?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
ext4: Clean up calls to ext4_get_group_desc()
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / fs / ext4 / super.c
blob91b98b58ccb9e3bb3a68137bd0ca3a38bb1acff5
1 /*
2 * linux/fs/ext4/super.c
3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19 #include <linux/module.h>
20 #include <linux/string.h>
21 #include <linux/fs.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/smp_lock.h>
30 #include <linux/buffer_head.h>
31 #include <linux/exportfs.h>
32 #include <linux/vfs.h>
33 #include <linux/random.h>
34 #include <linux/mount.h>
35 #include <linux/namei.h>
36 #include <linux/quotaops.h>
37 #include <linux/seq_file.h>
38 #include <linux/proc_fs.h>
39 #include <linux/ctype.h>
40 #include <linux/marker.h>
41 #include <linux/log2.h>
42 #include <linux/crc16.h>
43 #include <asm/uaccess.h>
44
45 #include "ext4.h"
46 #include "ext4_jbd2.h"
47 #include "xattr.h"
48 #include "acl.h"
49
50 static int default_mb_history_length = 1000;
51
52 module_param_named(default_mb_history_length, default_mb_history_length,
53 int, 0644);
54 MODULE_PARM_DESC(default_mb_history_length,
55 "Default number of entries saved for mb_history");
56
57 struct proc_dir_entry *ext4_proc_root;
58 static struct kset *ext4_kset;
59
60 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
61 unsigned long journal_devnum);
62 static int ext4_commit_super(struct super_block *sb, int sync);
63 static void ext4_mark_recovery_complete(struct super_block *sb,
64 struct ext4_super_block *es);
65 static void ext4_clear_journal_err(struct super_block *sb,
66 struct ext4_super_block *es);
67 static int ext4_sync_fs(struct super_block *sb, int wait);
68 static const char *ext4_decode_error(struct super_block *sb, int errno,
69 char nbuf[16]);
70 static int ext4_remount(struct super_block *sb, int *flags, char *data);
71 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
72 static int ext4_unfreeze(struct super_block *sb);
73 static void ext4_write_super(struct super_block *sb);
74 static int ext4_freeze(struct super_block *sb);
75
76
77 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
78 struct ext4_group_desc *bg)
79 {
80 return le32_to_cpu(bg->bg_block_bitmap_lo) |
81 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
82 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
83 }
84
85 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
86 struct ext4_group_desc *bg)
87 {
88 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
89 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
90 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
91 }
92
93 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
94 struct ext4_group_desc *bg)
95 {
96 return le32_to_cpu(bg->bg_inode_table_lo) |
97 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
98 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
99 }
100
101 __u32 ext4_free_blks_count(struct super_block *sb,
102 struct ext4_group_desc *bg)
103 {
104 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
105 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
106 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
107 }
108
109 __u32 ext4_free_inodes_count(struct super_block *sb,
110 struct ext4_group_desc *bg)
111 {
112 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
113 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
114 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
115 }
116
117 __u32 ext4_used_dirs_count(struct super_block *sb,
118 struct ext4_group_desc *bg)
119 {
120 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
121 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
122 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
123 }
124
125 __u32 ext4_itable_unused_count(struct super_block *sb,
126 struct ext4_group_desc *bg)
127 {
128 return le16_to_cpu(bg->bg_itable_unused_lo) |
129 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
130 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
131 }
132
133 void ext4_block_bitmap_set(struct super_block *sb,
134 struct ext4_group_desc *bg, ext4_fsblk_t blk)
135 {
136 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
137 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
138 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
139 }
140
141 void ext4_inode_bitmap_set(struct super_block *sb,
142 struct ext4_group_desc *bg, ext4_fsblk_t blk)
143 {
144 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
145 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
146 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
147 }
148
149 void ext4_inode_table_set(struct super_block *sb,
150 struct ext4_group_desc *bg, ext4_fsblk_t blk)
151 {
152 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
153 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
154 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
155 }
156
157 void ext4_free_blks_set(struct super_block *sb,
158 struct ext4_group_desc *bg, __u32 count)
159 {
160 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
161 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
162 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
163 }
164
165 void ext4_free_inodes_set(struct super_block *sb,
166 struct ext4_group_desc *bg, __u32 count)
167 {
168 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
169 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
170 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
171 }
172
173 void ext4_used_dirs_set(struct super_block *sb,
174 struct ext4_group_desc *bg, __u32 count)
175 {
176 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
177 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
178 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
179 }
180
181 void ext4_itable_unused_set(struct super_block *sb,
182 struct ext4_group_desc *bg, __u32 count)
183 {
184 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
185 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
186 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
187 }
188
189 /*
190 * Wrappers for jbd2_journal_start/end.
191 *
192 * The only special thing we need to do here is to make sure that all
193 * journal_end calls result in the superblock being marked dirty, so
194 * that sync() will call the filesystem's write_super callback if
195 * appropriate.
196 */
197 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
198 {
199 journal_t *journal;
200
201 if (sb->s_flags & MS_RDONLY)
202 return ERR_PTR(-EROFS);
203
204 /* Special case here: if the journal has aborted behind our
205 * backs (eg. EIO in the commit thread), then we still need to
206 * take the FS itself readonly cleanly. */
207 journal = EXT4_SB(sb)->s_journal;
208 if (journal) {
209 if (is_journal_aborted(journal)) {
210 ext4_abort(sb, __func__,
211 "Detected aborted journal");
212 return ERR_PTR(-EROFS);
213 }
214 return jbd2_journal_start(journal, nblocks);
215 }
216 /*
217 * We're not journaling, return the appropriate indication.
218 */
219 current->journal_info = EXT4_NOJOURNAL_HANDLE;
220 return current->journal_info;
221 }
222
223 /*
224 * The only special thing we need to do here is to make sure that all
225 * jbd2_journal_stop calls result in the superblock being marked dirty, so
226 * that sync() will call the filesystem's write_super callback if
227 * appropriate.
228 */
229 int __ext4_journal_stop(const char *where, handle_t *handle)
230 {
231 struct super_block *sb;
232 int err;
233 int rc;
234
235 if (!ext4_handle_valid(handle)) {
236 /*
237 * Do this here since we don't call jbd2_journal_stop() in
238 * no-journal mode.
239 */
240 current->journal_info = NULL;
241 return 0;
242 }
243 sb = handle->h_transaction->t_journal->j_private;
244 err = handle->h_err;
245 rc = jbd2_journal_stop(handle);
246
247 if (!err)
248 err = rc;
249 if (err)
250 __ext4_std_error(sb, where, err);
251 return err;
252 }
253
254 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
255 struct buffer_head *bh, handle_t *handle, int err)
256 {
257 char nbuf[16];
258 const char *errstr = ext4_decode_error(NULL, err, nbuf);
259
260 BUG_ON(!ext4_handle_valid(handle));
261
262 if (bh)
263 BUFFER_TRACE(bh, "abort");
264
265 if (!handle->h_err)
266 handle->h_err = err;
267
268 if (is_handle_aborted(handle))
269 return;
270
271 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
272 caller, errstr, err_fn);
273
274 jbd2_journal_abort_handle(handle);
275 }
276
277 /* Deal with the reporting of failure conditions on a filesystem such as
278 * inconsistencies detected or read IO failures.
279 *
280 * On ext2, we can store the error state of the filesystem in the
281 * superblock. That is not possible on ext4, because we may have other
282 * write ordering constraints on the superblock which prevent us from
283 * writing it out straight away; and given that the journal is about to
284 * be aborted, we can't rely on the current, or future, transactions to
285 * write out the superblock safely.
286 *
287 * We'll just use the jbd2_journal_abort() error code to record an error in
288 * the journal instead. On recovery, the journal will compain about
289 * that error until we've noted it down and cleared it.
290 */
291
292 static void ext4_handle_error(struct super_block *sb)
293 {
294 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
295
296 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
297 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
298
299 if (sb->s_flags & MS_RDONLY)
300 return;
301
302 if (!test_opt(sb, ERRORS_CONT)) {
303 journal_t *journal = EXT4_SB(sb)->s_journal;
304
305 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
306 if (journal)
307 jbd2_journal_abort(journal, -EIO);
308 }
309 if (test_opt(sb, ERRORS_RO)) {
310 printk(KERN_CRIT "Remounting filesystem read-only\n");
311 sb->s_flags |= MS_RDONLY;
312 }
313 ext4_commit_super(sb, 1);
314 if (test_opt(sb, ERRORS_PANIC))
315 panic("EXT4-fs (device %s): panic forced after error\n",
316 sb->s_id);
317 }
318
319 void ext4_error(struct super_block *sb, const char *function,
320 const char *fmt, ...)
321 {
322 va_list args;
323
324 va_start(args, fmt);
325 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
326 vprintk(fmt, args);
327 printk("\n");
328 va_end(args);
329
330 ext4_handle_error(sb);
331 }
332
333 static const char *ext4_decode_error(struct super_block *sb, int errno,
334 char nbuf[16])
335 {
336 char *errstr = NULL;
337
338 switch (errno) {
339 case -EIO:
340 errstr = "IO failure";
341 break;
342 case -ENOMEM:
343 errstr = "Out of memory";
344 break;
345 case -EROFS:
346 if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
347 errstr = "Journal has aborted";
348 else
349 errstr = "Readonly filesystem";
350 break;
351 default:
352 /* If the caller passed in an extra buffer for unknown
353 * errors, textualise them now. Else we just return
354 * NULL. */
355 if (nbuf) {
356 /* Check for truncated error codes... */
357 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
358 errstr = nbuf;
359 }
360 break;
361 }
362
363 return errstr;
364 }
365
366 /* __ext4_std_error decodes expected errors from journaling functions
367 * automatically and invokes the appropriate error response. */
368
369 void __ext4_std_error(struct super_block *sb, const char *function, int errno)
370 {
371 char nbuf[16];
372 const char *errstr;
373
374 /* Special case: if the error is EROFS, and we're not already
375 * inside a transaction, then there's really no point in logging
376 * an error. */
377 if (errno == -EROFS && journal_current_handle() == NULL &&
378 (sb->s_flags & MS_RDONLY))
379 return;
380
381 errstr = ext4_decode_error(sb, errno, nbuf);
382 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
383 sb->s_id, function, errstr);
384
385 ext4_handle_error(sb);
386 }
387
388 /*
389 * ext4_abort is a much stronger failure handler than ext4_error. The
390 * abort function may be used to deal with unrecoverable failures such
391 * as journal IO errors or ENOMEM at a critical moment in log management.
392 *
393 * We unconditionally force the filesystem into an ABORT|READONLY state,
394 * unless the error response on the fs has been set to panic in which
395 * case we take the easy way out and panic immediately.
396 */
397
398 void ext4_abort(struct super_block *sb, const char *function,
399 const char *fmt, ...)
400 {
401 va_list args;
402
403 printk(KERN_CRIT "ext4_abort called.\n");
404
405 va_start(args, fmt);
406 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
407 vprintk(fmt, args);
408 printk("\n");
409 va_end(args);
410
411 if (test_opt(sb, ERRORS_PANIC))
412 panic("EXT4-fs panic from previous error\n");
413
414 if (sb->s_flags & MS_RDONLY)
415 return;
416
417 printk(KERN_CRIT "Remounting filesystem read-only\n");
418 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
419 sb->s_flags |= MS_RDONLY;
420 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
421 if (EXT4_SB(sb)->s_journal)
422 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
423 }
424
425 void ext4_warning(struct super_block *sb, const char *function,
426 const char *fmt, ...)
427 {
428 va_list args;
429
430 va_start(args, fmt);
431 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
432 sb->s_id, function);
433 vprintk(fmt, args);
434 printk("\n");
435 va_end(args);
436 }
437
438 void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp,
439 const char *function, const char *fmt, ...)
440 __releases(bitlock)
441 __acquires(bitlock)
442 {
443 va_list args;
444 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
445
446 va_start(args, fmt);
447 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
448 vprintk(fmt, args);
449 printk("\n");
450 va_end(args);
451
452 if (test_opt(sb, ERRORS_CONT)) {
453 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
454 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
455 ext4_commit_super(sb, 0);
456 return;
457 }
458 ext4_unlock_group(sb, grp);
459 ext4_handle_error(sb);
460 /*
461 * We only get here in the ERRORS_RO case; relocking the group
462 * may be dangerous, but nothing bad will happen since the
463 * filesystem will have already been marked read/only and the
464 * journal has been aborted. We return 1 as a hint to callers
465 * who might what to use the return value from
466 * ext4_grp_locked_error() to distinguish beween the
467 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
468 * aggressively from the ext4 function in question, with a
469 * more appropriate error code.
470 */
471 ext4_lock_group(sb, grp);
472 return;
473 }
474
475
476 void ext4_update_dynamic_rev(struct super_block *sb)
477 {
478 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
479
480 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
481 return;
482
483 ext4_warning(sb, __func__,
484 "updating to rev %d because of new feature flag, "
485 "running e2fsck is recommended",
486 EXT4_DYNAMIC_REV);
487
488 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
489 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
490 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
491 /* leave es->s_feature_*compat flags alone */
492 /* es->s_uuid will be set by e2fsck if empty */
493
494 /*
495 * The rest of the superblock fields should be zero, and if not it
496 * means they are likely already in use, so leave them alone. We
497 * can leave it up to e2fsck to clean up any inconsistencies there.
498 */
499 }
500
501 /*
502 * Open the external journal device
503 */
504 static struct block_device *ext4_blkdev_get(dev_t dev)
505 {
506 struct block_device *bdev;
507 char b[BDEVNAME_SIZE];
508
509 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
510 if (IS_ERR(bdev))
511 goto fail;
512 return bdev;
513
514 fail:
515 printk(KERN_ERR "EXT4-fs: failed to open journal device %s: %ld\n",
516 __bdevname(dev, b), PTR_ERR(bdev));
517 return NULL;
518 }
519
520 /*
521 * Release the journal device
522 */
523 static int ext4_blkdev_put(struct block_device *bdev)
524 {
525 bd_release(bdev);
526 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
527 }
528
529 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
530 {
531 struct block_device *bdev;
532 int ret = -ENODEV;
533
534 bdev = sbi->journal_bdev;
535 if (bdev) {
536 ret = ext4_blkdev_put(bdev);
537 sbi->journal_bdev = NULL;
538 }
539 return ret;
540 }
541
542 static inline struct inode *orphan_list_entry(struct list_head *l)
543 {
544 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
545 }
546
547 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
548 {
549 struct list_head *l;
550
551 printk(KERN_ERR "sb orphan head is %d\n",
552 le32_to_cpu(sbi->s_es->s_last_orphan));
553
554 printk(KERN_ERR "sb_info orphan list:\n");
555 list_for_each(l, &sbi->s_orphan) {
556 struct inode *inode = orphan_list_entry(l);
557 printk(KERN_ERR " "
558 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
559 inode->i_sb->s_id, inode->i_ino, inode,
560 inode->i_mode, inode->i_nlink,
561 NEXT_ORPHAN(inode));
562 }
563 }
564
565 static void ext4_put_super(struct super_block *sb)
566 {
567 struct ext4_sb_info *sbi = EXT4_SB(sb);
568 struct ext4_super_block *es = sbi->s_es;
569 int i, err;
570
571 ext4_release_system_zone(sb);
572 ext4_mb_release(sb);
573 ext4_ext_release(sb);
574 ext4_xattr_put_super(sb);
575 if (sbi->s_journal) {
576 err = jbd2_journal_destroy(sbi->s_journal);
577 sbi->s_journal = NULL;
578 if (err < 0)
579 ext4_abort(sb, __func__,
580 "Couldn't clean up the journal");
581 }
582 if (!(sb->s_flags & MS_RDONLY)) {
583 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
584 es->s_state = cpu_to_le16(sbi->s_mount_state);
585 ext4_commit_super(sb, 1);
586 }
587 if (sbi->s_proc) {
588 remove_proc_entry(sb->s_id, ext4_proc_root);
589 }
590 kobject_del(&sbi->s_kobj);
591
592 for (i = 0; i < sbi->s_gdb_count; i++)
593 brelse(sbi->s_group_desc[i]);
594 kfree(sbi->s_group_desc);
595 if (is_vmalloc_addr(sbi->s_flex_groups))
596 vfree(sbi->s_flex_groups);
597 else
598 kfree(sbi->s_flex_groups);
599 percpu_counter_destroy(&sbi->s_freeblocks_counter);
600 percpu_counter_destroy(&sbi->s_freeinodes_counter);
601 percpu_counter_destroy(&sbi->s_dirs_counter);
602 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
603 brelse(sbi->s_sbh);
604 #ifdef CONFIG_QUOTA
605 for (i = 0; i < MAXQUOTAS; i++)
606 kfree(sbi->s_qf_names[i]);
607 #endif
608
609 /* Debugging code just in case the in-memory inode orphan list
610 * isn't empty. The on-disk one can be non-empty if we've
611 * detected an error and taken the fs readonly, but the
612 * in-memory list had better be clean by this point. */
613 if (!list_empty(&sbi->s_orphan))
614 dump_orphan_list(sb, sbi);
615 J_ASSERT(list_empty(&sbi->s_orphan));
616
617 invalidate_bdev(sb->s_bdev);
618 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
619 /*
620 * Invalidate the journal device's buffers. We don't want them
621 * floating about in memory - the physical journal device may
622 * hotswapped, and it breaks the `ro-after' testing code.
623 */
624 sync_blockdev(sbi->journal_bdev);
625 invalidate_bdev(sbi->journal_bdev);
626 ext4_blkdev_remove(sbi);
627 }
628 sb->s_fs_info = NULL;
629 /*
630 * Now that we are completely done shutting down the
631 * superblock, we need to actually destroy the kobject.
632 */
633 unlock_kernel();
634 unlock_super(sb);
635 kobject_put(&sbi->s_kobj);
636 wait_for_completion(&sbi->s_kobj_unregister);
637 lock_super(sb);
638 lock_kernel();
639 kfree(sbi->s_blockgroup_lock);
640 kfree(sbi);
641 return;
642 }
643
644 static struct kmem_cache *ext4_inode_cachep;
645
646 /*
647 * Called inside transaction, so use GFP_NOFS
648 */
649 static struct inode *ext4_alloc_inode(struct super_block *sb)
650 {
651 struct ext4_inode_info *ei;
652
653 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
654 if (!ei)
655 return NULL;
656 #ifdef CONFIG_EXT4_FS_POSIX_ACL
657 ei->i_acl = EXT4_ACL_NOT_CACHED;
658 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
659 #endif
660 ei->vfs_inode.i_version = 1;
661 ei->vfs_inode.i_data.writeback_index = 0;
662 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
663 INIT_LIST_HEAD(&ei->i_prealloc_list);
664 spin_lock_init(&ei->i_prealloc_lock);
665 /*
666 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
667 * therefore it can be null here. Don't check it, just initialize
668 * jinode.
669 */
670 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
671 ei->i_reserved_data_blocks = 0;
672 ei->i_reserved_meta_blocks = 0;
673 ei->i_allocated_meta_blocks = 0;
674 ei->i_delalloc_reserved_flag = 0;
675 spin_lock_init(&(ei->i_block_reservation_lock));
676 return &ei->vfs_inode;
677 }
678
679 static void ext4_destroy_inode(struct inode *inode)
680 {
681 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
682 printk("EXT4 Inode %p: orphan list check failed!\n",
683 EXT4_I(inode));
684 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
685 EXT4_I(inode), sizeof(struct ext4_inode_info),
686 true);
687 dump_stack();
688 }
689 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
690 }
691
692 static void init_once(void *foo)
693 {
694 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
695
696 INIT_LIST_HEAD(&ei->i_orphan);
697 #ifdef CONFIG_EXT4_FS_XATTR
698 init_rwsem(&ei->xattr_sem);
699 #endif
700 init_rwsem(&ei->i_data_sem);
701 inode_init_once(&ei->vfs_inode);
702 }
703
704 static int init_inodecache(void)
705 {
706 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
707 sizeof(struct ext4_inode_info),
708 0, (SLAB_RECLAIM_ACCOUNT|
709 SLAB_MEM_SPREAD),
710 init_once);
711 if (ext4_inode_cachep == NULL)
712 return -ENOMEM;
713 return 0;
714 }
715
716 static void destroy_inodecache(void)
717 {
718 kmem_cache_destroy(ext4_inode_cachep);
719 }
720
721 static void ext4_clear_inode(struct inode *inode)
722 {
723 #ifdef CONFIG_EXT4_FS_POSIX_ACL
724 if (EXT4_I(inode)->i_acl &&
725 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
726 posix_acl_release(EXT4_I(inode)->i_acl);
727 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
728 }
729 if (EXT4_I(inode)->i_default_acl &&
730 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
731 posix_acl_release(EXT4_I(inode)->i_default_acl);
732 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
733 }
734 #endif
735 ext4_discard_preallocations(inode);
736 if (EXT4_JOURNAL(inode))
737 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
738 &EXT4_I(inode)->jinode);
739 }
740
741 static inline void ext4_show_quota_options(struct seq_file *seq,
742 struct super_block *sb)
743 {
744 #if defined(CONFIG_QUOTA)
745 struct ext4_sb_info *sbi = EXT4_SB(sb);
746
747 if (sbi->s_jquota_fmt)
748 seq_printf(seq, ",jqfmt=%s",
749 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
750
751 if (sbi->s_qf_names[USRQUOTA])
752 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
753
754 if (sbi->s_qf_names[GRPQUOTA])
755 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
756
757 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
758 seq_puts(seq, ",usrquota");
759
760 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
761 seq_puts(seq, ",grpquota");
762 #endif
763 }
764
765 /*
766 * Show an option if
767 * - it's set to a non-default value OR
768 * - if the per-sb default is different from the global default
769 */
770 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
771 {
772 int def_errors;
773 unsigned long def_mount_opts;
774 struct super_block *sb = vfs->mnt_sb;
775 struct ext4_sb_info *sbi = EXT4_SB(sb);
776 struct ext4_super_block *es = sbi->s_es;
777
778 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
779 def_errors = le16_to_cpu(es->s_errors);
780
781 if (sbi->s_sb_block != 1)
782 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
783 if (test_opt(sb, MINIX_DF))
784 seq_puts(seq, ",minixdf");
785 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
786 seq_puts(seq, ",grpid");
787 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
788 seq_puts(seq, ",nogrpid");
789 if (sbi->s_resuid != EXT4_DEF_RESUID ||
790 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
791 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
792 }
793 if (sbi->s_resgid != EXT4_DEF_RESGID ||
794 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
795 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
796 }
797 if (test_opt(sb, ERRORS_RO)) {
798 if (def_errors == EXT4_ERRORS_PANIC ||
799 def_errors == EXT4_ERRORS_CONTINUE) {
800 seq_puts(seq, ",errors=remount-ro");
801 }
802 }
803 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
804 seq_puts(seq, ",errors=continue");
805 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
806 seq_puts(seq, ",errors=panic");
807 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
808 seq_puts(seq, ",nouid32");
809 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
810 seq_puts(seq, ",debug");
811 if (test_opt(sb, OLDALLOC))
812 seq_puts(seq, ",oldalloc");
813 #ifdef CONFIG_EXT4_FS_XATTR
814 if (test_opt(sb, XATTR_USER) &&
815 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
816 seq_puts(seq, ",user_xattr");
817 if (!test_opt(sb, XATTR_USER) &&
818 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
819 seq_puts(seq, ",nouser_xattr");
820 }
821 #endif
822 #ifdef CONFIG_EXT4_FS_POSIX_ACL
823 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
824 seq_puts(seq, ",acl");
825 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
826 seq_puts(seq, ",noacl");
827 #endif
828 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
829 seq_printf(seq, ",commit=%u",
830 (unsigned) (sbi->s_commit_interval / HZ));
831 }
832 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
833 seq_printf(seq, ",min_batch_time=%u",
834 (unsigned) sbi->s_min_batch_time);
835 }
836 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
837 seq_printf(seq, ",max_batch_time=%u",
838 (unsigned) sbi->s_min_batch_time);
839 }
840
841 /*
842 * We're changing the default of barrier mount option, so
843 * let's always display its mount state so it's clear what its
844 * status is.
845 */
846 seq_puts(seq, ",barrier=");
847 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
848 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
849 seq_puts(seq, ",journal_async_commit");
850 if (test_opt(sb, NOBH))
851 seq_puts(seq, ",nobh");
852 if (test_opt(sb, I_VERSION))
853 seq_puts(seq, ",i_version");
854 if (!test_opt(sb, DELALLOC))
855 seq_puts(seq, ",nodelalloc");
856
857
858 if (sbi->s_stripe)
859 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
860 /*
861 * journal mode get enabled in different ways
862 * So just print the value even if we didn't specify it
863 */
864 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
865 seq_puts(seq, ",data=journal");
866 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
867 seq_puts(seq, ",data=ordered");
868 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
869 seq_puts(seq, ",data=writeback");
870
871 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
872 seq_printf(seq, ",inode_readahead_blks=%u",
873 sbi->s_inode_readahead_blks);
874
875 if (test_opt(sb, DATA_ERR_ABORT))
876 seq_puts(seq, ",data_err=abort");
877
878 if (test_opt(sb, NO_AUTO_DA_ALLOC))
879 seq_puts(seq, ",noauto_da_alloc");
880
881 ext4_show_quota_options(seq, sb);
882 return 0;
883 }
884
885
886 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
887 u64 ino, u32 generation)
888 {
889 struct inode *inode;
890
891 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
892 return ERR_PTR(-ESTALE);
893 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
894 return ERR_PTR(-ESTALE);
895
896 /* iget isn't really right if the inode is currently unallocated!!
897 *
898 * ext4_read_inode will return a bad_inode if the inode had been
899 * deleted, so we should be safe.
900 *
901 * Currently we don't know the generation for parent directory, so
902 * a generation of 0 means "accept any"
903 */
904 inode = ext4_iget(sb, ino);
905 if (IS_ERR(inode))
906 return ERR_CAST(inode);
907 if (generation && inode->i_generation != generation) {
908 iput(inode);
909 return ERR_PTR(-ESTALE);
910 }
911
912 return inode;
913 }
914
915 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
916 int fh_len, int fh_type)
917 {
918 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
919 ext4_nfs_get_inode);
920 }
921
922 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
923 int fh_len, int fh_type)
924 {
925 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
926 ext4_nfs_get_inode);
927 }
928
929 /*
930 * Try to release metadata pages (indirect blocks, directories) which are
931 * mapped via the block device. Since these pages could have journal heads
932 * which would prevent try_to_free_buffers() from freeing them, we must use
933 * jbd2 layer's try_to_free_buffers() function to release them.
934 */
935 static int bdev_try_to_free_page(struct super_block *sb, struct page *page, gfp_t wait)
936 {
937 journal_t *journal = EXT4_SB(sb)->s_journal;
938
939 WARN_ON(PageChecked(page));
940 if (!page_has_buffers(page))
941 return 0;
942 if (journal)
943 return jbd2_journal_try_to_free_buffers(journal, page,
944 wait & ~__GFP_WAIT);
945 return try_to_free_buffers(page);
946 }
947
948 #ifdef CONFIG_QUOTA
949 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
950 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
951
952 static int ext4_write_dquot(struct dquot *dquot);
953 static int ext4_acquire_dquot(struct dquot *dquot);
954 static int ext4_release_dquot(struct dquot *dquot);
955 static int ext4_mark_dquot_dirty(struct dquot *dquot);
956 static int ext4_write_info(struct super_block *sb, int type);
957 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
958 char *path, int remount);
959 static int ext4_quota_on_mount(struct super_block *sb, int type);
960 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
961 size_t len, loff_t off);
962 static ssize_t ext4_quota_write(struct super_block *sb, int type,
963 const char *data, size_t len, loff_t off);
964
965 static struct dquot_operations ext4_quota_operations = {
966 .initialize = dquot_initialize,
967 .drop = dquot_drop,
968 .alloc_space = dquot_alloc_space,
969 .reserve_space = dquot_reserve_space,
970 .claim_space = dquot_claim_space,
971 .release_rsv = dquot_release_reserved_space,
972 .get_reserved_space = ext4_get_reserved_space,
973 .alloc_inode = dquot_alloc_inode,
974 .free_space = dquot_free_space,
975 .free_inode = dquot_free_inode,
976 .transfer = dquot_transfer,
977 .write_dquot = ext4_write_dquot,
978 .acquire_dquot = ext4_acquire_dquot,
979 .release_dquot = ext4_release_dquot,
980 .mark_dirty = ext4_mark_dquot_dirty,
981 .write_info = ext4_write_info,
982 .alloc_dquot = dquot_alloc,
983 .destroy_dquot = dquot_destroy,
984 };
985
986 static struct quotactl_ops ext4_qctl_operations = {
987 .quota_on = ext4_quota_on,
988 .quota_off = vfs_quota_off,
989 .quota_sync = vfs_quota_sync,
990 .get_info = vfs_get_dqinfo,
991 .set_info = vfs_set_dqinfo,
992 .get_dqblk = vfs_get_dqblk,
993 .set_dqblk = vfs_set_dqblk
994 };
995 #endif
996
997 static const struct super_operations ext4_sops = {
998 .alloc_inode = ext4_alloc_inode,
999 .destroy_inode = ext4_destroy_inode,
1000 .write_inode = ext4_write_inode,
1001 .dirty_inode = ext4_dirty_inode,
1002 .delete_inode = ext4_delete_inode,
1003 .put_super = ext4_put_super,
1004 .sync_fs = ext4_sync_fs,
1005 .freeze_fs = ext4_freeze,
1006 .unfreeze_fs = ext4_unfreeze,
1007 .statfs = ext4_statfs,
1008 .remount_fs = ext4_remount,
1009 .clear_inode = ext4_clear_inode,
1010 .show_options = ext4_show_options,
1011 #ifdef CONFIG_QUOTA
1012 .quota_read = ext4_quota_read,
1013 .quota_write = ext4_quota_write,
1014 #endif
1015 .bdev_try_to_free_page = bdev_try_to_free_page,
1016 };
1017
1018 static const struct super_operations ext4_nojournal_sops = {
1019 .alloc_inode = ext4_alloc_inode,
1020 .destroy_inode = ext4_destroy_inode,
1021 .write_inode = ext4_write_inode,
1022 .dirty_inode = ext4_dirty_inode,
1023 .delete_inode = ext4_delete_inode,
1024 .write_super = ext4_write_super,
1025 .put_super = ext4_put_super,
1026 .statfs = ext4_statfs,
1027 .remount_fs = ext4_remount,
1028 .clear_inode = ext4_clear_inode,
1029 .show_options = ext4_show_options,
1030 #ifdef CONFIG_QUOTA
1031 .quota_read = ext4_quota_read,
1032 .quota_write = ext4_quota_write,
1033 #endif
1034 .bdev_try_to_free_page = bdev_try_to_free_page,
1035 };
1036
1037 static const struct export_operations ext4_export_ops = {
1038 .fh_to_dentry = ext4_fh_to_dentry,
1039 .fh_to_parent = ext4_fh_to_parent,
1040 .get_parent = ext4_get_parent,
1041 };
1042
1043 enum {
1044 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1045 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1046 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1047 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1048 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1049 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1050 Opt_journal_update, Opt_journal_dev,
1051 Opt_journal_checksum, Opt_journal_async_commit,
1052 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1053 Opt_data_err_abort, Opt_data_err_ignore, Opt_mb_history_length,
1054 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1055 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
1056 Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err, Opt_resize,
1057 Opt_usrquota, Opt_grpquota, Opt_i_version,
1058 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1059 Opt_block_validity, Opt_noblock_validity,
1060 Opt_inode_readahead_blks, Opt_journal_ioprio
1061 };
1062
1063 static const match_table_t tokens = {
1064 {Opt_bsd_df, "bsddf"},
1065 {Opt_minix_df, "minixdf"},
1066 {Opt_grpid, "grpid"},
1067 {Opt_grpid, "bsdgroups"},
1068 {Opt_nogrpid, "nogrpid"},
1069 {Opt_nogrpid, "sysvgroups"},
1070 {Opt_resgid, "resgid=%u"},
1071 {Opt_resuid, "resuid=%u"},
1072 {Opt_sb, "sb=%u"},
1073 {Opt_err_cont, "errors=continue"},
1074 {Opt_err_panic, "errors=panic"},
1075 {Opt_err_ro, "errors=remount-ro"},
1076 {Opt_nouid32, "nouid32"},
1077 {Opt_debug, "debug"},
1078 {Opt_oldalloc, "oldalloc"},
1079 {Opt_orlov, "orlov"},
1080 {Opt_user_xattr, "user_xattr"},
1081 {Opt_nouser_xattr, "nouser_xattr"},
1082 {Opt_acl, "acl"},
1083 {Opt_noacl, "noacl"},
1084 {Opt_noload, "noload"},
1085 {Opt_nobh, "nobh"},
1086 {Opt_bh, "bh"},
1087 {Opt_commit, "commit=%u"},
1088 {Opt_min_batch_time, "min_batch_time=%u"},
1089 {Opt_max_batch_time, "max_batch_time=%u"},
1090 {Opt_journal_update, "journal=update"},
1091 {Opt_journal_dev, "journal_dev=%u"},
1092 {Opt_journal_checksum, "journal_checksum"},
1093 {Opt_journal_async_commit, "journal_async_commit"},
1094 {Opt_abort, "abort"},
1095 {Opt_data_journal, "data=journal"},
1096 {Opt_data_ordered, "data=ordered"},
1097 {Opt_data_writeback, "data=writeback"},
1098 {Opt_data_err_abort, "data_err=abort"},
1099 {Opt_data_err_ignore, "data_err=ignore"},
1100 {Opt_mb_history_length, "mb_history_length=%u"},
1101 {Opt_offusrjquota, "usrjquota="},
1102 {Opt_usrjquota, "usrjquota=%s"},
1103 {Opt_offgrpjquota, "grpjquota="},
1104 {Opt_grpjquota, "grpjquota=%s"},
1105 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1106 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1107 {Opt_grpquota, "grpquota"},
1108 {Opt_noquota, "noquota"},
1109 {Opt_quota, "quota"},
1110 {Opt_usrquota, "usrquota"},
1111 {Opt_barrier, "barrier=%u"},
1112 {Opt_barrier, "barrier"},
1113 {Opt_nobarrier, "nobarrier"},
1114 {Opt_i_version, "i_version"},
1115 {Opt_stripe, "stripe=%u"},
1116 {Opt_resize, "resize"},
1117 {Opt_delalloc, "delalloc"},
1118 {Opt_nodelalloc, "nodelalloc"},
1119 {Opt_block_validity, "block_validity"},
1120 {Opt_noblock_validity, "noblock_validity"},
1121 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1122 {Opt_journal_ioprio, "journal_ioprio=%u"},
1123 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1124 {Opt_auto_da_alloc, "auto_da_alloc"},
1125 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1126 {Opt_err, NULL},
1127 };
1128
1129 static ext4_fsblk_t get_sb_block(void **data)
1130 {
1131 ext4_fsblk_t sb_block;
1132 char *options = (char *) *data;
1133
1134 if (!options || strncmp(options, "sb=", 3) != 0)
1135 return 1; /* Default location */
1136 options += 3;
1137 /*todo: use simple_strtoll with >32bit ext4 */
1138 sb_block = simple_strtoul(options, &options, 0);
1139 if (*options && *options != ',') {
1140 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1141 (char *) *data);
1142 return 1;
1143 }
1144 if (*options == ',')
1145 options++;
1146 *data = (void *) options;
1147 return sb_block;
1148 }
1149
1150 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1151
1152 static int parse_options(char *options, struct super_block *sb,
1153 unsigned long *journal_devnum,
1154 unsigned int *journal_ioprio,
1155 ext4_fsblk_t *n_blocks_count, int is_remount)
1156 {
1157 struct ext4_sb_info *sbi = EXT4_SB(sb);
1158 char *p;
1159 substring_t args[MAX_OPT_ARGS];
1160 int data_opt = 0;
1161 int option;
1162 #ifdef CONFIG_QUOTA
1163 int qtype, qfmt;
1164 char *qname;
1165 #endif
1166
1167 if (!options)
1168 return 1;
1169
1170 while ((p = strsep(&options, ",")) != NULL) {
1171 int token;
1172 if (!*p)
1173 continue;
1174
1175 token = match_token(p, tokens, args);
1176 switch (token) {
1177 case Opt_bsd_df:
1178 clear_opt(sbi->s_mount_opt, MINIX_DF);
1179 break;
1180 case Opt_minix_df:
1181 set_opt(sbi->s_mount_opt, MINIX_DF);
1182 break;
1183 case Opt_grpid:
1184 set_opt(sbi->s_mount_opt, GRPID);
1185 break;
1186 case Opt_nogrpid:
1187 clear_opt(sbi->s_mount_opt, GRPID);
1188 break;
1189 case Opt_resuid:
1190 if (match_int(&args[0], &option))
1191 return 0;
1192 sbi->s_resuid = option;
1193 break;
1194 case Opt_resgid:
1195 if (match_int(&args[0], &option))
1196 return 0;
1197 sbi->s_resgid = option;
1198 break;
1199 case Opt_sb:
1200 /* handled by get_sb_block() instead of here */
1201 /* *sb_block = match_int(&args[0]); */
1202 break;
1203 case Opt_err_panic:
1204 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1205 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1206 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1207 break;
1208 case Opt_err_ro:
1209 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1210 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1211 set_opt(sbi->s_mount_opt, ERRORS_RO);
1212 break;
1213 case Opt_err_cont:
1214 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1215 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1216 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1217 break;
1218 case Opt_nouid32:
1219 set_opt(sbi->s_mount_opt, NO_UID32);
1220 break;
1221 case Opt_debug:
1222 set_opt(sbi->s_mount_opt, DEBUG);
1223 break;
1224 case Opt_oldalloc:
1225 set_opt(sbi->s_mount_opt, OLDALLOC);
1226 break;
1227 case Opt_orlov:
1228 clear_opt(sbi->s_mount_opt, OLDALLOC);
1229 break;
1230 #ifdef CONFIG_EXT4_FS_XATTR
1231 case Opt_user_xattr:
1232 set_opt(sbi->s_mount_opt, XATTR_USER);
1233 break;
1234 case Opt_nouser_xattr:
1235 clear_opt(sbi->s_mount_opt, XATTR_USER);
1236 break;
1237 #else
1238 case Opt_user_xattr:
1239 case Opt_nouser_xattr:
1240 printk(KERN_ERR "EXT4 (no)user_xattr options "
1241 "not supported\n");
1242 break;
1243 #endif
1244 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1245 case Opt_acl:
1246 set_opt(sbi->s_mount_opt, POSIX_ACL);
1247 break;
1248 case Opt_noacl:
1249 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1250 break;
1251 #else
1252 case Opt_acl:
1253 case Opt_noacl:
1254 printk(KERN_ERR "EXT4 (no)acl options "
1255 "not supported\n");
1256 break;
1257 #endif
1258 case Opt_journal_update:
1259 /* @@@ FIXME */
1260 /* Eventually we will want to be able to create
1261 a journal file here. For now, only allow the
1262 user to specify an existing inode to be the
1263 journal file. */
1264 if (is_remount) {
1265 printk(KERN_ERR "EXT4-fs: cannot specify "
1266 "journal on remount\n");
1267 return 0;
1268 }
1269 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1270 break;
1271 case Opt_journal_dev:
1272 if (is_remount) {
1273 printk(KERN_ERR "EXT4-fs: cannot specify "
1274 "journal on remount\n");
1275 return 0;
1276 }
1277 if (match_int(&args[0], &option))
1278 return 0;
1279 *journal_devnum = option;
1280 break;
1281 case Opt_journal_checksum:
1282 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1283 break;
1284 case Opt_journal_async_commit:
1285 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1286 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1287 break;
1288 case Opt_noload:
1289 set_opt(sbi->s_mount_opt, NOLOAD);
1290 break;
1291 case Opt_commit:
1292 if (match_int(&args[0], &option))
1293 return 0;
1294 if (option < 0)
1295 return 0;
1296 if (option == 0)
1297 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1298 sbi->s_commit_interval = HZ * option;
1299 break;
1300 case Opt_max_batch_time:
1301 if (match_int(&args[0], &option))
1302 return 0;
1303 if (option < 0)
1304 return 0;
1305 if (option == 0)
1306 option = EXT4_DEF_MAX_BATCH_TIME;
1307 sbi->s_max_batch_time = option;
1308 break;
1309 case Opt_min_batch_time:
1310 if (match_int(&args[0], &option))
1311 return 0;
1312 if (option < 0)
1313 return 0;
1314 sbi->s_min_batch_time = option;
1315 break;
1316 case Opt_data_journal:
1317 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1318 goto datacheck;
1319 case Opt_data_ordered:
1320 data_opt = EXT4_MOUNT_ORDERED_DATA;
1321 goto datacheck;
1322 case Opt_data_writeback:
1323 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1324 datacheck:
1325 if (is_remount) {
1326 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1327 != data_opt) {
1328 printk(KERN_ERR
1329 "EXT4-fs: cannot change data "
1330 "mode on remount\n");
1331 return 0;
1332 }
1333 } else {
1334 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1335 sbi->s_mount_opt |= data_opt;
1336 }
1337 break;
1338 case Opt_data_err_abort:
1339 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1340 break;
1341 case Opt_data_err_ignore:
1342 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1343 break;
1344 case Opt_mb_history_length:
1345 if (match_int(&args[0], &option))
1346 return 0;
1347 if (option < 0)
1348 return 0;
1349 sbi->s_mb_history_max = option;
1350 break;
1351 #ifdef CONFIG_QUOTA
1352 case Opt_usrjquota:
1353 qtype = USRQUOTA;
1354 goto set_qf_name;
1355 case Opt_grpjquota:
1356 qtype = GRPQUOTA;
1357 set_qf_name:
1358 if (sb_any_quota_loaded(sb) &&
1359 !sbi->s_qf_names[qtype]) {
1360 printk(KERN_ERR
1361 "EXT4-fs: Cannot change journaled "
1362 "quota options when quota turned on.\n");
1363 return 0;
1364 }
1365 qname = match_strdup(&args[0]);
1366 if (!qname) {
1367 printk(KERN_ERR
1368 "EXT4-fs: not enough memory for "
1369 "storing quotafile name.\n");
1370 return 0;
1371 }
1372 if (sbi->s_qf_names[qtype] &&
1373 strcmp(sbi->s_qf_names[qtype], qname)) {
1374 printk(KERN_ERR
1375 "EXT4-fs: %s quota file already "
1376 "specified.\n", QTYPE2NAME(qtype));
1377 kfree(qname);
1378 return 0;
1379 }
1380 sbi->s_qf_names[qtype] = qname;
1381 if (strchr(sbi->s_qf_names[qtype], '/')) {
1382 printk(KERN_ERR
1383 "EXT4-fs: quotafile must be on "
1384 "filesystem root.\n");
1385 kfree(sbi->s_qf_names[qtype]);
1386 sbi->s_qf_names[qtype] = NULL;
1387 return 0;
1388 }
1389 set_opt(sbi->s_mount_opt, QUOTA);
1390 break;
1391 case Opt_offusrjquota:
1392 qtype = USRQUOTA;
1393 goto clear_qf_name;
1394 case Opt_offgrpjquota:
1395 qtype = GRPQUOTA;
1396 clear_qf_name:
1397 if (sb_any_quota_loaded(sb) &&
1398 sbi->s_qf_names[qtype]) {
1399 printk(KERN_ERR "EXT4-fs: Cannot change "
1400 "journaled quota options when "
1401 "quota turned on.\n");
1402 return 0;
1403 }
1404 /*
1405 * The space will be released later when all options
1406 * are confirmed to be correct
1407 */
1408 sbi->s_qf_names[qtype] = NULL;
1409 break;
1410 case Opt_jqfmt_vfsold:
1411 qfmt = QFMT_VFS_OLD;
1412 goto set_qf_format;
1413 case Opt_jqfmt_vfsv0:
1414 qfmt = QFMT_VFS_V0;
1415 set_qf_format:
1416 if (sb_any_quota_loaded(sb) &&
1417 sbi->s_jquota_fmt != qfmt) {
1418 printk(KERN_ERR "EXT4-fs: Cannot change "
1419 "journaled quota options when "
1420 "quota turned on.\n");
1421 return 0;
1422 }
1423 sbi->s_jquota_fmt = qfmt;
1424 break;
1425 case Opt_quota:
1426 case Opt_usrquota:
1427 set_opt(sbi->s_mount_opt, QUOTA);
1428 set_opt(sbi->s_mount_opt, USRQUOTA);
1429 break;
1430 case Opt_grpquota:
1431 set_opt(sbi->s_mount_opt, QUOTA);
1432 set_opt(sbi->s_mount_opt, GRPQUOTA);
1433 break;
1434 case Opt_noquota:
1435 if (sb_any_quota_loaded(sb)) {
1436 printk(KERN_ERR "EXT4-fs: Cannot change quota "
1437 "options when quota turned on.\n");
1438 return 0;
1439 }
1440 clear_opt(sbi->s_mount_opt, QUOTA);
1441 clear_opt(sbi->s_mount_opt, USRQUOTA);
1442 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1443 break;
1444 #else
1445 case Opt_quota:
1446 case Opt_usrquota:
1447 case Opt_grpquota:
1448 printk(KERN_ERR
1449 "EXT4-fs: quota options not supported.\n");
1450 break;
1451 case Opt_usrjquota:
1452 case Opt_grpjquota:
1453 case Opt_offusrjquota:
1454 case Opt_offgrpjquota:
1455 case Opt_jqfmt_vfsold:
1456 case Opt_jqfmt_vfsv0:
1457 printk(KERN_ERR
1458 "EXT4-fs: journaled quota options not "
1459 "supported.\n");
1460 break;
1461 case Opt_noquota:
1462 break;
1463 #endif
1464 case Opt_abort:
1465 set_opt(sbi->s_mount_opt, ABORT);
1466 break;
1467 case Opt_nobarrier:
1468 clear_opt(sbi->s_mount_opt, BARRIER);
1469 break;
1470 case Opt_barrier:
1471 if (match_int(&args[0], &option)) {
1472 set_opt(sbi->s_mount_opt, BARRIER);
1473 break;
1474 }
1475 if (option)
1476 set_opt(sbi->s_mount_opt, BARRIER);
1477 else
1478 clear_opt(sbi->s_mount_opt, BARRIER);
1479 break;
1480 case Opt_ignore:
1481 break;
1482 case Opt_resize:
1483 if (!is_remount) {
1484 printk("EXT4-fs: resize option only available "
1485 "for remount\n");
1486 return 0;
1487 }
1488 if (match_int(&args[0], &option) != 0)
1489 return 0;
1490 *n_blocks_count = option;
1491 break;
1492 case Opt_nobh:
1493 set_opt(sbi->s_mount_opt, NOBH);
1494 break;
1495 case Opt_bh:
1496 clear_opt(sbi->s_mount_opt, NOBH);
1497 break;
1498 case Opt_i_version:
1499 set_opt(sbi->s_mount_opt, I_VERSION);
1500 sb->s_flags |= MS_I_VERSION;
1501 break;
1502 case Opt_nodelalloc:
1503 clear_opt(sbi->s_mount_opt, DELALLOC);
1504 break;
1505 case Opt_stripe:
1506 if (match_int(&args[0], &option))
1507 return 0;
1508 if (option < 0)
1509 return 0;
1510 sbi->s_stripe = option;
1511 break;
1512 case Opt_delalloc:
1513 set_opt(sbi->s_mount_opt, DELALLOC);
1514 break;
1515 case Opt_block_validity:
1516 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1517 break;
1518 case Opt_noblock_validity:
1519 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1520 break;
1521 case Opt_inode_readahead_blks:
1522 if (match_int(&args[0], &option))
1523 return 0;
1524 if (option < 0 || option > (1 << 30))
1525 return 0;
1526 if (!is_power_of_2(option)) {
1527 printk(KERN_ERR "EXT4-fs: inode_readahead_blks"
1528 " must be a power of 2\n");
1529 return 0;
1530 }
1531 sbi->s_inode_readahead_blks = option;
1532 break;
1533 case Opt_journal_ioprio:
1534 if (match_int(&args[0], &option))
1535 return 0;
1536 if (option < 0 || option > 7)
1537 break;
1538 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1539 option);
1540 break;
1541 case Opt_noauto_da_alloc:
1542 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1543 break;
1544 case Opt_auto_da_alloc:
1545 if (match_int(&args[0], &option)) {
1546 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1547 break;
1548 }
1549 if (option)
1550 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1551 else
1552 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1553 break;
1554 default:
1555 printk(KERN_ERR
1556 "EXT4-fs: Unrecognized mount option \"%s\" "
1557 "or missing value\n", p);
1558 return 0;
1559 }
1560 }
1561 #ifdef CONFIG_QUOTA
1562 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1563 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1564 sbi->s_qf_names[USRQUOTA])
1565 clear_opt(sbi->s_mount_opt, USRQUOTA);
1566
1567 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1568 sbi->s_qf_names[GRPQUOTA])
1569 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1570
1571 if ((sbi->s_qf_names[USRQUOTA] &&
1572 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1573 (sbi->s_qf_names[GRPQUOTA] &&
1574 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1575 printk(KERN_ERR "EXT4-fs: old and new quota "
1576 "format mixing.\n");
1577 return 0;
1578 }
1579
1580 if (!sbi->s_jquota_fmt) {
1581 printk(KERN_ERR "EXT4-fs: journaled quota format "
1582 "not specified.\n");
1583 return 0;
1584 }
1585 } else {
1586 if (sbi->s_jquota_fmt) {
1587 printk(KERN_ERR "EXT4-fs: journaled quota format "
1588 "specified with no journaling "
1589 "enabled.\n");
1590 return 0;
1591 }
1592 }
1593 #endif
1594 return 1;
1595 }
1596
1597 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1598 int read_only)
1599 {
1600 struct ext4_sb_info *sbi = EXT4_SB(sb);
1601 int res = 0;
1602
1603 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1604 printk(KERN_ERR "EXT4-fs warning: revision level too high, "
1605 "forcing read-only mode\n");
1606 res = MS_RDONLY;
1607 }
1608 if (read_only)
1609 return res;
1610 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1611 printk(KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1612 "running e2fsck is recommended\n");
1613 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1614 printk(KERN_WARNING
1615 "EXT4-fs warning: mounting fs with errors, "
1616 "running e2fsck is recommended\n");
1617 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1618 le16_to_cpu(es->s_mnt_count) >=
1619 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1620 printk(KERN_WARNING
1621 "EXT4-fs warning: maximal mount count reached, "
1622 "running e2fsck is recommended\n");
1623 else if (le32_to_cpu(es->s_checkinterval) &&
1624 (le32_to_cpu(es->s_lastcheck) +
1625 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1626 printk(KERN_WARNING
1627 "EXT4-fs warning: checktime reached, "
1628 "running e2fsck is recommended\n");
1629 if (!sbi->s_journal)
1630 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1631 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1632 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1633 le16_add_cpu(&es->s_mnt_count, 1);
1634 es->s_mtime = cpu_to_le32(get_seconds());
1635 ext4_update_dynamic_rev(sb);
1636 if (sbi->s_journal)
1637 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1638
1639 ext4_commit_super(sb, 1);
1640 if (test_opt(sb, DEBUG))
1641 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1642 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1643 sb->s_blocksize,
1644 sbi->s_groups_count,
1645 EXT4_BLOCKS_PER_GROUP(sb),
1646 EXT4_INODES_PER_GROUP(sb),
1647 sbi->s_mount_opt);
1648
1649 if (EXT4_SB(sb)->s_journal) {
1650 printk(KERN_INFO "EXT4 FS on %s, %s journal on %s\n",
1651 sb->s_id, EXT4_SB(sb)->s_journal->j_inode ? "internal" :
1652 "external", EXT4_SB(sb)->s_journal->j_devname);
1653 } else {
1654 printk(KERN_INFO "EXT4 FS on %s, no journal\n", sb->s_id);
1655 }
1656 return res;
1657 }
1658
1659 static int ext4_fill_flex_info(struct super_block *sb)
1660 {
1661 struct ext4_sb_info *sbi = EXT4_SB(sb);
1662 struct ext4_group_desc *gdp = NULL;
1663 ext4_group_t flex_group_count;
1664 ext4_group_t flex_group;
1665 int groups_per_flex = 0;
1666 size_t size;
1667 int i;
1668
1669 if (!sbi->s_es->s_log_groups_per_flex) {
1670 sbi->s_log_groups_per_flex = 0;
1671 return 1;
1672 }
1673
1674 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1675 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1676
1677 /* We allocate both existing and potentially added groups */
1678 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1679 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1680 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1681 size = flex_group_count * sizeof(struct flex_groups);
1682 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1683 if (sbi->s_flex_groups == NULL) {
1684 sbi->s_flex_groups = vmalloc(size);
1685 if (sbi->s_flex_groups)
1686 memset(sbi->s_flex_groups, 0, size);
1687 }
1688 if (sbi->s_flex_groups == NULL) {
1689 printk(KERN_ERR "EXT4-fs: not enough memory for "
1690 "%u flex groups\n", flex_group_count);
1691 goto failed;
1692 }
1693
1694 for (i = 0; i < sbi->s_groups_count; i++) {
1695 gdp = ext4_get_group_desc(sb, i, NULL);
1696
1697 flex_group = ext4_flex_group(sbi, i);
1698 atomic_set(&sbi->s_flex_groups[flex_group].free_inodes,
1699 ext4_free_inodes_count(sb, gdp));
1700 atomic_set(&sbi->s_flex_groups[flex_group].free_blocks,
1701 ext4_free_blks_count(sb, gdp));
1702 atomic_set(&sbi->s_flex_groups[flex_group].used_dirs,
1703 ext4_used_dirs_count(sb, gdp));
1704 }
1705
1706 return 1;
1707 failed:
1708 return 0;
1709 }
1710
1711 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1712 struct ext4_group_desc *gdp)
1713 {
1714 __u16 crc = 0;
1715
1716 if (sbi->s_es->s_feature_ro_compat &
1717 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1718 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1719 __le32 le_group = cpu_to_le32(block_group);
1720
1721 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1722 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1723 crc = crc16(crc, (__u8 *)gdp, offset);
1724 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1725 /* for checksum of struct ext4_group_desc do the rest...*/
1726 if ((sbi->s_es->s_feature_incompat &
1727 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1728 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1729 crc = crc16(crc, (__u8 *)gdp + offset,
1730 le16_to_cpu(sbi->s_es->s_desc_size) -
1731 offset);
1732 }
1733
1734 return cpu_to_le16(crc);
1735 }
1736
1737 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1738 struct ext4_group_desc *gdp)
1739 {
1740 if ((sbi->s_es->s_feature_ro_compat &
1741 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1742 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1743 return 0;
1744
1745 return 1;
1746 }
1747
1748 /* Called at mount-time, super-block is locked */
1749 static int ext4_check_descriptors(struct super_block *sb)
1750 {
1751 struct ext4_sb_info *sbi = EXT4_SB(sb);
1752 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1753 ext4_fsblk_t last_block;
1754 ext4_fsblk_t block_bitmap;
1755 ext4_fsblk_t inode_bitmap;
1756 ext4_fsblk_t inode_table;
1757 int flexbg_flag = 0;
1758 ext4_group_t i;
1759
1760 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1761 flexbg_flag = 1;
1762
1763 ext4_debug("Checking group descriptors");
1764
1765 for (i = 0; i < sbi->s_groups_count; i++) {
1766 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1767
1768 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1769 last_block = ext4_blocks_count(sbi->s_es) - 1;
1770 else
1771 last_block = first_block +
1772 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1773
1774 block_bitmap = ext4_block_bitmap(sb, gdp);
1775 if (block_bitmap < first_block || block_bitmap > last_block) {
1776 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1777 "Block bitmap for group %u not in group "
1778 "(block %llu)!\n", i, block_bitmap);
1779 return 0;
1780 }
1781 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1782 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1783 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1784 "Inode bitmap for group %u not in group "
1785 "(block %llu)!\n", i, inode_bitmap);
1786 return 0;
1787 }
1788 inode_table = ext4_inode_table(sb, gdp);
1789 if (inode_table < first_block ||
1790 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1791 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1792 "Inode table for group %u not in group "
1793 "(block %llu)!\n", i, inode_table);
1794 return 0;
1795 }
1796 ext4_lock_group(sb, i);
1797 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1798 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1799 "Checksum for group %u failed (%u!=%u)\n",
1800 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1801 gdp)), le16_to_cpu(gdp->bg_checksum));
1802 if (!(sb->s_flags & MS_RDONLY)) {
1803 ext4_unlock_group(sb, i);
1804 return 0;
1805 }
1806 }
1807 ext4_unlock_group(sb, i);
1808 if (!flexbg_flag)
1809 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1810 }
1811
1812 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1813 sbi->s_es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
1814 return 1;
1815 }
1816
1817 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1818 * the superblock) which were deleted from all directories, but held open by
1819 * a process at the time of a crash. We walk the list and try to delete these
1820 * inodes at recovery time (only with a read-write filesystem).
1821 *
1822 * In order to keep the orphan inode chain consistent during traversal (in
1823 * case of crash during recovery), we link each inode into the superblock
1824 * orphan list_head and handle it the same way as an inode deletion during
1825 * normal operation (which journals the operations for us).
1826 *
1827 * We only do an iget() and an iput() on each inode, which is very safe if we
1828 * accidentally point at an in-use or already deleted inode. The worst that
1829 * can happen in this case is that we get a "bit already cleared" message from
1830 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1831 * e2fsck was run on this filesystem, and it must have already done the orphan
1832 * inode cleanup for us, so we can safely abort without any further action.
1833 */
1834 static void ext4_orphan_cleanup(struct super_block *sb,
1835 struct ext4_super_block *es)
1836 {
1837 unsigned int s_flags = sb->s_flags;
1838 int nr_orphans = 0, nr_truncates = 0;
1839 #ifdef CONFIG_QUOTA
1840 int i;
1841 #endif
1842 if (!es->s_last_orphan) {
1843 jbd_debug(4, "no orphan inodes to clean up\n");
1844 return;
1845 }
1846
1847 if (bdev_read_only(sb->s_bdev)) {
1848 printk(KERN_ERR "EXT4-fs: write access "
1849 "unavailable, skipping orphan cleanup.\n");
1850 return;
1851 }
1852
1853 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1854 if (es->s_last_orphan)
1855 jbd_debug(1, "Errors on filesystem, "
1856 "clearing orphan list.\n");
1857 es->s_last_orphan = 0;
1858 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1859 return;
1860 }
1861
1862 if (s_flags & MS_RDONLY) {
1863 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
1864 sb->s_id);
1865 sb->s_flags &= ~MS_RDONLY;
1866 }
1867 #ifdef CONFIG_QUOTA
1868 /* Needed for iput() to work correctly and not trash data */
1869 sb->s_flags |= MS_ACTIVE;
1870 /* Turn on quotas so that they are updated correctly */
1871 for (i = 0; i < MAXQUOTAS; i++) {
1872 if (EXT4_SB(sb)->s_qf_names[i]) {
1873 int ret = ext4_quota_on_mount(sb, i);
1874 if (ret < 0)
1875 printk(KERN_ERR
1876 "EXT4-fs: Cannot turn on journaled "
1877 "quota: error %d\n", ret);
1878 }
1879 }
1880 #endif
1881
1882 while (es->s_last_orphan) {
1883 struct inode *inode;
1884
1885 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1886 if (IS_ERR(inode)) {
1887 es->s_last_orphan = 0;
1888 break;
1889 }
1890
1891 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1892 vfs_dq_init(inode);
1893 if (inode->i_nlink) {
1894 printk(KERN_DEBUG
1895 "%s: truncating inode %lu to %lld bytes\n",
1896 __func__, inode->i_ino, inode->i_size);
1897 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1898 inode->i_ino, inode->i_size);
1899 ext4_truncate(inode);
1900 nr_truncates++;
1901 } else {
1902 printk(KERN_DEBUG
1903 "%s: deleting unreferenced inode %lu\n",
1904 __func__, inode->i_ino);
1905 jbd_debug(2, "deleting unreferenced inode %lu\n",
1906 inode->i_ino);
1907 nr_orphans++;
1908 }
1909 iput(inode); /* The delete magic happens here! */
1910 }
1911
1912 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1913
1914 if (nr_orphans)
1915 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
1916 sb->s_id, PLURAL(nr_orphans));
1917 if (nr_truncates)
1918 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
1919 sb->s_id, PLURAL(nr_truncates));
1920 #ifdef CONFIG_QUOTA
1921 /* Turn quotas off */
1922 for (i = 0; i < MAXQUOTAS; i++) {
1923 if (sb_dqopt(sb)->files[i])
1924 vfs_quota_off(sb, i, 0);
1925 }
1926 #endif
1927 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1928 }
1929 /*
1930 * Maximal extent format file size.
1931 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1932 * extent format containers, within a sector_t, and within i_blocks
1933 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1934 * so that won't be a limiting factor.
1935 *
1936 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1937 */
1938 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1939 {
1940 loff_t res;
1941 loff_t upper_limit = MAX_LFS_FILESIZE;
1942
1943 /* small i_blocks in vfs inode? */
1944 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1945 /*
1946 * CONFIG_LBD is not enabled implies the inode
1947 * i_block represent total blocks in 512 bytes
1948 * 32 == size of vfs inode i_blocks * 8
1949 */
1950 upper_limit = (1LL << 32) - 1;
1951
1952 /* total blocks in file system block size */
1953 upper_limit >>= (blkbits - 9);
1954 upper_limit <<= blkbits;
1955 }
1956
1957 /* 32-bit extent-start container, ee_block */
1958 res = 1LL << 32;
1959 res <<= blkbits;
1960 res -= 1;
1961
1962 /* Sanity check against vm- & vfs- imposed limits */
1963 if (res > upper_limit)
1964 res = upper_limit;
1965
1966 return res;
1967 }
1968
1969 /*
1970 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1971 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1972 * We need to be 1 filesystem block less than the 2^48 sector limit.
1973 */
1974 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
1975 {
1976 loff_t res = EXT4_NDIR_BLOCKS;
1977 int meta_blocks;
1978 loff_t upper_limit;
1979 /* This is calculated to be the largest file size for a
1980 * dense, bitmapped file such that the total number of
1981 * sectors in the file, including data and all indirect blocks,
1982 * does not exceed 2^48 -1
1983 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1984 * total number of 512 bytes blocks of the file
1985 */
1986
1987 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1988 /*
1989 * !has_huge_files or CONFIG_LBD is not enabled
1990 * implies the inode i_block represent total blocks in
1991 * 512 bytes 32 == size of vfs inode i_blocks * 8
1992 */
1993 upper_limit = (1LL << 32) - 1;
1994
1995 /* total blocks in file system block size */
1996 upper_limit >>= (bits - 9);
1997
1998 } else {
1999 /*
2000 * We use 48 bit ext4_inode i_blocks
2001 * With EXT4_HUGE_FILE_FL set the i_blocks
2002 * represent total number of blocks in
2003 * file system block size
2004 */
2005 upper_limit = (1LL << 48) - 1;
2006
2007 }
2008
2009 /* indirect blocks */
2010 meta_blocks = 1;
2011 /* double indirect blocks */
2012 meta_blocks += 1 + (1LL << (bits-2));
2013 /* tripple indirect blocks */
2014 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2015
2016 upper_limit -= meta_blocks;
2017 upper_limit <<= bits;
2018
2019 res += 1LL << (bits-2);
2020 res += 1LL << (2*(bits-2));
2021 res += 1LL << (3*(bits-2));
2022 res <<= bits;
2023 if (res > upper_limit)
2024 res = upper_limit;
2025
2026 if (res > MAX_LFS_FILESIZE)
2027 res = MAX_LFS_FILESIZE;
2028
2029 return res;
2030 }
2031
2032 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2033 ext4_fsblk_t logical_sb_block, int nr)
2034 {
2035 struct ext4_sb_info *sbi = EXT4_SB(sb);
2036 ext4_group_t bg, first_meta_bg;
2037 int has_super = 0;
2038
2039 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2040
2041 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2042 nr < first_meta_bg)
2043 return logical_sb_block + nr + 1;
2044 bg = sbi->s_desc_per_block * nr;
2045 if (ext4_bg_has_super(sb, bg))
2046 has_super = 1;
2047 return (has_super + ext4_group_first_block_no(sb, bg));
2048 }
2049
2050 /**
2051 * ext4_get_stripe_size: Get the stripe size.
2052 * @sbi: In memory super block info
2053 *
2054 * If we have specified it via mount option, then
2055 * use the mount option value. If the value specified at mount time is
2056 * greater than the blocks per group use the super block value.
2057 * If the super block value is greater than blocks per group return 0.
2058 * Allocator needs it be less than blocks per group.
2059 *
2060 */
2061 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2062 {
2063 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2064 unsigned long stripe_width =
2065 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2066
2067 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2068 return sbi->s_stripe;
2069
2070 if (stripe_width <= sbi->s_blocks_per_group)
2071 return stripe_width;
2072
2073 if (stride <= sbi->s_blocks_per_group)
2074 return stride;
2075
2076 return 0;
2077 }
2078
2079 /* sysfs supprt */
2080
2081 struct ext4_attr {
2082 struct attribute attr;
2083 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2084 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2085 const char *, size_t);
2086 int offset;
2087 };
2088
2089 static int parse_strtoul(const char *buf,
2090 unsigned long max, unsigned long *value)
2091 {
2092 char *endp;
2093
2094 while (*buf && isspace(*buf))
2095 buf++;
2096 *value = simple_strtoul(buf, &endp, 0);
2097 while (*endp && isspace(*endp))
2098 endp++;
2099 if (*endp || *value > max)
2100 return -EINVAL;
2101
2102 return 0;
2103 }
2104
2105 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2106 struct ext4_sb_info *sbi,
2107 char *buf)
2108 {
2109 return snprintf(buf, PAGE_SIZE, "%llu\n",
2110 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2111 }
2112
2113 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2114 struct ext4_sb_info *sbi, char *buf)
2115 {
2116 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2117
2118 return snprintf(buf, PAGE_SIZE, "%lu\n",
2119 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2120 sbi->s_sectors_written_start) >> 1);
2121 }
2122
2123 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2124 struct ext4_sb_info *sbi, char *buf)
2125 {
2126 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2127
2128 return snprintf(buf, PAGE_SIZE, "%llu\n",
2129 sbi->s_kbytes_written +
2130 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2131 EXT4_SB(sb)->s_sectors_written_start) >> 1));
2132 }
2133
2134 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2135 struct ext4_sb_info *sbi,
2136 const char *buf, size_t count)
2137 {
2138 unsigned long t;
2139
2140 if (parse_strtoul(buf, 0x40000000, &t))
2141 return -EINVAL;
2142
2143 if (!is_power_of_2(t))
2144 return -EINVAL;
2145
2146 sbi->s_inode_readahead_blks = t;
2147 return count;
2148 }
2149
2150 static ssize_t sbi_ui_show(struct ext4_attr *a,
2151 struct ext4_sb_info *sbi, char *buf)
2152 {
2153 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2154
2155 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2156 }
2157
2158 static ssize_t sbi_ui_store(struct ext4_attr *a,
2159 struct ext4_sb_info *sbi,
2160 const char *buf, size_t count)
2161 {
2162 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2163 unsigned long t;
2164
2165 if (parse_strtoul(buf, 0xffffffff, &t))
2166 return -EINVAL;
2167 *ui = t;
2168 return count;
2169 }
2170
2171 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2172 static struct ext4_attr ext4_attr_##_name = { \
2173 .attr = {.name = __stringify(_name), .mode = _mode }, \
2174 .show = _show, \
2175 .store = _store, \
2176 .offset = offsetof(struct ext4_sb_info, _elname), \
2177 }
2178 #define EXT4_ATTR(name, mode, show, store) \
2179 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2180
2181 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2182 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2183 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2184 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2185 #define ATTR_LIST(name) &ext4_attr_##name.attr
2186
2187 EXT4_RO_ATTR(delayed_allocation_blocks);
2188 EXT4_RO_ATTR(session_write_kbytes);
2189 EXT4_RO_ATTR(lifetime_write_kbytes);
2190 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2191 inode_readahead_blks_store, s_inode_readahead_blks);
2192 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2193 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2194 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2195 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2196 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2197 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2198
2199 static struct attribute *ext4_attrs[] = {
2200 ATTR_LIST(delayed_allocation_blocks),
2201 ATTR_LIST(session_write_kbytes),
2202 ATTR_LIST(lifetime_write_kbytes),
2203 ATTR_LIST(inode_readahead_blks),
2204 ATTR_LIST(mb_stats),
2205 ATTR_LIST(mb_max_to_scan),
2206 ATTR_LIST(mb_min_to_scan),
2207 ATTR_LIST(mb_order2_req),
2208 ATTR_LIST(mb_stream_req),
2209 ATTR_LIST(mb_group_prealloc),
2210 NULL,
2211 };
2212
2213 static ssize_t ext4_attr_show(struct kobject *kobj,
2214 struct attribute *attr, char *buf)
2215 {
2216 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2217 s_kobj);
2218 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2219
2220 return a->show ? a->show(a, sbi, buf) : 0;
2221 }
2222
2223 static ssize_t ext4_attr_store(struct kobject *kobj,
2224 struct attribute *attr,
2225 const char *buf, size_t len)
2226 {
2227 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2228 s_kobj);
2229 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2230
2231 return a->store ? a->store(a, sbi, buf, len) : 0;
2232 }
2233
2234 static void ext4_sb_release(struct kobject *kobj)
2235 {
2236 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2237 s_kobj);
2238 complete(&sbi->s_kobj_unregister);
2239 }
2240
2241
2242 static struct sysfs_ops ext4_attr_ops = {
2243 .show = ext4_attr_show,
2244 .store = ext4_attr_store,
2245 };
2246
2247 static struct kobj_type ext4_ktype = {
2248 .default_attrs = ext4_attrs,
2249 .sysfs_ops = &ext4_attr_ops,
2250 .release = ext4_sb_release,
2251 };
2252
2253 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2254 __releases(kernel_lock)
2255 __acquires(kernel_lock)
2256
2257 {
2258 struct buffer_head *bh;
2259 struct ext4_super_block *es = NULL;
2260 struct ext4_sb_info *sbi;
2261 ext4_fsblk_t block;
2262 ext4_fsblk_t sb_block = get_sb_block(&data);
2263 ext4_fsblk_t logical_sb_block;
2264 unsigned long offset = 0;
2265 unsigned long journal_devnum = 0;
2266 unsigned long def_mount_opts;
2267 struct inode *root;
2268 char *cp;
2269 const char *descr;
2270 int ret = -EINVAL;
2271 int blocksize;
2272 unsigned int db_count;
2273 unsigned int i;
2274 int needs_recovery, has_huge_files;
2275 int features;
2276 __u64 blocks_count;
2277 int err;
2278 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2279
2280 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2281 if (!sbi)
2282 return -ENOMEM;
2283
2284 sbi->s_blockgroup_lock =
2285 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2286 if (!sbi->s_blockgroup_lock) {
2287 kfree(sbi);
2288 return -ENOMEM;
2289 }
2290 sb->s_fs_info = sbi;
2291 sbi->s_mount_opt = 0;
2292 sbi->s_resuid = EXT4_DEF_RESUID;
2293 sbi->s_resgid = EXT4_DEF_RESGID;
2294 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2295 sbi->s_sb_block = sb_block;
2296 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2297 sectors[1]);
2298
2299 unlock_kernel();
2300
2301 /* Cleanup superblock name */
2302 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2303 *cp = '!';
2304
2305 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2306 if (!blocksize) {
2307 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
2308 goto out_fail;
2309 }
2310
2311 /*
2312 * The ext4 superblock will not be buffer aligned for other than 1kB
2313 * block sizes. We need to calculate the offset from buffer start.
2314 */
2315 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2316 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2317 offset = do_div(logical_sb_block, blocksize);
2318 } else {
2319 logical_sb_block = sb_block;
2320 }
2321
2322 if (!(bh = sb_bread(sb, logical_sb_block))) {
2323 printk(KERN_ERR "EXT4-fs: unable to read superblock\n");
2324 goto out_fail;
2325 }
2326 /*
2327 * Note: s_es must be initialized as soon as possible because
2328 * some ext4 macro-instructions depend on its value
2329 */
2330 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2331 sbi->s_es = es;
2332 sb->s_magic = le16_to_cpu(es->s_magic);
2333 if (sb->s_magic != EXT4_SUPER_MAGIC)
2334 goto cantfind_ext4;
2335 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2336
2337 /* Set defaults before we parse the mount options */
2338 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2339 if (def_mount_opts & EXT4_DEFM_DEBUG)
2340 set_opt(sbi->s_mount_opt, DEBUG);
2341 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
2342 set_opt(sbi->s_mount_opt, GRPID);
2343 if (def_mount_opts & EXT4_DEFM_UID16)
2344 set_opt(sbi->s_mount_opt, NO_UID32);
2345 #ifdef CONFIG_EXT4_FS_XATTR
2346 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2347 set_opt(sbi->s_mount_opt, XATTR_USER);
2348 #endif
2349 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2350 if (def_mount_opts & EXT4_DEFM_ACL)
2351 set_opt(sbi->s_mount_opt, POSIX_ACL);
2352 #endif
2353 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2354 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2355 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2356 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2357 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2358 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2359
2360 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2361 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2362 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2363 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2364 else
2365 set_opt(sbi->s_mount_opt, ERRORS_RO);
2366
2367 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2368 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2369 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2370 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2371 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2372 sbi->s_mb_history_max = default_mb_history_length;
2373
2374 set_opt(sbi->s_mount_opt, BARRIER);
2375
2376 /*
2377 * enable delayed allocation by default
2378 * Use -o nodelalloc to turn it off
2379 */
2380 set_opt(sbi->s_mount_opt, DELALLOC);
2381
2382
2383 if (!parse_options((char *) data, sb, &journal_devnum,
2384 &journal_ioprio, NULL, 0))
2385 goto failed_mount;
2386
2387 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2388 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2389
2390 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2391 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2392 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2393 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2394 printk(KERN_WARNING
2395 "EXT4-fs warning: feature flags set on rev 0 fs, "
2396 "running e2fsck is recommended\n");
2397
2398 /*
2399 * Check feature flags regardless of the revision level, since we
2400 * previously didn't change the revision level when setting the flags,
2401 * so there is a chance incompat flags are set on a rev 0 filesystem.
2402 */
2403 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
2404 if (features) {
2405 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
2406 "unsupported optional features (%x).\n", sb->s_id,
2407 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2408 ~EXT4_FEATURE_INCOMPAT_SUPP));
2409 goto failed_mount;
2410 }
2411 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
2412 if (!(sb->s_flags & MS_RDONLY) && features) {
2413 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
2414 "unsupported optional features (%x).\n", sb->s_id,
2415 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2416 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2417 goto failed_mount;
2418 }
2419 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2420 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2421 if (has_huge_files) {
2422 /*
2423 * Large file size enabled file system can only be
2424 * mount if kernel is build with CONFIG_LBD
2425 */
2426 if (sizeof(root->i_blocks) < sizeof(u64) &&
2427 !(sb->s_flags & MS_RDONLY)) {
2428 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2429 "files cannot be mounted read-write "
2430 "without CONFIG_LBD.\n", sb->s_id);
2431 goto failed_mount;
2432 }
2433 }
2434 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2435
2436 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2437 blocksize > EXT4_MAX_BLOCK_SIZE) {
2438 printk(KERN_ERR
2439 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
2440 blocksize, sb->s_id);
2441 goto failed_mount;
2442 }
2443
2444 if (sb->s_blocksize != blocksize) {
2445
2446 /* Validate the filesystem blocksize */
2447 if (!sb_set_blocksize(sb, blocksize)) {
2448 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2449 blocksize);
2450 goto failed_mount;
2451 }
2452
2453 brelse(bh);
2454 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2455 offset = do_div(logical_sb_block, blocksize);
2456 bh = sb_bread(sb, logical_sb_block);
2457 if (!bh) {
2458 printk(KERN_ERR
2459 "EXT4-fs: Can't read superblock on 2nd try.\n");
2460 goto failed_mount;
2461 }
2462 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2463 sbi->s_es = es;
2464 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2465 printk(KERN_ERR
2466 "EXT4-fs: Magic mismatch, very weird !\n");
2467 goto failed_mount;
2468 }
2469 }
2470
2471 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2472 has_huge_files);
2473 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2474
2475 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2476 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2477 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2478 } else {
2479 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2480 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2481 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2482 (!is_power_of_2(sbi->s_inode_size)) ||
2483 (sbi->s_inode_size > blocksize)) {
2484 printk(KERN_ERR
2485 "EXT4-fs: unsupported inode size: %d\n",
2486 sbi->s_inode_size);
2487 goto failed_mount;
2488 }
2489 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2490 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2491 }
2492 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2493 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2494 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2495 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2496 !is_power_of_2(sbi->s_desc_size)) {
2497 printk(KERN_ERR
2498 "EXT4-fs: unsupported descriptor size %lu\n",
2499 sbi->s_desc_size);
2500 goto failed_mount;
2501 }
2502 } else
2503 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2504 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2505 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2506 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2507 goto cantfind_ext4;
2508 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2509 if (sbi->s_inodes_per_block == 0)
2510 goto cantfind_ext4;
2511 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2512 sbi->s_inodes_per_block;
2513 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2514 sbi->s_sbh = bh;
2515 sbi->s_mount_state = le16_to_cpu(es->s_state);
2516 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2517 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2518 for (i = 0; i < 4; i++)
2519 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2520 sbi->s_def_hash_version = es->s_def_hash_version;
2521 i = le32_to_cpu(es->s_flags);
2522 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2523 sbi->s_hash_unsigned = 3;
2524 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2525 #ifdef __CHAR_UNSIGNED__
2526 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2527 sbi->s_hash_unsigned = 3;
2528 #else
2529 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2530 #endif
2531 sb->s_dirt = 1;
2532 }
2533
2534 if (sbi->s_blocks_per_group > blocksize * 8) {
2535 printk(KERN_ERR
2536 "EXT4-fs: #blocks per group too big: %lu\n",
2537 sbi->s_blocks_per_group);
2538 goto failed_mount;
2539 }
2540 if (sbi->s_inodes_per_group > blocksize * 8) {
2541 printk(KERN_ERR
2542 "EXT4-fs: #inodes per group too big: %lu\n",
2543 sbi->s_inodes_per_group);
2544 goto failed_mount;
2545 }
2546
2547 if (ext4_blocks_count(es) >
2548 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
2549 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
2550 " too large to mount safely\n", sb->s_id);
2551 if (sizeof(sector_t) < 8)
2552 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
2553 "enabled\n");
2554 goto failed_mount;
2555 }
2556
2557 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2558 goto cantfind_ext4;
2559
2560 /* check blocks count against device size */
2561 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2562 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2563 printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu "
2564 "exceeds size of device (%llu blocks)\n",
2565 ext4_blocks_count(es), blocks_count);
2566 goto failed_mount;
2567 }
2568
2569 /*
2570 * It makes no sense for the first data block to be beyond the end
2571 * of the filesystem.
2572 */
2573 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2574 printk(KERN_WARNING "EXT4-fs: bad geometry: first data"
2575 "block %u is beyond end of filesystem (%llu)\n",
2576 le32_to_cpu(es->s_first_data_block),
2577 ext4_blocks_count(es));
2578 goto failed_mount;
2579 }
2580 blocks_count = (ext4_blocks_count(es) -
2581 le32_to_cpu(es->s_first_data_block) +
2582 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2583 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2584 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2585 printk(KERN_WARNING "EXT4-fs: groups count too large: %u "
2586 "(block count %llu, first data block %u, "
2587 "blocks per group %lu)\n", sbi->s_groups_count,
2588 ext4_blocks_count(es),
2589 le32_to_cpu(es->s_first_data_block),
2590 EXT4_BLOCKS_PER_GROUP(sb));
2591 goto failed_mount;
2592 }
2593 sbi->s_groups_count = blocks_count;
2594 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2595 EXT4_DESC_PER_BLOCK(sb);
2596 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2597 GFP_KERNEL);
2598 if (sbi->s_group_desc == NULL) {
2599 printk(KERN_ERR "EXT4-fs: not enough memory\n");
2600 goto failed_mount;
2601 }
2602
2603 #ifdef CONFIG_PROC_FS
2604 if (ext4_proc_root)
2605 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2606 #endif
2607
2608 bgl_lock_init(sbi->s_blockgroup_lock);
2609
2610 for (i = 0; i < db_count; i++) {
2611 block = descriptor_loc(sb, logical_sb_block, i);
2612 sbi->s_group_desc[i] = sb_bread(sb, block);
2613 if (!sbi->s_group_desc[i]) {
2614 printk(KERN_ERR "EXT4-fs: "
2615 "can't read group descriptor %d\n", i);
2616 db_count = i;
2617 goto failed_mount2;
2618 }
2619 }
2620 if (!ext4_check_descriptors(sb)) {
2621 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
2622 goto failed_mount2;
2623 }
2624 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2625 if (!ext4_fill_flex_info(sb)) {
2626 printk(KERN_ERR
2627 "EXT4-fs: unable to initialize "
2628 "flex_bg meta info!\n");
2629 goto failed_mount2;
2630 }
2631
2632 sbi->s_gdb_count = db_count;
2633 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2634 spin_lock_init(&sbi->s_next_gen_lock);
2635
2636 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2637 ext4_count_free_blocks(sb));
2638 if (!err) {
2639 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2640 ext4_count_free_inodes(sb));
2641 }
2642 if (!err) {
2643 err = percpu_counter_init(&sbi->s_dirs_counter,
2644 ext4_count_dirs(sb));
2645 }
2646 if (!err) {
2647 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2648 }
2649 if (err) {
2650 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2651 goto failed_mount3;
2652 }
2653
2654 sbi->s_stripe = ext4_get_stripe_size(sbi);
2655
2656 /*
2657 * set up enough so that it can read an inode
2658 */
2659 if (!test_opt(sb, NOLOAD) &&
2660 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2661 sb->s_op = &ext4_sops;
2662 else
2663 sb->s_op = &ext4_nojournal_sops;
2664 sb->s_export_op = &ext4_export_ops;
2665 sb->s_xattr = ext4_xattr_handlers;
2666 #ifdef CONFIG_QUOTA
2667 sb->s_qcop = &ext4_qctl_operations;
2668 sb->dq_op = &ext4_quota_operations;
2669 #endif
2670 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2671 mutex_init(&sbi->s_orphan_lock);
2672 mutex_init(&sbi->s_resize_lock);
2673
2674 sb->s_root = NULL;
2675
2676 needs_recovery = (es->s_last_orphan != 0 ||
2677 EXT4_HAS_INCOMPAT_FEATURE(sb,
2678 EXT4_FEATURE_INCOMPAT_RECOVER));
2679
2680 /*
2681 * The first inode we look at is the journal inode. Don't try
2682 * root first: it may be modified in the journal!
2683 */
2684 if (!test_opt(sb, NOLOAD) &&
2685 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2686 if (ext4_load_journal(sb, es, journal_devnum))
2687 goto failed_mount3;
2688 if (!(sb->s_flags & MS_RDONLY) &&
2689 EXT4_SB(sb)->s_journal->j_failed_commit) {
2690 printk(KERN_CRIT "EXT4-fs error (device %s): "
2691 "ext4_fill_super: Journal transaction "
2692 "%u is corrupt\n", sb->s_id,
2693 EXT4_SB(sb)->s_journal->j_failed_commit);
2694 if (test_opt(sb, ERRORS_RO)) {
2695 printk(KERN_CRIT
2696 "Mounting filesystem read-only\n");
2697 sb->s_flags |= MS_RDONLY;
2698 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2699 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2700 }
2701 if (test_opt(sb, ERRORS_PANIC)) {
2702 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2703 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2704 ext4_commit_super(sb, 1);
2705 goto failed_mount4;
2706 }
2707 }
2708 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2709 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2710 printk(KERN_ERR "EXT4-fs: required journal recovery "
2711 "suppressed and not mounted read-only\n");
2712 goto failed_mount4;
2713 } else {
2714 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2715 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2716 sbi->s_journal = NULL;
2717 needs_recovery = 0;
2718 goto no_journal;
2719 }
2720
2721 if (ext4_blocks_count(es) > 0xffffffffULL &&
2722 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2723 JBD2_FEATURE_INCOMPAT_64BIT)) {
2724 printk(KERN_ERR "EXT4-fs: Failed to set 64-bit journal feature\n");
2725 goto failed_mount4;
2726 }
2727
2728 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2729 jbd2_journal_set_features(sbi->s_journal,
2730 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2731 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2732 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2733 jbd2_journal_set_features(sbi->s_journal,
2734 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2735 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2736 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2737 } else {
2738 jbd2_journal_clear_features(sbi->s_journal,
2739 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2740 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2741 }
2742
2743 /* We have now updated the journal if required, so we can
2744 * validate the data journaling mode. */
2745 switch (test_opt(sb, DATA_FLAGS)) {
2746 case 0:
2747 /* No mode set, assume a default based on the journal
2748 * capabilities: ORDERED_DATA if the journal can
2749 * cope, else JOURNAL_DATA
2750 */
2751 if (jbd2_journal_check_available_features
2752 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2753 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2754 else
2755 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2756 break;
2757
2758 case EXT4_MOUNT_ORDERED_DATA:
2759 case EXT4_MOUNT_WRITEBACK_DATA:
2760 if (!jbd2_journal_check_available_features
2761 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2762 printk(KERN_ERR "EXT4-fs: Journal does not support "
2763 "requested data journaling mode\n");
2764 goto failed_mount4;
2765 }
2766 default:
2767 break;
2768 }
2769 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2770
2771 no_journal:
2772
2773 if (test_opt(sb, NOBH)) {
2774 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2775 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
2776 "its supported only with writeback mode\n");
2777 clear_opt(sbi->s_mount_opt, NOBH);
2778 }
2779 }
2780 /*
2781 * The jbd2_journal_load will have done any necessary log recovery,
2782 * so we can safely mount the rest of the filesystem now.
2783 */
2784
2785 root = ext4_iget(sb, EXT4_ROOT_INO);
2786 if (IS_ERR(root)) {
2787 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
2788 ret = PTR_ERR(root);
2789 goto failed_mount4;
2790 }
2791 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2792 iput(root);
2793 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
2794 goto failed_mount4;
2795 }
2796 sb->s_root = d_alloc_root(root);
2797 if (!sb->s_root) {
2798 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2799 iput(root);
2800 ret = -ENOMEM;
2801 goto failed_mount4;
2802 }
2803
2804 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2805
2806 /* determine the minimum size of new large inodes, if present */
2807 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2808 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2809 EXT4_GOOD_OLD_INODE_SIZE;
2810 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2811 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2812 if (sbi->s_want_extra_isize <
2813 le16_to_cpu(es->s_want_extra_isize))
2814 sbi->s_want_extra_isize =
2815 le16_to_cpu(es->s_want_extra_isize);
2816 if (sbi->s_want_extra_isize <
2817 le16_to_cpu(es->s_min_extra_isize))
2818 sbi->s_want_extra_isize =
2819 le16_to_cpu(es->s_min_extra_isize);
2820 }
2821 }
2822 /* Check if enough inode space is available */
2823 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2824 sbi->s_inode_size) {
2825 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2826 EXT4_GOOD_OLD_INODE_SIZE;
2827 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2828 "available.\n");
2829 }
2830
2831 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2832 printk(KERN_WARNING "EXT4-fs: Ignoring delalloc option - "
2833 "requested data journaling mode\n");
2834 clear_opt(sbi->s_mount_opt, DELALLOC);
2835 } else if (test_opt(sb, DELALLOC))
2836 printk(KERN_INFO "EXT4-fs: delayed allocation enabled\n");
2837
2838 err = ext4_setup_system_zone(sb);
2839 if (err) {
2840 printk(KERN_ERR "EXT4-fs: failed to initialize system "
2841 "zone (%d)\n", err);
2842 goto failed_mount4;
2843 }
2844
2845 ext4_ext_init(sb);
2846 err = ext4_mb_init(sb, needs_recovery);
2847 if (err) {
2848 printk(KERN_ERR "EXT4-fs: failed to initalize mballoc (%d)\n",
2849 err);
2850 goto failed_mount4;
2851 }
2852
2853 sbi->s_kobj.kset = ext4_kset;
2854 init_completion(&sbi->s_kobj_unregister);
2855 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2856 "%s", sb->s_id);
2857 if (err) {
2858 ext4_mb_release(sb);
2859 ext4_ext_release(sb);
2860 goto failed_mount4;
2861 };
2862
2863 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2864 ext4_orphan_cleanup(sb, es);
2865 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2866 if (needs_recovery) {
2867 printk(KERN_INFO "EXT4-fs: recovery complete.\n");
2868 ext4_mark_recovery_complete(sb, es);
2869 }
2870 if (EXT4_SB(sb)->s_journal) {
2871 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2872 descr = " journalled data mode";
2873 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2874 descr = " ordered data mode";
2875 else
2876 descr = " writeback data mode";
2877 } else
2878 descr = "out journal";
2879
2880 printk(KERN_INFO "EXT4-fs: mounted filesystem %s with%s\n",
2881 sb->s_id, descr);
2882
2883 lock_kernel();
2884 return 0;
2885
2886 cantfind_ext4:
2887 if (!silent)
2888 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
2889 sb->s_id);
2890 goto failed_mount;
2891
2892 failed_mount4:
2893 printk(KERN_ERR "EXT4-fs (device %s): mount failed\n", sb->s_id);
2894 ext4_release_system_zone(sb);
2895 if (sbi->s_journal) {
2896 jbd2_journal_destroy(sbi->s_journal);
2897 sbi->s_journal = NULL;
2898 }
2899 failed_mount3:
2900 if (sbi->s_flex_groups) {
2901 if (is_vmalloc_addr(sbi->s_flex_groups))
2902 vfree(sbi->s_flex_groups);
2903 else
2904 kfree(sbi->s_flex_groups);
2905 }
2906 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2907 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2908 percpu_counter_destroy(&sbi->s_dirs_counter);
2909 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2910 failed_mount2:
2911 for (i = 0; i < db_count; i++)
2912 brelse(sbi->s_group_desc[i]);
2913 kfree(sbi->s_group_desc);
2914 failed_mount:
2915 if (sbi->s_proc) {
2916 remove_proc_entry(sb->s_id, ext4_proc_root);
2917 }
2918 #ifdef CONFIG_QUOTA
2919 for (i = 0; i < MAXQUOTAS; i++)
2920 kfree(sbi->s_qf_names[i]);
2921 #endif
2922 ext4_blkdev_remove(sbi);
2923 brelse(bh);
2924 out_fail:
2925 sb->s_fs_info = NULL;
2926 kfree(sbi->s_blockgroup_lock);
2927 kfree(sbi);
2928 lock_kernel();
2929 return ret;
2930 }
2931
2932 /*
2933 * Setup any per-fs journal parameters now. We'll do this both on
2934 * initial mount, once the journal has been initialised but before we've
2935 * done any recovery; and again on any subsequent remount.
2936 */
2937 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2938 {
2939 struct ext4_sb_info *sbi = EXT4_SB(sb);
2940
2941 journal->j_commit_interval = sbi->s_commit_interval;
2942 journal->j_min_batch_time = sbi->s_min_batch_time;
2943 journal->j_max_batch_time = sbi->s_max_batch_time;
2944
2945 spin_lock(&journal->j_state_lock);
2946 if (test_opt(sb, BARRIER))
2947 journal->j_flags |= JBD2_BARRIER;
2948 else
2949 journal->j_flags &= ~JBD2_BARRIER;
2950 if (test_opt(sb, DATA_ERR_ABORT))
2951 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2952 else
2953 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
2954 spin_unlock(&journal->j_state_lock);
2955 }
2956
2957 static journal_t *ext4_get_journal(struct super_block *sb,
2958 unsigned int journal_inum)
2959 {
2960 struct inode *journal_inode;
2961 journal_t *journal;
2962
2963 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2964
2965 /* First, test for the existence of a valid inode on disk. Bad
2966 * things happen if we iget() an unused inode, as the subsequent
2967 * iput() will try to delete it. */
2968
2969 journal_inode = ext4_iget(sb, journal_inum);
2970 if (IS_ERR(journal_inode)) {
2971 printk(KERN_ERR "EXT4-fs: no journal found.\n");
2972 return NULL;
2973 }
2974 if (!journal_inode->i_nlink) {
2975 make_bad_inode(journal_inode);
2976 iput(journal_inode);
2977 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
2978 return NULL;
2979 }
2980
2981 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
2982 journal_inode, journal_inode->i_size);
2983 if (!S_ISREG(journal_inode->i_mode)) {
2984 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
2985 iput(journal_inode);
2986 return NULL;
2987 }
2988
2989 journal = jbd2_journal_init_inode(journal_inode);
2990 if (!journal) {
2991 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
2992 iput(journal_inode);
2993 return NULL;
2994 }
2995 journal->j_private = sb;
2996 ext4_init_journal_params(sb, journal);
2997 return journal;
2998 }
2999
3000 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3001 dev_t j_dev)
3002 {
3003 struct buffer_head *bh;
3004 journal_t *journal;
3005 ext4_fsblk_t start;
3006 ext4_fsblk_t len;
3007 int hblock, blocksize;
3008 ext4_fsblk_t sb_block;
3009 unsigned long offset;
3010 struct ext4_super_block *es;
3011 struct block_device *bdev;
3012
3013 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3014
3015 bdev = ext4_blkdev_get(j_dev);
3016 if (bdev == NULL)
3017 return NULL;
3018
3019 if (bd_claim(bdev, sb)) {
3020 printk(KERN_ERR
3021 "EXT4-fs: failed to claim external journal device.\n");
3022 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3023 return NULL;
3024 }
3025
3026 blocksize = sb->s_blocksize;
3027 hblock = bdev_hardsect_size(bdev);
3028 if (blocksize < hblock) {
3029 printk(KERN_ERR
3030 "EXT4-fs: blocksize too small for journal device.\n");
3031 goto out_bdev;
3032 }
3033
3034 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3035 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3036 set_blocksize(bdev, blocksize);
3037 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3038 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
3039 "external journal\n");
3040 goto out_bdev;
3041 }
3042
3043 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3044 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3045 !(le32_to_cpu(es->s_feature_incompat) &
3046 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3047 printk(KERN_ERR "EXT4-fs: external journal has "
3048 "bad superblock\n");
3049 brelse(bh);
3050 goto out_bdev;
3051 }
3052
3053 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3054 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
3055 brelse(bh);
3056 goto out_bdev;
3057 }
3058
3059 len = ext4_blocks_count(es);
3060 start = sb_block + 1;
3061 brelse(bh); /* we're done with the superblock */
3062
3063 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3064 start, len, blocksize);
3065 if (!journal) {
3066 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
3067 goto out_bdev;
3068 }
3069 journal->j_private = sb;
3070 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3071 wait_on_buffer(journal->j_sb_buffer);
3072 if (!buffer_uptodate(journal->j_sb_buffer)) {
3073 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
3074 goto out_journal;
3075 }
3076 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3077 printk(KERN_ERR "EXT4-fs: External journal has more than one "
3078 "user (unsupported) - %d\n",
3079 be32_to_cpu(journal->j_superblock->s_nr_users));
3080 goto out_journal;
3081 }
3082 EXT4_SB(sb)->journal_bdev = bdev;
3083 ext4_init_journal_params(sb, journal);
3084 return journal;
3085 out_journal:
3086 jbd2_journal_destroy(journal);
3087 out_bdev:
3088 ext4_blkdev_put(bdev);
3089 return NULL;
3090 }
3091
3092 static int ext4_load_journal(struct super_block *sb,
3093 struct ext4_super_block *es,
3094 unsigned long journal_devnum)
3095 {
3096 journal_t *journal;
3097 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3098 dev_t journal_dev;
3099 int err = 0;
3100 int really_read_only;
3101
3102 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3103
3104 if (journal_devnum &&
3105 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3106 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
3107 "numbers have changed\n");
3108 journal_dev = new_decode_dev(journal_devnum);
3109 } else
3110 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3111
3112 really_read_only = bdev_read_only(sb->s_bdev);
3113
3114 /*
3115 * Are we loading a blank journal or performing recovery after a
3116 * crash? For recovery, we need to check in advance whether we
3117 * can get read-write access to the device.
3118 */
3119
3120 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3121 if (sb->s_flags & MS_RDONLY) {
3122 printk(KERN_INFO "EXT4-fs: INFO: recovery "
3123 "required on readonly filesystem.\n");
3124 if (really_read_only) {
3125 printk(KERN_ERR "EXT4-fs: write access "
3126 "unavailable, cannot proceed.\n");
3127 return -EROFS;
3128 }
3129 printk(KERN_INFO "EXT4-fs: write access will "
3130 "be enabled during recovery.\n");
3131 }
3132 }
3133
3134 if (journal_inum && journal_dev) {
3135 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
3136 "and inode journals!\n");
3137 return -EINVAL;
3138 }
3139
3140 if (journal_inum) {
3141 if (!(journal = ext4_get_journal(sb, journal_inum)))
3142 return -EINVAL;
3143 } else {
3144 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3145 return -EINVAL;
3146 }
3147
3148 if (journal->j_flags & JBD2_BARRIER)
3149 printk(KERN_INFO "EXT4-fs: barriers enabled\n");
3150 else
3151 printk(KERN_INFO "EXT4-fs: barriers disabled\n");
3152
3153 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3154 err = jbd2_journal_update_format(journal);
3155 if (err) {
3156 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
3157 jbd2_journal_destroy(journal);
3158 return err;
3159 }
3160 }
3161
3162 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3163 err = jbd2_journal_wipe(journal, !really_read_only);
3164 if (!err)
3165 err = jbd2_journal_load(journal);
3166
3167 if (err) {
3168 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
3169 jbd2_journal_destroy(journal);
3170 return err;
3171 }
3172
3173 EXT4_SB(sb)->s_journal = journal;
3174 ext4_clear_journal_err(sb, es);
3175
3176 if (journal_devnum &&
3177 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3178 es->s_journal_dev = cpu_to_le32(journal_devnum);
3179
3180 /* Make sure we flush the recovery flag to disk. */
3181 ext4_commit_super(sb, 1);
3182 }
3183
3184 return 0;
3185 }
3186
3187 static int ext4_commit_super(struct super_block *sb, int sync)
3188 {
3189 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3190 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3191 int error = 0;
3192
3193 if (!sbh)
3194 return error;
3195 if (buffer_write_io_error(sbh)) {
3196 /*
3197 * Oh, dear. A previous attempt to write the
3198 * superblock failed. This could happen because the
3199 * USB device was yanked out. Or it could happen to
3200 * be a transient write error and maybe the block will
3201 * be remapped. Nothing we can do but to retry the
3202 * write and hope for the best.
3203 */
3204 printk(KERN_ERR "EXT4-fs: previous I/O error to "
3205 "superblock detected for %s.\n", sb->s_id);
3206 clear_buffer_write_io_error(sbh);
3207 set_buffer_uptodate(sbh);
3208 }
3209 es->s_wtime = cpu_to_le32(get_seconds());
3210 es->s_kbytes_written =
3211 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3212 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3213 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3214 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3215 &EXT4_SB(sb)->s_freeblocks_counter));
3216 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3217 &EXT4_SB(sb)->s_freeinodes_counter));
3218 sb->s_dirt = 0;
3219 BUFFER_TRACE(sbh, "marking dirty");
3220 mark_buffer_dirty(sbh);
3221 if (sync) {
3222 error = sync_dirty_buffer(sbh);
3223 if (error)
3224 return error;
3225
3226 error = buffer_write_io_error(sbh);
3227 if (error) {
3228 printk(KERN_ERR "EXT4-fs: I/O error while writing "
3229 "superblock for %s.\n", sb->s_id);
3230 clear_buffer_write_io_error(sbh);
3231 set_buffer_uptodate(sbh);
3232 }
3233 }
3234 return error;
3235 }
3236
3237
3238 /*
3239 * Have we just finished recovery? If so, and if we are mounting (or
3240 * remounting) the filesystem readonly, then we will end up with a
3241 * consistent fs on disk. Record that fact.
3242 */
3243 static void ext4_mark_recovery_complete(struct super_block *sb,
3244 struct ext4_super_block *es)
3245 {
3246 journal_t *journal = EXT4_SB(sb)->s_journal;
3247
3248 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3249 BUG_ON(journal != NULL);
3250 return;
3251 }
3252 jbd2_journal_lock_updates(journal);
3253 if (jbd2_journal_flush(journal) < 0)
3254 goto out;
3255
3256 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3257 sb->s_flags & MS_RDONLY) {
3258 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3259 ext4_commit_super(sb, 1);
3260 }
3261
3262 out:
3263 jbd2_journal_unlock_updates(journal);
3264 }
3265
3266 /*
3267 * If we are mounting (or read-write remounting) a filesystem whose journal
3268 * has recorded an error from a previous lifetime, move that error to the
3269 * main filesystem now.
3270 */
3271 static void ext4_clear_journal_err(struct super_block *sb,
3272 struct ext4_super_block *es)
3273 {
3274 journal_t *journal;
3275 int j_errno;
3276 const char *errstr;
3277
3278 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3279
3280 journal = EXT4_SB(sb)->s_journal;
3281
3282 /*
3283 * Now check for any error status which may have been recorded in the
3284 * journal by a prior ext4_error() or ext4_abort()
3285 */
3286
3287 j_errno = jbd2_journal_errno(journal);
3288 if (j_errno) {
3289 char nbuf[16];
3290
3291 errstr = ext4_decode_error(sb, j_errno, nbuf);
3292 ext4_warning(sb, __func__, "Filesystem error recorded "
3293 "from previous mount: %s", errstr);
3294 ext4_warning(sb, __func__, "Marking fs in need of "
3295 "filesystem check.");
3296
3297 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3298 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3299 ext4_commit_super(sb, 1);
3300
3301 jbd2_journal_clear_err(journal);
3302 }
3303 }
3304
3305 /*
3306 * Force the running and committing transactions to commit,
3307 * and wait on the commit.
3308 */
3309 int ext4_force_commit(struct super_block *sb)
3310 {
3311 journal_t *journal;
3312 int ret = 0;
3313
3314 if (sb->s_flags & MS_RDONLY)
3315 return 0;
3316
3317 journal = EXT4_SB(sb)->s_journal;
3318 if (journal)
3319 ret = ext4_journal_force_commit(journal);
3320
3321 return ret;
3322 }
3323
3324 static void ext4_write_super(struct super_block *sb)
3325 {
3326 ext4_commit_super(sb, 1);
3327 }
3328
3329 static int ext4_sync_fs(struct super_block *sb, int wait)
3330 {
3331 int ret = 0;
3332 tid_t target;
3333
3334 trace_mark(ext4_sync_fs, "dev %s wait %d", sb->s_id, wait);
3335 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
3336 if (wait)
3337 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
3338 }
3339 return ret;
3340 }
3341
3342 /*
3343 * LVM calls this function before a (read-only) snapshot is created. This
3344 * gives us a chance to flush the journal completely and mark the fs clean.
3345 */
3346 static int ext4_freeze(struct super_block *sb)
3347 {
3348 int error = 0;
3349 journal_t *journal;
3350
3351 if (sb->s_flags & MS_RDONLY)
3352 return 0;
3353
3354 journal = EXT4_SB(sb)->s_journal;
3355
3356 /* Now we set up the journal barrier. */
3357 jbd2_journal_lock_updates(journal);
3358
3359 /*
3360 * Don't clear the needs_recovery flag if we failed to flush
3361 * the journal.
3362 */
3363 error = jbd2_journal_flush(journal);
3364 if (error < 0) {
3365 out:
3366 jbd2_journal_unlock_updates(journal);
3367 return error;
3368 }
3369
3370 /* Journal blocked and flushed, clear needs_recovery flag. */
3371 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3372 error = ext4_commit_super(sb, 1);
3373 if (error)
3374 goto out;
3375 return 0;
3376 }
3377
3378 /*
3379 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3380 * flag here, even though the filesystem is not technically dirty yet.
3381 */
3382 static int ext4_unfreeze(struct super_block *sb)
3383 {
3384 if (sb->s_flags & MS_RDONLY)
3385 return 0;
3386
3387 lock_super(sb);
3388 /* Reset the needs_recovery flag before the fs is unlocked. */
3389 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3390 ext4_commit_super(sb, 1);
3391 unlock_super(sb);
3392 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3393 return 0;
3394 }
3395
3396 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3397 {
3398 struct ext4_super_block *es;
3399 struct ext4_sb_info *sbi = EXT4_SB(sb);
3400 ext4_fsblk_t n_blocks_count = 0;
3401 unsigned long old_sb_flags;
3402 struct ext4_mount_options old_opts;
3403 ext4_group_t g;
3404 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3405 int err;
3406 #ifdef CONFIG_QUOTA
3407 int i;
3408 #endif
3409
3410 /* Store the original options */
3411 old_sb_flags = sb->s_flags;
3412 old_opts.s_mount_opt = sbi->s_mount_opt;
3413 old_opts.s_resuid = sbi->s_resuid;
3414 old_opts.s_resgid = sbi->s_resgid;
3415 old_opts.s_commit_interval = sbi->s_commit_interval;
3416 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3417 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3418 #ifdef CONFIG_QUOTA
3419 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3420 for (i = 0; i < MAXQUOTAS; i++)
3421 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3422 #endif
3423 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3424 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3425
3426 /*
3427 * Allow the "check" option to be passed as a remount option.
3428 */
3429 if (!parse_options(data, sb, NULL, &journal_ioprio,
3430 &n_blocks_count, 1)) {
3431 err = -EINVAL;
3432 goto restore_opts;
3433 }
3434
3435 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
3436 ext4_abort(sb, __func__, "Abort forced by user");
3437
3438 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3439 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3440
3441 es = sbi->s_es;
3442
3443 if (sbi->s_journal) {
3444 ext4_init_journal_params(sb, sbi->s_journal);
3445 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3446 }
3447
3448 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3449 n_blocks_count > ext4_blocks_count(es)) {
3450 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
3451 err = -EROFS;
3452 goto restore_opts;
3453 }
3454
3455 if (*flags & MS_RDONLY) {
3456 /*
3457 * First of all, the unconditional stuff we have to do
3458 * to disable replay of the journal when we next remount
3459 */
3460 sb->s_flags |= MS_RDONLY;
3461
3462 /*
3463 * OK, test if we are remounting a valid rw partition
3464 * readonly, and if so set the rdonly flag and then
3465 * mark the partition as valid again.
3466 */
3467 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3468 (sbi->s_mount_state & EXT4_VALID_FS))
3469 es->s_state = cpu_to_le16(sbi->s_mount_state);
3470
3471 if (sbi->s_journal)
3472 ext4_mark_recovery_complete(sb, es);
3473 } else {
3474 int ret;
3475 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3476 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
3477 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3478 "remount RDWR because of unsupported "
3479 "optional features (%x).\n", sb->s_id,
3480 (le32_to_cpu(sbi->s_es->s_feature_ro_compat) &
3481 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3482 err = -EROFS;
3483 goto restore_opts;
3484 }
3485
3486 /*
3487 * Make sure the group descriptor checksums
3488 * are sane. If they aren't, refuse to
3489 * remount r/w.
3490 */
3491 for (g = 0; g < sbi->s_groups_count; g++) {
3492 struct ext4_group_desc *gdp =
3493 ext4_get_group_desc(sb, g, NULL);
3494
3495 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3496 printk(KERN_ERR
3497 "EXT4-fs: ext4_remount: "
3498 "Checksum for group %u failed (%u!=%u)\n",
3499 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3500 le16_to_cpu(gdp->bg_checksum));
3501 err = -EINVAL;
3502 goto restore_opts;
3503 }
3504 }
3505
3506 /*
3507 * If we have an unprocessed orphan list hanging
3508 * around from a previously readonly bdev mount,
3509 * require a full umount/remount for now.
3510 */
3511 if (es->s_last_orphan) {
3512 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3513 "remount RDWR because of unprocessed "
3514 "orphan inode list. Please "
3515 "umount/remount instead.\n",
3516 sb->s_id);
3517 err = -EINVAL;
3518 goto restore_opts;
3519 }
3520
3521 /*
3522 * Mounting a RDONLY partition read-write, so reread
3523 * and store the current valid flag. (It may have
3524 * been changed by e2fsck since we originally mounted
3525 * the partition.)
3526 */
3527 if (sbi->s_journal)
3528 ext4_clear_journal_err(sb, es);
3529 sbi->s_mount_state = le16_to_cpu(es->s_state);
3530 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3531 goto restore_opts;
3532 if (!ext4_setup_super(sb, es, 0))
3533 sb->s_flags &= ~MS_RDONLY;
3534 }
3535 }
3536 ext4_setup_system_zone(sb);
3537 if (sbi->s_journal == NULL)
3538 ext4_commit_super(sb, 1);
3539
3540 #ifdef CONFIG_QUOTA
3541 /* Release old quota file names */
3542 for (i = 0; i < MAXQUOTAS; i++)
3543 if (old_opts.s_qf_names[i] &&
3544 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3545 kfree(old_opts.s_qf_names[i]);
3546 #endif
3547 return 0;
3548 restore_opts:
3549 sb->s_flags = old_sb_flags;
3550 sbi->s_mount_opt = old_opts.s_mount_opt;
3551 sbi->s_resuid = old_opts.s_resuid;
3552 sbi->s_resgid = old_opts.s_resgid;
3553 sbi->s_commit_interval = old_opts.s_commit_interval;
3554 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3555 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3556 #ifdef CONFIG_QUOTA
3557 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3558 for (i = 0; i < MAXQUOTAS; i++) {
3559 if (sbi->s_qf_names[i] &&
3560 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3561 kfree(sbi->s_qf_names[i]);
3562 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3563 }
3564 #endif
3565 return err;
3566 }
3567
3568 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3569 {
3570 struct super_block *sb = dentry->d_sb;
3571 struct ext4_sb_info *sbi = EXT4_SB(sb);
3572 struct ext4_super_block *es = sbi->s_es;
3573 u64 fsid;
3574
3575 if (test_opt(sb, MINIX_DF)) {
3576 sbi->s_overhead_last = 0;
3577 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3578 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3579 ext4_fsblk_t overhead = 0;
3580
3581 /*
3582 * Compute the overhead (FS structures). This is constant
3583 * for a given filesystem unless the number of block groups
3584 * changes so we cache the previous value until it does.
3585 */
3586
3587 /*
3588 * All of the blocks before first_data_block are
3589 * overhead
3590 */
3591 overhead = le32_to_cpu(es->s_first_data_block);
3592
3593 /*
3594 * Add the overhead attributed to the superblock and
3595 * block group descriptors. If the sparse superblocks
3596 * feature is turned on, then not all groups have this.
3597 */
3598 for (i = 0; i < ngroups; i++) {
3599 overhead += ext4_bg_has_super(sb, i) +
3600 ext4_bg_num_gdb(sb, i);
3601 cond_resched();
3602 }
3603
3604 /*
3605 * Every block group has an inode bitmap, a block
3606 * bitmap, and an inode table.
3607 */
3608 overhead += ngroups * (2 + sbi->s_itb_per_group);
3609 sbi->s_overhead_last = overhead;
3610 smp_wmb();
3611 sbi->s_blocks_last = ext4_blocks_count(es);
3612 }
3613
3614 buf->f_type = EXT4_SUPER_MAGIC;
3615 buf->f_bsize = sb->s_blocksize;
3616 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3617 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3618 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3619 ext4_free_blocks_count_set(es, buf->f_bfree);
3620 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3621 if (buf->f_bfree < ext4_r_blocks_count(es))
3622 buf->f_bavail = 0;
3623 buf->f_files = le32_to_cpu(es->s_inodes_count);
3624 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3625 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
3626 buf->f_namelen = EXT4_NAME_LEN;
3627 fsid = le64_to_cpup((void *)es->s_uuid) ^
3628 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3629 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3630 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3631 return 0;
3632 }
3633
3634 /* Helper function for writing quotas on sync - we need to start transaction before quota file
3635 * is locked for write. Otherwise the are possible deadlocks:
3636 * Process 1 Process 2
3637 * ext4_create() quota_sync()
3638 * jbd2_journal_start() write_dquot()
3639 * vfs_dq_init() down(dqio_mutex)
3640 * down(dqio_mutex) jbd2_journal_start()
3641 *
3642 */
3643
3644 #ifdef CONFIG_QUOTA
3645
3646 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3647 {
3648 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3649 }
3650
3651 static int ext4_write_dquot(struct dquot *dquot)
3652 {
3653 int ret, err;
3654 handle_t *handle;
3655 struct inode *inode;
3656
3657 inode = dquot_to_inode(dquot);
3658 handle = ext4_journal_start(inode,
3659 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3660 if (IS_ERR(handle))
3661 return PTR_ERR(handle);
3662 ret = dquot_commit(dquot);
3663 err = ext4_journal_stop(handle);
3664 if (!ret)
3665 ret = err;
3666 return ret;
3667 }
3668
3669 static int ext4_acquire_dquot(struct dquot *dquot)
3670 {
3671 int ret, err;
3672 handle_t *handle;
3673
3674 handle = ext4_journal_start(dquot_to_inode(dquot),
3675 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3676 if (IS_ERR(handle))
3677 return PTR_ERR(handle);
3678 ret = dquot_acquire(dquot);
3679 err = ext4_journal_stop(handle);
3680 if (!ret)
3681 ret = err;
3682 return ret;
3683 }
3684
3685 static int ext4_release_dquot(struct dquot *dquot)
3686 {
3687 int ret, err;
3688 handle_t *handle;
3689
3690 handle = ext4_journal_start(dquot_to_inode(dquot),
3691 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3692 if (IS_ERR(handle)) {
3693 /* Release dquot anyway to avoid endless cycle in dqput() */
3694 dquot_release(dquot);
3695 return PTR_ERR(handle);
3696 }
3697 ret = dquot_release(dquot);
3698 err = ext4_journal_stop(handle);
3699 if (!ret)
3700 ret = err;
3701 return ret;
3702 }
3703
3704 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3705 {
3706 /* Are we journaling quotas? */
3707 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3708 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3709 dquot_mark_dquot_dirty(dquot);
3710 return ext4_write_dquot(dquot);
3711 } else {
3712 return dquot_mark_dquot_dirty(dquot);
3713 }
3714 }
3715
3716 static int ext4_write_info(struct super_block *sb, int type)
3717 {
3718 int ret, err;
3719 handle_t *handle;
3720
3721 /* Data block + inode block */
3722 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3723 if (IS_ERR(handle))
3724 return PTR_ERR(handle);
3725 ret = dquot_commit_info(sb, type);
3726 err = ext4_journal_stop(handle);
3727 if (!ret)
3728 ret = err;
3729 return ret;
3730 }
3731
3732 /*
3733 * Turn on quotas during mount time - we need to find
3734 * the quota file and such...
3735 */
3736 static int ext4_quota_on_mount(struct super_block *sb, int type)
3737 {
3738 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3739 EXT4_SB(sb)->s_jquota_fmt, type);
3740 }
3741
3742 /*
3743 * Standard function to be called on quota_on
3744 */
3745 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3746 char *name, int remount)
3747 {
3748 int err;
3749 struct path path;
3750
3751 if (!test_opt(sb, QUOTA))
3752 return -EINVAL;
3753 /* When remounting, no checks are needed and in fact, name is NULL */
3754 if (remount)
3755 return vfs_quota_on(sb, type, format_id, name, remount);
3756
3757 err = kern_path(name, LOOKUP_FOLLOW, &path);
3758 if (err)
3759 return err;
3760
3761 /* Quotafile not on the same filesystem? */
3762 if (path.mnt->mnt_sb != sb) {
3763 path_put(&path);
3764 return -EXDEV;
3765 }
3766 /* Journaling quota? */
3767 if (EXT4_SB(sb)->s_qf_names[type]) {
3768 /* Quotafile not in fs root? */
3769 if (path.dentry->d_parent != sb->s_root)
3770 printk(KERN_WARNING
3771 "EXT4-fs: Quota file not on filesystem root. "
3772 "Journaled quota will not work.\n");
3773 }
3774
3775 /*
3776 * When we journal data on quota file, we have to flush journal to see
3777 * all updates to the file when we bypass pagecache...
3778 */
3779 if (EXT4_SB(sb)->s_journal &&
3780 ext4_should_journal_data(path.dentry->d_inode)) {
3781 /*
3782 * We don't need to lock updates but journal_flush() could
3783 * otherwise be livelocked...
3784 */
3785 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3786 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3787 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3788 if (err) {
3789 path_put(&path);
3790 return err;
3791 }
3792 }
3793
3794 err = vfs_quota_on_path(sb, type, format_id, &path);
3795 path_put(&path);
3796 return err;
3797 }
3798
3799 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3800 * acquiring the locks... As quota files are never truncated and quota code
3801 * itself serializes the operations (and noone else should touch the files)
3802 * we don't have to be afraid of races */
3803 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3804 size_t len, loff_t off)
3805 {
3806 struct inode *inode = sb_dqopt(sb)->files[type];
3807 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3808 int err = 0;
3809 int offset = off & (sb->s_blocksize - 1);
3810 int tocopy;
3811 size_t toread;
3812 struct buffer_head *bh;
3813 loff_t i_size = i_size_read(inode);
3814
3815 if (off > i_size)
3816 return 0;
3817 if (off+len > i_size)
3818 len = i_size-off;
3819 toread = len;
3820 while (toread > 0) {
3821 tocopy = sb->s_blocksize - offset < toread ?
3822 sb->s_blocksize - offset : toread;
3823 bh = ext4_bread(NULL, inode, blk, 0, &err);
3824 if (err)
3825 return err;
3826 if (!bh) /* A hole? */
3827 memset(data, 0, tocopy);
3828 else
3829 memcpy(data, bh->b_data+offset, tocopy);
3830 brelse(bh);
3831 offset = 0;
3832 toread -= tocopy;
3833 data += tocopy;
3834 blk++;
3835 }
3836 return len;
3837 }
3838
3839 /* Write to quotafile (we know the transaction is already started and has
3840 * enough credits) */
3841 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3842 const char *data, size_t len, loff_t off)
3843 {
3844 struct inode *inode = sb_dqopt(sb)->files[type];
3845 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3846 int err = 0;
3847 int offset = off & (sb->s_blocksize - 1);
3848 int tocopy;
3849 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3850 size_t towrite = len;
3851 struct buffer_head *bh;
3852 handle_t *handle = journal_current_handle();
3853
3854 if (EXT4_SB(sb)->s_journal && !handle) {
3855 printk(KERN_WARNING "EXT4-fs: Quota write (off=%llu, len=%llu)"
3856 " cancelled because transaction is not started.\n",
3857 (unsigned long long)off, (unsigned long long)len);
3858 return -EIO;
3859 }
3860 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3861 while (towrite > 0) {
3862 tocopy = sb->s_blocksize - offset < towrite ?
3863 sb->s_blocksize - offset : towrite;
3864 bh = ext4_bread(handle, inode, blk, 1, &err);
3865 if (!bh)
3866 goto out;
3867 if (journal_quota) {
3868 err = ext4_journal_get_write_access(handle, bh);
3869 if (err) {
3870 brelse(bh);
3871 goto out;
3872 }
3873 }
3874 lock_buffer(bh);
3875 memcpy(bh->b_data+offset, data, tocopy);
3876 flush_dcache_page(bh->b_page);
3877 unlock_buffer(bh);
3878 if (journal_quota)
3879 err = ext4_handle_dirty_metadata(handle, NULL, bh);
3880 else {
3881 /* Always do at least ordered writes for quotas */
3882 err = ext4_jbd2_file_inode(handle, inode);
3883 mark_buffer_dirty(bh);
3884 }
3885 brelse(bh);
3886 if (err)
3887 goto out;
3888 offset = 0;
3889 towrite -= tocopy;
3890 data += tocopy;
3891 blk++;
3892 }
3893 out:
3894 if (len == towrite) {
3895 mutex_unlock(&inode->i_mutex);
3896 return err;
3897 }
3898 if (inode->i_size < off+len-towrite) {
3899 i_size_write(inode, off+len-towrite);
3900 EXT4_I(inode)->i_disksize = inode->i_size;
3901 }
3902 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3903 ext4_mark_inode_dirty(handle, inode);
3904 mutex_unlock(&inode->i_mutex);
3905 return len - towrite;
3906 }
3907
3908 #endif
3909
3910 static int ext4_get_sb(struct file_system_type *fs_type,
3911 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3912 {
3913 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3914 }
3915
3916 static struct file_system_type ext4_fs_type = {
3917 .owner = THIS_MODULE,
3918 .name = "ext4",
3919 .get_sb = ext4_get_sb,
3920 .kill_sb = kill_block_super,
3921 .fs_flags = FS_REQUIRES_DEV,
3922 };
3923
3924 #ifdef CONFIG_EXT4DEV_COMPAT
3925 static int ext4dev_get_sb(struct file_system_type *fs_type,
3926 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3927 {
3928 printk(KERN_WARNING "EXT4-fs: Update your userspace programs "
3929 "to mount using ext4\n");
3930 printk(KERN_WARNING "EXT4-fs: ext4dev backwards compatibility "
3931 "will go away by 2.6.31\n");
3932 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3933 }
3934
3935 static struct file_system_type ext4dev_fs_type = {
3936 .owner = THIS_MODULE,
3937 .name = "ext4dev",
3938 .get_sb = ext4dev_get_sb,
3939 .kill_sb = kill_block_super,
3940 .fs_flags = FS_REQUIRES_DEV,
3941 };
3942 MODULE_ALIAS("ext4dev");
3943 #endif
3944
3945 static int __init init_ext4_fs(void)
3946 {
3947 int err;
3948
3949 err = init_ext4_system_zone();
3950 if (err)
3951 return err;
3952 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
3953 if (!ext4_kset)
3954 goto out4;
3955 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
3956 err = init_ext4_mballoc();
3957 if (err)
3958 goto out3;
3959
3960 err = init_ext4_xattr();
3961 if (err)
3962 goto out2;
3963 err = init_inodecache();
3964 if (err)
3965 goto out1;
3966 err = register_filesystem(&ext4_fs_type);
3967 if (err)
3968 goto out;
3969 #ifdef CONFIG_EXT4DEV_COMPAT
3970 err = register_filesystem(&ext4dev_fs_type);
3971 if (err) {
3972 unregister_filesystem(&ext4_fs_type);
3973 goto out;
3974 }
3975 #endif
3976 return 0;
3977 out:
3978 destroy_inodecache();
3979 out1:
3980 exit_ext4_xattr();
3981 out2:
3982 exit_ext4_mballoc();
3983 out3:
3984 remove_proc_entry("fs/ext4", NULL);
3985 kset_unregister(ext4_kset);
3986 out4:
3987 exit_ext4_system_zone();
3988 return err;
3989 }
3990
3991 static void __exit exit_ext4_fs(void)
3992 {
3993 unregister_filesystem(&ext4_fs_type);
3994 #ifdef CONFIG_EXT4DEV_COMPAT
3995 unregister_filesystem(&ext4dev_fs_type);
3996 #endif
3997 destroy_inodecache();
3998 exit_ext4_xattr();
3999 exit_ext4_mballoc();
4000 remove_proc_entry("fs/ext4", NULL);
4001 kset_unregister(ext4_kset);
4002 exit_ext4_system_zone();
4003 }
4004
4005 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4006 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4007 MODULE_LICENSE("GPL");
4008 module_init(init_ext4_fs)
4009 module_exit(exit_ext4_fs)
Linux 2.6 ibm-acpi/thinkpad-acpi driver git tree