?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
ext4: Make printk's consistently prefixed with "EXT4-fs: "
[linux-2.6/mini2440.git] / fs / ext4 / super.c
blobb69d0920386598202cf8b5e4d143de431470ccfc
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/jbd2.h>
24 #include <linux/slab.h>
25 #include <linux/init.h>
26 #include <linux/blkdev.h>
27 #include <linux/parser.h>
28 #include <linux/smp_lock.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/proc_fs.h>
38 #include <linux/marker.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <asm/uaccess.h>
42
43 #include "ext4.h"
44 #include "ext4_jbd2.h"
45 #include "xattr.h"
46 #include "acl.h"
47 #include "namei.h"
48 #include "group.h"
49
50 struct proc_dir_entry *ext4_proc_root;
51
52 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
53 unsigned long journal_devnum);
54 static void ext4_commit_super(struct super_block *sb,
55 struct ext4_super_block *es, int sync);
56 static void ext4_mark_recovery_complete(struct super_block *sb,
57 struct ext4_super_block *es);
58 static void ext4_clear_journal_err(struct super_block *sb,
59 struct ext4_super_block *es);
60 static int ext4_sync_fs(struct super_block *sb, int wait);
61 static const char *ext4_decode_error(struct super_block *sb, int errno,
62 char nbuf[16]);
63 static int ext4_remount(struct super_block *sb, int *flags, char *data);
64 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
65 static void ext4_unlockfs(struct super_block *sb);
66 static void ext4_write_super(struct super_block *sb);
67 static void ext4_write_super_lockfs(struct super_block *sb);
68
69
70 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
71 struct ext4_group_desc *bg)
72 {
73 return le32_to_cpu(bg->bg_block_bitmap_lo) |
74 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
75 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
76 }
77
78 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
79 struct ext4_group_desc *bg)
80 {
81 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
82 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
83 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
84 }
85
86 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
87 struct ext4_group_desc *bg)
88 {
89 return le32_to_cpu(bg->bg_inode_table_lo) |
90 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
91 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
92 }
93
94 __u32 ext4_free_blks_count(struct super_block *sb,
95 struct ext4_group_desc *bg)
96 {
97 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
98 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
99 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
100 }
101
102 __u32 ext4_free_inodes_count(struct super_block *sb,
103 struct ext4_group_desc *bg)
104 {
105 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
106 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
107 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
108 }
109
110 __u32 ext4_used_dirs_count(struct super_block *sb,
111 struct ext4_group_desc *bg)
112 {
113 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
114 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
115 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
116 }
117
118 __u32 ext4_itable_unused_count(struct super_block *sb,
119 struct ext4_group_desc *bg)
120 {
121 return le16_to_cpu(bg->bg_itable_unused_lo) |
122 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
123 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
124 }
125
126 void ext4_block_bitmap_set(struct super_block *sb,
127 struct ext4_group_desc *bg, ext4_fsblk_t blk)
128 {
129 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
130 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
131 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
132 }
133
134 void ext4_inode_bitmap_set(struct super_block *sb,
135 struct ext4_group_desc *bg, ext4_fsblk_t blk)
136 {
137 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
138 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
139 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
140 }
141
142 void ext4_inode_table_set(struct super_block *sb,
143 struct ext4_group_desc *bg, ext4_fsblk_t blk)
144 {
145 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
146 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
147 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
148 }
149
150 void ext4_free_blks_set(struct super_block *sb,
151 struct ext4_group_desc *bg, __u32 count)
152 {
153 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
154 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
155 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
156 }
157
158 void ext4_free_inodes_set(struct super_block *sb,
159 struct ext4_group_desc *bg, __u32 count)
160 {
161 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
162 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
163 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
164 }
165
166 void ext4_used_dirs_set(struct super_block *sb,
167 struct ext4_group_desc *bg, __u32 count)
168 {
169 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
170 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
171 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
172 }
173
174 void ext4_itable_unused_set(struct super_block *sb,
175 struct ext4_group_desc *bg, __u32 count)
176 {
177 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
178 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
179 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
180 }
181
182 /*
183 * Wrappers for jbd2_journal_start/end.
184 *
185 * The only special thing we need to do here is to make sure that all
186 * journal_end calls result in the superblock being marked dirty, so
187 * that sync() will call the filesystem's write_super callback if
188 * appropriate.
189 */
190 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
191 {
192 journal_t *journal;
193
194 if (sb->s_flags & MS_RDONLY)
195 return ERR_PTR(-EROFS);
196
197 /* Special case here: if the journal has aborted behind our
198 * backs (eg. EIO in the commit thread), then we still need to
199 * take the FS itself readonly cleanly. */
200 journal = EXT4_SB(sb)->s_journal;
201 if (journal) {
202 if (is_journal_aborted(journal)) {
203 ext4_abort(sb, __func__,
204 "Detected aborted journal");
205 return ERR_PTR(-EROFS);
206 }
207 return jbd2_journal_start(journal, nblocks);
208 }
209 /*
210 * We're not journaling, return the appropriate indication.
211 */
212 current->journal_info = EXT4_NOJOURNAL_HANDLE;
213 return current->journal_info;
214 }
215
216 /*
217 * The only special thing we need to do here is to make sure that all
218 * jbd2_journal_stop calls result in the superblock being marked dirty, so
219 * that sync() will call the filesystem's write_super callback if
220 * appropriate.
221 */
222 int __ext4_journal_stop(const char *where, handle_t *handle)
223 {
224 struct super_block *sb;
225 int err;
226 int rc;
227
228 if (!ext4_handle_valid(handle)) {
229 /*
230 * Do this here since we don't call jbd2_journal_stop() in
231 * no-journal mode.
232 */
233 current->journal_info = NULL;
234 return 0;
235 }
236 sb = handle->h_transaction->t_journal->j_private;
237 err = handle->h_err;
238 rc = jbd2_journal_stop(handle);
239
240 if (!err)
241 err = rc;
242 if (err)
243 __ext4_std_error(sb, where, err);
244 return err;
245 }
246
247 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
248 struct buffer_head *bh, handle_t *handle, int err)
249 {
250 char nbuf[16];
251 const char *errstr = ext4_decode_error(NULL, err, nbuf);
252
253 BUG_ON(!ext4_handle_valid(handle));
254
255 if (bh)
256 BUFFER_TRACE(bh, "abort");
257
258 if (!handle->h_err)
259 handle->h_err = err;
260
261 if (is_handle_aborted(handle))
262 return;
263
264 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
265 caller, errstr, err_fn);
266
267 jbd2_journal_abort_handle(handle);
268 }
269
270 /* Deal with the reporting of failure conditions on a filesystem such as
271 * inconsistencies detected or read IO failures.
272 *
273 * On ext2, we can store the error state of the filesystem in the
274 * superblock. That is not possible on ext4, because we may have other
275 * write ordering constraints on the superblock which prevent us from
276 * writing it out straight away; and given that the journal is about to
277 * be aborted, we can't rely on the current, or future, transactions to
278 * write out the superblock safely.
279 *
280 * We'll just use the jbd2_journal_abort() error code to record an error in
281 * the journal instead. On recovery, the journal will compain about
282 * that error until we've noted it down and cleared it.
283 */
284
285 static void ext4_handle_error(struct super_block *sb)
286 {
287 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
288
289 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
290 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
291
292 if (sb->s_flags & MS_RDONLY)
293 return;
294
295 if (!test_opt(sb, ERRORS_CONT)) {
296 journal_t *journal = EXT4_SB(sb)->s_journal;
297
298 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
299 if (journal)
300 jbd2_journal_abort(journal, -EIO);
301 }
302 if (test_opt(sb, ERRORS_RO)) {
303 printk(KERN_CRIT "Remounting filesystem read-only\n");
304 sb->s_flags |= MS_RDONLY;
305 }
306 ext4_commit_super(sb, es, 1);
307 if (test_opt(sb, ERRORS_PANIC))
308 panic("EXT4-fs (device %s): panic forced after error\n",
309 sb->s_id);
310 }
311
312 void ext4_error(struct super_block *sb, const char *function,
313 const char *fmt, ...)
314 {
315 va_list args;
316
317 va_start(args, fmt);
318 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
319 vprintk(fmt, args);
320 printk("\n");
321 va_end(args);
322
323 ext4_handle_error(sb);
324 }
325
326 static const char *ext4_decode_error(struct super_block *sb, int errno,
327 char nbuf[16])
328 {
329 char *errstr = NULL;
330
331 switch (errno) {
332 case -EIO:
333 errstr = "IO failure";
334 break;
335 case -ENOMEM:
336 errstr = "Out of memory";
337 break;
338 case -EROFS:
339 if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
340 errstr = "Journal has aborted";
341 else
342 errstr = "Readonly filesystem";
343 break;
344 default:
345 /* If the caller passed in an extra buffer for unknown
346 * errors, textualise them now. Else we just return
347 * NULL. */
348 if (nbuf) {
349 /* Check for truncated error codes... */
350 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
351 errstr = nbuf;
352 }
353 break;
354 }
355
356 return errstr;
357 }
358
359 /* __ext4_std_error decodes expected errors from journaling functions
360 * automatically and invokes the appropriate error response. */
361
362 void __ext4_std_error(struct super_block *sb, const char *function, int errno)
363 {
364 char nbuf[16];
365 const char *errstr;
366
367 /* Special case: if the error is EROFS, and we're not already
368 * inside a transaction, then there's really no point in logging
369 * an error. */
370 if (errno == -EROFS && journal_current_handle() == NULL &&
371 (sb->s_flags & MS_RDONLY))
372 return;
373
374 errstr = ext4_decode_error(sb, errno, nbuf);
375 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
376 sb->s_id, function, errstr);
377
378 ext4_handle_error(sb);
379 }
380
381 /*
382 * ext4_abort is a much stronger failure handler than ext4_error. The
383 * abort function may be used to deal with unrecoverable failures such
384 * as journal IO errors or ENOMEM at a critical moment in log management.
385 *
386 * We unconditionally force the filesystem into an ABORT|READONLY state,
387 * unless the error response on the fs has been set to panic in which
388 * case we take the easy way out and panic immediately.
389 */
390
391 void ext4_abort(struct super_block *sb, const char *function,
392 const char *fmt, ...)
393 {
394 va_list args;
395
396 printk(KERN_CRIT "ext4_abort called.\n");
397
398 va_start(args, fmt);
399 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
400 vprintk(fmt, args);
401 printk("\n");
402 va_end(args);
403
404 if (test_opt(sb, ERRORS_PANIC))
405 panic("EXT4-fs panic from previous error\n");
406
407 if (sb->s_flags & MS_RDONLY)
408 return;
409
410 printk(KERN_CRIT "Remounting filesystem read-only\n");
411 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
412 sb->s_flags |= MS_RDONLY;
413 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
414 if (EXT4_SB(sb)->s_journal)
415 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
416 }
417
418 void ext4_warning(struct super_block *sb, const char *function,
419 const char *fmt, ...)
420 {
421 va_list args;
422
423 va_start(args, fmt);
424 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
425 sb->s_id, function);
426 vprintk(fmt, args);
427 printk("\n");
428 va_end(args);
429 }
430
431 void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp,
432 const char *function, const char *fmt, ...)
433 __releases(bitlock)
434 __acquires(bitlock)
435 {
436 va_list args;
437 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
438
439 va_start(args, fmt);
440 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
441 vprintk(fmt, args);
442 printk("\n");
443 va_end(args);
444
445 if (test_opt(sb, ERRORS_CONT)) {
446 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
447 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
448 ext4_commit_super(sb, es, 0);
449 return;
450 }
451 ext4_unlock_group(sb, grp);
452 ext4_handle_error(sb);
453 /*
454 * We only get here in the ERRORS_RO case; relocking the group
455 * may be dangerous, but nothing bad will happen since the
456 * filesystem will have already been marked read/only and the
457 * journal has been aborted. We return 1 as a hint to callers
458 * who might what to use the return value from
459 * ext4_grp_locked_error() to distinguish beween the
460 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
461 * aggressively from the ext4 function in question, with a
462 * more appropriate error code.
463 */
464 ext4_lock_group(sb, grp);
465 return;
466 }
467
468
469 void ext4_update_dynamic_rev(struct super_block *sb)
470 {
471 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
472
473 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
474 return;
475
476 ext4_warning(sb, __func__,
477 "updating to rev %d because of new feature flag, "
478 "running e2fsck is recommended",
479 EXT4_DYNAMIC_REV);
480
481 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
482 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
483 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
484 /* leave es->s_feature_*compat flags alone */
485 /* es->s_uuid will be set by e2fsck if empty */
486
487 /*
488 * The rest of the superblock fields should be zero, and if not it
489 * means they are likely already in use, so leave them alone. We
490 * can leave it up to e2fsck to clean up any inconsistencies there.
491 */
492 }
493
494 /*
495 * Open the external journal device
496 */
497 static struct block_device *ext4_blkdev_get(dev_t dev)
498 {
499 struct block_device *bdev;
500 char b[BDEVNAME_SIZE];
501
502 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
503 if (IS_ERR(bdev))
504 goto fail;
505 return bdev;
506
507 fail:
508 printk(KERN_ERR "EXT4-fs: failed to open journal device %s: %ld\n",
509 __bdevname(dev, b), PTR_ERR(bdev));
510 return NULL;
511 }
512
513 /*
514 * Release the journal device
515 */
516 static int ext4_blkdev_put(struct block_device *bdev)
517 {
518 bd_release(bdev);
519 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
520 }
521
522 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
523 {
524 struct block_device *bdev;
525 int ret = -ENODEV;
526
527 bdev = sbi->journal_bdev;
528 if (bdev) {
529 ret = ext4_blkdev_put(bdev);
530 sbi->journal_bdev = NULL;
531 }
532 return ret;
533 }
534
535 static inline struct inode *orphan_list_entry(struct list_head *l)
536 {
537 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
538 }
539
540 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
541 {
542 struct list_head *l;
543
544 printk(KERN_ERR "sb orphan head is %d\n",
545 le32_to_cpu(sbi->s_es->s_last_orphan));
546
547 printk(KERN_ERR "sb_info orphan list:\n");
548 list_for_each(l, &sbi->s_orphan) {
549 struct inode *inode = orphan_list_entry(l);
550 printk(KERN_ERR " "
551 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
552 inode->i_sb->s_id, inode->i_ino, inode,
553 inode->i_mode, inode->i_nlink,
554 NEXT_ORPHAN(inode));
555 }
556 }
557
558 static void ext4_put_super(struct super_block *sb)
559 {
560 struct ext4_sb_info *sbi = EXT4_SB(sb);
561 struct ext4_super_block *es = sbi->s_es;
562 int i, err;
563
564 ext4_mb_release(sb);
565 ext4_ext_release(sb);
566 ext4_xattr_put_super(sb);
567 if (sbi->s_journal) {
568 err = jbd2_journal_destroy(sbi->s_journal);
569 sbi->s_journal = NULL;
570 if (err < 0)
571 ext4_abort(sb, __func__,
572 "Couldn't clean up the journal");
573 }
574 if (!(sb->s_flags & MS_RDONLY)) {
575 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
576 es->s_state = cpu_to_le16(sbi->s_mount_state);
577 ext4_commit_super(sb, es, 1);
578 }
579 if (sbi->s_proc) {
580 remove_proc_entry("inode_readahead_blks", sbi->s_proc);
581 remove_proc_entry(sb->s_id, ext4_proc_root);
582 }
583
584 for (i = 0; i < sbi->s_gdb_count; i++)
585 brelse(sbi->s_group_desc[i]);
586 kfree(sbi->s_group_desc);
587 kfree(sbi->s_flex_groups);
588 percpu_counter_destroy(&sbi->s_freeblocks_counter);
589 percpu_counter_destroy(&sbi->s_freeinodes_counter);
590 percpu_counter_destroy(&sbi->s_dirs_counter);
591 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
592 brelse(sbi->s_sbh);
593 #ifdef CONFIG_QUOTA
594 for (i = 0; i < MAXQUOTAS; i++)
595 kfree(sbi->s_qf_names[i]);
596 #endif
597
598 /* Debugging code just in case the in-memory inode orphan list
599 * isn't empty. The on-disk one can be non-empty if we've
600 * detected an error and taken the fs readonly, but the
601 * in-memory list had better be clean by this point. */
602 if (!list_empty(&sbi->s_orphan))
603 dump_orphan_list(sb, sbi);
604 J_ASSERT(list_empty(&sbi->s_orphan));
605
606 invalidate_bdev(sb->s_bdev);
607 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
608 /*
609 * Invalidate the journal device's buffers. We don't want them
610 * floating about in memory - the physical journal device may
611 * hotswapped, and it breaks the `ro-after' testing code.
612 */
613 sync_blockdev(sbi->journal_bdev);
614 invalidate_bdev(sbi->journal_bdev);
615 ext4_blkdev_remove(sbi);
616 }
617 sb->s_fs_info = NULL;
618 kfree(sbi);
619 return;
620 }
621
622 static struct kmem_cache *ext4_inode_cachep;
623
624 /*
625 * Called inside transaction, so use GFP_NOFS
626 */
627 static struct inode *ext4_alloc_inode(struct super_block *sb)
628 {
629 struct ext4_inode_info *ei;
630
631 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
632 if (!ei)
633 return NULL;
634 #ifdef CONFIG_EXT4_FS_POSIX_ACL
635 ei->i_acl = EXT4_ACL_NOT_CACHED;
636 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
637 #endif
638 ei->vfs_inode.i_version = 1;
639 ei->vfs_inode.i_data.writeback_index = 0;
640 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
641 INIT_LIST_HEAD(&ei->i_prealloc_list);
642 spin_lock_init(&ei->i_prealloc_lock);
643 /*
644 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
645 * therefore it can be null here. Don't check it, just initialize
646 * jinode.
647 */
648 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
649 ei->i_reserved_data_blocks = 0;
650 ei->i_reserved_meta_blocks = 0;
651 ei->i_allocated_meta_blocks = 0;
652 ei->i_delalloc_reserved_flag = 0;
653 spin_lock_init(&(ei->i_block_reservation_lock));
654 return &ei->vfs_inode;
655 }
656
657 static void ext4_destroy_inode(struct inode *inode)
658 {
659 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
660 printk("EXT4 Inode %p: orphan list check failed!\n",
661 EXT4_I(inode));
662 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
663 EXT4_I(inode), sizeof(struct ext4_inode_info),
664 true);
665 dump_stack();
666 }
667 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
668 }
669
670 static void init_once(void *foo)
671 {
672 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
673
674 INIT_LIST_HEAD(&ei->i_orphan);
675 #ifdef CONFIG_EXT4_FS_XATTR
676 init_rwsem(&ei->xattr_sem);
677 #endif
678 init_rwsem(&ei->i_data_sem);
679 inode_init_once(&ei->vfs_inode);
680 }
681
682 static int init_inodecache(void)
683 {
684 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
685 sizeof(struct ext4_inode_info),
686 0, (SLAB_RECLAIM_ACCOUNT|
687 SLAB_MEM_SPREAD),
688 init_once);
689 if (ext4_inode_cachep == NULL)
690 return -ENOMEM;
691 return 0;
692 }
693
694 static void destroy_inodecache(void)
695 {
696 kmem_cache_destroy(ext4_inode_cachep);
697 }
698
699 static void ext4_clear_inode(struct inode *inode)
700 {
701 #ifdef CONFIG_EXT4_FS_POSIX_ACL
702 if (EXT4_I(inode)->i_acl &&
703 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
704 posix_acl_release(EXT4_I(inode)->i_acl);
705 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
706 }
707 if (EXT4_I(inode)->i_default_acl &&
708 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
709 posix_acl_release(EXT4_I(inode)->i_default_acl);
710 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
711 }
712 #endif
713 ext4_discard_preallocations(inode);
714 if (EXT4_JOURNAL(inode))
715 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
716 &EXT4_I(inode)->jinode);
717 }
718
719 static inline void ext4_show_quota_options(struct seq_file *seq,
720 struct super_block *sb)
721 {
722 #if defined(CONFIG_QUOTA)
723 struct ext4_sb_info *sbi = EXT4_SB(sb);
724
725 if (sbi->s_jquota_fmt)
726 seq_printf(seq, ",jqfmt=%s",
727 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
728
729 if (sbi->s_qf_names[USRQUOTA])
730 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
731
732 if (sbi->s_qf_names[GRPQUOTA])
733 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
734
735 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
736 seq_puts(seq, ",usrquota");
737
738 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
739 seq_puts(seq, ",grpquota");
740 #endif
741 }
742
743 /*
744 * Show an option if
745 * - it's set to a non-default value OR
746 * - if the per-sb default is different from the global default
747 */
748 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
749 {
750 int def_errors;
751 unsigned long def_mount_opts;
752 struct super_block *sb = vfs->mnt_sb;
753 struct ext4_sb_info *sbi = EXT4_SB(sb);
754 struct ext4_super_block *es = sbi->s_es;
755
756 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
757 def_errors = le16_to_cpu(es->s_errors);
758
759 if (sbi->s_sb_block != 1)
760 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
761 if (test_opt(sb, MINIX_DF))
762 seq_puts(seq, ",minixdf");
763 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
764 seq_puts(seq, ",grpid");
765 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
766 seq_puts(seq, ",nogrpid");
767 if (sbi->s_resuid != EXT4_DEF_RESUID ||
768 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
769 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
770 }
771 if (sbi->s_resgid != EXT4_DEF_RESGID ||
772 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
773 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
774 }
775 if (test_opt(sb, ERRORS_RO)) {
776 if (def_errors == EXT4_ERRORS_PANIC ||
777 def_errors == EXT4_ERRORS_CONTINUE) {
778 seq_puts(seq, ",errors=remount-ro");
779 }
780 }
781 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
782 seq_puts(seq, ",errors=continue");
783 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
784 seq_puts(seq, ",errors=panic");
785 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
786 seq_puts(seq, ",nouid32");
787 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
788 seq_puts(seq, ",debug");
789 if (test_opt(sb, OLDALLOC))
790 seq_puts(seq, ",oldalloc");
791 #ifdef CONFIG_EXT4_FS_XATTR
792 if (test_opt(sb, XATTR_USER) &&
793 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
794 seq_puts(seq, ",user_xattr");
795 if (!test_opt(sb, XATTR_USER) &&
796 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
797 seq_puts(seq, ",nouser_xattr");
798 }
799 #endif
800 #ifdef CONFIG_EXT4_FS_POSIX_ACL
801 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
802 seq_puts(seq, ",acl");
803 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
804 seq_puts(seq, ",noacl");
805 #endif
806 if (!test_opt(sb, RESERVATION))
807 seq_puts(seq, ",noreservation");
808 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
809 seq_printf(seq, ",commit=%u",
810 (unsigned) (sbi->s_commit_interval / HZ));
811 }
812 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
813 seq_printf(seq, ",min_batch_time=%u",
814 (unsigned) sbi->s_min_batch_time);
815 }
816 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
817 seq_printf(seq, ",max_batch_time=%u",
818 (unsigned) sbi->s_min_batch_time);
819 }
820
821 /*
822 * We're changing the default of barrier mount option, so
823 * let's always display its mount state so it's clear what its
824 * status is.
825 */
826 seq_puts(seq, ",barrier=");
827 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
828 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
829 seq_puts(seq, ",journal_async_commit");
830 if (test_opt(sb, NOBH))
831 seq_puts(seq, ",nobh");
832 if (!test_opt(sb, EXTENTS))
833 seq_puts(seq, ",noextents");
834 if (test_opt(sb, I_VERSION))
835 seq_puts(seq, ",i_version");
836 if (!test_opt(sb, DELALLOC))
837 seq_puts(seq, ",nodelalloc");
838
839
840 if (sbi->s_stripe)
841 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
842 /*
843 * journal mode get enabled in different ways
844 * So just print the value even if we didn't specify it
845 */
846 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
847 seq_puts(seq, ",data=journal");
848 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
849 seq_puts(seq, ",data=ordered");
850 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
851 seq_puts(seq, ",data=writeback");
852
853 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
854 seq_printf(seq, ",inode_readahead_blks=%u",
855 sbi->s_inode_readahead_blks);
856
857 if (test_opt(sb, DATA_ERR_ABORT))
858 seq_puts(seq, ",data_err=abort");
859
860 ext4_show_quota_options(seq, sb);
861 return 0;
862 }
863
864
865 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
866 u64 ino, u32 generation)
867 {
868 struct inode *inode;
869
870 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
871 return ERR_PTR(-ESTALE);
872 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
873 return ERR_PTR(-ESTALE);
874
875 /* iget isn't really right if the inode is currently unallocated!!
876 *
877 * ext4_read_inode will return a bad_inode if the inode had been
878 * deleted, so we should be safe.
879 *
880 * Currently we don't know the generation for parent directory, so
881 * a generation of 0 means "accept any"
882 */
883 inode = ext4_iget(sb, ino);
884 if (IS_ERR(inode))
885 return ERR_CAST(inode);
886 if (generation && inode->i_generation != generation) {
887 iput(inode);
888 return ERR_PTR(-ESTALE);
889 }
890
891 return inode;
892 }
893
894 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
895 int fh_len, int fh_type)
896 {
897 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
898 ext4_nfs_get_inode);
899 }
900
901 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
902 int fh_len, int fh_type)
903 {
904 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
905 ext4_nfs_get_inode);
906 }
907
908 /*
909 * Try to release metadata pages (indirect blocks, directories) which are
910 * mapped via the block device. Since these pages could have journal heads
911 * which would prevent try_to_free_buffers() from freeing them, we must use
912 * jbd2 layer's try_to_free_buffers() function to release them.
913 */
914 static int bdev_try_to_free_page(struct super_block *sb, struct page *page, gfp_t wait)
915 {
916 journal_t *journal = EXT4_SB(sb)->s_journal;
917
918 WARN_ON(PageChecked(page));
919 if (!page_has_buffers(page))
920 return 0;
921 if (journal)
922 return jbd2_journal_try_to_free_buffers(journal, page,
923 wait & ~__GFP_WAIT);
924 return try_to_free_buffers(page);
925 }
926
927 #ifdef CONFIG_QUOTA
928 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
929 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
930
931 static int ext4_dquot_initialize(struct inode *inode, int type);
932 static int ext4_dquot_drop(struct inode *inode);
933 static int ext4_write_dquot(struct dquot *dquot);
934 static int ext4_acquire_dquot(struct dquot *dquot);
935 static int ext4_release_dquot(struct dquot *dquot);
936 static int ext4_mark_dquot_dirty(struct dquot *dquot);
937 static int ext4_write_info(struct super_block *sb, int type);
938 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
939 char *path, int remount);
940 static int ext4_quota_on_mount(struct super_block *sb, int type);
941 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
942 size_t len, loff_t off);
943 static ssize_t ext4_quota_write(struct super_block *sb, int type,
944 const char *data, size_t len, loff_t off);
945
946 static struct dquot_operations ext4_quota_operations = {
947 .initialize = ext4_dquot_initialize,
948 .drop = ext4_dquot_drop,
949 .alloc_space = dquot_alloc_space,
950 .alloc_inode = dquot_alloc_inode,
951 .free_space = dquot_free_space,
952 .free_inode = dquot_free_inode,
953 .transfer = dquot_transfer,
954 .write_dquot = ext4_write_dquot,
955 .acquire_dquot = ext4_acquire_dquot,
956 .release_dquot = ext4_release_dquot,
957 .mark_dirty = ext4_mark_dquot_dirty,
958 .write_info = ext4_write_info
959 };
960
961 static struct quotactl_ops ext4_qctl_operations = {
962 .quota_on = ext4_quota_on,
963 .quota_off = vfs_quota_off,
964 .quota_sync = vfs_quota_sync,
965 .get_info = vfs_get_dqinfo,
966 .set_info = vfs_set_dqinfo,
967 .get_dqblk = vfs_get_dqblk,
968 .set_dqblk = vfs_set_dqblk
969 };
970 #endif
971
972 static const struct super_operations ext4_sops = {
973 .alloc_inode = ext4_alloc_inode,
974 .destroy_inode = ext4_destroy_inode,
975 .write_inode = ext4_write_inode,
976 .dirty_inode = ext4_dirty_inode,
977 .delete_inode = ext4_delete_inode,
978 .put_super = ext4_put_super,
979 .write_super = ext4_write_super,
980 .sync_fs = ext4_sync_fs,
981 .write_super_lockfs = ext4_write_super_lockfs,
982 .unlockfs = ext4_unlockfs,
983 .statfs = ext4_statfs,
984 .remount_fs = ext4_remount,
985 .clear_inode = ext4_clear_inode,
986 .show_options = ext4_show_options,
987 #ifdef CONFIG_QUOTA
988 .quota_read = ext4_quota_read,
989 .quota_write = ext4_quota_write,
990 #endif
991 .bdev_try_to_free_page = bdev_try_to_free_page,
992 };
993
994 static const struct export_operations ext4_export_ops = {
995 .fh_to_dentry = ext4_fh_to_dentry,
996 .fh_to_parent = ext4_fh_to_parent,
997 .get_parent = ext4_get_parent,
998 };
999
1000 enum {
1001 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1002 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1003 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1004 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1005 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
1006 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1007 Opt_journal_update, Opt_journal_dev,
1008 Opt_journal_checksum, Opt_journal_async_commit,
1009 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1010 Opt_data_err_abort, Opt_data_err_ignore,
1011 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1012 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
1013 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
1014 Opt_grpquota, Opt_extents, Opt_noextents, Opt_i_version,
1015 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1016 Opt_inode_readahead_blks, Opt_journal_ioprio
1017 };
1018
1019 static const match_table_t tokens = {
1020 {Opt_bsd_df, "bsddf"},
1021 {Opt_minix_df, "minixdf"},
1022 {Opt_grpid, "grpid"},
1023 {Opt_grpid, "bsdgroups"},
1024 {Opt_nogrpid, "nogrpid"},
1025 {Opt_nogrpid, "sysvgroups"},
1026 {Opt_resgid, "resgid=%u"},
1027 {Opt_resuid, "resuid=%u"},
1028 {Opt_sb, "sb=%u"},
1029 {Opt_err_cont, "errors=continue"},
1030 {Opt_err_panic, "errors=panic"},
1031 {Opt_err_ro, "errors=remount-ro"},
1032 {Opt_nouid32, "nouid32"},
1033 {Opt_debug, "debug"},
1034 {Opt_oldalloc, "oldalloc"},
1035 {Opt_orlov, "orlov"},
1036 {Opt_user_xattr, "user_xattr"},
1037 {Opt_nouser_xattr, "nouser_xattr"},
1038 {Opt_acl, "acl"},
1039 {Opt_noacl, "noacl"},
1040 {Opt_reservation, "reservation"},
1041 {Opt_noreservation, "noreservation"},
1042 {Opt_noload, "noload"},
1043 {Opt_nobh, "nobh"},
1044 {Opt_bh, "bh"},
1045 {Opt_commit, "commit=%u"},
1046 {Opt_min_batch_time, "min_batch_time=%u"},
1047 {Opt_max_batch_time, "max_batch_time=%u"},
1048 {Opt_journal_update, "journal=update"},
1049 {Opt_journal_dev, "journal_dev=%u"},
1050 {Opt_journal_checksum, "journal_checksum"},
1051 {Opt_journal_async_commit, "journal_async_commit"},
1052 {Opt_abort, "abort"},
1053 {Opt_data_journal, "data=journal"},
1054 {Opt_data_ordered, "data=ordered"},
1055 {Opt_data_writeback, "data=writeback"},
1056 {Opt_data_err_abort, "data_err=abort"},
1057 {Opt_data_err_ignore, "data_err=ignore"},
1058 {Opt_offusrjquota, "usrjquota="},
1059 {Opt_usrjquota, "usrjquota=%s"},
1060 {Opt_offgrpjquota, "grpjquota="},
1061 {Opt_grpjquota, "grpjquota=%s"},
1062 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1063 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1064 {Opt_grpquota, "grpquota"},
1065 {Opt_noquota, "noquota"},
1066 {Opt_quota, "quota"},
1067 {Opt_usrquota, "usrquota"},
1068 {Opt_barrier, "barrier=%u"},
1069 {Opt_extents, "extents"},
1070 {Opt_noextents, "noextents"},
1071 {Opt_i_version, "i_version"},
1072 {Opt_stripe, "stripe=%u"},
1073 {Opt_resize, "resize"},
1074 {Opt_delalloc, "delalloc"},
1075 {Opt_nodelalloc, "nodelalloc"},
1076 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1077 {Opt_journal_ioprio, "journal_ioprio=%u"},
1078 {Opt_err, NULL},
1079 };
1080
1081 static ext4_fsblk_t get_sb_block(void **data)
1082 {
1083 ext4_fsblk_t sb_block;
1084 char *options = (char *) *data;
1085
1086 if (!options || strncmp(options, "sb=", 3) != 0)
1087 return 1; /* Default location */
1088 options += 3;
1089 /*todo: use simple_strtoll with >32bit ext4 */
1090 sb_block = simple_strtoul(options, &options, 0);
1091 if (*options && *options != ',') {
1092 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1093 (char *) *data);
1094 return 1;
1095 }
1096 if (*options == ',')
1097 options++;
1098 *data = (void *) options;
1099 return sb_block;
1100 }
1101
1102 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1103
1104 static int parse_options(char *options, struct super_block *sb,
1105 unsigned long *journal_devnum,
1106 unsigned int *journal_ioprio,
1107 ext4_fsblk_t *n_blocks_count, int is_remount)
1108 {
1109 struct ext4_sb_info *sbi = EXT4_SB(sb);
1110 char *p;
1111 substring_t args[MAX_OPT_ARGS];
1112 int data_opt = 0;
1113 int option;
1114 #ifdef CONFIG_QUOTA
1115 int qtype, qfmt;
1116 char *qname;
1117 #endif
1118 ext4_fsblk_t last_block;
1119
1120 if (!options)
1121 return 1;
1122
1123 while ((p = strsep(&options, ",")) != NULL) {
1124 int token;
1125 if (!*p)
1126 continue;
1127
1128 token = match_token(p, tokens, args);
1129 switch (token) {
1130 case Opt_bsd_df:
1131 clear_opt(sbi->s_mount_opt, MINIX_DF);
1132 break;
1133 case Opt_minix_df:
1134 set_opt(sbi->s_mount_opt, MINIX_DF);
1135 break;
1136 case Opt_grpid:
1137 set_opt(sbi->s_mount_opt, GRPID);
1138 break;
1139 case Opt_nogrpid:
1140 clear_opt(sbi->s_mount_opt, GRPID);
1141 break;
1142 case Opt_resuid:
1143 if (match_int(&args[0], &option))
1144 return 0;
1145 sbi->s_resuid = option;
1146 break;
1147 case Opt_resgid:
1148 if (match_int(&args[0], &option))
1149 return 0;
1150 sbi->s_resgid = option;
1151 break;
1152 case Opt_sb:
1153 /* handled by get_sb_block() instead of here */
1154 /* *sb_block = match_int(&args[0]); */
1155 break;
1156 case Opt_err_panic:
1157 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1158 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1159 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1160 break;
1161 case Opt_err_ro:
1162 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1163 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1164 set_opt(sbi->s_mount_opt, ERRORS_RO);
1165 break;
1166 case Opt_err_cont:
1167 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1168 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1169 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1170 break;
1171 case Opt_nouid32:
1172 set_opt(sbi->s_mount_opt, NO_UID32);
1173 break;
1174 case Opt_debug:
1175 set_opt(sbi->s_mount_opt, DEBUG);
1176 break;
1177 case Opt_oldalloc:
1178 set_opt(sbi->s_mount_opt, OLDALLOC);
1179 break;
1180 case Opt_orlov:
1181 clear_opt(sbi->s_mount_opt, OLDALLOC);
1182 break;
1183 #ifdef CONFIG_EXT4_FS_XATTR
1184 case Opt_user_xattr:
1185 set_opt(sbi->s_mount_opt, XATTR_USER);
1186 break;
1187 case Opt_nouser_xattr:
1188 clear_opt(sbi->s_mount_opt, XATTR_USER);
1189 break;
1190 #else
1191 case Opt_user_xattr:
1192 case Opt_nouser_xattr:
1193 printk(KERN_ERR "EXT4 (no)user_xattr options "
1194 "not supported\n");
1195 break;
1196 #endif
1197 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1198 case Opt_acl:
1199 set_opt(sbi->s_mount_opt, POSIX_ACL);
1200 break;
1201 case Opt_noacl:
1202 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1203 break;
1204 #else
1205 case Opt_acl:
1206 case Opt_noacl:
1207 printk(KERN_ERR "EXT4 (no)acl options "
1208 "not supported\n");
1209 break;
1210 #endif
1211 case Opt_reservation:
1212 set_opt(sbi->s_mount_opt, RESERVATION);
1213 break;
1214 case Opt_noreservation:
1215 clear_opt(sbi->s_mount_opt, RESERVATION);
1216 break;
1217 case Opt_journal_update:
1218 /* @@@ FIXME */
1219 /* Eventually we will want to be able to create
1220 a journal file here. For now, only allow the
1221 user to specify an existing inode to be the
1222 journal file. */
1223 if (is_remount) {
1224 printk(KERN_ERR "EXT4-fs: cannot specify "
1225 "journal on remount\n");
1226 return 0;
1227 }
1228 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1229 break;
1230 case Opt_journal_dev:
1231 if (is_remount) {
1232 printk(KERN_ERR "EXT4-fs: cannot specify "
1233 "journal on remount\n");
1234 return 0;
1235 }
1236 if (match_int(&args[0], &option))
1237 return 0;
1238 *journal_devnum = option;
1239 break;
1240 case Opt_journal_checksum:
1241 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1242 break;
1243 case Opt_journal_async_commit:
1244 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1245 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1246 break;
1247 case Opt_noload:
1248 set_opt(sbi->s_mount_opt, NOLOAD);
1249 break;
1250 case Opt_commit:
1251 if (match_int(&args[0], &option))
1252 return 0;
1253 if (option < 0)
1254 return 0;
1255 if (option == 0)
1256 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1257 sbi->s_commit_interval = HZ * option;
1258 break;
1259 case Opt_max_batch_time:
1260 if (match_int(&args[0], &option))
1261 return 0;
1262 if (option < 0)
1263 return 0;
1264 if (option == 0)
1265 option = EXT4_DEF_MAX_BATCH_TIME;
1266 sbi->s_max_batch_time = option;
1267 break;
1268 case Opt_min_batch_time:
1269 if (match_int(&args[0], &option))
1270 return 0;
1271 if (option < 0)
1272 return 0;
1273 sbi->s_min_batch_time = option;
1274 break;
1275 case Opt_data_journal:
1276 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1277 goto datacheck;
1278 case Opt_data_ordered:
1279 data_opt = EXT4_MOUNT_ORDERED_DATA;
1280 goto datacheck;
1281 case Opt_data_writeback:
1282 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1283 datacheck:
1284 if (is_remount) {
1285 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1286 != data_opt) {
1287 printk(KERN_ERR
1288 "EXT4-fs: cannot change data "
1289 "mode on remount\n");
1290 return 0;
1291 }
1292 } else {
1293 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1294 sbi->s_mount_opt |= data_opt;
1295 }
1296 break;
1297 case Opt_data_err_abort:
1298 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1299 break;
1300 case Opt_data_err_ignore:
1301 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1302 break;
1303 #ifdef CONFIG_QUOTA
1304 case Opt_usrjquota:
1305 qtype = USRQUOTA;
1306 goto set_qf_name;
1307 case Opt_grpjquota:
1308 qtype = GRPQUOTA;
1309 set_qf_name:
1310 if ((sb_any_quota_enabled(sb) ||
1311 sb_any_quota_suspended(sb)) &&
1312 !sbi->s_qf_names[qtype]) {
1313 printk(KERN_ERR
1314 "EXT4-fs: Cannot change journaled "
1315 "quota options when quota turned on.\n");
1316 return 0;
1317 }
1318 qname = match_strdup(&args[0]);
1319 if (!qname) {
1320 printk(KERN_ERR
1321 "EXT4-fs: not enough memory for "
1322 "storing quotafile name.\n");
1323 return 0;
1324 }
1325 if (sbi->s_qf_names[qtype] &&
1326 strcmp(sbi->s_qf_names[qtype], qname)) {
1327 printk(KERN_ERR
1328 "EXT4-fs: %s quota file already "
1329 "specified.\n", QTYPE2NAME(qtype));
1330 kfree(qname);
1331 return 0;
1332 }
1333 sbi->s_qf_names[qtype] = qname;
1334 if (strchr(sbi->s_qf_names[qtype], '/')) {
1335 printk(KERN_ERR
1336 "EXT4-fs: quotafile must be on "
1337 "filesystem root.\n");
1338 kfree(sbi->s_qf_names[qtype]);
1339 sbi->s_qf_names[qtype] = NULL;
1340 return 0;
1341 }
1342 set_opt(sbi->s_mount_opt, QUOTA);
1343 break;
1344 case Opt_offusrjquota:
1345 qtype = USRQUOTA;
1346 goto clear_qf_name;
1347 case Opt_offgrpjquota:
1348 qtype = GRPQUOTA;
1349 clear_qf_name:
1350 if ((sb_any_quota_enabled(sb) ||
1351 sb_any_quota_suspended(sb)) &&
1352 sbi->s_qf_names[qtype]) {
1353 printk(KERN_ERR "EXT4-fs: Cannot change "
1354 "journaled quota options when "
1355 "quota turned on.\n");
1356 return 0;
1357 }
1358 /*
1359 * The space will be released later when all options
1360 * are confirmed to be correct
1361 */
1362 sbi->s_qf_names[qtype] = NULL;
1363 break;
1364 case Opt_jqfmt_vfsold:
1365 qfmt = QFMT_VFS_OLD;
1366 goto set_qf_format;
1367 case Opt_jqfmt_vfsv0:
1368 qfmt = QFMT_VFS_V0;
1369 set_qf_format:
1370 if ((sb_any_quota_enabled(sb) ||
1371 sb_any_quota_suspended(sb)) &&
1372 sbi->s_jquota_fmt != qfmt) {
1373 printk(KERN_ERR "EXT4-fs: Cannot change "
1374 "journaled quota options when "
1375 "quota turned on.\n");
1376 return 0;
1377 }
1378 sbi->s_jquota_fmt = qfmt;
1379 break;
1380 case Opt_quota:
1381 case Opt_usrquota:
1382 set_opt(sbi->s_mount_opt, QUOTA);
1383 set_opt(sbi->s_mount_opt, USRQUOTA);
1384 break;
1385 case Opt_grpquota:
1386 set_opt(sbi->s_mount_opt, QUOTA);
1387 set_opt(sbi->s_mount_opt, GRPQUOTA);
1388 break;
1389 case Opt_noquota:
1390 if (sb_any_quota_enabled(sb)) {
1391 printk(KERN_ERR "EXT4-fs: Cannot change quota "
1392 "options when quota turned on.\n");
1393 return 0;
1394 }
1395 clear_opt(sbi->s_mount_opt, QUOTA);
1396 clear_opt(sbi->s_mount_opt, USRQUOTA);
1397 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1398 break;
1399 #else
1400 case Opt_quota:
1401 case Opt_usrquota:
1402 case Opt_grpquota:
1403 printk(KERN_ERR
1404 "EXT4-fs: quota options not supported.\n");
1405 break;
1406 case Opt_usrjquota:
1407 case Opt_grpjquota:
1408 case Opt_offusrjquota:
1409 case Opt_offgrpjquota:
1410 case Opt_jqfmt_vfsold:
1411 case Opt_jqfmt_vfsv0:
1412 printk(KERN_ERR
1413 "EXT4-fs: journaled quota options not "
1414 "supported.\n");
1415 break;
1416 case Opt_noquota:
1417 break;
1418 #endif
1419 case Opt_abort:
1420 set_opt(sbi->s_mount_opt, ABORT);
1421 break;
1422 case Opt_barrier:
1423 if (match_int(&args[0], &option))
1424 return 0;
1425 if (option)
1426 set_opt(sbi->s_mount_opt, BARRIER);
1427 else
1428 clear_opt(sbi->s_mount_opt, BARRIER);
1429 break;
1430 case Opt_ignore:
1431 break;
1432 case Opt_resize:
1433 if (!is_remount) {
1434 printk("EXT4-fs: resize option only available "
1435 "for remount\n");
1436 return 0;
1437 }
1438 if (match_int(&args[0], &option) != 0)
1439 return 0;
1440 *n_blocks_count = option;
1441 break;
1442 case Opt_nobh:
1443 set_opt(sbi->s_mount_opt, NOBH);
1444 break;
1445 case Opt_bh:
1446 clear_opt(sbi->s_mount_opt, NOBH);
1447 break;
1448 case Opt_extents:
1449 if (!EXT4_HAS_INCOMPAT_FEATURE(sb,
1450 EXT4_FEATURE_INCOMPAT_EXTENTS)) {
1451 ext4_warning(sb, __func__,
1452 "extents feature not enabled "
1453 "on this filesystem, use tune2fs");
1454 return 0;
1455 }
1456 set_opt(sbi->s_mount_opt, EXTENTS);
1457 break;
1458 case Opt_noextents:
1459 /*
1460 * When e2fsprogs support resizing an already existing
1461 * ext3 file system to greater than 2**32 we need to
1462 * add support to block allocator to handle growing
1463 * already existing block mapped inode so that blocks
1464 * allocated for them fall within 2**32
1465 */
1466 last_block = ext4_blocks_count(sbi->s_es) - 1;
1467 if (last_block > 0xffffffffULL) {
1468 printk(KERN_ERR "EXT4-fs: Filesystem too "
1469 "large to mount with "
1470 "-o noextents options\n");
1471 return 0;
1472 }
1473 clear_opt(sbi->s_mount_opt, EXTENTS);
1474 break;
1475 case Opt_i_version:
1476 set_opt(sbi->s_mount_opt, I_VERSION);
1477 sb->s_flags |= MS_I_VERSION;
1478 break;
1479 case Opt_nodelalloc:
1480 clear_opt(sbi->s_mount_opt, DELALLOC);
1481 break;
1482 case Opt_stripe:
1483 if (match_int(&args[0], &option))
1484 return 0;
1485 if (option < 0)
1486 return 0;
1487 sbi->s_stripe = option;
1488 break;
1489 case Opt_delalloc:
1490 set_opt(sbi->s_mount_opt, DELALLOC);
1491 break;
1492 case Opt_inode_readahead_blks:
1493 if (match_int(&args[0], &option))
1494 return 0;
1495 if (option < 0 || option > (1 << 30))
1496 return 0;
1497 sbi->s_inode_readahead_blks = option;
1498 break;
1499 case Opt_journal_ioprio:
1500 if (match_int(&args[0], &option))
1501 return 0;
1502 if (option < 0 || option > 7)
1503 break;
1504 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1505 option);
1506 break;
1507 default:
1508 printk(KERN_ERR
1509 "EXT4-fs: Unrecognized mount option \"%s\" "
1510 "or missing value\n", p);
1511 return 0;
1512 }
1513 }
1514 #ifdef CONFIG_QUOTA
1515 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1516 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1517 sbi->s_qf_names[USRQUOTA])
1518 clear_opt(sbi->s_mount_opt, USRQUOTA);
1519
1520 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1521 sbi->s_qf_names[GRPQUOTA])
1522 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1523
1524 if ((sbi->s_qf_names[USRQUOTA] &&
1525 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1526 (sbi->s_qf_names[GRPQUOTA] &&
1527 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1528 printk(KERN_ERR "EXT4-fs: old and new quota "
1529 "format mixing.\n");
1530 return 0;
1531 }
1532
1533 if (!sbi->s_jquota_fmt) {
1534 printk(KERN_ERR "EXT4-fs: journaled quota format "
1535 "not specified.\n");
1536 return 0;
1537 }
1538 } else {
1539 if (sbi->s_jquota_fmt) {
1540 printk(KERN_ERR "EXT4-fs: journaled quota format "
1541 "specified with no journaling "
1542 "enabled.\n");
1543 return 0;
1544 }
1545 }
1546 #endif
1547 return 1;
1548 }
1549
1550 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1551 int read_only)
1552 {
1553 struct ext4_sb_info *sbi = EXT4_SB(sb);
1554 int res = 0;
1555
1556 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1557 printk(KERN_ERR "EXT4-fs warning: revision level too high, "
1558 "forcing read-only mode\n");
1559 res = MS_RDONLY;
1560 }
1561 if (read_only)
1562 return res;
1563 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1564 printk(KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1565 "running e2fsck is recommended\n");
1566 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1567 printk(KERN_WARNING
1568 "EXT4-fs warning: mounting fs with errors, "
1569 "running e2fsck is recommended\n");
1570 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1571 le16_to_cpu(es->s_mnt_count) >=
1572 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1573 printk(KERN_WARNING
1574 "EXT4-fs warning: maximal mount count reached, "
1575 "running e2fsck is recommended\n");
1576 else if (le32_to_cpu(es->s_checkinterval) &&
1577 (le32_to_cpu(es->s_lastcheck) +
1578 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1579 printk(KERN_WARNING
1580 "EXT4-fs warning: checktime reached, "
1581 "running e2fsck is recommended\n");
1582 if (!sbi->s_journal)
1583 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1584 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1585 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1586 le16_add_cpu(&es->s_mnt_count, 1);
1587 es->s_mtime = cpu_to_le32(get_seconds());
1588 ext4_update_dynamic_rev(sb);
1589 if (sbi->s_journal)
1590 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1591
1592 ext4_commit_super(sb, es, 1);
1593 if (test_opt(sb, DEBUG))
1594 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1595 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1596 sb->s_blocksize,
1597 sbi->s_groups_count,
1598 EXT4_BLOCKS_PER_GROUP(sb),
1599 EXT4_INODES_PER_GROUP(sb),
1600 sbi->s_mount_opt);
1601
1602 if (EXT4_SB(sb)->s_journal) {
1603 printk(KERN_INFO "EXT4 FS on %s, %s journal on %s\n",
1604 sb->s_id, EXT4_SB(sb)->s_journal->j_inode ? "internal" :
1605 "external", EXT4_SB(sb)->s_journal->j_devname);
1606 } else {
1607 printk(KERN_INFO "EXT4 FS on %s, no journal\n", sb->s_id);
1608 }
1609 return res;
1610 }
1611
1612 static int ext4_fill_flex_info(struct super_block *sb)
1613 {
1614 struct ext4_sb_info *sbi = EXT4_SB(sb);
1615 struct ext4_group_desc *gdp = NULL;
1616 struct buffer_head *bh;
1617 ext4_group_t flex_group_count;
1618 ext4_group_t flex_group;
1619 int groups_per_flex = 0;
1620 int i;
1621
1622 if (!sbi->s_es->s_log_groups_per_flex) {
1623 sbi->s_log_groups_per_flex = 0;
1624 return 1;
1625 }
1626
1627 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1628 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1629
1630 /* We allocate both existing and potentially added groups */
1631 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1632 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1633 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1634 sbi->s_flex_groups = kzalloc(flex_group_count *
1635 sizeof(struct flex_groups), GFP_KERNEL);
1636 if (sbi->s_flex_groups == NULL) {
1637 printk(KERN_ERR "EXT4-fs: not enough memory for "
1638 "%u flex groups\n", flex_group_count);
1639 goto failed;
1640 }
1641
1642 for (i = 0; i < sbi->s_groups_count; i++) {
1643 gdp = ext4_get_group_desc(sb, i, &bh);
1644
1645 flex_group = ext4_flex_group(sbi, i);
1646 sbi->s_flex_groups[flex_group].free_inodes +=
1647 ext4_free_inodes_count(sb, gdp);
1648 sbi->s_flex_groups[flex_group].free_blocks +=
1649 ext4_free_blks_count(sb, gdp);
1650 }
1651
1652 return 1;
1653 failed:
1654 return 0;
1655 }
1656
1657 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1658 struct ext4_group_desc *gdp)
1659 {
1660 __u16 crc = 0;
1661
1662 if (sbi->s_es->s_feature_ro_compat &
1663 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1664 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1665 __le32 le_group = cpu_to_le32(block_group);
1666
1667 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1668 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1669 crc = crc16(crc, (__u8 *)gdp, offset);
1670 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1671 /* for checksum of struct ext4_group_desc do the rest...*/
1672 if ((sbi->s_es->s_feature_incompat &
1673 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1674 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1675 crc = crc16(crc, (__u8 *)gdp + offset,
1676 le16_to_cpu(sbi->s_es->s_desc_size) -
1677 offset);
1678 }
1679
1680 return cpu_to_le16(crc);
1681 }
1682
1683 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1684 struct ext4_group_desc *gdp)
1685 {
1686 if ((sbi->s_es->s_feature_ro_compat &
1687 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1688 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1689 return 0;
1690
1691 return 1;
1692 }
1693
1694 /* Called at mount-time, super-block is locked */
1695 static int ext4_check_descriptors(struct super_block *sb)
1696 {
1697 struct ext4_sb_info *sbi = EXT4_SB(sb);
1698 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1699 ext4_fsblk_t last_block;
1700 ext4_fsblk_t block_bitmap;
1701 ext4_fsblk_t inode_bitmap;
1702 ext4_fsblk_t inode_table;
1703 int flexbg_flag = 0;
1704 ext4_group_t i;
1705
1706 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1707 flexbg_flag = 1;
1708
1709 ext4_debug("Checking group descriptors");
1710
1711 for (i = 0; i < sbi->s_groups_count; i++) {
1712 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1713
1714 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1715 last_block = ext4_blocks_count(sbi->s_es) - 1;
1716 else
1717 last_block = first_block +
1718 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1719
1720 block_bitmap = ext4_block_bitmap(sb, gdp);
1721 if (block_bitmap < first_block || block_bitmap > last_block) {
1722 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1723 "Block bitmap for group %u not in group "
1724 "(block %llu)!\n", i, block_bitmap);
1725 return 0;
1726 }
1727 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1728 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1729 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1730 "Inode bitmap for group %u not in group "
1731 "(block %llu)!\n", i, inode_bitmap);
1732 return 0;
1733 }
1734 inode_table = ext4_inode_table(sb, gdp);
1735 if (inode_table < first_block ||
1736 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1737 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1738 "Inode table for group %u not in group "
1739 "(block %llu)!\n", i, inode_table);
1740 return 0;
1741 }
1742 spin_lock(sb_bgl_lock(sbi, i));
1743 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1744 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1745 "Checksum for group %u failed (%u!=%u)\n",
1746 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1747 gdp)), le16_to_cpu(gdp->bg_checksum));
1748 if (!(sb->s_flags & MS_RDONLY)) {
1749 spin_unlock(sb_bgl_lock(sbi, i));
1750 return 0;
1751 }
1752 }
1753 spin_unlock(sb_bgl_lock(sbi, i));
1754 if (!flexbg_flag)
1755 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1756 }
1757
1758 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1759 sbi->s_es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
1760 return 1;
1761 }
1762
1763 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1764 * the superblock) which were deleted from all directories, but held open by
1765 * a process at the time of a crash. We walk the list and try to delete these
1766 * inodes at recovery time (only with a read-write filesystem).
1767 *
1768 * In order to keep the orphan inode chain consistent during traversal (in
1769 * case of crash during recovery), we link each inode into the superblock
1770 * orphan list_head and handle it the same way as an inode deletion during
1771 * normal operation (which journals the operations for us).
1772 *
1773 * We only do an iget() and an iput() on each inode, which is very safe if we
1774 * accidentally point at an in-use or already deleted inode. The worst that
1775 * can happen in this case is that we get a "bit already cleared" message from
1776 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1777 * e2fsck was run on this filesystem, and it must have already done the orphan
1778 * inode cleanup for us, so we can safely abort without any further action.
1779 */
1780 static void ext4_orphan_cleanup(struct super_block *sb,
1781 struct ext4_super_block *es)
1782 {
1783 unsigned int s_flags = sb->s_flags;
1784 int nr_orphans = 0, nr_truncates = 0;
1785 #ifdef CONFIG_QUOTA
1786 int i;
1787 #endif
1788 if (!es->s_last_orphan) {
1789 jbd_debug(4, "no orphan inodes to clean up\n");
1790 return;
1791 }
1792
1793 if (bdev_read_only(sb->s_bdev)) {
1794 printk(KERN_ERR "EXT4-fs: write access "
1795 "unavailable, skipping orphan cleanup.\n");
1796 return;
1797 }
1798
1799 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1800 if (es->s_last_orphan)
1801 jbd_debug(1, "Errors on filesystem, "
1802 "clearing orphan list.\n");
1803 es->s_last_orphan = 0;
1804 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1805 return;
1806 }
1807
1808 if (s_flags & MS_RDONLY) {
1809 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
1810 sb->s_id);
1811 sb->s_flags &= ~MS_RDONLY;
1812 }
1813 #ifdef CONFIG_QUOTA
1814 /* Needed for iput() to work correctly and not trash data */
1815 sb->s_flags |= MS_ACTIVE;
1816 /* Turn on quotas so that they are updated correctly */
1817 for (i = 0; i < MAXQUOTAS; i++) {
1818 if (EXT4_SB(sb)->s_qf_names[i]) {
1819 int ret = ext4_quota_on_mount(sb, i);
1820 if (ret < 0)
1821 printk(KERN_ERR
1822 "EXT4-fs: Cannot turn on journaled "
1823 "quota: error %d\n", ret);
1824 }
1825 }
1826 #endif
1827
1828 while (es->s_last_orphan) {
1829 struct inode *inode;
1830
1831 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1832 if (IS_ERR(inode)) {
1833 es->s_last_orphan = 0;
1834 break;
1835 }
1836
1837 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1838 DQUOT_INIT(inode);
1839 if (inode->i_nlink) {
1840 printk(KERN_DEBUG
1841 "%s: truncating inode %lu to %lld bytes\n",
1842 __func__, inode->i_ino, inode->i_size);
1843 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1844 inode->i_ino, inode->i_size);
1845 ext4_truncate(inode);
1846 nr_truncates++;
1847 } else {
1848 printk(KERN_DEBUG
1849 "%s: deleting unreferenced inode %lu\n",
1850 __func__, inode->i_ino);
1851 jbd_debug(2, "deleting unreferenced inode %lu\n",
1852 inode->i_ino);
1853 nr_orphans++;
1854 }
1855 iput(inode); /* The delete magic happens here! */
1856 }
1857
1858 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1859
1860 if (nr_orphans)
1861 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
1862 sb->s_id, PLURAL(nr_orphans));
1863 if (nr_truncates)
1864 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
1865 sb->s_id, PLURAL(nr_truncates));
1866 #ifdef CONFIG_QUOTA
1867 /* Turn quotas off */
1868 for (i = 0; i < MAXQUOTAS; i++) {
1869 if (sb_dqopt(sb)->files[i])
1870 vfs_quota_off(sb, i, 0);
1871 }
1872 #endif
1873 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1874 }
1875 /*
1876 * Maximal extent format file size.
1877 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1878 * extent format containers, within a sector_t, and within i_blocks
1879 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1880 * so that won't be a limiting factor.
1881 *
1882 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1883 */
1884 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1885 {
1886 loff_t res;
1887 loff_t upper_limit = MAX_LFS_FILESIZE;
1888
1889 /* small i_blocks in vfs inode? */
1890 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1891 /*
1892 * CONFIG_LBD is not enabled implies the inode
1893 * i_block represent total blocks in 512 bytes
1894 * 32 == size of vfs inode i_blocks * 8
1895 */
1896 upper_limit = (1LL << 32) - 1;
1897
1898 /* total blocks in file system block size */
1899 upper_limit >>= (blkbits - 9);
1900 upper_limit <<= blkbits;
1901 }
1902
1903 /* 32-bit extent-start container, ee_block */
1904 res = 1LL << 32;
1905 res <<= blkbits;
1906 res -= 1;
1907
1908 /* Sanity check against vm- & vfs- imposed limits */
1909 if (res > upper_limit)
1910 res = upper_limit;
1911
1912 return res;
1913 }
1914
1915 /*
1916 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1917 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1918 * We need to be 1 filesystem block less than the 2^48 sector limit.
1919 */
1920 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
1921 {
1922 loff_t res = EXT4_NDIR_BLOCKS;
1923 int meta_blocks;
1924 loff_t upper_limit;
1925 /* This is calculated to be the largest file size for a
1926 * dense, bitmapped file such that the total number of
1927 * sectors in the file, including data and all indirect blocks,
1928 * does not exceed 2^48 -1
1929 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1930 * total number of 512 bytes blocks of the file
1931 */
1932
1933 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1934 /*
1935 * !has_huge_files or CONFIG_LBD is not enabled
1936 * implies the inode i_block represent total blocks in
1937 * 512 bytes 32 == size of vfs inode i_blocks * 8
1938 */
1939 upper_limit = (1LL << 32) - 1;
1940
1941 /* total blocks in file system block size */
1942 upper_limit >>= (bits - 9);
1943
1944 } else {
1945 /*
1946 * We use 48 bit ext4_inode i_blocks
1947 * With EXT4_HUGE_FILE_FL set the i_blocks
1948 * represent total number of blocks in
1949 * file system block size
1950 */
1951 upper_limit = (1LL << 48) - 1;
1952
1953 }
1954
1955 /* indirect blocks */
1956 meta_blocks = 1;
1957 /* double indirect blocks */
1958 meta_blocks += 1 + (1LL << (bits-2));
1959 /* tripple indirect blocks */
1960 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1961
1962 upper_limit -= meta_blocks;
1963 upper_limit <<= bits;
1964
1965 res += 1LL << (bits-2);
1966 res += 1LL << (2*(bits-2));
1967 res += 1LL << (3*(bits-2));
1968 res <<= bits;
1969 if (res > upper_limit)
1970 res = upper_limit;
1971
1972 if (res > MAX_LFS_FILESIZE)
1973 res = MAX_LFS_FILESIZE;
1974
1975 return res;
1976 }
1977
1978 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
1979 ext4_fsblk_t logical_sb_block, int nr)
1980 {
1981 struct ext4_sb_info *sbi = EXT4_SB(sb);
1982 ext4_group_t bg, first_meta_bg;
1983 int has_super = 0;
1984
1985 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1986
1987 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
1988 nr < first_meta_bg)
1989 return logical_sb_block + nr + 1;
1990 bg = sbi->s_desc_per_block * nr;
1991 if (ext4_bg_has_super(sb, bg))
1992 has_super = 1;
1993 return (has_super + ext4_group_first_block_no(sb, bg));
1994 }
1995
1996 /**
1997 * ext4_get_stripe_size: Get the stripe size.
1998 * @sbi: In memory super block info
1999 *
2000 * If we have specified it via mount option, then
2001 * use the mount option value. If the value specified at mount time is
2002 * greater than the blocks per group use the super block value.
2003 * If the super block value is greater than blocks per group return 0.
2004 * Allocator needs it be less than blocks per group.
2005 *
2006 */
2007 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2008 {
2009 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2010 unsigned long stripe_width =
2011 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2012
2013 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2014 return sbi->s_stripe;
2015
2016 if (stripe_width <= sbi->s_blocks_per_group)
2017 return stripe_width;
2018
2019 if (stride <= sbi->s_blocks_per_group)
2020 return stride;
2021
2022 return 0;
2023 }
2024
2025 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2026 __releases(kernel_lock)
2027 __acquires(kernel_lock)
2028
2029 {
2030 struct buffer_head *bh;
2031 struct ext4_super_block *es = NULL;
2032 struct ext4_sb_info *sbi;
2033 ext4_fsblk_t block;
2034 ext4_fsblk_t sb_block = get_sb_block(&data);
2035 ext4_fsblk_t logical_sb_block;
2036 unsigned long offset = 0;
2037 unsigned long journal_devnum = 0;
2038 unsigned long def_mount_opts;
2039 struct inode *root;
2040 char *cp;
2041 const char *descr;
2042 int ret = -EINVAL;
2043 int blocksize;
2044 unsigned int db_count;
2045 unsigned int i;
2046 int needs_recovery, has_huge_files;
2047 int features;
2048 __u64 blocks_count;
2049 int err;
2050 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2051
2052 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2053 if (!sbi)
2054 return -ENOMEM;
2055 sb->s_fs_info = sbi;
2056 sbi->s_mount_opt = 0;
2057 sbi->s_resuid = EXT4_DEF_RESUID;
2058 sbi->s_resgid = EXT4_DEF_RESGID;
2059 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2060 sbi->s_sb_block = sb_block;
2061
2062 unlock_kernel();
2063
2064 /* Cleanup superblock name */
2065 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2066 *cp = '!';
2067
2068 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2069 if (!blocksize) {
2070 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
2071 goto out_fail;
2072 }
2073
2074 /*
2075 * The ext4 superblock will not be buffer aligned for other than 1kB
2076 * block sizes. We need to calculate the offset from buffer start.
2077 */
2078 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2079 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2080 offset = do_div(logical_sb_block, blocksize);
2081 } else {
2082 logical_sb_block = sb_block;
2083 }
2084
2085 if (!(bh = sb_bread(sb, logical_sb_block))) {
2086 printk(KERN_ERR "EXT4-fs: unable to read superblock\n");
2087 goto out_fail;
2088 }
2089 /*
2090 * Note: s_es must be initialized as soon as possible because
2091 * some ext4 macro-instructions depend on its value
2092 */
2093 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2094 sbi->s_es = es;
2095 sb->s_magic = le16_to_cpu(es->s_magic);
2096 if (sb->s_magic != EXT4_SUPER_MAGIC)
2097 goto cantfind_ext4;
2098
2099 /* Set defaults before we parse the mount options */
2100 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2101 if (def_mount_opts & EXT4_DEFM_DEBUG)
2102 set_opt(sbi->s_mount_opt, DEBUG);
2103 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
2104 set_opt(sbi->s_mount_opt, GRPID);
2105 if (def_mount_opts & EXT4_DEFM_UID16)
2106 set_opt(sbi->s_mount_opt, NO_UID32);
2107 #ifdef CONFIG_EXT4_FS_XATTR
2108 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2109 set_opt(sbi->s_mount_opt, XATTR_USER);
2110 #endif
2111 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2112 if (def_mount_opts & EXT4_DEFM_ACL)
2113 set_opt(sbi->s_mount_opt, POSIX_ACL);
2114 #endif
2115 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2116 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2117 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2118 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2119 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2120 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2121
2122 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2123 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2124 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2125 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2126 else
2127 set_opt(sbi->s_mount_opt, ERRORS_RO);
2128
2129 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2130 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2131 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2132 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2133 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2134
2135 set_opt(sbi->s_mount_opt, RESERVATION);
2136 set_opt(sbi->s_mount_opt, BARRIER);
2137
2138 /*
2139 * turn on extents feature by default in ext4 filesystem
2140 * only if feature flag already set by mkfs or tune2fs.
2141 * Use -o noextents to turn it off
2142 */
2143 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
2144 set_opt(sbi->s_mount_opt, EXTENTS);
2145 else
2146 ext4_warning(sb, __func__,
2147 "extents feature not enabled on this filesystem, "
2148 "use tune2fs.");
2149
2150 /*
2151 * enable delayed allocation by default
2152 * Use -o nodelalloc to turn it off
2153 */
2154 set_opt(sbi->s_mount_opt, DELALLOC);
2155
2156
2157 if (!parse_options((char *) data, sb, &journal_devnum,
2158 &journal_ioprio, NULL, 0))
2159 goto failed_mount;
2160
2161 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2162 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2163
2164 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2165 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2166 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2167 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2168 printk(KERN_WARNING
2169 "EXT4-fs warning: feature flags set on rev 0 fs, "
2170 "running e2fsck is recommended\n");
2171
2172 /*
2173 * Check feature flags regardless of the revision level, since we
2174 * previously didn't change the revision level when setting the flags,
2175 * so there is a chance incompat flags are set on a rev 0 filesystem.
2176 */
2177 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
2178 if (features) {
2179 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
2180 "unsupported optional features (%x).\n", sb->s_id,
2181 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2182 ~EXT4_FEATURE_INCOMPAT_SUPP));
2183 goto failed_mount;
2184 }
2185 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
2186 if (!(sb->s_flags & MS_RDONLY) && features) {
2187 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
2188 "unsupported optional features (%x).\n", sb->s_id,
2189 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2190 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2191 goto failed_mount;
2192 }
2193 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2194 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2195 if (has_huge_files) {
2196 /*
2197 * Large file size enabled file system can only be
2198 * mount if kernel is build with CONFIG_LBD
2199 */
2200 if (sizeof(root->i_blocks) < sizeof(u64) &&
2201 !(sb->s_flags & MS_RDONLY)) {
2202 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2203 "files cannot be mounted read-write "
2204 "without CONFIG_LBD.\n", sb->s_id);
2205 goto failed_mount;
2206 }
2207 }
2208 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2209
2210 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2211 blocksize > EXT4_MAX_BLOCK_SIZE) {
2212 printk(KERN_ERR
2213 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
2214 blocksize, sb->s_id);
2215 goto failed_mount;
2216 }
2217
2218 if (sb->s_blocksize != blocksize) {
2219
2220 /* Validate the filesystem blocksize */
2221 if (!sb_set_blocksize(sb, blocksize)) {
2222 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2223 blocksize);
2224 goto failed_mount;
2225 }
2226
2227 brelse(bh);
2228 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2229 offset = do_div(logical_sb_block, blocksize);
2230 bh = sb_bread(sb, logical_sb_block);
2231 if (!bh) {
2232 printk(KERN_ERR
2233 "EXT4-fs: Can't read superblock on 2nd try.\n");
2234 goto failed_mount;
2235 }
2236 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2237 sbi->s_es = es;
2238 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2239 printk(KERN_ERR
2240 "EXT4-fs: Magic mismatch, very weird !\n");
2241 goto failed_mount;
2242 }
2243 }
2244
2245 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2246 has_huge_files);
2247 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2248
2249 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2250 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2251 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2252 } else {
2253 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2254 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2255 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2256 (!is_power_of_2(sbi->s_inode_size)) ||
2257 (sbi->s_inode_size > blocksize)) {
2258 printk(KERN_ERR
2259 "EXT4-fs: unsupported inode size: %d\n",
2260 sbi->s_inode_size);
2261 goto failed_mount;
2262 }
2263 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2264 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2265 }
2266 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2267 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2268 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2269 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2270 !is_power_of_2(sbi->s_desc_size)) {
2271 printk(KERN_ERR
2272 "EXT4-fs: unsupported descriptor size %lu\n",
2273 sbi->s_desc_size);
2274 goto failed_mount;
2275 }
2276 } else
2277 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2278 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2279 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2280 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2281 goto cantfind_ext4;
2282 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2283 if (sbi->s_inodes_per_block == 0)
2284 goto cantfind_ext4;
2285 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2286 sbi->s_inodes_per_block;
2287 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2288 sbi->s_sbh = bh;
2289 sbi->s_mount_state = le16_to_cpu(es->s_state);
2290 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2291 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2292 for (i = 0; i < 4; i++)
2293 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2294 sbi->s_def_hash_version = es->s_def_hash_version;
2295 i = le32_to_cpu(es->s_flags);
2296 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2297 sbi->s_hash_unsigned = 3;
2298 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2299 #ifdef __CHAR_UNSIGNED__
2300 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2301 sbi->s_hash_unsigned = 3;
2302 #else
2303 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2304 #endif
2305 sb->s_dirt = 1;
2306 }
2307
2308 if (sbi->s_blocks_per_group > blocksize * 8) {
2309 printk(KERN_ERR
2310 "EXT4-fs: #blocks per group too big: %lu\n",
2311 sbi->s_blocks_per_group);
2312 goto failed_mount;
2313 }
2314 if (sbi->s_inodes_per_group > blocksize * 8) {
2315 printk(KERN_ERR
2316 "EXT4-fs: #inodes per group too big: %lu\n",
2317 sbi->s_inodes_per_group);
2318 goto failed_mount;
2319 }
2320
2321 if (ext4_blocks_count(es) >
2322 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
2323 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
2324 " too large to mount safely\n", sb->s_id);
2325 if (sizeof(sector_t) < 8)
2326 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
2327 "enabled\n");
2328 goto failed_mount;
2329 }
2330
2331 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2332 goto cantfind_ext4;
2333
2334 /*
2335 * It makes no sense for the first data block to be beyond the end
2336 * of the filesystem.
2337 */
2338 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2339 printk(KERN_WARNING "EXT4-fs: bad geometry: first data"
2340 "block %u is beyond end of filesystem (%llu)\n",
2341 le32_to_cpu(es->s_first_data_block),
2342 ext4_blocks_count(es));
2343 goto failed_mount;
2344 }
2345 blocks_count = (ext4_blocks_count(es) -
2346 le32_to_cpu(es->s_first_data_block) +
2347 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2348 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2349 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2350 printk(KERN_WARNING "EXT4-fs: groups count too large: %u "
2351 "(block count %llu, first data block %u, "
2352 "blocks per group %lu)\n", sbi->s_groups_count,
2353 ext4_blocks_count(es),
2354 le32_to_cpu(es->s_first_data_block),
2355 EXT4_BLOCKS_PER_GROUP(sb));
2356 goto failed_mount;
2357 }
2358 sbi->s_groups_count = blocks_count;
2359 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2360 EXT4_DESC_PER_BLOCK(sb);
2361 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2362 GFP_KERNEL);
2363 if (sbi->s_group_desc == NULL) {
2364 printk(KERN_ERR "EXT4-fs: not enough memory\n");
2365 goto failed_mount;
2366 }
2367
2368 #ifdef CONFIG_PROC_FS
2369 if (ext4_proc_root)
2370 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2371
2372 if (sbi->s_proc)
2373 proc_create_data("inode_readahead_blks", 0644, sbi->s_proc,
2374 &ext4_ui_proc_fops,
2375 &sbi->s_inode_readahead_blks);
2376 #endif
2377
2378 bgl_lock_init(&sbi->s_blockgroup_lock);
2379
2380 for (i = 0; i < db_count; i++) {
2381 block = descriptor_loc(sb, logical_sb_block, i);
2382 sbi->s_group_desc[i] = sb_bread(sb, block);
2383 if (!sbi->s_group_desc[i]) {
2384 printk(KERN_ERR "EXT4-fs: "
2385 "can't read group descriptor %d\n", i);
2386 db_count = i;
2387 goto failed_mount2;
2388 }
2389 }
2390 if (!ext4_check_descriptors(sb)) {
2391 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
2392 goto failed_mount2;
2393 }
2394 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2395 if (!ext4_fill_flex_info(sb)) {
2396 printk(KERN_ERR
2397 "EXT4-fs: unable to initialize "
2398 "flex_bg meta info!\n");
2399 goto failed_mount2;
2400 }
2401
2402 sbi->s_gdb_count = db_count;
2403 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2404 spin_lock_init(&sbi->s_next_gen_lock);
2405
2406 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2407 ext4_count_free_blocks(sb));
2408 if (!err) {
2409 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2410 ext4_count_free_inodes(sb));
2411 }
2412 if (!err) {
2413 err = percpu_counter_init(&sbi->s_dirs_counter,
2414 ext4_count_dirs(sb));
2415 }
2416 if (!err) {
2417 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2418 }
2419 if (err) {
2420 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2421 goto failed_mount3;
2422 }
2423
2424 sbi->s_stripe = ext4_get_stripe_size(sbi);
2425
2426 /*
2427 * set up enough so that it can read an inode
2428 */
2429 sb->s_op = &ext4_sops;
2430 sb->s_export_op = &ext4_export_ops;
2431 sb->s_xattr = ext4_xattr_handlers;
2432 #ifdef CONFIG_QUOTA
2433 sb->s_qcop = &ext4_qctl_operations;
2434 sb->dq_op = &ext4_quota_operations;
2435 #endif
2436 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2437
2438 sb->s_root = NULL;
2439
2440 needs_recovery = (es->s_last_orphan != 0 ||
2441 EXT4_HAS_INCOMPAT_FEATURE(sb,
2442 EXT4_FEATURE_INCOMPAT_RECOVER));
2443
2444 /*
2445 * The first inode we look at is the journal inode. Don't try
2446 * root first: it may be modified in the journal!
2447 */
2448 if (!test_opt(sb, NOLOAD) &&
2449 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2450 if (ext4_load_journal(sb, es, journal_devnum))
2451 goto failed_mount3;
2452 if (!(sb->s_flags & MS_RDONLY) &&
2453 EXT4_SB(sb)->s_journal->j_failed_commit) {
2454 printk(KERN_CRIT "EXT4-fs error (device %s): "
2455 "ext4_fill_super: Journal transaction "
2456 "%u is corrupt\n", sb->s_id,
2457 EXT4_SB(sb)->s_journal->j_failed_commit);
2458 if (test_opt(sb, ERRORS_RO)) {
2459 printk(KERN_CRIT
2460 "Mounting filesystem read-only\n");
2461 sb->s_flags |= MS_RDONLY;
2462 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2463 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2464 }
2465 if (test_opt(sb, ERRORS_PANIC)) {
2466 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2467 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2468 ext4_commit_super(sb, es, 1);
2469 goto failed_mount4;
2470 }
2471 }
2472 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2473 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2474 printk(KERN_ERR "EXT4-fs: required journal recovery "
2475 "suppressed and not mounted read-only\n");
2476 goto failed_mount4;
2477 } else {
2478 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2479 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2480 sbi->s_journal = NULL;
2481 needs_recovery = 0;
2482 goto no_journal;
2483 }
2484
2485 if (ext4_blocks_count(es) > 0xffffffffULL &&
2486 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2487 JBD2_FEATURE_INCOMPAT_64BIT)) {
2488 printk(KERN_ERR "EXT4-fs: Failed to set 64-bit journal feature\n");
2489 goto failed_mount4;
2490 }
2491
2492 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2493 jbd2_journal_set_features(sbi->s_journal,
2494 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2495 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2496 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2497 jbd2_journal_set_features(sbi->s_journal,
2498 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2499 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2500 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2501 } else {
2502 jbd2_journal_clear_features(sbi->s_journal,
2503 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2504 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2505 }
2506
2507 /* We have now updated the journal if required, so we can
2508 * validate the data journaling mode. */
2509 switch (test_opt(sb, DATA_FLAGS)) {
2510 case 0:
2511 /* No mode set, assume a default based on the journal
2512 * capabilities: ORDERED_DATA if the journal can
2513 * cope, else JOURNAL_DATA
2514 */
2515 if (jbd2_journal_check_available_features
2516 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2517 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2518 else
2519 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2520 break;
2521
2522 case EXT4_MOUNT_ORDERED_DATA:
2523 case EXT4_MOUNT_WRITEBACK_DATA:
2524 if (!jbd2_journal_check_available_features
2525 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2526 printk(KERN_ERR "EXT4-fs: Journal does not support "
2527 "requested data journaling mode\n");
2528 goto failed_mount4;
2529 }
2530 default:
2531 break;
2532 }
2533 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2534
2535 no_journal:
2536
2537 if (test_opt(sb, NOBH)) {
2538 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2539 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
2540 "its supported only with writeback mode\n");
2541 clear_opt(sbi->s_mount_opt, NOBH);
2542 }
2543 }
2544 /*
2545 * The jbd2_journal_load will have done any necessary log recovery,
2546 * so we can safely mount the rest of the filesystem now.
2547 */
2548
2549 root = ext4_iget(sb, EXT4_ROOT_INO);
2550 if (IS_ERR(root)) {
2551 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
2552 ret = PTR_ERR(root);
2553 goto failed_mount4;
2554 }
2555 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2556 iput(root);
2557 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
2558 goto failed_mount4;
2559 }
2560 sb->s_root = d_alloc_root(root);
2561 if (!sb->s_root) {
2562 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2563 iput(root);
2564 ret = -ENOMEM;
2565 goto failed_mount4;
2566 }
2567
2568 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2569
2570 /* determine the minimum size of new large inodes, if present */
2571 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2572 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2573 EXT4_GOOD_OLD_INODE_SIZE;
2574 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2575 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2576 if (sbi->s_want_extra_isize <
2577 le16_to_cpu(es->s_want_extra_isize))
2578 sbi->s_want_extra_isize =
2579 le16_to_cpu(es->s_want_extra_isize);
2580 if (sbi->s_want_extra_isize <
2581 le16_to_cpu(es->s_min_extra_isize))
2582 sbi->s_want_extra_isize =
2583 le16_to_cpu(es->s_min_extra_isize);
2584 }
2585 }
2586 /* Check if enough inode space is available */
2587 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2588 sbi->s_inode_size) {
2589 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2590 EXT4_GOOD_OLD_INODE_SIZE;
2591 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2592 "available.\n");
2593 }
2594
2595 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2596 printk(KERN_WARNING "EXT4-fs: Ignoring delalloc option - "
2597 "requested data journaling mode\n");
2598 clear_opt(sbi->s_mount_opt, DELALLOC);
2599 } else if (test_opt(sb, DELALLOC))
2600 printk(KERN_INFO "EXT4-fs: delayed allocation enabled\n");
2601
2602 ext4_ext_init(sb);
2603 err = ext4_mb_init(sb, needs_recovery);
2604 if (err) {
2605 printk(KERN_ERR "EXT4-fs: failed to initalize mballoc (%d)\n",
2606 err);
2607 goto failed_mount4;
2608 }
2609
2610 /*
2611 * akpm: core read_super() calls in here with the superblock locked.
2612 * That deadlocks, because orphan cleanup needs to lock the superblock
2613 * in numerous places. Here we just pop the lock - it's relatively
2614 * harmless, because we are now ready to accept write_super() requests,
2615 * and aviro says that's the only reason for hanging onto the
2616 * superblock lock.
2617 */
2618 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2619 ext4_orphan_cleanup(sb, es);
2620 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2621 if (needs_recovery) {
2622 printk(KERN_INFO "EXT4-fs: recovery complete.\n");
2623 ext4_mark_recovery_complete(sb, es);
2624 }
2625 if (EXT4_SB(sb)->s_journal) {
2626 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2627 descr = " journalled data mode";
2628 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2629 descr = " ordered data mode";
2630 else
2631 descr = " writeback data mode";
2632 } else
2633 descr = "out journal";
2634
2635 printk(KERN_INFO "EXT4-fs: mounted filesystem %s with%s\n",
2636 sb->s_id, descr);
2637
2638 lock_kernel();
2639 return 0;
2640
2641 cantfind_ext4:
2642 if (!silent)
2643 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
2644 sb->s_id);
2645 goto failed_mount;
2646
2647 failed_mount4:
2648 printk(KERN_ERR "EXT4-fs (device %s): mount failed\n", sb->s_id);
2649 if (sbi->s_journal) {
2650 jbd2_journal_destroy(sbi->s_journal);
2651 sbi->s_journal = NULL;
2652 }
2653 failed_mount3:
2654 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2655 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2656 percpu_counter_destroy(&sbi->s_dirs_counter);
2657 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2658 failed_mount2:
2659 for (i = 0; i < db_count; i++)
2660 brelse(sbi->s_group_desc[i]);
2661 kfree(sbi->s_group_desc);
2662 failed_mount:
2663 if (sbi->s_proc) {
2664 remove_proc_entry("inode_readahead_blks", sbi->s_proc);
2665 remove_proc_entry(sb->s_id, ext4_proc_root);
2666 }
2667 #ifdef CONFIG_QUOTA
2668 for (i = 0; i < MAXQUOTAS; i++)
2669 kfree(sbi->s_qf_names[i]);
2670 #endif
2671 ext4_blkdev_remove(sbi);
2672 brelse(bh);
2673 out_fail:
2674 sb->s_fs_info = NULL;
2675 kfree(sbi);
2676 lock_kernel();
2677 return ret;
2678 }
2679
2680 /*
2681 * Setup any per-fs journal parameters now. We'll do this both on
2682 * initial mount, once the journal has been initialised but before we've
2683 * done any recovery; and again on any subsequent remount.
2684 */
2685 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2686 {
2687 struct ext4_sb_info *sbi = EXT4_SB(sb);
2688
2689 journal->j_commit_interval = sbi->s_commit_interval;
2690 journal->j_min_batch_time = sbi->s_min_batch_time;
2691 journal->j_max_batch_time = sbi->s_max_batch_time;
2692
2693 spin_lock(&journal->j_state_lock);
2694 if (test_opt(sb, BARRIER))
2695 journal->j_flags |= JBD2_BARRIER;
2696 else
2697 journal->j_flags &= ~JBD2_BARRIER;
2698 if (test_opt(sb, DATA_ERR_ABORT))
2699 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2700 else
2701 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
2702 spin_unlock(&journal->j_state_lock);
2703 }
2704
2705 static journal_t *ext4_get_journal(struct super_block *sb,
2706 unsigned int journal_inum)
2707 {
2708 struct inode *journal_inode;
2709 journal_t *journal;
2710
2711 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2712
2713 /* First, test for the existence of a valid inode on disk. Bad
2714 * things happen if we iget() an unused inode, as the subsequent
2715 * iput() will try to delete it. */
2716
2717 journal_inode = ext4_iget(sb, journal_inum);
2718 if (IS_ERR(journal_inode)) {
2719 printk(KERN_ERR "EXT4-fs: no journal found.\n");
2720 return NULL;
2721 }
2722 if (!journal_inode->i_nlink) {
2723 make_bad_inode(journal_inode);
2724 iput(journal_inode);
2725 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
2726 return NULL;
2727 }
2728
2729 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
2730 journal_inode, journal_inode->i_size);
2731 if (!S_ISREG(journal_inode->i_mode)) {
2732 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
2733 iput(journal_inode);
2734 return NULL;
2735 }
2736
2737 journal = jbd2_journal_init_inode(journal_inode);
2738 if (!journal) {
2739 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
2740 iput(journal_inode);
2741 return NULL;
2742 }
2743 journal->j_private = sb;
2744 ext4_init_journal_params(sb, journal);
2745 return journal;
2746 }
2747
2748 static journal_t *ext4_get_dev_journal(struct super_block *sb,
2749 dev_t j_dev)
2750 {
2751 struct buffer_head *bh;
2752 journal_t *journal;
2753 ext4_fsblk_t start;
2754 ext4_fsblk_t len;
2755 int hblock, blocksize;
2756 ext4_fsblk_t sb_block;
2757 unsigned long offset;
2758 struct ext4_super_block *es;
2759 struct block_device *bdev;
2760
2761 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2762
2763 bdev = ext4_blkdev_get(j_dev);
2764 if (bdev == NULL)
2765 return NULL;
2766
2767 if (bd_claim(bdev, sb)) {
2768 printk(KERN_ERR
2769 "EXT4-fs: failed to claim external journal device.\n");
2770 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
2771 return NULL;
2772 }
2773
2774 blocksize = sb->s_blocksize;
2775 hblock = bdev_hardsect_size(bdev);
2776 if (blocksize < hblock) {
2777 printk(KERN_ERR
2778 "EXT4-fs: blocksize too small for journal device.\n");
2779 goto out_bdev;
2780 }
2781
2782 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2783 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
2784 set_blocksize(bdev, blocksize);
2785 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2786 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
2787 "external journal\n");
2788 goto out_bdev;
2789 }
2790
2791 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2792 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
2793 !(le32_to_cpu(es->s_feature_incompat) &
2794 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2795 printk(KERN_ERR "EXT4-fs: external journal has "
2796 "bad superblock\n");
2797 brelse(bh);
2798 goto out_bdev;
2799 }
2800
2801 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2802 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
2803 brelse(bh);
2804 goto out_bdev;
2805 }
2806
2807 len = ext4_blocks_count(es);
2808 start = sb_block + 1;
2809 brelse(bh); /* we're done with the superblock */
2810
2811 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
2812 start, len, blocksize);
2813 if (!journal) {
2814 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
2815 goto out_bdev;
2816 }
2817 journal->j_private = sb;
2818 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2819 wait_on_buffer(journal->j_sb_buffer);
2820 if (!buffer_uptodate(journal->j_sb_buffer)) {
2821 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
2822 goto out_journal;
2823 }
2824 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2825 printk(KERN_ERR "EXT4-fs: External journal has more than one "
2826 "user (unsupported) - %d\n",
2827 be32_to_cpu(journal->j_superblock->s_nr_users));
2828 goto out_journal;
2829 }
2830 EXT4_SB(sb)->journal_bdev = bdev;
2831 ext4_init_journal_params(sb, journal);
2832 return journal;
2833 out_journal:
2834 jbd2_journal_destroy(journal);
2835 out_bdev:
2836 ext4_blkdev_put(bdev);
2837 return NULL;
2838 }
2839
2840 static int ext4_load_journal(struct super_block *sb,
2841 struct ext4_super_block *es,
2842 unsigned long journal_devnum)
2843 {
2844 journal_t *journal;
2845 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2846 dev_t journal_dev;
2847 int err = 0;
2848 int really_read_only;
2849
2850 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2851
2852 if (journal_devnum &&
2853 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2854 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
2855 "numbers have changed\n");
2856 journal_dev = new_decode_dev(journal_devnum);
2857 } else
2858 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2859
2860 really_read_only = bdev_read_only(sb->s_bdev);
2861
2862 /*
2863 * Are we loading a blank journal or performing recovery after a
2864 * crash? For recovery, we need to check in advance whether we
2865 * can get read-write access to the device.
2866 */
2867
2868 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2869 if (sb->s_flags & MS_RDONLY) {
2870 printk(KERN_INFO "EXT4-fs: INFO: recovery "
2871 "required on readonly filesystem.\n");
2872 if (really_read_only) {
2873 printk(KERN_ERR "EXT4-fs: write access "
2874 "unavailable, cannot proceed.\n");
2875 return -EROFS;
2876 }
2877 printk(KERN_INFO "EXT4-fs: write access will "
2878 "be enabled during recovery.\n");
2879 }
2880 }
2881
2882 if (journal_inum && journal_dev) {
2883 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
2884 "and inode journals!\n");
2885 return -EINVAL;
2886 }
2887
2888 if (journal_inum) {
2889 if (!(journal = ext4_get_journal(sb, journal_inum)))
2890 return -EINVAL;
2891 } else {
2892 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
2893 return -EINVAL;
2894 }
2895
2896 if (journal->j_flags & JBD2_BARRIER)
2897 printk(KERN_INFO "EXT4-fs: barriers enabled\n");
2898 else
2899 printk(KERN_INFO "EXT4-fs: barriers disabled\n");
2900
2901 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2902 err = jbd2_journal_update_format(journal);
2903 if (err) {
2904 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
2905 jbd2_journal_destroy(journal);
2906 return err;
2907 }
2908 }
2909
2910 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
2911 err = jbd2_journal_wipe(journal, !really_read_only);
2912 if (!err)
2913 err = jbd2_journal_load(journal);
2914
2915 if (err) {
2916 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
2917 jbd2_journal_destroy(journal);
2918 return err;
2919 }
2920
2921 EXT4_SB(sb)->s_journal = journal;
2922 ext4_clear_journal_err(sb, es);
2923
2924 if (journal_devnum &&
2925 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2926 es->s_journal_dev = cpu_to_le32(journal_devnum);
2927 sb->s_dirt = 1;
2928
2929 /* Make sure we flush the recovery flag to disk. */
2930 ext4_commit_super(sb, es, 1);
2931 }
2932
2933 return 0;
2934 }
2935
2936 static void ext4_commit_super(struct super_block *sb,
2937 struct ext4_super_block *es, int sync)
2938 {
2939 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
2940
2941 if (!sbh)
2942 return;
2943 if (buffer_write_io_error(sbh)) {
2944 /*
2945 * Oh, dear. A previous attempt to write the
2946 * superblock failed. This could happen because the
2947 * USB device was yanked out. Or it could happen to
2948 * be a transient write error and maybe the block will
2949 * be remapped. Nothing we can do but to retry the
2950 * write and hope for the best.
2951 */
2952 printk(KERN_ERR "EXT4-fs: previous I/O error to "
2953 "superblock detected for %s.\n", sb->s_id);
2954 clear_buffer_write_io_error(sbh);
2955 set_buffer_uptodate(sbh);
2956 }
2957 es->s_wtime = cpu_to_le32(get_seconds());
2958 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
2959 &EXT4_SB(sb)->s_freeblocks_counter));
2960 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
2961 &EXT4_SB(sb)->s_freeinodes_counter));
2962
2963 BUFFER_TRACE(sbh, "marking dirty");
2964 mark_buffer_dirty(sbh);
2965 if (sync) {
2966 sync_dirty_buffer(sbh);
2967 if (buffer_write_io_error(sbh)) {
2968 printk(KERN_ERR "EXT4-fs: I/O error while writing "
2969 "superblock for %s.\n", sb->s_id);
2970 clear_buffer_write_io_error(sbh);
2971 set_buffer_uptodate(sbh);
2972 }
2973 }
2974 }
2975
2976
2977 /*
2978 * Have we just finished recovery? If so, and if we are mounting (or
2979 * remounting) the filesystem readonly, then we will end up with a
2980 * consistent fs on disk. Record that fact.
2981 */
2982 static void ext4_mark_recovery_complete(struct super_block *sb,
2983 struct ext4_super_block *es)
2984 {
2985 journal_t *journal = EXT4_SB(sb)->s_journal;
2986
2987 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2988 BUG_ON(journal != NULL);
2989 return;
2990 }
2991 jbd2_journal_lock_updates(journal);
2992 if (jbd2_journal_flush(journal) < 0)
2993 goto out;
2994
2995 lock_super(sb);
2996 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
2997 sb->s_flags & MS_RDONLY) {
2998 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2999 sb->s_dirt = 0;
3000 ext4_commit_super(sb, es, 1);
3001 }
3002 unlock_super(sb);
3003
3004 out:
3005 jbd2_journal_unlock_updates(journal);
3006 }
3007
3008 /*
3009 * If we are mounting (or read-write remounting) a filesystem whose journal
3010 * has recorded an error from a previous lifetime, move that error to the
3011 * main filesystem now.
3012 */
3013 static void ext4_clear_journal_err(struct super_block *sb,
3014 struct ext4_super_block *es)
3015 {
3016 journal_t *journal;
3017 int j_errno;
3018 const char *errstr;
3019
3020 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3021
3022 journal = EXT4_SB(sb)->s_journal;
3023
3024 /*
3025 * Now check for any error status which may have been recorded in the
3026 * journal by a prior ext4_error() or ext4_abort()
3027 */
3028
3029 j_errno = jbd2_journal_errno(journal);
3030 if (j_errno) {
3031 char nbuf[16];
3032
3033 errstr = ext4_decode_error(sb, j_errno, nbuf);
3034 ext4_warning(sb, __func__, "Filesystem error recorded "
3035 "from previous mount: %s", errstr);
3036 ext4_warning(sb, __func__, "Marking fs in need of "
3037 "filesystem check.");
3038
3039 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3040 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3041 ext4_commit_super(sb, es, 1);
3042
3043 jbd2_journal_clear_err(journal);
3044 }
3045 }
3046
3047 /*
3048 * Force the running and committing transactions to commit,
3049 * and wait on the commit.
3050 */
3051 int ext4_force_commit(struct super_block *sb)
3052 {
3053 journal_t *journal;
3054 int ret = 0;
3055
3056 if (sb->s_flags & MS_RDONLY)
3057 return 0;
3058
3059 journal = EXT4_SB(sb)->s_journal;
3060 if (journal) {
3061 sb->s_dirt = 0;
3062 ret = ext4_journal_force_commit(journal);
3063 }
3064
3065 return ret;
3066 }
3067
3068 /*
3069 * Ext4 always journals updates to the superblock itself, so we don't
3070 * have to propagate any other updates to the superblock on disk at this
3071 * point. (We can probably nuke this function altogether, and remove
3072 * any mention to sb->s_dirt in all of fs/ext4; eventual cleanup...)
3073 */
3074 static void ext4_write_super(struct super_block *sb)
3075 {
3076 if (EXT4_SB(sb)->s_journal) {
3077 if (mutex_trylock(&sb->s_lock) != 0)
3078 BUG();
3079 sb->s_dirt = 0;
3080 } else {
3081 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
3082 }
3083 }
3084
3085 static int ext4_sync_fs(struct super_block *sb, int wait)
3086 {
3087 int ret = 0;
3088
3089 trace_mark(ext4_sync_fs, "dev %s wait %d", sb->s_id, wait);
3090 sb->s_dirt = 0;
3091 if (EXT4_SB(sb)->s_journal) {
3092 if (wait)
3093 ret = ext4_force_commit(sb);
3094 else
3095 jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, NULL);
3096 } else {
3097 ext4_commit_super(sb, EXT4_SB(sb)->s_es, wait);
3098 }
3099 return ret;
3100 }
3101
3102 /*
3103 * LVM calls this function before a (read-only) snapshot is created. This
3104 * gives us a chance to flush the journal completely and mark the fs clean.
3105 */
3106 static void ext4_write_super_lockfs(struct super_block *sb)
3107 {
3108 sb->s_dirt = 0;
3109
3110 if (!(sb->s_flags & MS_RDONLY)) {
3111 journal_t *journal = EXT4_SB(sb)->s_journal;
3112
3113 if (journal) {
3114 /* Now we set up the journal barrier. */
3115 jbd2_journal_lock_updates(journal);
3116
3117 /*
3118 * We don't want to clear needs_recovery flag when we
3119 * failed to flush the journal.
3120 */
3121 if (jbd2_journal_flush(journal) < 0)
3122 return;
3123 }
3124
3125 /* Journal blocked and flushed, clear needs_recovery flag. */
3126 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3127 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
3128 }
3129 }
3130
3131 /*
3132 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3133 * flag here, even though the filesystem is not technically dirty yet.
3134 */
3135 static void ext4_unlockfs(struct super_block *sb)
3136 {
3137 if (EXT4_SB(sb)->s_journal && !(sb->s_flags & MS_RDONLY)) {
3138 lock_super(sb);
3139 /* Reser the needs_recovery flag before the fs is unlocked. */
3140 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3141 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
3142 unlock_super(sb);
3143 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3144 }
3145 }
3146
3147 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3148 {
3149 struct ext4_super_block *es;
3150 struct ext4_sb_info *sbi = EXT4_SB(sb);
3151 ext4_fsblk_t n_blocks_count = 0;
3152 unsigned long old_sb_flags;
3153 struct ext4_mount_options old_opts;
3154 ext4_group_t g;
3155 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3156 int err;
3157 #ifdef CONFIG_QUOTA
3158 int i;
3159 #endif
3160
3161 /* Store the original options */
3162 old_sb_flags = sb->s_flags;
3163 old_opts.s_mount_opt = sbi->s_mount_opt;
3164 old_opts.s_resuid = sbi->s_resuid;
3165 old_opts.s_resgid = sbi->s_resgid;
3166 old_opts.s_commit_interval = sbi->s_commit_interval;
3167 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3168 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3169 #ifdef CONFIG_QUOTA
3170 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3171 for (i = 0; i < MAXQUOTAS; i++)
3172 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3173 #endif
3174 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3175 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3176
3177 /*
3178 * Allow the "check" option to be passed as a remount option.
3179 */
3180 if (!parse_options(data, sb, NULL, &journal_ioprio,
3181 &n_blocks_count, 1)) {
3182 err = -EINVAL;
3183 goto restore_opts;
3184 }
3185
3186 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
3187 ext4_abort(sb, __func__, "Abort forced by user");
3188
3189 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3190 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3191
3192 es = sbi->s_es;
3193
3194 if (sbi->s_journal) {
3195 ext4_init_journal_params(sb, sbi->s_journal);
3196 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3197 }
3198
3199 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3200 n_blocks_count > ext4_blocks_count(es)) {
3201 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
3202 err = -EROFS;
3203 goto restore_opts;
3204 }
3205
3206 if (*flags & MS_RDONLY) {
3207 /*
3208 * First of all, the unconditional stuff we have to do
3209 * to disable replay of the journal when we next remount
3210 */
3211 sb->s_flags |= MS_RDONLY;
3212
3213 /*
3214 * OK, test if we are remounting a valid rw partition
3215 * readonly, and if so set the rdonly flag and then
3216 * mark the partition as valid again.
3217 */
3218 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3219 (sbi->s_mount_state & EXT4_VALID_FS))
3220 es->s_state = cpu_to_le16(sbi->s_mount_state);
3221
3222 /*
3223 * We have to unlock super so that we can wait for
3224 * transactions.
3225 */
3226 if (sbi->s_journal) {
3227 unlock_super(sb);
3228 ext4_mark_recovery_complete(sb, es);
3229 lock_super(sb);
3230 }
3231 } else {
3232 int ret;
3233 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3234 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
3235 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3236 "remount RDWR because of unsupported "
3237 "optional features (%x).\n", sb->s_id,
3238 (le32_to_cpu(sbi->s_es->s_feature_ro_compat) &
3239 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3240 err = -EROFS;
3241 goto restore_opts;
3242 }
3243
3244 /*
3245 * Make sure the group descriptor checksums
3246 * are sane. If they aren't, refuse to
3247 * remount r/w.
3248 */
3249 for (g = 0; g < sbi->s_groups_count; g++) {
3250 struct ext4_group_desc *gdp =
3251 ext4_get_group_desc(sb, g, NULL);
3252
3253 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3254 printk(KERN_ERR
3255 "EXT4-fs: ext4_remount: "
3256 "Checksum for group %u failed (%u!=%u)\n",
3257 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3258 le16_to_cpu(gdp->bg_checksum));
3259 err = -EINVAL;
3260 goto restore_opts;
3261 }
3262 }
3263
3264 /*
3265 * If we have an unprocessed orphan list hanging
3266 * around from a previously readonly bdev mount,
3267 * require a full umount/remount for now.
3268 */
3269 if (es->s_last_orphan) {
3270 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3271 "remount RDWR because of unprocessed "
3272 "orphan inode list. Please "
3273 "umount/remount instead.\n",
3274 sb->s_id);
3275 err = -EINVAL;
3276 goto restore_opts;
3277 }
3278
3279 /*
3280 * Mounting a RDONLY partition read-write, so reread
3281 * and store the current valid flag. (It may have
3282 * been changed by e2fsck since we originally mounted
3283 * the partition.)
3284 */
3285 if (sbi->s_journal)
3286 ext4_clear_journal_err(sb, es);
3287 sbi->s_mount_state = le16_to_cpu(es->s_state);
3288 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3289 goto restore_opts;
3290 if (!ext4_setup_super(sb, es, 0))
3291 sb->s_flags &= ~MS_RDONLY;
3292 }
3293 }
3294 if (sbi->s_journal == NULL)
3295 ext4_commit_super(sb, es, 1);
3296
3297 #ifdef CONFIG_QUOTA
3298 /* Release old quota file names */
3299 for (i = 0; i < MAXQUOTAS; i++)
3300 if (old_opts.s_qf_names[i] &&
3301 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3302 kfree(old_opts.s_qf_names[i]);
3303 #endif
3304 return 0;
3305 restore_opts:
3306 sb->s_flags = old_sb_flags;
3307 sbi->s_mount_opt = old_opts.s_mount_opt;
3308 sbi->s_resuid = old_opts.s_resuid;
3309 sbi->s_resgid = old_opts.s_resgid;
3310 sbi->s_commit_interval = old_opts.s_commit_interval;
3311 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3312 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3313 #ifdef CONFIG_QUOTA
3314 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3315 for (i = 0; i < MAXQUOTAS; i++) {
3316 if (sbi->s_qf_names[i] &&
3317 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3318 kfree(sbi->s_qf_names[i]);
3319 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3320 }
3321 #endif
3322 return err;
3323 }
3324
3325 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3326 {
3327 struct super_block *sb = dentry->d_sb;
3328 struct ext4_sb_info *sbi = EXT4_SB(sb);
3329 struct ext4_super_block *es = sbi->s_es;
3330 u64 fsid;
3331
3332 if (test_opt(sb, MINIX_DF)) {
3333 sbi->s_overhead_last = 0;
3334 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3335 ext4_group_t ngroups = sbi->s_groups_count, i;
3336 ext4_fsblk_t overhead = 0;
3337 smp_rmb();
3338
3339 /*
3340 * Compute the overhead (FS structures). This is constant
3341 * for a given filesystem unless the number of block groups
3342 * changes so we cache the previous value until it does.
3343 */
3344
3345 /*
3346 * All of the blocks before first_data_block are
3347 * overhead
3348 */
3349 overhead = le32_to_cpu(es->s_first_data_block);
3350
3351 /*
3352 * Add the overhead attributed to the superblock and
3353 * block group descriptors. If the sparse superblocks
3354 * feature is turned on, then not all groups have this.
3355 */
3356 for (i = 0; i < ngroups; i++) {
3357 overhead += ext4_bg_has_super(sb, i) +
3358 ext4_bg_num_gdb(sb, i);
3359 cond_resched();
3360 }
3361
3362 /*
3363 * Every block group has an inode bitmap, a block
3364 * bitmap, and an inode table.
3365 */
3366 overhead += ngroups * (2 + sbi->s_itb_per_group);
3367 sbi->s_overhead_last = overhead;
3368 smp_wmb();
3369 sbi->s_blocks_last = ext4_blocks_count(es);
3370 }
3371
3372 buf->f_type = EXT4_SUPER_MAGIC;
3373 buf->f_bsize = sb->s_blocksize;
3374 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3375 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3376 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3377 ext4_free_blocks_count_set(es, buf->f_bfree);
3378 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3379 if (buf->f_bfree < ext4_r_blocks_count(es))
3380 buf->f_bavail = 0;
3381 buf->f_files = le32_to_cpu(es->s_inodes_count);
3382 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3383 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
3384 buf->f_namelen = EXT4_NAME_LEN;
3385 fsid = le64_to_cpup((void *)es->s_uuid) ^
3386 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3387 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3388 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3389 return 0;
3390 }
3391
3392 /* Helper function for writing quotas on sync - we need to start transaction before quota file
3393 * is locked for write. Otherwise the are possible deadlocks:
3394 * Process 1 Process 2
3395 * ext4_create() quota_sync()
3396 * jbd2_journal_start() write_dquot()
3397 * DQUOT_INIT() down(dqio_mutex)
3398 * down(dqio_mutex) jbd2_journal_start()
3399 *
3400 */
3401
3402 #ifdef CONFIG_QUOTA
3403
3404 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3405 {
3406 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3407 }
3408
3409 static int ext4_dquot_initialize(struct inode *inode, int type)
3410 {
3411 handle_t *handle;
3412 int ret, err;
3413
3414 /* We may create quota structure so we need to reserve enough blocks */
3415 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
3416 if (IS_ERR(handle))
3417 return PTR_ERR(handle);
3418 ret = dquot_initialize(inode, type);
3419 err = ext4_journal_stop(handle);
3420 if (!ret)
3421 ret = err;
3422 return ret;
3423 }
3424
3425 static int ext4_dquot_drop(struct inode *inode)
3426 {
3427 handle_t *handle;
3428 int ret, err;
3429
3430 /* We may delete quota structure so we need to reserve enough blocks */
3431 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
3432 if (IS_ERR(handle)) {
3433 /*
3434 * We call dquot_drop() anyway to at least release references
3435 * to quota structures so that umount does not hang.
3436 */
3437 dquot_drop(inode);
3438 return PTR_ERR(handle);
3439 }
3440 ret = dquot_drop(inode);
3441 err = ext4_journal_stop(handle);
3442 if (!ret)
3443 ret = err;
3444 return ret;
3445 }
3446
3447 static int ext4_write_dquot(struct dquot *dquot)
3448 {
3449 int ret, err;
3450 handle_t *handle;
3451 struct inode *inode;
3452
3453 inode = dquot_to_inode(dquot);
3454 handle = ext4_journal_start(inode,
3455 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3456 if (IS_ERR(handle))
3457 return PTR_ERR(handle);
3458 ret = dquot_commit(dquot);
3459 err = ext4_journal_stop(handle);
3460 if (!ret)
3461 ret = err;
3462 return ret;
3463 }
3464
3465 static int ext4_acquire_dquot(struct dquot *dquot)
3466 {
3467 int ret, err;
3468 handle_t *handle;
3469
3470 handle = ext4_journal_start(dquot_to_inode(dquot),
3471 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3472 if (IS_ERR(handle))
3473 return PTR_ERR(handle);
3474 ret = dquot_acquire(dquot);
3475 err = ext4_journal_stop(handle);
3476 if (!ret)
3477 ret = err;
3478 return ret;
3479 }
3480
3481 static int ext4_release_dquot(struct dquot *dquot)
3482 {
3483 int ret, err;
3484 handle_t *handle;
3485
3486 handle = ext4_journal_start(dquot_to_inode(dquot),
3487 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3488 if (IS_ERR(handle)) {
3489 /* Release dquot anyway to avoid endless cycle in dqput() */
3490 dquot_release(dquot);
3491 return PTR_ERR(handle);
3492 }
3493 ret = dquot_release(dquot);
3494 err = ext4_journal_stop(handle);
3495 if (!ret)
3496 ret = err;
3497 return ret;
3498 }
3499
3500 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3501 {
3502 /* Are we journaling quotas? */
3503 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3504 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3505 dquot_mark_dquot_dirty(dquot);
3506 return ext4_write_dquot(dquot);
3507 } else {
3508 return dquot_mark_dquot_dirty(dquot);
3509 }
3510 }
3511
3512 static int ext4_write_info(struct super_block *sb, int type)
3513 {
3514 int ret, err;
3515 handle_t *handle;
3516
3517 /* Data block + inode block */
3518 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3519 if (IS_ERR(handle))
3520 return PTR_ERR(handle);
3521 ret = dquot_commit_info(sb, type);
3522 err = ext4_journal_stop(handle);
3523 if (!ret)
3524 ret = err;
3525 return ret;
3526 }
3527
3528 /*
3529 * Turn on quotas during mount time - we need to find
3530 * the quota file and such...
3531 */
3532 static int ext4_quota_on_mount(struct super_block *sb, int type)
3533 {
3534 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3535 EXT4_SB(sb)->s_jquota_fmt, type);
3536 }
3537
3538 /*
3539 * Standard function to be called on quota_on
3540 */
3541 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3542 char *name, int remount)
3543 {
3544 int err;
3545 struct path path;
3546
3547 if (!test_opt(sb, QUOTA))
3548 return -EINVAL;
3549 /* When remounting, no checks are needed and in fact, name is NULL */
3550 if (remount)
3551 return vfs_quota_on(sb, type, format_id, name, remount);
3552
3553 err = kern_path(name, LOOKUP_FOLLOW, &path);
3554 if (err)
3555 return err;
3556
3557 /* Quotafile not on the same filesystem? */
3558 if (path.mnt->mnt_sb != sb) {
3559 path_put(&path);
3560 return -EXDEV;
3561 }
3562 /* Journaling quota? */
3563 if (EXT4_SB(sb)->s_qf_names[type]) {
3564 /* Quotafile not in fs root? */
3565 if (path.dentry->d_parent != sb->s_root)
3566 printk(KERN_WARNING
3567 "EXT4-fs: Quota file not on filesystem root. "
3568 "Journaled quota will not work.\n");
3569 }
3570
3571 /*
3572 * When we journal data on quota file, we have to flush journal to see
3573 * all updates to the file when we bypass pagecache...
3574 */
3575 if (EXT4_SB(sb)->s_journal &&
3576 ext4_should_journal_data(path.dentry->d_inode)) {
3577 /*
3578 * We don't need to lock updates but journal_flush() could
3579 * otherwise be livelocked...
3580 */
3581 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3582 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3583 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3584 if (err) {
3585 path_put(&path);
3586 return err;
3587 }
3588 }
3589
3590 err = vfs_quota_on_path(sb, type, format_id, &path);
3591 path_put(&path);
3592 return err;
3593 }
3594
3595 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3596 * acquiring the locks... As quota files are never truncated and quota code
3597 * itself serializes the operations (and noone else should touch the files)
3598 * we don't have to be afraid of races */
3599 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3600 size_t len, loff_t off)
3601 {
3602 struct inode *inode = sb_dqopt(sb)->files[type];
3603 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3604 int err = 0;
3605 int offset = off & (sb->s_blocksize - 1);
3606 int tocopy;
3607 size_t toread;
3608 struct buffer_head *bh;
3609 loff_t i_size = i_size_read(inode);
3610
3611 if (off > i_size)
3612 return 0;
3613 if (off+len > i_size)
3614 len = i_size-off;
3615 toread = len;
3616 while (toread > 0) {
3617 tocopy = sb->s_blocksize - offset < toread ?
3618 sb->s_blocksize - offset : toread;
3619 bh = ext4_bread(NULL, inode, blk, 0, &err);
3620 if (err)
3621 return err;
3622 if (!bh) /* A hole? */
3623 memset(data, 0, tocopy);
3624 else
3625 memcpy(data, bh->b_data+offset, tocopy);
3626 brelse(bh);
3627 offset = 0;
3628 toread -= tocopy;
3629 data += tocopy;
3630 blk++;
3631 }
3632 return len;
3633 }
3634
3635 /* Write to quotafile (we know the transaction is already started and has
3636 * enough credits) */
3637 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3638 const char *data, size_t len, loff_t off)
3639 {
3640 struct inode *inode = sb_dqopt(sb)->files[type];
3641 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3642 int err = 0;
3643 int offset = off & (sb->s_blocksize - 1);
3644 int tocopy;
3645 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3646 size_t towrite = len;
3647 struct buffer_head *bh;
3648 handle_t *handle = journal_current_handle();
3649
3650 if (EXT4_SB(sb)->s_journal && !handle) {
3651 printk(KERN_WARNING "EXT4-fs: Quota write (off=%llu, len=%llu)"
3652 " cancelled because transaction is not started.\n",
3653 (unsigned long long)off, (unsigned long long)len);
3654 return -EIO;
3655 }
3656 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3657 while (towrite > 0) {
3658 tocopy = sb->s_blocksize - offset < towrite ?
3659 sb->s_blocksize - offset : towrite;
3660 bh = ext4_bread(handle, inode, blk, 1, &err);
3661 if (!bh)
3662 goto out;
3663 if (journal_quota) {
3664 err = ext4_journal_get_write_access(handle, bh);
3665 if (err) {
3666 brelse(bh);
3667 goto out;
3668 }
3669 }
3670 lock_buffer(bh);
3671 memcpy(bh->b_data+offset, data, tocopy);
3672 flush_dcache_page(bh->b_page);
3673 unlock_buffer(bh);
3674 if (journal_quota)
3675 err = ext4_handle_dirty_metadata(handle, NULL, bh);
3676 else {
3677 /* Always do at least ordered writes for quotas */
3678 err = ext4_jbd2_file_inode(handle, inode);
3679 mark_buffer_dirty(bh);
3680 }
3681 brelse(bh);
3682 if (err)
3683 goto out;
3684 offset = 0;
3685 towrite -= tocopy;
3686 data += tocopy;
3687 blk++;
3688 }
3689 out:
3690 if (len == towrite) {
3691 mutex_unlock(&inode->i_mutex);
3692 return err;
3693 }
3694 if (inode->i_size < off+len-towrite) {
3695 i_size_write(inode, off+len-towrite);
3696 EXT4_I(inode)->i_disksize = inode->i_size;
3697 }
3698 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3699 ext4_mark_inode_dirty(handle, inode);
3700 mutex_unlock(&inode->i_mutex);
3701 return len - towrite;
3702 }
3703
3704 #endif
3705
3706 static int ext4_get_sb(struct file_system_type *fs_type,
3707 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3708 {
3709 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3710 }
3711
3712 #ifdef CONFIG_PROC_FS
3713 static int ext4_ui_proc_show(struct seq_file *m, void *v)
3714 {
3715 unsigned int *p = m->private;
3716
3717 seq_printf(m, "%u\n", *p);
3718 return 0;
3719 }
3720
3721 static int ext4_ui_proc_open(struct inode *inode, struct file *file)
3722 {
3723 return single_open(file, ext4_ui_proc_show, PDE(inode)->data);
3724 }
3725
3726 static ssize_t ext4_ui_proc_write(struct file *file, const char __user *buf,
3727 size_t cnt, loff_t *ppos)
3728 {
3729 unsigned long *p = PDE(file->f_path.dentry->d_inode)->data;
3730 char str[32];
3731
3732 if (cnt >= sizeof(str))
3733 return -EINVAL;
3734 if (copy_from_user(str, buf, cnt))
3735 return -EFAULT;
3736
3737 *p = simple_strtoul(str, NULL, 0);
3738 return cnt;
3739 }
3740
3741 const struct file_operations ext4_ui_proc_fops = {
3742 .owner = THIS_MODULE,
3743 .open = ext4_ui_proc_open,
3744 .read = seq_read,
3745 .llseek = seq_lseek,
3746 .release = single_release,
3747 .write = ext4_ui_proc_write,
3748 };
3749 #endif
3750
3751 static struct file_system_type ext4_fs_type = {
3752 .owner = THIS_MODULE,
3753 .name = "ext4",
3754 .get_sb = ext4_get_sb,
3755 .kill_sb = kill_block_super,
3756 .fs_flags = FS_REQUIRES_DEV,
3757 };
3758
3759 #ifdef CONFIG_EXT4DEV_COMPAT
3760 static int ext4dev_get_sb(struct file_system_type *fs_type,
3761 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3762 {
3763 printk(KERN_WARNING "EXT4-fs: Update your userspace programs "
3764 "to mount using ext4\n");
3765 printk(KERN_WARNING "EXT4-fs: ext4dev backwards compatibility "
3766 "will go away by 2.6.31\n");
3767 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3768 }
3769
3770 static struct file_system_type ext4dev_fs_type = {
3771 .owner = THIS_MODULE,
3772 .name = "ext4dev",
3773 .get_sb = ext4dev_get_sb,
3774 .kill_sb = kill_block_super,
3775 .fs_flags = FS_REQUIRES_DEV,
3776 };
3777 MODULE_ALIAS("ext4dev");
3778 #endif
3779
3780 static int __init init_ext4_fs(void)
3781 {
3782 int err;
3783
3784 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
3785 err = init_ext4_mballoc();
3786 if (err)
3787 return err;
3788
3789 err = init_ext4_xattr();
3790 if (err)
3791 goto out2;
3792 err = init_inodecache();
3793 if (err)
3794 goto out1;
3795 err = register_filesystem(&ext4_fs_type);
3796 if (err)
3797 goto out;
3798 #ifdef CONFIG_EXT4DEV_COMPAT
3799 err = register_filesystem(&ext4dev_fs_type);
3800 if (err) {
3801 unregister_filesystem(&ext4_fs_type);
3802 goto out;
3803 }
3804 #endif
3805 return 0;
3806 out:
3807 destroy_inodecache();
3808 out1:
3809 exit_ext4_xattr();
3810 out2:
3811 exit_ext4_mballoc();
3812 return err;
3813 }
3814
3815 static void __exit exit_ext4_fs(void)
3816 {
3817 unregister_filesystem(&ext4_fs_type);
3818 #ifdef CONFIG_EXT4DEV_COMPAT
3819 unregister_filesystem(&ext4dev_fs_type);
3820 #endif
3821 destroy_inodecache();
3822 exit_ext4_xattr();
3823 exit_ext4_mballoc();
3824 remove_proc_entry("fs/ext4", NULL);
3825 }
3826
3827 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3828 MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
3829 MODULE_LICENSE("GPL");
3830 module_init(init_ext4_fs)
3831 module_exit(exit_ext4_fs)
Support for the Chinese Samsung S3C2440 based development boards