_attr_store(struct kobject *kobj,
2283 struct attribute *attr,
2284 const char *buf, size_t len)
2285 {
2286 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2287 s_kobj);
2288 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2289
2290 return a->store ? a->store(a, sbi, buf, len) : 0;
2291 }
2292
2293 static void ext4_sb_release(struct kobject *kobj)
2294 {
2295 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2296 s_kobj);
2297 complete(&sbi->s_kobj_unregister);
2298 }
2299
2300
2301 static struct sysfs_ops ext4_attr_ops = {
2302 .show = ext4_attr_show,
2303 .store = ext4_attr_store,
2304 };
2305
2306 static struct kobj_type ext4_ktype = {
2307 .default_attrs = ext4_attrs,
2308 .sysfs_ops = &ext4_attr_ops,
2309 .release = ext4_sb_release,
2310 };
2311
2312 /*
2313 * Check whether this filesystem can be mounted based on
2314 * the features present and the RDONLY/RDWR mount requested.
2315 * Returns 1 if this filesystem can be mounted as requested,
2316 * 0 if it cannot be.
2317 */
2318 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2319 {
2320 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2321 ext4_msg(sb, KERN_ERR,
2322 "Couldn't mount because of "
2323 "unsupported optional features (%x)",
2324 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2325 ~EXT4_FEATURE_INCOMPAT_SUPP));
2326 return 0;
2327 }
2328
2329 if (readonly)
2330 return 1;
2331
2332 /* Check that feature set is OK for a read-write mount */
2333 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2334 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2335 "unsupported optional features (%x)",
2336 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2337 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2338 return 0;
2339 }
2340 /*
2341 * Large file size enabled file system can only be mounted
2342 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2343 */
2344 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2345 if (sizeof(blkcnt_t) < sizeof(u64)) {
2346 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2347 "cannot be mounted RDWR without "
2348 "CONFIG_LBDAF");
2349 return 0;
2350 }
2351 }
2352 return 1;
2353 }
2354
2355 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2356 __releases(kernel_lock)
2357 __acquires(kernel_lock)
2358 {
2359 struct buffer_head *bh;
2360 struct ext4_super_block *es = NULL;
2361 struct ext4_sb_info *sbi;
2362 ext4_fsblk_t block;
2363 ext4_fsblk_t sb_block = get_sb_block(&data);
2364 ext4_fsblk_t logical_sb_block;
2365 unsigned long offset = 0;
2366 unsigned long journal_devnum = 0;
2367 unsigned long def_mount_opts;
2368 struct inode *root;
2369 char *cp;
2370 const char *descr;
2371 int ret = -EINVAL;
2372 int blocksize;
2373 unsigned int db_count;
2374 unsigned int i;
2375 int needs_recovery, has_huge_files;
2376 __u64 blocks_count;
2377 int err;
2378 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2379
2380 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2381 if (!sbi)
2382 return -ENOMEM;
2383
2384 sbi->s_blockgroup_lock =
2385 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2386 if (!sbi->s_blockgroup_lock) {
2387 kfree(sbi);
2388 return -ENOMEM;
2389 }
2390 sb->s_fs_info = sbi;
2391 sbi->s_mount_opt = 0;
2392 sbi->s_resuid = EXT4_DEF_RESUID;
2393 sbi->s_resgid = EXT4_DEF_RESGID;
2394 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2395 sbi->s_sb_block = sb_block;
2396 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2397 sectors[1]);
2398
2399 unlock_kernel();
2400
2401 /* Cleanup superblock name */
2402 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2403 *cp = '!';
2404
2405 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2406 if (!blocksize) {
2407 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
2408 goto out_fail;
2409 }
2410
2411 /*
2412 * The ext4 superblock will not be buffer aligned for other than 1kB
2413 * block sizes. We need to calculate the offset from buffer start.
2414 */
2415 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2416 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2417 offset = do_div(logical_sb_block, blocksize);
2418 } else {
2419 logical_sb_block = sb_block;
2420 }
2421
2422 if (!(bh = sb_bread(sb, logical_sb_block))) {
2423 ext4_msg(sb, KERN_ERR, "unable to read superblock");
2424 goto out_fail;
2425 }
2426 /*
2427 * Note: s_es must be initialized as soon as possible because
2428 * some ext4 macro-instructions depend on its value
2429 */
2430 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2431 sbi->s_es = es;
2432 sb->s_magic = le16_to_cpu(es->s_magic);
2433 if (sb->s_magic != EXT4_SUPER_MAGIC)
2434 goto cantfind_ext4;
2435 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2436
2437 /* Set defaults before we parse the mount options */
2438 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2439 if (def_mount_opts & EXT4_DEFM_DEBUG)
2440 set_opt(sbi->s_mount_opt, DEBUG);
2441 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
2442 set_opt(sbi->s_mount_opt, GRPID);
2443 if (def_mount_opts & EXT4_DEFM_UID16)
2444 set_opt(sbi->s_mount_opt, NO_UID32);
2445 #ifdef CONFIG_EXT4_FS_XATTR
2446 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2447 set_opt(sbi->s_mount_opt, XATTR_USER);
2448 #endif
2449 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2450 if (def_mount_opts & EXT4_DEFM_ACL)
2451 set_opt(sbi->s_mount_opt, POSIX_ACL);
2452 #endif
2453 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2454 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2455 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2456 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2457 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2458 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2459
2460 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2461 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2462 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2463 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2464 else
2465 set_opt(sbi->s_mount_opt, ERRORS_RO);
2466
2467 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2468 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2469 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2470 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2471 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2472 sbi->s_mb_history_max = default_mb_history_length;
2473
2474 set_opt(sbi->s_mount_opt, BARRIER);
2475
2476 /*
2477 * enable delayed allocation by default
2478 * Use -o nodelalloc to turn it off
2479 */
2480 set_opt(sbi->s_mount_opt, DELALLOC);
2481
2482 if (!parse_options((char *) data, sb, &journal_devnum,
2483 &journal_ioprio, NULL, 0))
2484 goto failed_mount;
2485
2486 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2487 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2488
2489 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2490 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2491 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2492 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2493 ext4_msg(sb, KERN_WARNING,
2494 "feature flags set on rev 0 fs, "
2495 "running e2fsck is recommended");
2496
2497 /*
2498 * Check feature flags regardless of the revision level, since we
2499 * previously didn't change the revision level when setting the flags,
2500 * so there is a chance incompat flags are set on a rev 0 filesystem.
2501 */
2502 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
2503 goto failed_mount;
2504
2505 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2506
2507 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2508 blocksize > EXT4_MAX_BLOCK_SIZE) {
2509 ext4_msg(sb, KERN_ERR,
2510 "Unsupported filesystem blocksize %d", blocksize);
2511 goto failed_mount;
2512 }
2513
2514 if (sb->s_blocksize != blocksize) {
2515 /* Validate the filesystem blocksize */
2516 if (!sb_set_blocksize(sb, blocksize)) {
2517 ext4_msg(sb, KERN_ERR, "bad block size %d",
2518 blocksize);
2519 goto failed_mount;
2520 }
2521
2522 brelse(bh);
2523 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2524 offset = do_div(logical_sb_block, blocksize);
2525 bh = sb_bread(sb, logical_sb_block);
2526 if (!bh) {
2527 ext4_msg(sb, KERN_ERR,
2528 "Can't read superblock on 2nd try");
2529 goto failed_mount;
2530 }
2531 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2532 sbi->s_es = es;
2533 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2534 ext4_msg(sb, KERN_ERR,
2535 "Magic mismatch, very weird!");
2536 goto failed_mount;
2537 }
2538 }
2539
2540 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2541 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2542 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2543 has_huge_files);
2544 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2545
2546 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2547 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2548 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2549 } else {
2550 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2551 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2552 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2553 (!is_power_of_2(sbi->s_inode_size)) ||
2554 (sbi->s_inode_size > blocksize)) {
2555 ext4_msg(sb, KERN_ERR,
2556 "unsupported inode size: %d",
2557 sbi->s_inode_size);
2558 goto failed_mount;
2559 }
2560 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2561 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2562 }
2563
2564 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2565 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2566 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2567 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2568 !is_power_of_2(sbi->s_desc_size)) {
2569 ext4_msg(sb, KERN_ERR,
2570 "unsupported descriptor size %lu",
2571 sbi->s_desc_size);
2572 goto failed_mount;
2573 }
2574 } else
2575 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2576
2577 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2578 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2579 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2580 goto cantfind_ext4;
2581
2582 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2583 if (sbi->s_inodes_per_block == 0)
2584 goto cantfind_ext4;
2585 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2586 sbi->s_inodes_per_block;
2587 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2588 sbi->s_sbh = bh;
2589 sbi->s_mount_state = le16_to_cpu(es->s_state);
2590 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2591 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2592
2593 for (i = 0; i < 4; i++)
2594 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2595 sbi->s_def_hash_version = es->s_def_hash_version;
2596 i = le32_to_cpu(es->s_flags);
2597 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2598 sbi->s_hash_unsigned = 3;
2599 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2600 #ifdef __CHAR_UNSIGNED__
2601 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2602 sbi->s_hash_unsigned = 3;
2603 #else
2604 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2605 #endif
2606 sb->s_dirt = 1;
2607 }
2608
2609 if (sbi->s_blocks_per_group > blocksize * 8) {
2610 ext4_msg(sb, KERN_ERR,
2611 "#blocks per group too big: %lu",
2612 sbi->s_blocks_per_group);
2613 goto failed_mount;
2614 }
2615 if (sbi->s_inodes_per_group > blocksize * 8) {
2616 ext4_msg(sb, KERN_ERR,
2617 "#inodes per group too big: %lu",
2618 sbi->s_inodes_per_group);
2619 goto failed_mount;
2620 }
2621
2622 /*
2623 * Test whether we have more sectors than will fit in sector_t,
2624 * and whether the max offset is addressable by the page cache.
2625 */
2626 if ((ext4_blocks_count(es) >
2627 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2628 (ext4_blocks_count(es) >
2629 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
2630 ext4_msg(sb, KERN_ERR, "filesystem"
2631 " too large to mount safely on this system");
2632 if (sizeof(sector_t) < 8)
2633 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
2634 ret = -EFBIG;
2635 goto failed_mount;
2636 }
2637
2638 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2639 goto cantfind_ext4;
2640
2641 /* check blocks count against device size */
2642 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2643 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2644 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2645 "exceeds size of device (%llu blocks)",
2646 ext4_blocks_count(es), blocks_count);
2647 goto failed_mount;
2648 }
2649
2650 /*
2651 * It makes no sense for the first data block to be beyond the end
2652 * of the filesystem.
2653 */
2654 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2655 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2656 "block %u is beyond end of filesystem (%llu)",
2657 le32_to_cpu(es->s_first_data_block),
2658 ext4_blocks_count(es));
2659 goto failed_mount;
2660 }
2661 blocks_count = (ext4_blocks_count(es) -
2662 le32_to_cpu(es->s_first_data_block) +
2663 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2664 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2665 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2666 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
2667 "(block count %llu, first data block %u, "
2668 "blocks per group %lu)", sbi->s_groups_count,
2669 ext4_blocks_count(es),
2670 le32_to_cpu(es->s_first_data_block),
2671 EXT4_BLOCKS_PER_GROUP(sb));
2672 goto failed_mount;
2673 }
2674 sbi->s_groups_count = blocks_count;
2675 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
2676 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
2677 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2678 EXT4_DESC_PER_BLOCK(sb);
2679 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2680 GFP_KERNEL);
2681 if (sbi->s_group_desc == NULL) {
2682 ext4_msg(sb, KERN_ERR, "not enough memory");
2683 goto failed_mount;
2684 }
2685
2686 #ifdef CONFIG_PROC_FS
2687 if (ext4_proc_root)
2688 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2689 #endif
2690
2691 bgl_lock_init(sbi->s_blockgroup_lock);
2692
2693 for (i = 0; i < db_count; i++) {
2694 block = descriptor_loc(sb, logical_sb_block, i);
2695 sbi->s_group_desc[i] = sb_bread(sb, block);
2696 if (!sbi->s_group_desc[i]) {
2697 ext4_msg(sb, KERN_ERR,
2698 "can't read group descriptor %d", i);
2699 db_count = i;
2700 goto failed_mount2;
2701 }
2702 }
2703 if (!ext4_check_descriptors(sb)) {
2704 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
2705 goto failed_mount2;
2706 }
2707 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2708 if (!ext4_fill_flex_info(sb)) {
2709 ext4_msg(sb, KERN_ERR,
2710 "unable to initialize "
2711 "flex_bg meta info!");
2712 goto failed_mount2;
2713 }
2714
2715 sbi->s_gdb_count = db_count;
2716 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2717 spin_lock_init(&sbi->s_next_gen_lock);
2718
2719 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2720 ext4_count_free_blocks(sb));
2721 if (!err) {
2722 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2723 ext4_count_free_inodes(sb));
2724 }
2725 if (!err) {
2726 err = percpu_counter_init(&sbi->s_dirs_counter,
2727 ext4_count_dirs(sb));
2728 }
2729 if (!err) {
2730 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2731 }
2732 if (err) {
2733 ext4_msg(sb, KERN_ERR, "insufficient memory");
2734 goto failed_mount3;
2735 }
2736
2737 sbi->s_stripe = ext4_get_stripe_size(sbi);
2738 sbi->s_max_writeback_mb_bump = 128;
2739
2740 /*
2741 * set up enough so that it can read an inode
2742 */
2743 if (!test_opt(sb, NOLOAD) &&
2744 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2745 sb->s_op = &ext4_sops;
2746 else
2747 sb->s_op = &ext4_nojournal_sops;
2748 sb->s_export_op = &ext4_export_ops;
2749 sb->s_xattr = ext4_xattr_handlers;
2750 #ifdef CONFIG_QUOTA
2751 sb->s_qcop = &ext4_qctl_operations;
2752 sb->dq_op = &ext4_quota_operations;
2753 #endif
2754 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2755 mutex_init(&sbi->s_orphan_lock);
2756 mutex_init(&sbi->s_resize_lock);
2757
2758 sb->s_root = NULL;
2759
2760 needs_recovery = (es->s_last_orphan != 0 ||
2761 EXT4_HAS_INCOMPAT_FEATURE(sb,
2762 EXT4_FEATURE_INCOMPAT_RECOVER));
2763
2764 /*
2765 * The first inode we look at is the journal inode. Don't try
2766 * root first: it may be modified in the journal!
2767 */
2768 if (!test_opt(sb, NOLOAD) &&
2769 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2770 if (ext4_load_journal(sb, es, journal_devnum))
2771 goto failed_mount3;
2772 if (!(sb->s_flags & MS_RDONLY) &&
2773 EXT4_SB(sb)->s_journal->j_failed_commit) {
2774 ext4_msg(sb, KERN_CRIT, "error: "
2775 "ext4_fill_super: Journal transaction "
2776 "%u is corrupt",
2777 EXT4_SB(sb)->s_journal->j_failed_commit);
2778 if (test_opt(sb, ERRORS_RO)) {
2779 ext4_msg(sb, KERN_CRIT,
2780 "Mounting filesystem read-only");
2781 sb->s_flags |= MS_RDONLY;
2782 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2783 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2784 }
2785 if (test_opt(sb, ERRORS_PANIC)) {
2786 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2787 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2788 ext4_commit_super(sb, 1);
2789 goto failed_mount4;
2790 }
2791 }
2792 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2793 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2794 ext4_msg(sb, KERN_ERR, "required journal recovery "
2795 "suppressed and not mounted read-only");
2796 goto failed_mount4;
2797 } else {
2798 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2799 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2800 sbi->s_journal = NULL;
2801 needs_recovery = 0;
2802 goto no_journal;
2803 }
2804
2805 if (ext4_blocks_count(es) > 0xffffffffULL &&
2806 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2807 JBD2_FEATURE_INCOMPAT_64BIT)) {
2808 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
2809 goto failed_mount4;
2810 }
2811
2812 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2813 jbd2_journal_set_features(sbi->s_journal,
2814 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2815 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2816 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2817 jbd2_journal_set_features(sbi->s_journal,
2818 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2819 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2820 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2821 } else {
2822 jbd2_journal_clear_features(sbi->s_journal,
2823 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2824 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2825 }
2826
2827 /* We have now updated the journal if required, so we can
2828 * validate the data journaling mode. */
2829 switch (test_opt(sb, DATA_FLAGS)) {
2830 case 0:
2831 /* No mode set, assume a default based on the journal
2832 * capabilities: ORDERED_DATA if the journal can
2833 * cope, else JOURNAL_DATA
2834 */
2835 if (jbd2_journal_check_available_features
2836 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2837 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2838 else
2839 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2840 break;
2841
2842 case EXT4_MOUNT_ORDERED_DATA:
2843 case EXT4_MOUNT_WRITEBACK_DATA:
2844 if (!jbd2_journal_check_available_features
2845 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2846 ext4_msg(sb, KERN_ERR, "Journal does not support "
2847 "requested data journaling mode");
2848 goto failed_mount4;
2849 }
2850 default:
2851 break;
2852 }
2853 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2854
2855 no_journal:
2856
2857 if (test_opt(sb, NOBH)) {
2858 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2859 ext4_msg(sb, KERN_WARNING, "Ignoring nobh option - "
2860 "its supported only with writeback mode");
2861 clear_opt(sbi->s_mount_opt, NOBH);
2862 }
2863 }
2864 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
2865 if (!EXT4_SB(sb)->dio_unwritten_wq) {
2866 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
2867 goto failed_mount_wq;
2868 }
2869
2870 /*
2871 * The jbd2_journal_load will have done any necessary log recovery,
2872 * so we can safely mount the rest of the filesystem now.
2873 */
2874
2875 root = ext4_iget(sb, EXT4_ROOT_INO);
2876 if (IS_ERR(root)) {
2877 ext4_msg(sb, KERN_ERR, "get root inode failed");
2878 ret = PTR_ERR(root);
2879 goto failed_mount4;
2880 }
2881 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2882 iput(root);
2883 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
2884 goto failed_mount4;
2885 }
2886 sb->s_root = d_alloc_root(root);
2887 if (!sb->s_root) {
2888 ext4_msg(sb, KERN_ERR, "get root dentry failed");
2889 iput(root);
2890 ret = -ENOMEM;
2891 goto failed_mount4;
2892 }
2893
2894 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2895
2896 /* determine the minimum size of new large inodes, if present */
2897 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2898 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2899 EXT4_GOOD_OLD_INODE_SIZE;
2900 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2901 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2902 if (sbi->s_want_extra_isize <
2903 le16_to_cpu(es->s_want_extra_isize))
2904 sbi->s_want_extra_isize =
2905 le16_to_cpu(es->s_want_extra_isize);
2906 if (sbi->s_want_extra_isize <
2907 le16_to_cpu(es->s_min_extra_isize))
2908 sbi->s_want_extra_isize =
2909 le16_to_cpu(es->s_min_extra_isize);
2910 }
2911 }
2912 /* Check if enough inode space is available */
2913 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2914 sbi->s_inode_size) {
2915 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2916 EXT4_GOOD_OLD_INODE_SIZE;
2917 ext4_msg(sb, KERN_INFO, "required extra inode space not"
2918 "available");
2919 }
2920
2921 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2922 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
2923 "requested data journaling mode");
2924 clear_opt(sbi->s_mount_opt, DELALLOC);
2925 } else if (test_opt(sb, DELALLOC))
2926 ext4_msg(sb, KERN_INFO, "delayed allocation enabled");
2927
2928 err = ext4_setup_system_zone(sb);
2929 if (err) {
2930 ext4_msg(sb, KERN_ERR, "failed to initialize system "
2931 "zone (%d)\n", err);
2932 goto failed_mount4;
2933 }
2934
2935 ext4_ext_init(sb);
2936 err = ext4_mb_init(sb, needs_recovery);
2937 if (err) {
2938 ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
2939 err);
2940 goto failed_mount4;
2941 }
2942
2943 sbi->s_kobj.kset = ext4_kset;
2944 init_completion(&sbi->s_kobj_unregister);
2945 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2946 "%s", sb->s_id);
2947 if (err) {
2948 ext4_mb_release(sb);
2949 ext4_ext_release(sb);
2950 goto failed_mount4;
2951 };
2952
2953 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2954 ext4_orphan_cleanup(sb, es);
2955 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2956 if (needs_recovery) {
2957 ext4_msg(sb, KERN_INFO, "recovery complete");
2958 ext4_mark_recovery_complete(sb, es);
2959 }
2960 if (EXT4_SB(sb)->s_journal) {
2961 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2962 descr = " journalled data mode";
2963 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2964 descr = " ordered data mode";
2965 else
2966 descr = " writeback data mode";
2967 } else
2968 descr = "out journal";
2969
2970 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s", descr);
2971
2972 lock_kernel();
2973 return 0;
2974
2975 cantfind_ext4:
2976 if (!silent)
2977 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
2978 goto failed_mount;
2979
2980 failed_mount4:
2981 ext4_msg(sb, KERN_ERR, "mount failed");
2982 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
2983 failed_mount_wq:
2984 ext4_release_system_zone(sb);
2985 if (sbi->s_journal) {
2986 jbd2_journal_destroy(sbi->s_journal);
2987 sbi->s_journal = NULL;
2988 }
2989 failed_mount3:
2990 if (sbi->s_flex_groups) {
2991 if (is_vmalloc_addr(sbi->s_flex_groups))
2992 vfree(sbi->s_flex_groups);
2993 else
2994 kfree(sbi->s_flex_groups);
2995 }
2996 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2997 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2998 percpu_counter_destroy(&sbi->s_dirs_counter);
2999 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
3000 failed_mount2:
3001 for (i = 0; i < db_count; i++)
3002 brelse(sbi->s_group_desc[i]);
3003 kfree(sbi->s_group_desc);
3004 failed_mount:
3005 if (sbi->s_proc) {
3006 remove_proc_entry(sb->s_id, ext4_proc_root);
3007 }
3008 #ifdef CONFIG_QUOTA
3009 for (i = 0; i < MAXQUOTAS; i++)
3010 kfree(sbi->s_qf_names[i]);
3011 #endif
3012 ext4_blkdev_remove(sbi);
3013 brelse(bh);
3014 out_fail:
3015 sb->s_fs_info = NULL;
3016 kfree(sbi->s_blockgroup_lock);
3017 kfree(sbi);
3018 lock_kernel();
3019 return ret;
3020 }
3021
3022 /*
3023 * Setup any per-fs journal parameters now. We'll do this both on
3024 * initial mount, once the journal has been initialised but before we've
3025 * done any recovery; and again on any subsequent remount.
3026 */
3027 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3028 {
3029 struct ext4_sb_info *sbi = EXT4_SB(sb);
3030
3031 journal->j_commit_interval = sbi->s_commit_interval;
3032 journal->j_min_batch_time = sbi->s_min_batch_time;
3033 journal->j_max_batch_time = sbi->s_max_batch_time;
3034
3035 spin_lock(&journal->j_state_lock);
3036 if (test_opt(sb, BARRIER))
3037 journal->j_flags |= JBD2_BARRIER;
3038 else
3039 journal->j_flags &= ~JBD2_BARRIER;
3040 if (test_opt(sb, DATA_ERR_ABORT))
3041 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3042 else
3043 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3044 spin_unlock(&journal->j_state_lock);
3045 }
3046
3047 static journal_t *ext4_get_journal(struct super_block *sb,
3048 unsigned int journal_inum)
3049 {
3050 struct inode *journal_inode;
3051 journal_t *journal;
3052
3053 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3054
3055 /* First, test for the existence of a valid inode on disk. Bad
3056 * things happen if we iget() an unused inode, as the subsequent
3057 * iput() will try to delete it. */
3058
3059 journal_inode = ext4_iget(sb, journal_inum);
3060 if (IS_ERR(journal_inode)) {
3061 ext4_msg(sb, KERN_ERR, "no journal found");
3062 return NULL;
3063 }
3064 if (!journal_inode->i_nlink) {
3065 make_bad_inode(journal_inode);
3066 iput(journal_inode);
3067 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3068 return NULL;
3069 }
3070
3071 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3072 journal_inode, journal_inode->i_size);
3073 if (!S_ISREG(journal_inode->i_mode)) {
3074 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3075 iput(journal_inode);
3076 return NULL;
3077 }
3078
3079 journal = jbd2_journal_init_inode(journal_inode);
3080 if (!journal) {
3081 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3082 iput(journal_inode);
3083 return NULL;
3084 }
3085 journal->j_private = sb;
3086 ext4_init_journal_params(sb, journal);
3087 return journal;
3088 }
3089
3090 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3091 dev_t j_dev)
3092 {
3093 struct buffer_head *bh;
3094 journal_t *journal;
3095 ext4_fsblk_t start;
3096 ext4_fsblk_t len;
3097 int hblock, blocksize;
3098 ext4_fsblk_t sb_block;
3099 unsigned long offset;
3100 struct ext4_super_block *es;
3101 struct block_device *bdev;
3102
3103 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3104
3105 bdev = ext4_blkdev_get(j_dev, sb);
3106 if (bdev == NULL)
3107 return NULL;
3108
3109 if (bd_claim(bdev, sb)) {
3110 ext4_msg(sb, KERN_ERR,
3111 "failed to claim external journal device");
3112 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3113 return NULL;
3114 }
3115
3116 blocksize = sb->s_blocksize;
3117 hblock = bdev_logical_block_size(bdev);
3118 if (blocksize < hblock) {
3119 ext4_msg(sb, KERN_ERR,
3120 "blocksize too small for journal device");
3121 goto out_bdev;
3122 }
3123
3124 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3125 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3126 set_blocksize(bdev, blocksize);
3127 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3128 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3129 "external journal");
3130 goto out_bdev;
3131 }
3132
3133 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3134 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3135 !(le32_to_cpu(es->s_feature_incompat) &
3136 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3137 ext4_msg(sb, KERN_ERR, "external journal has "
3138 "bad superblock");
3139 brelse(bh);
3140 goto out_bdev;
3141 }
3142
3143 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3144 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3145 brelse(bh);
3146 goto out_bdev;
3147 }
3148
3149 len = ext4_blocks_count(es);
3150 start = sb_block + 1;
3151 brelse(bh); /* we're done with the superblock */
3152
3153 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3154 start, len, blocksize);
3155 if (!journal) {
3156 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3157 goto out_bdev;
3158 }
3159 journal->j_private = sb;
3160 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3161 wait_on_buffer(journal->j_sb_buffer);
3162 if (!buffer_uptodate(journal->j_sb_buffer)) {
3163 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3164 goto out_journal;
3165 }
3166 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3167 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3168 "user (unsupported) - %d",
3169 be32_to_cpu(journal->j_superblock->s_nr_users));
3170 goto out_journal;
3171 }
3172 EXT4_SB(sb)->journal_bdev = bdev;
3173 ext4_init_journal_params(sb, journal);
3174 return journal;
3175
3176 out_journal:
3177 jbd2_journal_destroy(journal);
3178 out_bdev:
3179 ext4_blkdev_put(bdev);
3180 return NULL;
3181 }
3182
3183 static int ext4_load_journal(struct super_block *sb,
3184 struct ext4_super_block *es,
3185 unsigned long journal_devnum)
3186 {
3187 journal_t *journal;
3188 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3189 dev_t journal_dev;
3190 int err = 0;
3191 int really_read_only;
3192
3193 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3194
3195 if (journal_devnum &&
3196 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3197 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3198 "numbers have changed");
3199 journal_dev = new_decode_dev(journal_devnum);
3200 } else
3201 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3202
3203 really_read_only = bdev_read_only(sb->s_bdev);
3204
3205 /*
3206 * Are we loading a blank journal or performing recovery after a
3207 * crash? For recovery, we need to check in advance whether we
3208 * can get read-write access to the device.
3209 */
3210 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3211 if (sb->s_flags & MS_RDONLY) {
3212 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3213 "required on readonly filesystem");
3214 if (really_read_only) {
3215 ext4_msg(sb, KERN_ERR, "write access "
3216 "unavailable, cannot proceed");
3217 return -EROFS;
3218 }
3219 ext4_msg(sb, KERN_INFO, "write access will "
3220 "be enabled during recovery");
3221 }
3222 }
3223
3224 if (journal_inum && journal_dev) {
3225 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3226 "and inode journals!");
3227 return -EINVAL;
3228 }
3229
3230 if (journal_inum) {
3231 if (!(journal = ext4_get_journal(sb, journal_inum)))
3232 return -EINVAL;
3233 } else {
3234 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3235 return -EINVAL;
3236 }
3237
3238 if (journal->j_flags & JBD2_BARRIER)
3239 ext4_msg(sb, KERN_INFO, "barriers enabled");
3240 else
3241 ext4_msg(sb, KERN_INFO, "barriers disabled");
3242
3243 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3244 err = jbd2_journal_update_format(journal);
3245 if (err) {
3246 ext4_msg(sb, KERN_ERR, "error updating journal");
3247 jbd2_journal_destroy(journal);
3248 return err;
3249 }
3250 }
3251
3252 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3253 err = jbd2_journal_wipe(journal, !really_read_only);
3254 if (!err)
3255 err = jbd2_journal_load(journal);
3256
3257 if (err) {
3258 ext4_msg(sb, KERN_ERR, "error loading journal");
3259 jbd2_journal_destroy(journal);
3260 return err;
3261 }
3262
3263 EXT4_SB(sb)->s_journal = journal;
3264 ext4_clear_journal_err(sb, es);
3265
3266 if (journal_devnum &&
3267 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3268 es->s_journal_dev = cpu_to_le32(journal_devnum);
3269
3270 /* Make sure we flush the recovery flag to disk. */
3271 ext4_commit_super(sb, 1);
3272 }
3273
3274 return 0;
3275 }
3276
3277 static int ext4_commit_super(struct super_block *sb, int sync)
3278 {
3279 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3280 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3281 int error = 0;
3282
3283 if (!sbh)
3284 return error;
3285 if (buffer_write_io_error(sbh)) {
3286 /*
3287 * Oh, dear. A previous attempt to write the
3288 * superblock failed. This could happen because the
3289 * USB device was yanked out. Or it could happen to
3290 * be a transient write error and maybe the block will
3291 * be remapped. Nothing we can do but to retry the
3292 * write and hope for the best.
3293 */
3294 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3295 "superblock detected");
3296 clear_buffer_write_io_error(sbh);
3297 set_buffer_uptodate(sbh);
3298 }
3299 /*
3300 * If the file system is mounted read-only, don't update the
3301 * superblock write time. This avoids updating the superblock
3302 * write time when we are mounting the root file system
3303 * read/only but we need to replay the journal; at that point,
3304 * for people who are east of GMT and who make their clock
3305 * tick in localtime for Windows bug-for-bug compatibility,
3306 * the clock is set in the future, and this will cause e2fsck
3307 * to complain and force a full file system check.
3308 */
3309 if (!(sb->s_flags & MS_RDONLY))
3310 es->s_wtime = cpu_to_le32(get_seconds());
3311 es->s_kbytes_written =
3312 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3313 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3314 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3315 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3316 &EXT4_SB(sb)->s_freeblocks_counter));
3317 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3318 &EXT4_SB(sb)->s_freeinodes_counter));
3319 sb->s_dirt = 0;
3320 BUFFER_TRACE(sbh, "marking dirty");
3321 mark_buffer_dirty(sbh);
3322 if (sync) {
3323 error = sync_dirty_buffer(sbh);
3324 if (error)
3325 return error;
3326
3327 error = buffer_write_io_error(sbh);
3328 if (error) {
3329 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3330 "superblock");
3331 clear_buffer_write_io_error(sbh);
3332 set_buffer_uptodate(sbh);
3333 }
3334 }
3335 return error;
3336 }
3337
3338 /*
3339 * Have we just finished recovery? If so, and if we are mounting (or
3340 * remounting) the filesystem readonly, then we will end up with a
3341 * consistent fs on disk. Record that fact.
3342 */
3343 static void ext4_mark_recovery_complete(struct super_block *sb,
3344 struct ext4_super_block *es)
3345 {
3346 journal_t *journal = EXT4_SB(sb)->s_journal;
3347
3348 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3349 BUG_ON(journal != NULL);
3350 return;
3351 }
3352 jbd2_journal_lock_updates(journal);
3353 if (jbd2_journal_flush(journal) < 0)
3354 goto out;
3355
3356 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3357 sb->s_flags & MS_RDONLY) {
3358 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3359 ext4_commit_super(sb, 1);
3360 }
3361
3362 out:
3363 jbd2_journal_unlock_updates(journal);
3364 }
3365
3366 /*
3367 * If we are mounting (or read-write remounting) a filesystem whose journal
3368 * has recorded an error from a previous lifetime, move that error to the
3369 * main filesystem now.
3370 */
3371 static void ext4_clear_journal_err(struct super_block *sb,
3372 struct ext4_super_block *es)
3373 {
3374 journal_t *journal;
3375 int j_errno;
3376 const char *errstr;
3377
3378 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3379
3380 journal = EXT4_SB(sb)->s_journal;
3381
3382 /*
3383 * Now check for any error status which may have been recorded in the
3384 * journal by a prior ext4_error() or ext4_abort()
3385 */
3386
3387 j_errno = jbd2_journal_errno(journal);
3388 if (j_errno) {
3389 char nbuf[16];
3390
3391 errstr = ext4_decode_error(sb, j_errno, nbuf);
3392 ext4_warning(sb, __func__, "Filesystem error recorded "
3393 "from previous mount: %s", errstr);
3394 ext4_warning(sb, __func__, "Marking fs in need of "
3395 "filesystem check.");
3396
3397 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3398 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3399 ext4_commit_super(sb, 1);
3400
3401 jbd2_journal_clear_err(journal);
3402 }
3403 }
3404
3405 /*
3406 * Force the running and committing transactions to commit,
3407 * and wait on the commit.
3408 */
3409 int ext4_force_commit(struct super_block *sb)
3410 {
3411 journal_t *journal;
3412 int ret = 0;
3413
3414 if (sb->s_flags & MS_RDONLY)
3415 return 0;
3416
3417 journal = EXT4_SB(sb)->s_journal;
3418 if (journal)
3419 ret = ext4_journal_force_commit(journal);
3420
3421 return ret;
3422 }
3423
3424 static void ext4_write_super(struct super_block *sb)
3425 {
3426 lock_super(sb);
3427 ext4_commit_super(sb, 1);
3428 unlock_super(sb);
3429 }
3430
3431 static int ext4_sync_fs(struct super_block *sb, int wait)
3432 {
3433 int ret = 0;
3434 tid_t target;
3435 struct ext4_sb_info *sbi = EXT4_SB(sb);
3436
3437 trace_ext4_sync_fs(sb, wait);
3438 flush_workqueue(sbi->dio_unwritten_wq);
3439 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
3440 if (wait)
3441 jbd2_log_wait_commit(sbi->s_journal, target);
3442 }
3443 return ret;
3444 }
3445
3446 /*
3447 * LVM calls this function before a (read-only) snapshot is created. This
3448 * gives us a chance to flush the journal completely and mark the fs clean.
3449 */
3450 static int ext4_freeze(struct super_block *sb)
3451 {
3452 int error = 0;
3453 journal_t *journal;
3454
3455 if (sb->s_flags & MS_RDONLY)
3456 return 0;
3457
3458 journal = EXT4_SB(sb)->s_journal;
3459
3460 /* Now we set up the journal barrier. */
3461 jbd2_journal_lock_updates(journal);
3462
3463 /*
3464 * Don't clear the needs_recovery flag if we failed to flush
3465 * the journal.
3466 */
3467 error = jbd2_journal_flush(journal);
3468 if (error < 0) {
3469 out:
3470 jbd2_journal_unlock_updates(journal);
3471 return error;
3472 }
3473
3474 /* Journal blocked and flushed, clear needs_recovery flag. */
3475 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3476 error = ext4_commit_super(sb, 1);
3477 if (error)
3478 goto out;
3479 return 0;
3480 }
3481
3482 /*
3483 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3484 * flag here, even though the filesystem is not technically dirty yet.
3485 */
3486 static int ext4_unfreeze(struct super_block *sb)
3487 {
3488 if (sb->s_flags & MS_RDONLY)
3489 return 0;
3490
3491 lock_super(sb);
3492 /* Reset the needs_recovery flag before the fs is unlocked. */
3493 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3494 ext4_commit_super(sb, 1);
3495 unlock_super(sb);
3496 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3497 return 0;
3498 }
3499
3500 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3501 {
3502 struct ext4_super_block *es;
3503 struct ext4_sb_info *sbi = EXT4_SB(sb);
3504 ext4_fsblk_t n_blocks_count = 0;
3505 unsigned long old_sb_flags;
3506 struct ext4_mount_options old_opts;
3507 ext4_group_t g;
3508 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3509 int err;
3510 #ifdef CONFIG_QUOTA
3511 int i;
3512 #endif
3513
3514 lock_kernel();
3515
3516 /* Store the original options */
3517 lock_super(sb);
3518 old_sb_flags = sb->s_flags;
3519 old_opts.s_mount_opt = sbi->s_mount_opt;
3520 old_opts.s_resuid = sbi->s_resuid;
3521 old_opts.s_resgid = sbi->s_resgid;
3522 old_opts.s_commit_interval = sbi->s_commit_interval;
3523 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3524 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3525 #ifdef CONFIG_QUOTA
3526 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3527 for (i = 0; i < MAXQUOTAS; i++)
3528 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3529 #endif
3530 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3531 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3532
3533 /*
3534 * Allow the "check" option to be passed as a remount option.
3535 */
3536 if (!parse_options(data, sb, NULL, &journal_ioprio,
3537 &n_blocks_count, 1)) {
3538 err = -EINVAL;
3539 goto restore_opts;
3540 }
3541
3542 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
3543 ext4_abort(sb, __func__, "Abort forced by user");
3544
3545 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3546 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3547
3548 es = sbi->s_es;
3549
3550 if (sbi->s_journal) {
3551 ext4_init_journal_params(sb, sbi->s_journal);
3552 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3553 }
3554
3555 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3556 n_blocks_count > ext4_blocks_count(es)) {
3557 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
3558 err = -EROFS;
3559 goto restore_opts;
3560 }
3561
3562 if (*flags & MS_RDONLY) {
3563 /*
3564 * First of all, the unconditional stuff we have to do
3565 * to disable replay of the journal when we next remount
3566 */
3567 sb->s_flags |= MS_RDONLY;
3568
3569 /*
3570 * OK, test if we are remounting a valid rw partition
3571 * readonly, and if so set the rdonly flag and then
3572 * mark the partition as valid again.
3573 */
3574 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3575 (sbi->s_mount_state & EXT4_VALID_FS))
3576 es->s_state = cpu_to_le16(sbi->s_mount_state);
3577
3578 if (sbi->s_journal)
3579 ext4_mark_recovery_complete(sb, es);
3580 } else {
3581 /* Make sure we can mount this feature set readwrite */
3582 if (!ext4_feature_set_ok(sb, 0)) {
3583 err = -EROFS;
3584 goto restore_opts;
3585 }
3586 /*
3587 * Make sure the group descriptor checksums
3588 * are sane. If they aren't, refuse to remount r/w.
3589 */
3590 for (g = 0; g < sbi->s_groups_count; g++) {
3591 struct ext4_group_desc *gdp =
3592 ext4_get_group_desc(sb, g, NULL);
3593
3594 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3595 ext4_msg(sb, KERN_ERR,
3596 "ext4_remount: Checksum for group %u failed (%u!=%u)",
3597 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3598 le16_to_cpu(gdp->bg_checksum));
3599 err = -EINVAL;
3600 goto restore_opts;
3601 }
3602 }
3603
3604 /*
3605 * If we have an unprocessed orphan list hanging
3606 * around from a previously readonly bdev mount,
3607 * require a full umount/remount for now.
3608 */
3609 if (es->s_last_orphan) {
3610 ext4_msg(sb, KERN_WARNING, "Couldn't "
3611 "remount RDWR because of unprocessed "
3612 "orphan inode list. Please "
3613 "umount/remount instead");
3614 err = -EINVAL;
3615 goto restore_opts;
3616 }
3617
3618 /*
3619 * Mounting a RDONLY partition read-write, so reread
3620 * and store the current valid flag. (It may have
3621 * been changed by e2fsck since we originally mounted
3622 * the partition.)
3623 */
3624 if (sbi->s_journal)
3625 ext4_clear_journal_err(sb, es);
3626 sbi->s_mount_state = le16_to_cpu(es->s_state);
3627 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3628 goto restore_opts;
3629 if (!ext4_setup_super(sb, es, 0))
3630 sb->s_flags &= ~MS_RDONLY;
3631 }
3632 }
3633 ext4_setup_system_zone(sb);
3634 if (sbi->s_journal == NULL)
3635 ext4_commit_super(sb, 1);
3636
3637 #ifdef CONFIG_QUOTA
3638 /* Release old quota file names */
3639 for (i = 0; i < MAXQUOTAS; i++)
3640 if (old_opts.s_qf_names[i] &&
3641 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3642 kfree(old_opts.s_qf_names[i]);
3643 #endif
3644 unlock_super(sb);
3645 unlock_kernel();
3646 return 0;
3647
3648 restore_opts:
3649 sb->s_flags = old_sb_flags;
3650 sbi->s_mount_opt = old_opts.s_mount_opt;
3651 sbi->s_resuid = old_opts.s_resuid;
3652 sbi->s_resgid = old_opts.s_resgid;
3653 sbi->s_commit_interval = old_opts.s_commit_interval;
3654 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3655 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3656 #ifdef CONFIG_QUOTA
3657 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3658 for (i = 0; i < MAXQUOTAS; i++) {
3659 if (sbi->s_qf_names[i] &&
3660 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3661 kfree(sbi->s_qf_names[i]);
3662 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3663 }
3664 #endif
3665 unlock_super(sb);
3666 unlock_kernel();
3667 return err;
3668 }
3669
3670 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3671 {
3672 struct super_block *sb = dentry->d_sb;
3673 struct ext4_sb_info *sbi = EXT4_SB(sb);
3674 struct ext4_super_block *es = sbi->s_es;
3675 u64 fsid;
3676
3677 if (test_opt(sb, MINIX_DF)) {
3678 sbi->s_overhead_last = 0;
3679 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3680 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3681 ext4_fsblk_t overhead = 0;
3682
3683 /*
3684 * Compute the overhead (FS structures). This is constant
3685 * for a given filesystem unless the number of block groups
3686 * changes so we cache the previous value until it does.
3687 */
3688
3689 /*
3690 * All of the blocks before first_data_block are
3691 * overhead
3692 */
3693 overhead = le32_to_cpu(es->s_first_data_block);
3694
3695 /*
3696 * Add the overhead attributed to the superblock and
3697 * block group descriptors. If the sparse superblocks
3698 * feature is turned on, then not all groups have this.
3699 */
3700 for (i = 0; i < ngroups; i++) {
3701 overhead += ext4_bg_has_super(sb, i) +
3702 ext4_bg_num_gdb(sb, i);
3703 cond_resched();
3704 }
3705
3706 /*
3707 * Every block group has an inode bitmap, a block
3708 * bitmap, and an inode table.
3709 */
3710 overhead += ngroups * (2 + sbi->s_itb_per_group);
3711 sbi->s_overhead_last = overhead;
3712 smp_wmb();
3713 sbi->s_blocks_last = ext4_blocks_count(es);
3714 }
3715
3716 buf->f_type = EXT4_SUPER_MAGIC;
3717 buf->f_bsize = sb->s_blocksize;
3718 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3719 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3720 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3721 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3722 if (buf->f_bfree < ext4_r_blocks_count(es))
3723 buf->f_bavail = 0;
3724 buf->f_files = le32_to_cpu(es->s_inodes_count);
3725 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3726 buf->f_namelen = EXT4_NAME_LEN;
3727 fsid = le64_to_cpup((void *)es->s_uuid) ^
3728 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3729 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3730 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3731
3732 return 0;
3733 }
3734
3735 /* Helper function for writing quotas on sync - we need to start transaction
3736 * before quota file is locked for write. Otherwise the are possible deadlocks:
3737 * Process 1 Process 2
3738 * ext4_create() quota_sync()
3739 * jbd2_journal_start() write_dquot()
3740 * vfs_dq_init() down(dqio_mutex)
3741 * down(dqio_mutex) jbd2_journal_start()
3742 *
3743 */
3744
3745 #ifdef CONFIG_QUOTA
3746
3747 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3748 {
3749 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3750 }
3751
3752 static int ext4_write_dquot(struct dquot *dquot)
3753 {
3754 int ret, err;
3755 handle_t *handle;
3756 struct inode *inode;
3757
3758 inode = dquot_to_inode(dquot);
3759 handle = ext4_journal_start(inode,
3760 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3761 if (IS_ERR(handle))
3762 return PTR_ERR(handle);
3763 ret = dquot_commit(dquot);
3764 err = ext4_journal_stop(handle);
3765 if (!ret)
3766 ret = err;
3767 return ret;
3768 }
3769
3770 static int ext4_acquire_dquot(struct dquot *dquot)
3771 {
3772 int ret, err;
3773 handle_t *handle;
3774
3775 handle = ext4_journal_start(dquot_to_inode(dquot),
3776 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3777 if (IS_ERR(handle))
3778 return PTR_ERR(handle);
3779 ret = dquot_acquire(dquot);
3780 err = ext4_journal_stop(handle);
3781 if (!ret)
3782 ret = err;
3783 return ret;
3784 }
3785
3786 static int ext4_release_dquot(struct dquot *dquot)
3787 {
3788 int ret, err;
3789 handle_t *handle;
3790
3791 handle = ext4_journal_start(dquot_to_inode(dquot),
3792 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3793 if (IS_ERR(handle)) {
3794 /* Release dquot anyway to avoid endless cycle in dqput() */
3795 dquot_release(dquot);
3796 return PTR_ERR(handle);
3797 }
3798 ret = dquot_release(dquot);
3799 err = ext4_journal_stop(handle);
3800 if (!ret)
3801 ret = err;
3802 return ret;
3803 }
3804
3805 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3806 {
3807 /* Are we journaling quotas? */
3808 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3809 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3810 dquot_mark_dquot_dirty(dquot);
3811 return ext4_write_dquot(dquot);
3812 } else {
3813 return dquot_mark_dquot_dirty(dquot);
3814 }
3815 }
3816
3817 static int ext4_write_info(struct super_block *sb, int type)
3818 {
3819 int ret, err;
3820 handle_t *handle;
3821
3822 /* Data block + inode block */
3823 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3824 if (IS_ERR(handle))
3825 return PTR_ERR(handle);
3826 ret = dquot_commit_info(sb, type);
3827 err = ext4_journal_stop(handle);
3828 if (!ret)
3829 ret = err;
3830 return ret;
3831 }
3832
3833 /*
3834 * Turn on quotas during mount time - we need to find
3835 * the quota file and such...
3836 */
3837 static int ext4_quota_on_mount(struct super_block *sb, int type)
3838 {
3839 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3840 EXT4_SB(sb)->s_jquota_fmt, type);
3841 }
3842
3843 /*
3844 * Standard function to be called on quota_on
3845 */
3846 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3847 char *name, int remount)
3848 {
3849 int err;
3850 struct path path;
3851
3852 if (!test_opt(sb, QUOTA))
3853 return -EINVAL;
3854 /* When remounting, no checks are needed and in fact, name is NULL */
3855 if (remount)
3856 return vfs_quota_on(sb, type, format_id, name, remount);
3857
3858 err = kern_path(name, LOOKUP_FOLLOW, &path);
3859 if (err)
3860 return err;
3861
3862 /* Quotafile not on the same filesystem? */
3863 if (path.mnt->mnt_sb != sb) {
3864 path_put(&path);
3865 return -EXDEV;
3866 }
3867 /* Journaling quota? */
3868 if (EXT4_SB(sb)->s_qf_names[type]) {
3869 /* Quotafile not in fs root? */
3870 if (path.dentry->d_parent != sb->s_root)
3871 ext4_msg(sb, KERN_WARNING,
3872 "Quota file not on filesystem root. "
3873 "Journaled quota will not work");
3874 }
3875
3876 /*
3877 * When we journal data on quota file, we have to flush journal to see
3878 * all updates to the file when we bypass pagecache...
3879 */
3880 if (EXT4_SB(sb)->s_journal &&
3881 ext4_should_journal_data(path.dentry->d_inode)) {
3882 /*
3883 * We don't need to lock updates but journal_flush() could
3884 * otherwise be livelocked...
3885 */
3886 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3887 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3888 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3889 if (err) {
3890 path_put(&path);
3891 return err;
3892 }
3893 }
3894
3895 err = vfs_quota_on_path(sb, type, format_id, &path);
3896 path_put(&path);
3897 return err;
3898 }
3899
3900 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3901 * acquiring the locks... As quota files are never truncated and quota code
3902 * itself serializes the operations (and noone else should touch the files)
3903 * we don't have to be afraid of races */
3904 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3905 size_t len, loff_t off)
3906 {
3907 struct inode *inode = sb_dqopt(sb)->files[type];
3908 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3909 int err = 0;
3910 int offset = off & (sb->s_blocksize - 1);
3911 int tocopy;
3912 size_t toread;
3913 struct buffer_head *bh;
3914 loff_t i_size = i_size_read(inode);
3915
3916 if (off > i_size)
3917 return 0;
3918 if (off+len > i_size)
3919 len = i_size-off;
3920 toread = len;
3921 while (toread > 0) {
3922 tocopy = sb->s_blocksize - offset < toread ?
3923 sb->s_blocksize - offset : toread;
3924 bh = ext4_bread(NULL, inode, blk, 0, &err);
3925 if (err)
3926 return err;
3927 if (!bh) /* A hole? */
3928 memset(data, 0, tocopy);
3929 else
3930 memcpy(data, bh->b_data+offset, tocopy);
3931 brelse(bh);
3932 offset = 0;
3933 toread -= tocopy;
3934 data += tocopy;
3935 blk++;
3936 }
3937 return len;
3938 }
3939
3940 /* Write to quotafile (we know the transaction is already started and has
3941 * enough credits) */
3942 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3943 const char *data, size_t len, loff_t off)
3944 {
3945 struct inode *inode = sb_dqopt(sb)->files[type];
3946 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3947 int err = 0;
3948 int offset = off & (sb->s_blocksize - 1);
3949 int tocopy;
3950 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3951 size_t towrite = len;
3952 struct buffer_head *bh;
3953 handle_t *handle = journal_current_handle();
3954
3955 if (EXT4_SB(sb)->s_journal && !handle) {
3956 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
3957 " cancelled because transaction is not started",
3958 (unsigned long long)off, (unsigned long long)len);
3959 return -EIO;
3960 }
3961 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3962 while (towrite > 0) {
3963 tocopy = sb->s_blocksize - offset < towrite ?
3964 sb->s_blocksize - offset : towrite;
3965 bh = ext4_bread(handle, inode, blk, 1, &err);
3966 if (!bh)
3967 goto out;
3968 if (journal_quota) {
3969 err = ext4_journal_get_write_access(handle, bh);
3970 if (err) {
3971 brelse(bh);
3972 goto out;
3973 }
3974 }
3975 lock_buffer(bh);
3976 memcpy(bh->b_data+offset, data, tocopy);
3977 flush_dcache_page(bh->b_page);
3978 unlock_buffer(bh);
3979 if (journal_quota)
3980 err = ext4_handle_dirty_metadata(handle, NULL, bh);
3981 else {
3982 /* Always do at least ordered writes for quotas */
3983 err = ext4_jbd2_file_inode(handle, inode);
3984 mark_buffer_dirty(bh);
3985 }
3986 brelse(bh);
3987 if (err)
3988 goto out;
3989 offset = 0;
3990 towrite -= tocopy;
3991 data += tocopy;
3992 blk++;
3993 }
3994 out:
3995 if (len == towrite) {
3996 mutex_unlock(&inode->i_mutex);
3997 return err;
3998 }
3999 if (inode->i_size < off+len-towrite) {
4000 i_size_write(inode, off+len-towrite);
4001 EXT4_I(inode)->i_disksize = inode->i_size;
4002 }
4003 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4004 ext4_mark_inode_dirty(handle, inode);
4005 mutex_unlock(&inode->i_mutex);
4006 return len - towrite;
4007 }
4008
4009 #endif
4010
4011 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
4012 const char *dev_name, void *data, struct vfsmount *mnt)
4013 {
4014 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
4015 }
4016
4017 static struct file_system_type ext4_fs_type = {
4018 .owner = THIS_MODULE,
4019 .name = "ext4",
4020 .get_sb = ext4_get_sb,
4021 .kill_sb = kill_block_super,
4022 .fs_flags = FS_REQUIRES_DEV,
4023 };
4024
4025 #ifdef CONFIG_EXT4DEV_COMPAT
4026 static int ext4dev_get_sb(struct file_system_type *fs_type, int flags,
4027 const char *dev_name, void *data,struct vfsmount *mnt)
4028 {
4029 printk(KERN_WARNING "EXT4-fs (%s): Update your userspace programs "
4030 "to mount using ext4\n", dev_name);
4031 printk(KERN_WARNING "EXT4-fs (%s): ext4dev backwards compatibility "
4032 "will go away by 2.6.31\n", dev_name);
4033 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
4034 }
4035
4036 static struct file_system_type ext4dev_fs_type = {
4037 .owner = THIS_MODULE,
4038 .name = "ext4dev",
4039 .get_sb = ext4dev_get_sb,
4040 .kill_sb = kill_block_super,
4041 .fs_flags = FS_REQUIRES_DEV,
4042 };
4043 MODULE_ALIAS("ext4dev");
4044 #endif
4045
4046 static int __init init_ext4_fs(void)
4047 {
4048 int err;
4049
4050 err = init_ext4_system_zone();
4051 if (err)
4052 return err;
4053 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4054 if (!ext4_kset)
4055 goto out4;
4056 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
4057 err = init_ext4_mballoc();
4058 if (err)
4059 goto out3;
4060
4061 err = init_ext4_xattr();
4062 if (err)
4063 goto out2;
4064 err = init_inodecache();
4065 if (err)
4066 goto out1;
4067 err = register_filesystem(&ext4_fs_type);
4068 if (err)
4069 goto out;
4070 #ifdef CONFIG_EXT4DEV_COMPAT
4071 err = register_filesystem(&ext4dev_fs_type);
4072 if (err) {
4073 unregister_filesystem(&ext4_fs_type);
4074 goto out;
4075 }
4076 #endif
4077 return 0;
4078 out:
4079 destroy_inodecache();
4080 out1:
4081 exit_ext4_xattr();
4082 out2:
4083 exit_ext4_mballoc();
4084 out3:
4085 remove_proc_entry("fs/ext4", NULL);
4086 kset_unregister(ext4_kset);
4087 out4:
4088 exit_ext4_system_zone();
4089 return err;
4090 }
4091
4092 static void __exit exit_ext4_fs(void)
4093 {
4094 unregister_filesystem(&ext4_fs_type);
4095 #ifdef CONFIG_EXT4DEV_COMPAT
4096 unregister_filesystem(&ext4dev_fs_type);
4097 #endif
4098 destroy_inodecache();
4099 exit_ext4_xattr();
4100 exit_ext4_mballoc();
4101 remove_proc_entry("fs/ext4", NULL);
4102 kset_unregister(ext4_kset);
4103 exit_ext4_system_zone();
4104 }
4105
4106 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4107 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4108 MODULE_LICENSE("GPL");
4109 module_init(init_ext4_fs)
4110 module_exit(exit_ext4_fs)
Linux 2.6 ibm-acpi/thinkpad-acpi driver git tree