?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
futex: Fix wakeup race by setting TASK_INTERRUPTIBLE before queue_me()
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / fs / ext4 / super.c
bloba6b1ab7347287e120abd4c3fe4a8de79a284e40c
1 /*
2 * linux/fs/ext4/super.c
3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19 #include <linux/module.h>
20 #include <linux/string.h>
21 #include <linux/fs.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/smp_lock.h>
30 #include <linux/buffer_head.h>
31 #include <linux/exportfs.h>
32 #include <linux/vfs.h>
33 #include <linux/random.h>
34 #include <linux/mount.h>
35 #include <linux/namei.h>
36 #include <linux/quotaops.h>
37 #include <linux/seq_file.h>
38 #include <linux/proc_fs.h>
39 #include <linux/ctype.h>
40 #include <linux/log2.h>
41 #include <linux/crc16.h>
42 #include <asm/uaccess.h>
43
44 #include "ext4.h"
45 #include "ext4_jbd2.h"
46 #include "xattr.h"
47 #include "acl.h"
48 #include "mballoc.h"
49
50 #define CREATE_TRACE_POINTS
51 #include <trace/events/ext4.h>
52
53 static int default_mb_history_length = 1000;
54
55 module_param_named(default_mb_history_length, default_mb_history_length,
56 int, 0644);
57 MODULE_PARM_DESC(default_mb_history_length,
58 "Default number of entries saved for mb_history");
59
60 struct proc_dir_entry *ext4_proc_root;
61 static struct kset *ext4_kset;
62
63 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
64 unsigned long journal_devnum);
65 static int ext4_commit_super(struct super_block *sb, int sync);
66 static void ext4_mark_recovery_complete(struct super_block *sb,
67 struct ext4_super_block *es);
68 static void ext4_clear_journal_err(struct super_block *sb,
69 struct ext4_super_block *es);
70 static int ext4_sync_fs(struct super_block *sb, int wait);
71 static const char *ext4_decode_error(struct super_block *sb, int errno,
72 char nbuf[16]);
73 static int ext4_remount(struct super_block *sb, int *flags, char *data);
74 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
75 static int ext4_unfreeze(struct super_block *sb);
76 static void ext4_write_super(struct super_block *sb);
77 static int ext4_freeze(struct super_block *sb);
78
79
80 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
81 struct ext4_group_desc *bg)
82 {
83 return le32_to_cpu(bg->bg_block_bitmap_lo) |
84 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
85 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
86 }
87
88 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
89 struct ext4_group_desc *bg)
90 {
91 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
92 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
93 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
94 }
95
96 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
97 struct ext4_group_desc *bg)
98 {
99 return le32_to_cpu(bg->bg_inode_table_lo) |
100 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
101 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
102 }
103
104 __u32 ext4_free_blks_count(struct super_block *sb,
105 struct ext4_group_desc *bg)
106 {
107 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
108 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
109 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
110 }
111
112 __u32 ext4_free_inodes_count(struct super_block *sb,
113 struct ext4_group_desc *bg)
114 {
115 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
116 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
117 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
118 }
119
120 __u32 ext4_used_dirs_count(struct super_block *sb,
121 struct ext4_group_desc *bg)
122 {
123 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
124 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
125 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
126 }
127
128 __u32 ext4_itable_unused_count(struct super_block *sb,
129 struct ext4_group_desc *bg)
130 {
131 return le16_to_cpu(bg->bg_itable_unused_lo) |
132 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
133 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
134 }
135
136 void ext4_block_bitmap_set(struct super_block *sb,
137 struct ext4_group_desc *bg, ext4_fsblk_t blk)
138 {
139 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
140 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
141 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
142 }
143
144 void ext4_inode_bitmap_set(struct super_block *sb,
145 struct ext4_group_desc *bg, ext4_fsblk_t blk)
146 {
147 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
148 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
149 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
150 }
151
152 void ext4_inode_table_set(struct super_block *sb,
153 struct ext4_group_desc *bg, ext4_fsblk_t blk)
154 {
155 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
156 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
157 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
158 }
159
160 void ext4_free_blks_set(struct super_block *sb,
161 struct ext4_group_desc *bg, __u32 count)
162 {
163 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
164 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
165 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
166 }
167
168 void ext4_free_inodes_set(struct super_block *sb,
169 struct ext4_group_desc *bg, __u32 count)
170 {
171 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
172 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
173 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
174 }
175
176 void ext4_used_dirs_set(struct super_block *sb,
177 struct ext4_group_desc *bg, __u32 count)
178 {
179 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
180 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
181 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
182 }
183
184 void ext4_itable_unused_set(struct super_block *sb,
185 struct ext4_group_desc *bg, __u32 count)
186 {
187 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
188 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
189 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
190 }
191
192 /*
193 * Wrappers for jbd2_journal_start/end.
194 *
195 * The only special thing we need to do here is to make sure that all
196 * journal_end calls result in the superblock being marked dirty, so
197 * that sync() will call the filesystem's write_super callback if
198 * appropriate.
199 */
200 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
201 {
202 journal_t *journal;
203
204 if (sb->s_flags & MS_RDONLY)
205 return ERR_PTR(-EROFS);
206
207 /* Special case here: if the journal has aborted behind our
208 * backs (eg. EIO in the commit thread), then we still need to
209 * take the FS itself readonly cleanly. */
210 journal = EXT4_SB(sb)->s_journal;
211 if (journal) {
212 if (is_journal_aborted(journal)) {
213 ext4_abort(sb, __func__, "Detected aborted journal");
214 return ERR_PTR(-EROFS);
215 }
216 return jbd2_journal_start(journal, nblocks);
217 }
218 /*
219 * We're not journaling, return the appropriate indication.
220 */
221 current->journal_info = EXT4_NOJOURNAL_HANDLE;
222 return current->journal_info;
223 }
224
225 /*
226 * The only special thing we need to do here is to make sure that all
227 * jbd2_journal_stop calls result in the superblock being marked dirty, so
228 * that sync() will call the filesystem's write_super callback if
229 * appropriate.
230 */
231 int __ext4_journal_stop(const char *where, handle_t *handle)
232 {
233 struct super_block *sb;
234 int err;
235 int rc;
236
237 if (!ext4_handle_valid(handle)) {
238 /*
239 * Do this here since we don't call jbd2_journal_stop() in
240 * no-journal mode.
241 */
242 current->journal_info = NULL;
243 return 0;
244 }
245 sb = handle->h_transaction->t_journal->j_private;
246 err = handle->h_err;
247 rc = jbd2_journal_stop(handle);
248
249 if (!err)
250 err = rc;
251 if (err)
252 __ext4_std_error(sb, where, err);
253 return err;
254 }
255
256 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
257 struct buffer_head *bh, handle_t *handle, int err)
258 {
259 char nbuf[16];
260 const char *errstr = ext4_decode_error(NULL, err, nbuf);
261
262 BUG_ON(!ext4_handle_valid(handle));
263
264 if (bh)
265 BUFFER_TRACE(bh, "abort");
266
267 if (!handle->h_err)
268 handle->h_err = err;
269
270 if (is_handle_aborted(handle))
271 return;
272
273 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
274 caller, errstr, err_fn);
275
276 jbd2_journal_abort_handle(handle);
277 }
278
279 /* Deal with the reporting of failure conditions on a filesystem such as
280 * inconsistencies detected or read IO failures.
281 *
282 * On ext2, we can store the error state of the filesystem in the
283 * superblock. That is not possible on ext4, because we may have other
284 * write ordering constraints on the superblock which prevent us from
285 * writing it out straight away; and given that the journal is about to
286 * be aborted, we can't rely on the current, or future, transactions to
287 * write out the superblock safely.
288 *
289 * We'll just use the jbd2_journal_abort() error code to record an error in
290 * the journal instead. On recovery, the journal will compain about
291 * that error until we've noted it down and cleared it.
292 */
293
294 static void ext4_handle_error(struct super_block *sb)
295 {
296 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
297
298 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
299 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
300
301 if (sb->s_flags & MS_RDONLY)
302 return;
303
304 if (!test_opt(sb, ERRORS_CONT)) {
305 journal_t *journal = EXT4_SB(sb)->s_journal;
306
307 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
308 if (journal)
309 jbd2_journal_abort(journal, -EIO);
310 }
311 if (test_opt(sb, ERRORS_RO)) {
312 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
313 sb->s_flags |= MS_RDONLY;
314 }
315 ext4_commit_super(sb, 1);
316 if (test_opt(sb, ERRORS_PANIC))
317 panic("EXT4-fs (device %s): panic forced after error\n",
318 sb->s_id);
319 }
320
321 void ext4_error(struct super_block *sb, const char *function,
322 const char *fmt, ...)
323 {
324 va_list args;
325
326 va_start(args, fmt);
327 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
328 vprintk(fmt, args);
329 printk("\n");
330 va_end(args);
331
332 ext4_handle_error(sb);
333 }
334
335 static const char *ext4_decode_error(struct super_block *sb, int errno,
336 char nbuf[16])
337 {
338 char *errstr = NULL;
339
340 switch (errno) {
341 case -EIO:
342 errstr = "IO failure";
343 break;
344 case -ENOMEM:
345 errstr = "Out of memory";
346 break;
347 case -EROFS:
348 if (!sb || (EXT4_SB(sb)->s_journal &&
349 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
350 errstr = "Journal has aborted";
351 else
352 errstr = "Readonly filesystem";
353 break;
354 default:
355 /* If the caller passed in an extra buffer for unknown
356 * errors, textualise them now. Else we just return
357 * NULL. */
358 if (nbuf) {
359 /* Check for truncated error codes... */
360 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
361 errstr = nbuf;
362 }
363 break;
364 }
365
366 return errstr;
367 }
368
369 /* __ext4_std_error decodes expected errors from journaling functions
370 * automatically and invokes the appropriate error response. */
371
372 void __ext4_std_error(struct super_block *sb, const char *function, int errno)
373 {
374 char nbuf[16];
375 const char *errstr;
376
377 /* Special case: if the error is EROFS, and we're not already
378 * inside a transaction, then there's really no point in logging
379 * an error. */
380 if (errno == -EROFS && journal_current_handle() == NULL &&
381 (sb->s_flags & MS_RDONLY))
382 return;
383
384 errstr = ext4_decode_error(sb, errno, nbuf);
385 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
386 sb->s_id, function, errstr);
387
388 ext4_handle_error(sb);
389 }
390
391 /*
392 * ext4_abort is a much stronger failure handler than ext4_error. The
393 * abort function may be used to deal with unrecoverable failures such
394 * as journal IO errors or ENOMEM at a critical moment in log management.
395 *
396 * We unconditionally force the filesystem into an ABORT|READONLY state,
397 * unless the error response on the fs has been set to panic in which
398 * case we take the easy way out and panic immediately.
399 */
400
401 void ext4_abort(struct super_block *sb, const char *function,
402 const char *fmt, ...)
403 {
404 va_list args;
405
406 va_start(args, fmt);
407 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
408 vprintk(fmt, args);
409 printk("\n");
410 va_end(args);
411
412 if (test_opt(sb, ERRORS_PANIC))
413 panic("EXT4-fs panic from previous error\n");
414
415 if (sb->s_flags & MS_RDONLY)
416 return;
417
418 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
419 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
420 sb->s_flags |= MS_RDONLY;
421 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
422 if (EXT4_SB(sb)->s_journal)
423 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
424 }
425
426 void ext4_msg (struct super_block * sb, const char *prefix,
427 const char *fmt, ...)
428 {
429 va_list args;
430
431 va_start(args, fmt);
432 printk("%sEXT4-fs (%s): ", prefix, sb->s_id);
433 vprintk(fmt, args);
434 printk("\n");
435 va_end(args);
436 }
437
438 void ext4_warning(struct super_block *sb, const char *function,
439 const char *fmt, ...)
440 {
441 va_list args;
442
443 va_start(args, fmt);
444 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
445 sb->s_id, function);
446 vprintk(fmt, args);
447 printk("\n");
448 va_end(args);
449 }
450
451 void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp,
452 const char *function, const char *fmt, ...)
453 __releases(bitlock)
454 __acquires(bitlock)
455 {
456 va_list args;
457 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
458
459 va_start(args, fmt);
460 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
461 vprintk(fmt, args);
462 printk("\n");
463 va_end(args);
464
465 if (test_opt(sb, ERRORS_CONT)) {
466 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
467 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
468 ext4_commit_super(sb, 0);
469 return;
470 }
471 ext4_unlock_group(sb, grp);
472 ext4_handle_error(sb);
473 /*
474 * We only get here in the ERRORS_RO case; relocking the group
475 * may be dangerous, but nothing bad will happen since the
476 * filesystem will have already been marked read/only and the
477 * journal has been aborted. We return 1 as a hint to callers
478 * who might what to use the return value from
479 * ext4_grp_locked_error() to distinguish beween the
480 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
481 * aggressively from the ext4 function in question, with a
482 * more appropriate error code.
483 */
484 ext4_lock_group(sb, grp);
485 return;
486 }
487
488 void ext4_update_dynamic_rev(struct super_block *sb)
489 {
490 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
491
492 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
493 return;
494
495 ext4_warning(sb, __func__,
496 "updating to rev %d because of new feature flag, "
497 "running e2fsck is recommended",
498 EXT4_DYNAMIC_REV);
499
500 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
501 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
502 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
503 /* leave es->s_feature_*compat flags alone */
504 /* es->s_uuid will be set by e2fsck if empty */
505
506 /*
507 * The rest of the superblock fields should be zero, and if not it
508 * means they are likely already in use, so leave them alone. We
509 * can leave it up to e2fsck to clean up any inconsistencies there.
510 */
511 }
512
513 /*
514 * Open the external journal device
515 */
516 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
517 {
518 struct block_device *bdev;
519 char b[BDEVNAME_SIZE];
520
521 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
522 if (IS_ERR(bdev))
523 goto fail;
524 return bdev;
525
526 fail:
527 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
528 __bdevname(dev, b), PTR_ERR(bdev));
529 return NULL;
530 }
531
532 /*
533 * Release the journal device
534 */
535 static int ext4_blkdev_put(struct block_device *bdev)
536 {
537 bd_release(bdev);
538 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
539 }
540
541 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
542 {
543 struct block_device *bdev;
544 int ret = -ENODEV;
545
546 bdev = sbi->journal_bdev;
547 if (bdev) {
548 ret = ext4_blkdev_put(bdev);
549 sbi->journal_bdev = NULL;
550 }
551 return ret;
552 }
553
554 static inline struct inode *orphan_list_entry(struct list_head *l)
555 {
556 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
557 }
558
559 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
560 {
561 struct list_head *l;
562
563 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
564 le32_to_cpu(sbi->s_es->s_last_orphan));
565
566 printk(KERN_ERR "sb_info orphan list:\n");
567 list_for_each(l, &sbi->s_orphan) {
568 struct inode *inode = orphan_list_entry(l);
569 printk(KERN_ERR " "
570 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
571 inode->i_sb->s_id, inode->i_ino, inode,
572 inode->i_mode, inode->i_nlink,
573 NEXT_ORPHAN(inode));
574 }
575 }
576
577 static void ext4_put_super(struct super_block *sb)
578 {
579 struct ext4_sb_info *sbi = EXT4_SB(sb);
580 struct ext4_super_block *es = sbi->s_es;
581 int i, err;
582
583 lock_super(sb);
584 lock_kernel();
585 if (sb->s_dirt)
586 ext4_commit_super(sb, 1);
587
588 ext4_release_system_zone(sb);
589 ext4_mb_release(sb);
590 ext4_ext_release(sb);
591 ext4_xattr_put_super(sb);
592 if (sbi->s_journal) {
593 err = jbd2_journal_destroy(sbi->s_journal);
594 sbi->s_journal = NULL;
595 if (err < 0)
596 ext4_abort(sb, __func__,
597 "Couldn't clean up the journal");
598 }
599 if (!(sb->s_flags & MS_RDONLY)) {
600 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
601 es->s_state = cpu_to_le16(sbi->s_mount_state);
602 ext4_commit_super(sb, 1);
603 }
604 if (sbi->s_proc) {
605 remove_proc_entry(sb->s_id, ext4_proc_root);
606 }
607 kobject_del(&sbi->s_kobj);
608
609 for (i = 0; i < sbi->s_gdb_count; i++)
610 brelse(sbi->s_group_desc[i]);
611 kfree(sbi->s_group_desc);
612 if (is_vmalloc_addr(sbi->s_flex_groups))
613 vfree(sbi->s_flex_groups);
614 else
615 kfree(sbi->s_flex_groups);
616 percpu_counter_destroy(&sbi->s_freeblocks_counter);
617 percpu_counter_destroy(&sbi->s_freeinodes_counter);
618 percpu_counter_destroy(&sbi->s_dirs_counter);
619 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
620 brelse(sbi->s_sbh);
621 #ifdef CONFIG_QUOTA
622 for (i = 0; i < MAXQUOTAS; i++)
623 kfree(sbi->s_qf_names[i]);
624 #endif
625
626 /* Debugging code just in case the in-memory inode orphan list
627 * isn't empty. The on-disk one can be non-empty if we've
628 * detected an error and taken the fs readonly, but the
629 * in-memory list had better be clean by this point. */
630 if (!list_empty(&sbi->s_orphan))
631 dump_orphan_list(sb, sbi);
632 J_ASSERT(list_empty(&sbi->s_orphan));
633
634 invalidate_bdev(sb->s_bdev);
635 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
636 /*
637 * Invalidate the journal device's buffers. We don't want them
638 * floating about in memory - the physical journal device may
639 * hotswapped, and it breaks the `ro-after' testing code.
640 */
641 sync_blockdev(sbi->journal_bdev);
642 invalidate_bdev(sbi->journal_bdev);
643 ext4_blkdev_remove(sbi);
644 }
645 sb->s_fs_info = NULL;
646 /*
647 * Now that we are completely done shutting down the
648 * superblock, we need to actually destroy the kobject.
649 */
650 unlock_kernel();
651 unlock_super(sb);
652 kobject_put(&sbi->s_kobj);
653 wait_for_completion(&sbi->s_kobj_unregister);
654 kfree(sbi->s_blockgroup_lock);
655 kfree(sbi);
656 }
657
658 static struct kmem_cache *ext4_inode_cachep;
659
660 /*
661 * Called inside transaction, so use GFP_NOFS
662 */
663 static struct inode *ext4_alloc_inode(struct super_block *sb)
664 {
665 struct ext4_inode_info *ei;
666
667 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
668 if (!ei)
669 return NULL;
670
671 ei->vfs_inode.i_version = 1;
672 ei->vfs_inode.i_data.writeback_index = 0;
673 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
674 INIT_LIST_HEAD(&ei->i_prealloc_list);
675 spin_lock_init(&ei->i_prealloc_lock);
676 /*
677 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
678 * therefore it can be null here. Don't check it, just initialize
679 * jinode.
680 */
681 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
682 ei->i_reserved_data_blocks = 0;
683 ei->i_reserved_meta_blocks = 0;
684 ei->i_allocated_meta_blocks = 0;
685 ei->i_delalloc_reserved_flag = 0;
686 spin_lock_init(&(ei->i_block_reservation_lock));
687
688 return &ei->vfs_inode;
689 }
690
691 static void ext4_destroy_inode(struct inode *inode)
692 {
693 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
694 ext4_msg(inode->i_sb, KERN_ERR,
695 "Inode %lu (%p): orphan list check failed!",
696 inode->i_ino, EXT4_I(inode));
697 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
698 EXT4_I(inode), sizeof(struct ext4_inode_info),
699 true);
700 dump_stack();
701 }
702 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
703 }
704
705 static void init_once(void *foo)
706 {
707 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
708
709 INIT_LIST_HEAD(&ei->i_orphan);
710 #ifdef CONFIG_EXT4_FS_XATTR
711 init_rwsem(&ei->xattr_sem);
712 #endif
713 init_rwsem(&ei->i_data_sem);
714 inode_init_once(&ei->vfs_inode);
715 }
716
717 static int init_inodecache(void)
718 {
719 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
720 sizeof(struct ext4_inode_info),
721 0, (SLAB_RECLAIM_ACCOUNT|
722 SLAB_MEM_SPREAD),
723 init_once);
724 if (ext4_inode_cachep == NULL)
725 return -ENOMEM;
726 return 0;
727 }
728
729 static void destroy_inodecache(void)
730 {
731 kmem_cache_destroy(ext4_inode_cachep);
732 }
733
734 static void ext4_clear_inode(struct inode *inode)
735 {
736 ext4_discard_preallocations(inode);
737 if (EXT4_JOURNAL(inode))
738 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
739 &EXT4_I(inode)->jinode);
740 }
741
742 static inline void ext4_show_quota_options(struct seq_file *seq,
743 struct super_block *sb)
744 {
745 #if defined(CONFIG_QUOTA)
746 struct ext4_sb_info *sbi = EXT4_SB(sb);
747
748 if (sbi->s_jquota_fmt)
749 seq_printf(seq, ",jqfmt=%s",
750 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
751
752 if (sbi->s_qf_names[USRQUOTA])
753 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
754
755 if (sbi->s_qf_names[GRPQUOTA])
756 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
757
758 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
759 seq_puts(seq, ",usrquota");
760
761 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
762 seq_puts(seq, ",grpquota");
763 #endif
764 }
765
766 /*
767 * Show an option if
768 * - it's set to a non-default value OR
769 * - if the per-sb default is different from the global default
770 */
771 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
772 {
773 int def_errors;
774 unsigned long def_mount_opts;
775 struct super_block *sb = vfs->mnt_sb;
776 struct ext4_sb_info *sbi = EXT4_SB(sb);
777 struct ext4_super_block *es = sbi->s_es;
778
779 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
780 def_errors = le16_to_cpu(es->s_errors);
781
782 if (sbi->s_sb_block != 1)
783 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
784 if (test_opt(sb, MINIX_DF))
785 seq_puts(seq, ",minixdf");
786 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
787 seq_puts(seq, ",grpid");
788 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
789 seq_puts(seq, ",nogrpid");
790 if (sbi->s_resuid != EXT4_DEF_RESUID ||
791 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
792 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
793 }
794 if (sbi->s_resgid != EXT4_DEF_RESGID ||
795 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
796 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
797 }
798 if (test_opt(sb, ERRORS_RO)) {
799 if (def_errors == EXT4_ERRORS_PANIC ||
800 def_errors == EXT4_ERRORS_CONTINUE) {
801 seq_puts(seq, ",errors=remount-ro");
802 }
803 }
804 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
805 seq_puts(seq, ",errors=continue");
806 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
807 seq_puts(seq, ",errors=panic");
808 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
809 seq_puts(seq, ",nouid32");
810 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
811 seq_puts(seq, ",debug");
812 if (test_opt(sb, OLDALLOC))
813 seq_puts(seq, ",oldalloc");
814 #ifdef CONFIG_EXT4_FS_XATTR
815 if (test_opt(sb, XATTR_USER) &&
816 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
817 seq_puts(seq, ",user_xattr");
818 if (!test_opt(sb, XATTR_USER) &&
819 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
820 seq_puts(seq, ",nouser_xattr");
821 }
822 #endif
823 #ifdef CONFIG_EXT4_FS_POSIX_ACL
824 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
825 seq_puts(seq, ",acl");
826 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
827 seq_puts(seq, ",noacl");
828 #endif
829 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
830 seq_printf(seq, ",commit=%u",
831 (unsigned) (sbi->s_commit_interval / HZ));
832 }
833 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
834 seq_printf(seq, ",min_batch_time=%u",
835 (unsigned) sbi->s_min_batch_time);
836 }
837 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
838 seq_printf(seq, ",max_batch_time=%u",
839 (unsigned) sbi->s_min_batch_time);
840 }
841
842 /*
843 * We're changing the default of barrier mount option, so
844 * let's always display its mount state so it's clear what its
845 * status is.
846 */
847 seq_puts(seq, ",barrier=");
848 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
849 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
850 seq_puts(seq, ",journal_async_commit");
851 if (test_opt(sb, NOBH))
852 seq_puts(seq, ",nobh");
853 if (test_opt(sb, I_VERSION))
854 seq_puts(seq, ",i_version");
855 if (!test_opt(sb, DELALLOC))
856 seq_puts(seq, ",nodelalloc");
857
858
859 if (sbi->s_stripe)
860 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
861 /*
862 * journal mode get enabled in different ways
863 * So just print the value even if we didn't specify it
864 */
865 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
866 seq_puts(seq, ",data=journal");
867 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
868 seq_puts(seq, ",data=ordered");
869 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
870 seq_puts(seq, ",data=writeback");
871
872 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
873 seq_printf(seq, ",inode_readahead_blks=%u",
874 sbi->s_inode_readahead_blks);
875
876 if (test_opt(sb, DATA_ERR_ABORT))
877 seq_puts(seq, ",data_err=abort");
878
879 if (test_opt(sb, NO_AUTO_DA_ALLOC))
880 seq_puts(seq, ",noauto_da_alloc");
881
882 ext4_show_quota_options(seq, sb);
883
884 return 0;
885 }
886
887 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
888 u64 ino, u32 generation)
889 {
890 struct inode *inode;
891
892 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
893 return ERR_PTR(-ESTALE);
894 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
895 return ERR_PTR(-ESTALE);
896
897 /* iget isn't really right if the inode is currently unallocated!!
898 *
899 * ext4_read_inode will return a bad_inode if the inode had been
900 * deleted, so we should be safe.
901 *
902 * Currently we don't know the generation for parent directory, so
903 * a generation of 0 means "accept any"
904 */
905 inode = ext4_iget(sb, ino);
906 if (IS_ERR(inode))
907 return ERR_CAST(inode);
908 if (generation && inode->i_generation != generation) {
909 iput(inode);
910 return ERR_PTR(-ESTALE);
911 }
912
913 return inode;
914 }
915
916 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
917 int fh_len, int fh_type)
918 {
919 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
920 ext4_nfs_get_inode);
921 }
922
923 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
924 int fh_len, int fh_type)
925 {
926 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
927 ext4_nfs_get_inode);
928 }
929
930 /*
931 * Try to release metadata pages (indirect blocks, directories) which are
932 * mapped via the block device. Since these pages could have journal heads
933 * which would prevent try_to_free_buffers() from freeing them, we must use
934 * jbd2 layer's try_to_free_buffers() function to release them.
935 */
936 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
937 gfp_t wait)
938 {
939 journal_t *journal = EXT4_SB(sb)->s_journal;
940
941 WARN_ON(PageChecked(page));
942 if (!page_has_buffers(page))
943 return 0;
944 if (journal)
945 return jbd2_journal_try_to_free_buffers(journal, page,
946 wait & ~__GFP_WAIT);
947 return try_to_free_buffers(page);
948 }
949
950 #ifdef CONFIG_QUOTA
951 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
952 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
953
954 static int ext4_write_dquot(struct dquot *dquot);
955 static int ext4_acquire_dquot(struct dquot *dquot);
956 static int ext4_release_dquot(struct dquot *dquot);
957 static int ext4_mark_dquot_dirty(struct dquot *dquot);
958 static int ext4_write_info(struct super_block *sb, int type);
959 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
960 char *path, int remount);
961 static int ext4_quota_on_mount(struct super_block *sb, int type);
962 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
963 size_t len, loff_t off);
964 static ssize_t ext4_quota_write(struct super_block *sb, int type,
965 const char *data, size_t len, loff_t off);
966
967 static struct dquot_operations ext4_quota_operations = {
968 .initialize = dquot_initialize,
969 .drop = dquot_drop,
970 .alloc_space = dquot_alloc_space,
971 .reserve_space = dquot_reserve_space,
972 .claim_space = dquot_claim_space,
973 .release_rsv = dquot_release_reserved_space,
974 .get_reserved_space = ext4_get_reserved_space,
975 .alloc_inode = dquot_alloc_inode,
976 .free_space = dquot_free_space,
977 .free_inode = dquot_free_inode,
978 .transfer = dquot_transfer,
979 .write_dquot = ext4_write_dquot,
980 .acquire_dquot = ext4_acquire_dquot,
981 .release_dquot = ext4_release_dquot,
982 .mark_dirty = ext4_mark_dquot_dirty,
983 .write_info = ext4_write_info,
984 .alloc_dquot = dquot_alloc,
985 .destroy_dquot = dquot_destroy,
986 };
987
988 static struct quotactl_ops ext4_qctl_operations = {
989 .quota_on = ext4_quota_on,
990 .quota_off = vfs_quota_off,
991 .quota_sync = vfs_quota_sync,
992 .get_info = vfs_get_dqinfo,
993 .set_info = vfs_set_dqinfo,
994 .get_dqblk = vfs_get_dqblk,
995 .set_dqblk = vfs_set_dqblk
996 };
997 #endif
998
999 static const struct super_operations ext4_sops = {
1000 .alloc_inode = ext4_alloc_inode,
1001 .destroy_inode = ext4_destroy_inode,
1002 .write_inode = ext4_write_inode,
1003 .dirty_inode = ext4_dirty_inode,
1004 .delete_inode = ext4_delete_inode,
1005 .put_super = ext4_put_super,
1006 .sync_fs = ext4_sync_fs,
1007 .freeze_fs = ext4_freeze,
1008 .unfreeze_fs = ext4_unfreeze,
1009 .statfs = ext4_statfs,
1010 .remount_fs = ext4_remount,
1011 .clear_inode = ext4_clear_inode,
1012 .show_options = ext4_show_options,
1013 #ifdef CONFIG_QUOTA
1014 .quota_read = ext4_quota_read,
1015 .quota_write = ext4_quota_write,
1016 #endif
1017 .bdev_try_to_free_page = bdev_try_to_free_page,
1018 };
1019
1020 static const struct super_operations ext4_nojournal_sops = {
1021 .alloc_inode = ext4_alloc_inode,
1022 .destroy_inode = ext4_destroy_inode,
1023 .write_inode = ext4_write_inode,
1024 .dirty_inode = ext4_dirty_inode,
1025 .delete_inode = ext4_delete_inode,
1026 .write_super = ext4_write_super,
1027 .put_super = ext4_put_super,
1028 .statfs = ext4_statfs,
1029 .remount_fs = ext4_remount,
1030 .clear_inode = ext4_clear_inode,
1031 .show_options = ext4_show_options,
1032 #ifdef CONFIG_QUOTA
1033 .quota_read = ext4_quota_read,
1034 .quota_write = ext4_quota_write,
1035 #endif
1036 .bdev_try_to_free_page = bdev_try_to_free_page,
1037 };
1038
1039 static const struct export_operations ext4_export_ops = {
1040 .fh_to_dentry = ext4_fh_to_dentry,
1041 .fh_to_parent = ext4_fh_to_parent,
1042 .get_parent = ext4_get_parent,
1043 };
1044
1045 enum {
1046 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1047 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1048 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1049 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1050 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1051 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1052 Opt_journal_update, Opt_journal_dev,
1053 Opt_journal_checksum, Opt_journal_async_commit,
1054 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1055 Opt_data_err_abort, Opt_data_err_ignore, Opt_mb_history_length,
1056 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1057 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
1058 Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err, Opt_resize,
1059 Opt_usrquota, Opt_grpquota, Opt_i_version,
1060 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1061 Opt_block_validity, Opt_noblock_validity,
1062 Opt_inode_readahead_blks, Opt_journal_ioprio
1063 };
1064
1065 static const match_table_t tokens = {
1066 {Opt_bsd_df, "bsddf"},
1067 {Opt_minix_df, "minixdf"},
1068 {Opt_grpid, "grpid"},
1069 {Opt_grpid, "bsdgroups"},
1070 {Opt_nogrpid, "nogrpid"},
1071 {Opt_nogrpid, "sysvgroups"},
1072 {Opt_resgid, "resgid=%u"},
1073 {Opt_resuid, "resuid=%u"},
1074 {Opt_sb, "sb=%u"},
1075 {Opt_err_cont, "errors=continue"},
1076 {Opt_err_panic, "errors=panic"},
1077 {Opt_err_ro, "errors=remount-ro"},
1078 {Opt_nouid32, "nouid32"},
1079 {Opt_debug, "debug"},
1080 {Opt_oldalloc, "oldalloc"},
1081 {Opt_orlov, "orlov"},
1082 {Opt_user_xattr, "user_xattr"},
1083 {Opt_nouser_xattr, "nouser_xattr"},
1084 {Opt_acl, "acl"},
1085 {Opt_noacl, "noacl"},
1086 {Opt_noload, "noload"},
1087 {Opt_nobh, "nobh"},
1088 {Opt_bh, "bh"},
1089 {Opt_commit, "commit=%u"},
1090 {Opt_min_batch_time, "min_batch_time=%u"},
1091 {Opt_max_batch_time, "max_batch_time=%u"},
1092 {Opt_journal_update, "journal=update"},
1093 {Opt_journal_dev, "journal_dev=%u"},
1094 {Opt_journal_checksum, "journal_checksum"},
1095 {Opt_journal_async_commit, "journal_async_commit"},
1096 {Opt_abort, "abort"},
1097 {Opt_data_journal, "data=journal"},
1098 {Opt_data_ordered, "data=ordered"},
1099 {Opt_data_writeback, "data=writeback"},
1100 {Opt_data_err_abort, "data_err=abort"},
1101 {Opt_data_err_ignore, "data_err=ignore"},
1102 {Opt_mb_history_length, "mb_history_length=%u"},
1103 {Opt_offusrjquota, "usrjquota="},
1104 {Opt_usrjquota, "usrjquota=%s"},
1105 {Opt_offgrpjquota, "grpjquota="},
1106 {Opt_grpjquota, "grpjquota=%s"},
1107 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1108 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1109 {Opt_grpquota, "grpquota"},
1110 {Opt_noquota, "noquota"},
1111 {Opt_quota, "quota"},
1112 {Opt_usrquota, "usrquota"},
1113 {Opt_barrier, "barrier=%u"},
1114 {Opt_barrier, "barrier"},
1115 {Opt_nobarrier, "nobarrier"},
1116 {Opt_i_version, "i_version"},
1117 {Opt_stripe, "stripe=%u"},
1118 {Opt_resize, "resize"},
1119 {Opt_delalloc, "delalloc"},
1120 {Opt_nodelalloc, "nodelalloc"},
1121 {Opt_block_validity, "block_validity"},
1122 {Opt_noblock_validity, "noblock_validity"},
1123 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1124 {Opt_journal_ioprio, "journal_ioprio=%u"},
1125 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1126 {Opt_auto_da_alloc, "auto_da_alloc"},
1127 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1128 {Opt_err, NULL},
1129 };
1130
1131 static ext4_fsblk_t get_sb_block(void **data)
1132 {
1133 ext4_fsblk_t sb_block;
1134 char *options = (char *) *data;
1135
1136 if (!options || strncmp(options, "sb=", 3) != 0)
1137 return 1; /* Default location */
1138
1139 options += 3;
1140 /* TODO: use simple_strtoll with >32bit ext4 */
1141 sb_block = simple_strtoul(options, &options, 0);
1142 if (*options && *options != ',') {
1143 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1144 (char *) *data);
1145 return 1;
1146 }
1147 if (*options == ',')
1148 options++;
1149 *data = (void *) options;
1150
1151 return sb_block;
1152 }
1153
1154 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1155
1156 static int parse_options(char *options, struct super_block *sb,
1157 unsigned long *journal_devnum,
1158 unsigned int *journal_ioprio,
1159 ext4_fsblk_t *n_blocks_count, int is_remount)
1160 {
1161 struct ext4_sb_info *sbi = EXT4_SB(sb);
1162 char *p;
1163 substring_t args[MAX_OPT_ARGS];
1164 int data_opt = 0;
1165 int option;
1166 #ifdef CONFIG_QUOTA
1167 int qtype, qfmt;
1168 char *qname;
1169 #endif
1170
1171 if (!options)
1172 return 1;
1173
1174 while ((p = strsep(&options, ",")) != NULL) {
1175 int token;
1176 if (!*p)
1177 continue;
1178
1179 token = match_token(p, tokens, args);
1180 switch (token) {
1181 case Opt_bsd_df:
1182 clear_opt(sbi->s_mount_opt, MINIX_DF);
1183 break;
1184 case Opt_minix_df:
1185 set_opt(sbi->s_mount_opt, MINIX_DF);
1186 break;
1187 case Opt_grpid:
1188 set_opt(sbi->s_mount_opt, GRPID);
1189 break;
1190 case Opt_nogrpid:
1191 clear_opt(sbi->s_mount_opt, GRPID);
1192 break;
1193 case Opt_resuid:
1194 if (match_int(&args[0], &option))
1195 return 0;
1196 sbi->s_resuid = option;
1197 break;
1198 case Opt_resgid:
1199 if (match_int(&args[0], &option))
1200 return 0;
1201 sbi->s_resgid = option;
1202 break;
1203 case Opt_sb:
1204 /* handled by get_sb_block() instead of here */
1205 /* *sb_block = match_int(&args[0]); */
1206 break;
1207 case Opt_err_panic:
1208 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1209 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1210 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1211 break;
1212 case Opt_err_ro:
1213 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1214 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1215 set_opt(sbi->s_mount_opt, ERRORS_RO);
1216 break;
1217 case Opt_err_cont:
1218 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1219 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1220 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1221 break;
1222 case Opt_nouid32:
1223 set_opt(sbi->s_mount_opt, NO_UID32);
1224 break;
1225 case Opt_debug:
1226 set_opt(sbi->s_mount_opt, DEBUG);
1227 break;
1228 case Opt_oldalloc:
1229 set_opt(sbi->s_mount_opt, OLDALLOC);
1230 break;
1231 case Opt_orlov:
1232 clear_opt(sbi->s_mount_opt, OLDALLOC);
1233 break;
1234 #ifdef CONFIG_EXT4_FS_XATTR
1235 case Opt_user_xattr:
1236 set_opt(sbi->s_mount_opt, XATTR_USER);
1237 break;
1238 case Opt_nouser_xattr:
1239 clear_opt(sbi->s_mount_opt, XATTR_USER);
1240 break;
1241 #else
1242 case Opt_user_xattr:
1243 case Opt_nouser_xattr:
1244 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1245 break;
1246 #endif
1247 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1248 case Opt_acl:
1249 set_opt(sbi->s_mount_opt, POSIX_ACL);
1250 break;
1251 case Opt_noacl:
1252 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1253 break;
1254 #else
1255 case Opt_acl:
1256 case Opt_noacl:
1257 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1258 break;
1259 #endif
1260 case Opt_journal_update:
1261 /* @@@ FIXME */
1262 /* Eventually we will want to be able to create
1263 a journal file here. For now, only allow the
1264 user to specify an existing inode to be the
1265 journal file. */
1266 if (is_remount) {
1267 ext4_msg(sb, KERN_ERR,
1268 "Cannot specify journal on remount");
1269 return 0;
1270 }
1271 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1272 break;
1273 case Opt_journal_dev:
1274 if (is_remount) {
1275 ext4_msg(sb, KERN_ERR,
1276 "Cannot specify journal on remount");
1277 return 0;
1278 }
1279 if (match_int(&args[0], &option))
1280 return 0;
1281 *journal_devnum = option;
1282 break;
1283 case Opt_journal_checksum:
1284 break; /* Kept for backwards compatibility */
1285 case Opt_journal_async_commit:
1286 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
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 ext4_msg(sb, KERN_ERR,
1329 "Cannot change data mode on remount");
1330 return 0;
1331 }
1332 } else {
1333 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1334 sbi->s_mount_opt |= data_opt;
1335 }
1336 break;
1337 case Opt_data_err_abort:
1338 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1339 break;
1340 case Opt_data_err_ignore:
1341 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1342 break;
1343 case Opt_mb_history_length:
1344 if (match_int(&args[0], &option))
1345 return 0;
1346 if (option < 0)
1347 return 0;
1348 sbi->s_mb_history_max = option;
1349 break;
1350 #ifdef CONFIG_QUOTA
1351 case Opt_usrjquota:
1352 qtype = USRQUOTA;
1353 goto set_qf_name;
1354 case Opt_grpjquota:
1355 qtype = GRPQUOTA;
1356 set_qf_name:
1357 if (sb_any_quota_loaded(sb) &&
1358 !sbi->s_qf_names[qtype]) {
1359 ext4_msg(sb, KERN_ERR,
1360 "Cannot change journaled "
1361 "quota options when quota turned on");
1362 return 0;
1363 }
1364 qname = match_strdup(&args[0]);
1365 if (!qname) {
1366 ext4_msg(sb, KERN_ERR,
1367 "Not enough memory for "
1368 "storing quotafile name");
1369 return 0;
1370 }
1371 if (sbi->s_qf_names[qtype] &&
1372 strcmp(sbi->s_qf_names[qtype], qname)) {
1373 ext4_msg(sb, KERN_ERR,
1374 "%s quota file already "
1375 "specified", QTYPE2NAME(qtype));
1376 kfree(qname);
1377 return 0;
1378 }
1379 sbi->s_qf_names[qtype] = qname;
1380 if (strchr(sbi->s_qf_names[qtype], '/')) {
1381 ext4_msg(sb, KERN_ERR,
1382 "quotafile must be on "
1383 "filesystem root");
1384 kfree(sbi->s_qf_names[qtype]);
1385 sbi->s_qf_names[qtype] = NULL;
1386 return 0;
1387 }
1388 set_opt(sbi->s_mount_opt, QUOTA);
1389 break;
1390 case Opt_offusrjquota:
1391 qtype = USRQUOTA;
1392 goto clear_qf_name;
1393 case Opt_offgrpjquota:
1394 qtype = GRPQUOTA;
1395 clear_qf_name:
1396 if (sb_any_quota_loaded(sb) &&
1397 sbi->s_qf_names[qtype]) {
1398 ext4_msg(sb, KERN_ERR, "Cannot change "
1399 "journaled quota options when "
1400 "quota turned on");
1401 return 0;
1402 }
1403 /*
1404 * The space will be released later when all options
1405 * are confirmed to be correct
1406 */
1407 sbi->s_qf_names[qtype] = NULL;
1408 break;
1409 case Opt_jqfmt_vfsold:
1410 qfmt = QFMT_VFS_OLD;
1411 goto set_qf_format;
1412 case Opt_jqfmt_vfsv0:
1413 qfmt = QFMT_VFS_V0;
1414 set_qf_format:
1415 if (sb_any_quota_loaded(sb) &&
1416 sbi->s_jquota_fmt != qfmt) {
1417 ext4_msg(sb, KERN_ERR, "Cannot change "
1418 "journaled quota options when "
1419 "quota turned on");
1420 return 0;
1421 }
1422 sbi->s_jquota_fmt = qfmt;
1423 break;
1424 case Opt_quota:
1425 case Opt_usrquota:
1426 set_opt(sbi->s_mount_opt, QUOTA);
1427 set_opt(sbi->s_mount_opt, USRQUOTA);
1428 break;
1429 case Opt_grpquota:
1430 set_opt(sbi->s_mount_opt, QUOTA);
1431 set_opt(sbi->s_mount_opt, GRPQUOTA);
1432 break;
1433 case Opt_noquota:
1434 if (sb_any_quota_loaded(sb)) {
1435 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1436 "options when quota turned on");
1437 return 0;
1438 }
1439 clear_opt(sbi->s_mount_opt, QUOTA);
1440 clear_opt(sbi->s_mount_opt, USRQUOTA);
1441 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1442 break;
1443 #else
1444 case Opt_quota:
1445 case Opt_usrquota:
1446 case Opt_grpquota:
1447 ext4_msg(sb, KERN_ERR,
1448 "quota options not supported");
1449 break;
1450 case Opt_usrjquota:
1451 case Opt_grpjquota:
1452 case Opt_offusrjquota:
1453 case Opt_offgrpjquota:
1454 case Opt_jqfmt_vfsold:
1455 case Opt_jqfmt_vfsv0:
1456 ext4_msg(sb, KERN_ERR,
1457 "journaled quota options not supported");
1458 break;
1459 case Opt_noquota:
1460 break;
1461 #endif
1462 case Opt_abort:
1463 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1464 break;
1465 case Opt_nobarrier:
1466 clear_opt(sbi->s_mount_opt, BARRIER);
1467 break;
1468 case Opt_barrier:
1469 if (match_int(&args[0], &option)) {
1470 set_opt(sbi->s_mount_opt, BARRIER);
1471 break;
1472 }
1473 if (option)
1474 set_opt(sbi->s_mount_opt, BARRIER);
1475 else
1476 clear_opt(sbi->s_mount_opt, BARRIER);
1477 break;
1478 case Opt_ignore:
1479 break;
1480 case Opt_resize:
1481 if (!is_remount) {
1482 ext4_msg(sb, KERN_ERR,
1483 "resize option only available "
1484 "for remount");
1485 return 0;
1486 }
1487 if (match_int(&args[0], &option) != 0)
1488 return 0;
1489 *n_blocks_count = option;
1490 break;
1491 case Opt_nobh:
1492 set_opt(sbi->s_mount_opt, NOBH);
1493 break;
1494 case Opt_bh:
1495 clear_opt(sbi->s_mount_opt, NOBH);
1496 break;
1497 case Opt_i_version:
1498 set_opt(sbi->s_mount_opt, I_VERSION);
1499 sb->s_flags |= MS_I_VERSION;
1500 break;
1501 case Opt_nodelalloc:
1502 clear_opt(sbi->s_mount_opt, DELALLOC);
1503 break;
1504 case Opt_stripe:
1505 if (match_int(&args[0], &option))
1506 return 0;
1507 if (option < 0)
1508 return 0;
1509 sbi->s_stripe = option;
1510 break;
1511 case Opt_delalloc:
1512 set_opt(sbi->s_mount_opt, DELALLOC);
1513 break;
1514 case Opt_block_validity:
1515 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1516 break;
1517 case Opt_noblock_validity:
1518 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1519 break;
1520 case Opt_inode_readahead_blks:
1521 if (match_int(&args[0], &option))
1522 return 0;
1523 if (option < 0 || option > (1 << 30))
1524 return 0;
1525 if (!is_power_of_2(option)) {
1526 ext4_msg(sb, KERN_ERR,
1527 "EXT4-fs: inode_readahead_blks"
1528 " must be a power of 2");
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 ext4_msg(sb, KERN_ERR,
1556 "Unrecognized mount option \"%s\" "
1557 "or missing value", 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 ext4_msg(sb, KERN_ERR, "old and new quota "
1576 "format mixing");
1577 return 0;
1578 }
1579
1580 if (!sbi->s_jquota_fmt) {
1581 ext4_msg(sb, KERN_ERR, "journaled quota format "
1582 "not specified");
1583 return 0;
1584 }
1585 } else {
1586 if (sbi->s_jquota_fmt) {
1587 ext4_msg(sb, KERN_ERR, "journaled quota format "
1588 "specified with no journaling "
1589 "enabled");
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 ext4_msg(sb, KERN_ERR, "revision level too high, "
1605 "forcing read-only mode");
1606 res = MS_RDONLY;
1607 }
1608 if (read_only)
1609 return res;
1610 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1611 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1612 "running e2fsck is recommended");
1613 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1614 ext4_msg(sb, KERN_WARNING,
1615 "warning: mounting fs with errors, "
1616 "running e2fsck is recommended");
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 ext4_msg(sb, KERN_WARNING,
1621 "warning: maximal mount count reached, "
1622 "running e2fsck is recommended");
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 ext4_msg(sb, KERN_WARNING,
1627 "warning: checktime reached, "
1628 "running e2fsck is recommended");
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=%04x]\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 ext4_msg(sb, KERN_INFO, "%s journal on %s",
1651 EXT4_SB(sb)->s_journal->j_inode ? "internal" :
1652 "external", EXT4_SB(sb)->s_journal->j_devname);
1653 } else {
1654 ext4_msg(sb, KERN_INFO, "no journal");
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 ext4_msg(sb, KERN_ERR, "not enough memory for "
1690 "%u flex groups", 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_add(ext4_free_inodes_count(sb, gdp),
1699 &sbi->s_flex_groups[flex_group].free_inodes);
1700 atomic_add(ext4_free_blks_count(sb, gdp),
1701 &sbi->s_flex_groups[flex_group].free_blocks);
1702 atomic_add(ext4_used_dirs_count(sb, gdp),
1703 &sbi->s_flex_groups[flex_group].used_dirs);
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 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1777 "Block bitmap for group %u not in group "
1778 "(block %llu)!", 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 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1784 "Inode bitmap for group %u not in group "
1785 "(block %llu)!", 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 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1792 "Inode table for group %u not in group "
1793 "(block %llu)!", i, inode_table);
1794 return 0;
1795 }
1796 ext4_lock_group(sb, i);
1797 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1798 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1799 "Checksum for group %u failed (%u!=%u)",
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 ext4_msg(sb, KERN_ERR, "write access "
1849 "unavailable, skipping orphan cleanup");
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 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1864 sb->s_flags &= ~MS_RDONLY;
1865 }
1866 #ifdef CONFIG_QUOTA
1867 /* Needed for iput() to work correctly and not trash data */
1868 sb->s_flags |= MS_ACTIVE;
1869 /* Turn on quotas so that they are updated correctly */
1870 for (i = 0; i < MAXQUOTAS; i++) {
1871 if (EXT4_SB(sb)->s_qf_names[i]) {
1872 int ret = ext4_quota_on_mount(sb, i);
1873 if (ret < 0)
1874 ext4_msg(sb, KERN_ERR,
1875 "Cannot turn on journaled "
1876 "quota: error %d", ret);
1877 }
1878 }
1879 #endif
1880
1881 while (es->s_last_orphan) {
1882 struct inode *inode;
1883
1884 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1885 if (IS_ERR(inode)) {
1886 es->s_last_orphan = 0;
1887 break;
1888 }
1889
1890 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1891 vfs_dq_init(inode);
1892 if (inode->i_nlink) {
1893 ext4_msg(sb, KERN_DEBUG,
1894 "%s: truncating inode %lu to %lld bytes",
1895 __func__, inode->i_ino, inode->i_size);
1896 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1897 inode->i_ino, inode->i_size);
1898 ext4_truncate(inode);
1899 nr_truncates++;
1900 } else {
1901 ext4_msg(sb, KERN_DEBUG,
1902 "%s: deleting unreferenced inode %lu",
1903 __func__, inode->i_ino);
1904 jbd_debug(2, "deleting unreferenced inode %lu\n",
1905 inode->i_ino);
1906 nr_orphans++;
1907 }
1908 iput(inode); /* The delete magic happens here! */
1909 }
1910
1911 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1912
1913 if (nr_orphans)
1914 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1915 PLURAL(nr_orphans));
1916 if (nr_truncates)
1917 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1918 PLURAL(nr_truncates));
1919 #ifdef CONFIG_QUOTA
1920 /* Turn quotas off */
1921 for (i = 0; i < MAXQUOTAS; i++) {
1922 if (sb_dqopt(sb)->files[i])
1923 vfs_quota_off(sb, i, 0);
1924 }
1925 #endif
1926 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1927 }
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_LBDAF 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 dense, block
1980 * mapped file such that the file's total number of 512-byte sectors,
1981 * including data and all indirect blocks, does not exceed (2^48 - 1).
1982 *
1983 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
1984 * number of 512-byte sectors of the file.
1985 */
1986
1987 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1988 /*
1989 * !has_huge_files or CONFIG_LBDAF not enabled implies that
1990 * the inode i_block field represents total file blocks in
1991 * 2^32 512-byte sectors == 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
2048 return (has_super + ext4_group_first_block_no(sb, bg));
2049 }
2050
2051 /**
2052 * ext4_get_stripe_size: Get the stripe size.
2053 * @sbi: In memory super block info
2054 *
2055 * If we have specified it via mount option, then
2056 * use the mount option value. If the value specified at mount time is
2057 * greater than the blocks per group use the super block value.
2058 * If the super block value is greater than blocks per group return 0.
2059 * Allocator needs it be less than blocks per group.
2060 *
2061 */
2062 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2063 {
2064 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2065 unsigned long stripe_width =
2066 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2067
2068 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2069 return sbi->s_stripe;
2070
2071 if (stripe_width <= sbi->s_blocks_per_group)
2072 return stripe_width;
2073
2074 if (stride <= sbi->s_blocks_per_group)
2075 return stride;
2076
2077 return 0;
2078 }
2079
2080 /* sysfs supprt */
2081
2082 struct ext4_attr {
2083 struct attribute attr;
2084 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2085 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2086 const char *, size_t);
2087 int offset;
2088 };
2089
2090 static int parse_strtoul(const char *buf,
2091 unsigned long max, unsigned long *value)
2092 {
2093 char *endp;
2094
2095 while (*buf && isspace(*buf))
2096 buf++;
2097 *value = simple_strtoul(buf, &endp, 0);
2098 while (*endp && isspace(*endp))
2099 endp++;
2100 if (*endp || *value > max)
2101 return -EINVAL;
2102
2103 return 0;
2104 }
2105
2106 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2107 struct ext4_sb_info *sbi,
2108 char *buf)
2109 {
2110 return snprintf(buf, PAGE_SIZE, "%llu\n",
2111 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2112 }
2113
2114 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2115 struct ext4_sb_info *sbi, char *buf)
2116 {
2117 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2118
2119 return snprintf(buf, PAGE_SIZE, "%lu\n",
2120 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2121 sbi->s_sectors_written_start) >> 1);
2122 }
2123
2124 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2125 struct ext4_sb_info *sbi, char *buf)
2126 {
2127 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2128
2129 return snprintf(buf, PAGE_SIZE, "%llu\n",
2130 sbi->s_kbytes_written +
2131 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2132 EXT4_SB(sb)->s_sectors_written_start) >> 1));
2133 }
2134
2135 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2136 struct ext4_sb_info *sbi,
2137 const char *buf, size_t count)
2138 {
2139 unsigned long t;
2140
2141 if (parse_strtoul(buf, 0x40000000, &t))
2142 return -EINVAL;
2143
2144 if (!is_power_of_2(t))
2145 return -EINVAL;
2146
2147 sbi->s_inode_readahead_blks = t;
2148 return count;
2149 }
2150
2151 static ssize_t sbi_ui_show(struct ext4_attr *a,
2152 struct ext4_sb_info *sbi, char *buf)
2153 {
2154 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2155
2156 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2157 }
2158
2159 static ssize_t sbi_ui_store(struct ext4_attr *a,
2160 struct ext4_sb_info *sbi,
2161 const char *buf, size_t count)
2162 {
2163 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2164 unsigned long t;
2165
2166 if (parse_strtoul(buf, 0xffffffff, &t))
2167 return -EINVAL;
2168 *ui = t;
2169 return count;
2170 }
2171
2172 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2173 static struct ext4_attr ext4_attr_##_name = { \
2174 .attr = {.name = __stringify(_name), .mode = _mode }, \
2175 .show = _show, \
2176 .store = _store, \
2177 .offset = offsetof(struct ext4_sb_info, _elname), \
2178 }
2179 #define EXT4_ATTR(name, mode, show, store) \
2180 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2181
2182 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2183 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2184 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2185 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2186 #define ATTR_LIST(name) &ext4_attr_##name.attr
2187
2188 EXT4_RO_ATTR(delayed_allocation_blocks);
2189 EXT4_RO_ATTR(session_write_kbytes);
2190 EXT4_RO_ATTR(lifetime_write_kbytes);
2191 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2192 inode_readahead_blks_store, s_inode_readahead_blks);
2193 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2194 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2195 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2196 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2197 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2198 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2199 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2200
2201 static struct attribute *ext4_attrs[] = {
2202 ATTR_LIST(delayed_allocation_blocks),
2203 ATTR_LIST(session_write_kbytes),
2204 ATTR_LIST(lifetime_write_kbytes),
2205 ATTR_LIST(inode_readahead_blks),
2206 ATTR_LIST(inode_goal),
2207 ATTR_LIST(mb_stats),
2208 ATTR_LIST(mb_max_to_scan),
2209 ATTR_LIST(mb_min_to_scan),
2210 ATTR_LIST(mb_order2_req),
2211 ATTR_LIST(mb_stream_req),
2212 ATTR_LIST(mb_group_prealloc),
2213 NULL,
2214 };
2215
2216 static ssize_t ext4_attr_show(struct kobject *kobj,
2217 struct attribute *attr, char *buf)
2218 {
2219 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2220 s_kobj);
2221 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2222
2223 return a->show ? a->show(a, sbi, buf) : 0;
2224 }
2225
2226 static ssize_t ext4_attr_store(struct kobject *kobj,
2227 struct attribute *attr,
2228 const char *buf, size_t len)
2229 {
2230 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2231 s_kobj);
2232 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2233
2234 return a->store ? a->store(a, sbi, buf, len) : 0;
2235 }
2236
2237 static void ext4_sb_release(struct kobject *kobj)
2238 {
2239 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2240 s_kobj);
2241 complete(&sbi->s_kobj_unregister);
2242 }
2243
2244
2245 static struct sysfs_ops ext4_attr_ops = {
2246 .show = ext4_attr_show,
2247 .store = ext4_attr_store,
2248 };
2249
2250 static struct kobj_type ext4_ktype = {
2251 .default_attrs = ext4_attrs,
2252 .sysfs_ops = &ext4_attr_ops,
2253 .release = ext4_sb_release,
2254 };
2255
2256 /*
2257 * Check whether this filesystem can be mounted based on
2258 * the features present and the RDONLY/RDWR mount requested.
2259 * Returns 1 if this filesystem can be mounted as requested,
2260 * 0 if it cannot be.
2261 */
2262 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2263 {
2264 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2265 ext4_msg(sb, KERN_ERR,
2266 "Couldn't mount because of "
2267 "unsupported optional features (%x)",
2268 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2269 ~EXT4_FEATURE_INCOMPAT_SUPP));
2270 return 0;
2271 }
2272
2273 if (readonly)
2274 return 1;
2275
2276 /* Check that feature set is OK for a read-write mount */
2277 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2278 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2279 "unsupported optional features (%x)",
2280 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2281 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2282 return 0;
2283 }
2284 /*
2285 * Large file size enabled file system can only be mounted
2286 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2287 */
2288 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2289 if (sizeof(blkcnt_t) < sizeof(u64)) {
2290 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2291 "cannot be mounted RDWR without "
2292 "CONFIG_LBDAF");
2293 return 0;
2294 }
2295 }
2296 return 1;
2297 }
2298
2299 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2300 __releases(kernel_lock)
2301 __acquires(kernel_lock)
2302 {
2303 struct buffer_head *bh;
2304 struct ext4_super_block *es = NULL;
2305 struct ext4_sb_info *sbi;
2306 ext4_fsblk_t block;
2307 ext4_fsblk_t sb_block = get_sb_block(&data);
2308 ext4_fsblk_t logical_sb_block;
2309 unsigned long offset = 0;
2310 unsigned long journal_devnum = 0;
2311 unsigned long def_mount_opts;
2312 struct inode *root;
2313 char *cp;
2314 const char *descr;
2315 int ret = -EINVAL;
2316 int blocksize;
2317 unsigned int db_count;
2318 unsigned int i;
2319 int needs_recovery, has_huge_files;
2320 __u64 blocks_count;
2321 int err;
2322 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2323
2324 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2325 if (!sbi)
2326 return -ENOMEM;
2327
2328 sbi->s_blockgroup_lock =
2329 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2330 if (!sbi->s_blockgroup_lock) {
2331 kfree(sbi);
2332 return -ENOMEM;
2333 }
2334 sb->s_fs_info = sbi;
2335 sbi->s_mount_opt = 0;
2336 sbi->s_resuid = EXT4_DEF_RESUID;
2337 sbi->s_resgid = EXT4_DEF_RESGID;
2338 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2339 sbi->s_sb_block = sb_block;
2340 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2341 sectors[1]);
2342
2343 unlock_kernel();
2344
2345 /* Cleanup superblock name */
2346 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2347 *cp = '!';
2348
2349 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2350 if (!blocksize) {
2351 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
2352 goto out_fail;
2353 }
2354
2355 /*
2356 * The ext4 superblock will not be buffer aligned for other than 1kB
2357 * block sizes. We need to calculate the offset from buffer start.
2358 */
2359 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2360 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2361 offset = do_div(logical_sb_block, blocksize);
2362 } else {
2363 logical_sb_block = sb_block;
2364 }
2365
2366 if (!(bh = sb_bread(sb, logical_sb_block))) {
2367 ext4_msg(sb, KERN_ERR, "unable to read superblock");
2368 goto out_fail;
2369 }
2370 /*
2371 * Note: s_es must be initialized as soon as possible because
2372 * some ext4 macro-instructions depend on its value
2373 */
2374 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2375 sbi->s_es = es;
2376 sb->s_magic = le16_to_cpu(es->s_magic);
2377 if (sb->s_magic != EXT4_SUPER_MAGIC)
2378 goto cantfind_ext4;
2379 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2380
2381 /* Set defaults before we parse the mount options */
2382 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2383 if (def_mount_opts & EXT4_DEFM_DEBUG)
2384 set_opt(sbi->s_mount_opt, DEBUG);
2385 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
2386 set_opt(sbi->s_mount_opt, GRPID);
2387 if (def_mount_opts & EXT4_DEFM_UID16)
2388 set_opt(sbi->s_mount_opt, NO_UID32);
2389 #ifdef CONFIG_EXT4_FS_XATTR
2390 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2391 set_opt(sbi->s_mount_opt, XATTR_USER);
2392 #endif
2393 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2394 if (def_mount_opts & EXT4_DEFM_ACL)
2395 set_opt(sbi->s_mount_opt, POSIX_ACL);
2396 #endif
2397 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2398 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2399 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2400 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2401 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2402 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2403
2404 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2405 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2406 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2407 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2408 else
2409 set_opt(sbi->s_mount_opt, ERRORS_RO);
2410
2411 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2412 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2413 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2414 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2415 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2416 sbi->s_mb_history_max = default_mb_history_length;
2417
2418 set_opt(sbi->s_mount_opt, BARRIER);
2419
2420 /*
2421 * enable delayed allocation by default
2422 * Use -o nodelalloc to turn it off
2423 */
2424 set_opt(sbi->s_mount_opt, DELALLOC);
2425
2426 if (!parse_options((char *) data, sb, &journal_devnum,
2427 &journal_ioprio, NULL, 0))
2428 goto failed_mount;
2429
2430 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2431 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2432
2433 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2434 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2435 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2436 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2437 ext4_msg(sb, KERN_WARNING,
2438 "feature flags set on rev 0 fs, "
2439 "running e2fsck is recommended");
2440
2441 /*
2442 * Check feature flags regardless of the revision level, since we
2443 * previously didn't change the revision level when setting the flags,
2444 * so there is a chance incompat flags are set on a rev 0 filesystem.
2445 */
2446 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
2447 goto failed_mount;
2448
2449 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2450
2451 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2452 blocksize > EXT4_MAX_BLOCK_SIZE) {
2453 ext4_msg(sb, KERN_ERR,
2454 "Unsupported filesystem blocksize %d", blocksize);
2455 goto failed_mount;
2456 }
2457
2458 if (sb->s_blocksize != blocksize) {
2459 /* Validate the filesystem blocksize */
2460 if (!sb_set_blocksize(sb, blocksize)) {
2461 ext4_msg(sb, KERN_ERR, "bad block size %d",
2462 blocksize);
2463 goto failed_mount;
2464 }
2465
2466 brelse(bh);
2467 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2468 offset = do_div(logical_sb_block, blocksize);
2469 bh = sb_bread(sb, logical_sb_block);
2470 if (!bh) {
2471 ext4_msg(sb, KERN_ERR,
2472 "Can't read superblock on 2nd try");
2473 goto failed_mount;
2474 }
2475 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2476 sbi->s_es = es;
2477 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2478 ext4_msg(sb, KERN_ERR,
2479 "Magic mismatch, very weird!");
2480 goto failed_mount;
2481 }
2482 }
2483
2484 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2485 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2486 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2487 has_huge_files);
2488 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2489
2490 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2491 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2492 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2493 } else {
2494 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2495 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2496 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2497 (!is_power_of_2(sbi->s_inode_size)) ||
2498 (sbi->s_inode_size > blocksize)) {
2499 ext4_msg(sb, KERN_ERR,
2500 "unsupported inode size: %d",
2501 sbi->s_inode_size);
2502 goto failed_mount;
2503 }
2504 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2505 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2506 }
2507
2508 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2509 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2510 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2511 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2512 !is_power_of_2(sbi->s_desc_size)) {
2513 ext4_msg(sb, KERN_ERR,
2514 "unsupported descriptor size %lu",
2515 sbi->s_desc_size);
2516 goto failed_mount;
2517 }
2518 } else
2519 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2520
2521 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2522 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2523 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2524 goto cantfind_ext4;
2525
2526 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2527 if (sbi->s_inodes_per_block == 0)
2528 goto cantfind_ext4;
2529 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2530 sbi->s_inodes_per_block;
2531 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2532 sbi->s_sbh = bh;
2533 sbi->s_mount_state = le16_to_cpu(es->s_state);
2534 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2535 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2536
2537 for (i = 0; i < 4; i++)
2538 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2539 sbi->s_def_hash_version = es->s_def_hash_version;
2540 i = le32_to_cpu(es->s_flags);
2541 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2542 sbi->s_hash_unsigned = 3;
2543 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2544 #ifdef __CHAR_UNSIGNED__
2545 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2546 sbi->s_hash_unsigned = 3;
2547 #else
2548 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2549 #endif
2550 sb->s_dirt = 1;
2551 }
2552
2553 if (sbi->s_blocks_per_group > blocksize * 8) {
2554 ext4_msg(sb, KERN_ERR,
2555 "#blocks per group too big: %lu",
2556 sbi->s_blocks_per_group);
2557 goto failed_mount;
2558 }
2559 if (sbi->s_inodes_per_group > blocksize * 8) {
2560 ext4_msg(sb, KERN_ERR,
2561 "#inodes per group too big: %lu",
2562 sbi->s_inodes_per_group);
2563 goto failed_mount;
2564 }
2565
2566 /*
2567 * Test whether we have more sectors than will fit in sector_t,
2568 * and whether the max offset is addressable by the page cache.
2569 */
2570 if ((ext4_blocks_count(es) >
2571 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2572 (ext4_blocks_count(es) >
2573 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
2574 ext4_msg(sb, KERN_ERR, "filesystem"
2575 " too large to mount safely on this system");
2576 if (sizeof(sector_t) < 8)
2577 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
2578 ret = -EFBIG;
2579 goto failed_mount;
2580 }
2581
2582 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2583 goto cantfind_ext4;
2584
2585 /* check blocks count against device size */
2586 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2587 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2588 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2589 "exceeds size of device (%llu blocks)",
2590 ext4_blocks_count(es), blocks_count);
2591 goto failed_mount;
2592 }
2593
2594 /*
2595 * It makes no sense for the first data block to be beyond the end
2596 * of the filesystem.
2597 */
2598 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2599 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2600 "block %u is beyond end of filesystem (%llu)",
2601 le32_to_cpu(es->s_first_data_block),
2602 ext4_blocks_count(es));
2603 goto failed_mount;
2604 }
2605 blocks_count = (ext4_blocks_count(es) -
2606 le32_to_cpu(es->s_first_data_block) +
2607 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2608 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2609 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2610 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
2611 "(block count %llu, first data block %u, "
2612 "blocks per group %lu)", sbi->s_groups_count,
2613 ext4_blocks_count(es),
2614 le32_to_cpu(es->s_first_data_block),
2615 EXT4_BLOCKS_PER_GROUP(sb));
2616 goto failed_mount;
2617 }
2618 sbi->s_groups_count = blocks_count;
2619 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
2620 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
2621 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2622 EXT4_DESC_PER_BLOCK(sb);
2623 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2624 GFP_KERNEL);
2625 if (sbi->s_group_desc == NULL) {
2626 ext4_msg(sb, KERN_ERR, "not enough memory");
2627 goto failed_mount;
2628 }
2629
2630 #ifdef CONFIG_PROC_FS
2631 if (ext4_proc_root)
2632 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2633 #endif
2634
2635 bgl_lock_init(sbi->s_blockgroup_lock);
2636
2637 for (i = 0; i < db_count; i++) {
2638 block = descriptor_loc(sb, logical_sb_block, i);
2639 sbi->s_group_desc[i] = sb_bread(sb, block);
2640 if (!sbi->s_group_desc[i]) {
2641 ext4_msg(sb, KERN_ERR,
2642 "can't read group descriptor %d", i);
2643 db_count = i;
2644 goto failed_mount2;
2645 }
2646 }
2647 if (!ext4_check_descriptors(sb)) {
2648 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
2649 goto failed_mount2;
2650 }
2651 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2652 if (!ext4_fill_flex_info(sb)) {
2653 ext4_msg(sb, KERN_ERR,
2654 "unable to initialize "
2655 "flex_bg meta info!");
2656 goto failed_mount2;
2657 }
2658
2659 sbi->s_gdb_count = db_count;
2660 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2661 spin_lock_init(&sbi->s_next_gen_lock);
2662
2663 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2664 ext4_count_free_blocks(sb));
2665 if (!err) {
2666 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2667 ext4_count_free_inodes(sb));
2668 }
2669 if (!err) {
2670 err = percpu_counter_init(&sbi->s_dirs_counter,
2671 ext4_count_dirs(sb));
2672 }
2673 if (!err) {
2674 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2675 }
2676 if (err) {
2677 ext4_msg(sb, KERN_ERR, "insufficient memory");
2678 goto failed_mount3;
2679 }
2680
2681 sbi->s_stripe = ext4_get_stripe_size(sbi);
2682
2683 /*
2684 * set up enough so that it can read an inode
2685 */
2686 if (!test_opt(sb, NOLOAD) &&
2687 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2688 sb->s_op = &ext4_sops;
2689 else
2690 sb->s_op = &ext4_nojournal_sops;
2691 sb->s_export_op = &ext4_export_ops;
2692 sb->s_xattr = ext4_xattr_handlers;
2693 #ifdef CONFIG_QUOTA
2694 sb->s_qcop = &ext4_qctl_operations;
2695 sb->dq_op = &ext4_quota_operations;
2696 #endif
2697 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2698 mutex_init(&sbi->s_orphan_lock);
2699 mutex_init(&sbi->s_resize_lock);
2700
2701 sb->s_root = NULL;
2702
2703 needs_recovery = (es->s_last_orphan != 0 ||
2704 EXT4_HAS_INCOMPAT_FEATURE(sb,
2705 EXT4_FEATURE_INCOMPAT_RECOVER));
2706
2707 /*
2708 * The first inode we look at is the journal inode. Don't try
2709 * root first: it may be modified in the journal!
2710 */
2711 if (!test_opt(sb, NOLOAD) &&
2712 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2713 if (ext4_load_journal(sb, es, journal_devnum))
2714 goto failed_mount3;
2715 if (!(sb->s_flags & MS_RDONLY) &&
2716 EXT4_SB(sb)->s_journal->j_failed_commit) {
2717 ext4_msg(sb, KERN_CRIT, "error: "
2718 "ext4_fill_super: Journal transaction "
2719 "%u is corrupt",
2720 EXT4_SB(sb)->s_journal->j_failed_commit);
2721 if (test_opt(sb, ERRORS_RO)) {
2722 ext4_msg(sb, KERN_CRIT,
2723 "Mounting filesystem read-only");
2724 sb->s_flags |= MS_RDONLY;
2725 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2726 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2727 }
2728 if (test_opt(sb, ERRORS_PANIC)) {
2729 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2730 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2731 ext4_commit_super(sb, 1);
2732 goto failed_mount4;
2733 }
2734 }
2735 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2736 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2737 ext4_msg(sb, KERN_ERR, "required journal recovery "
2738 "suppressed and not mounted read-only");
2739 goto failed_mount4;
2740 } else {
2741 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2742 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2743 sbi->s_journal = NULL;
2744 needs_recovery = 0;
2745 goto no_journal;
2746 }
2747
2748 if (ext4_blocks_count(es) > 0xffffffffULL &&
2749 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2750 JBD2_FEATURE_INCOMPAT_64BIT)) {
2751 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
2752 goto failed_mount4;
2753 }
2754
2755 jbd2_journal_set_features(sbi->s_journal,
2756 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2757 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
2758 jbd2_journal_set_features(sbi->s_journal, 0, 0,
2759 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2760 else
2761 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2762 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2763
2764 /* We have now updated the journal if required, so we can
2765 * validate the data journaling mode. */
2766 switch (test_opt(sb, DATA_FLAGS)) {
2767 case 0:
2768 /* No mode set, assume a default based on the journal
2769 * capabilities: ORDERED_DATA if the journal can
2770 * cope, else JOURNAL_DATA
2771 */
2772 if (jbd2_journal_check_available_features
2773 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2774 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2775 else
2776 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2777 break;
2778
2779 case EXT4_MOUNT_ORDERED_DATA:
2780 case EXT4_MOUNT_WRITEBACK_DATA:
2781 if (!jbd2_journal_check_available_features
2782 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2783 ext4_msg(sb, KERN_ERR, "Journal does not support "
2784 "requested data journaling mode");
2785 goto failed_mount4;
2786 }
2787 default:
2788 break;
2789 }
2790 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2791
2792 no_journal:
2793
2794 if (test_opt(sb, NOBH)) {
2795 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2796 ext4_msg(sb, KERN_WARNING, "Ignoring nobh option - "
2797 "its supported only with writeback mode");
2798 clear_opt(sbi->s_mount_opt, NOBH);
2799 }
2800 }
2801 /*
2802 * The jbd2_journal_load will have done any necessary log recovery,
2803 * so we can safely mount the rest of the filesystem now.
2804 */
2805
2806 root = ext4_iget(sb, EXT4_ROOT_INO);
2807 if (IS_ERR(root)) {
2808 ext4_msg(sb, KERN_ERR, "get root inode failed");
2809 ret = PTR_ERR(root);
2810 goto failed_mount4;
2811 }
2812 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2813 iput(root);
2814 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
2815 goto failed_mount4;
2816 }
2817 sb->s_root = d_alloc_root(root);
2818 if (!sb->s_root) {
2819 ext4_msg(sb, KERN_ERR, "get root dentry failed");
2820 iput(root);
2821 ret = -ENOMEM;
2822 goto failed_mount4;
2823 }
2824
2825 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2826
2827 /* determine the minimum size of new large inodes, if present */
2828 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2829 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2830 EXT4_GOOD_OLD_INODE_SIZE;
2831 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2832 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2833 if (sbi->s_want_extra_isize <
2834 le16_to_cpu(es->s_want_extra_isize))
2835 sbi->s_want_extra_isize =
2836 le16_to_cpu(es->s_want_extra_isize);
2837 if (sbi->s_want_extra_isize <
2838 le16_to_cpu(es->s_min_extra_isize))
2839 sbi->s_want_extra_isize =
2840 le16_to_cpu(es->s_min_extra_isize);
2841 }
2842 }
2843 /* Check if enough inode space is available */
2844 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2845 sbi->s_inode_size) {
2846 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2847 EXT4_GOOD_OLD_INODE_SIZE;
2848 ext4_msg(sb, KERN_INFO, "required extra inode space not"
2849 "available");
2850 }
2851
2852 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2853 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
2854 "requested data journaling mode");
2855 clear_opt(sbi->s_mount_opt, DELALLOC);
2856 } else if (test_opt(sb, DELALLOC))
2857 ext4_msg(sb, KERN_INFO, "delayed allocation enabled");
2858
2859 err = ext4_setup_system_zone(sb);
2860 if (err) {
2861 ext4_msg(sb, KERN_ERR, "failed to initialize system "
2862 "zone (%d)\n", err);
2863 goto failed_mount4;
2864 }
2865
2866 ext4_ext_init(sb);
2867 err = ext4_mb_init(sb, needs_recovery);
2868 if (err) {
2869 ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
2870 err);
2871 goto failed_mount4;
2872 }
2873
2874 sbi->s_kobj.kset = ext4_kset;
2875 init_completion(&sbi->s_kobj_unregister);
2876 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2877 "%s", sb->s_id);
2878 if (err) {
2879 ext4_mb_release(sb);
2880 ext4_ext_release(sb);
2881 goto failed_mount4;
2882 };
2883
2884 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2885 ext4_orphan_cleanup(sb, es);
2886 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2887 if (needs_recovery) {
2888 ext4_msg(sb, KERN_INFO, "recovery complete");
2889 ext4_mark_recovery_complete(sb, es);
2890 }
2891 if (EXT4_SB(sb)->s_journal) {
2892 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2893 descr = " journalled data mode";
2894 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2895 descr = " ordered data mode";
2896 else
2897 descr = " writeback data mode";
2898 } else
2899 descr = "out journal";
2900
2901 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s", descr);
2902
2903 lock_kernel();
2904 return 0;
2905
2906 cantfind_ext4:
2907 if (!silent)
2908 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
2909 goto failed_mount;
2910
2911 failed_mount4:
2912 ext4_msg(sb, KERN_ERR, "mount failed");
2913 ext4_release_system_zone(sb);
2914 if (sbi->s_journal) {
2915 jbd2_journal_destroy(sbi->s_journal);
2916 sbi->s_journal = NULL;
2917 }
2918 failed_mount3:
2919 if (sbi->s_flex_groups) {
2920 if (is_vmalloc_addr(sbi->s_flex_groups))
2921 vfree(sbi->s_flex_groups);
2922 else
2923 kfree(sbi->s_flex_groups);
2924 }
2925 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2926 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2927 percpu_counter_destroy(&sbi->s_dirs_counter);
2928 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2929 failed_mount2:
2930 for (i = 0; i < db_count; i++)
2931 brelse(sbi->s_group_desc[i]);
2932 kfree(sbi->s_group_desc);
2933 failed_mount:
2934 if (sbi->s_proc) {
2935 remove_proc_entry(sb->s_id, ext4_proc_root);
2936 }
2937 #ifdef CONFIG_QUOTA
2938 for (i = 0; i < MAXQUOTAS; i++)
2939 kfree(sbi->s_qf_names[i]);
2940 #endif
2941 ext4_blkdev_remove(sbi);
2942 brelse(bh);
2943 out_fail:
2944 sb->s_fs_info = NULL;
2945 kfree(sbi->s_blockgroup_lock);
2946 kfree(sbi);
2947 lock_kernel();
2948 return ret;
2949 }
2950
2951 /*
2952 * Setup any per-fs journal parameters now. We'll do this both on
2953 * initial mount, once the journal has been initialised but before we've
2954 * done any recovery; and again on any subsequent remount.
2955 */
2956 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2957 {
2958 struct ext4_sb_info *sbi = EXT4_SB(sb);
2959
2960 journal->j_commit_interval = sbi->s_commit_interval;
2961 journal->j_min_batch_time = sbi->s_min_batch_time;
2962 journal->j_max_batch_time = sbi->s_max_batch_time;
2963
2964 spin_lock(&journal->j_state_lock);
2965 if (test_opt(sb, BARRIER))
2966 journal->j_flags |= JBD2_BARRIER;
2967 else
2968 journal->j_flags &= ~JBD2_BARRIER;
2969 if (test_opt(sb, DATA_ERR_ABORT))
2970 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2971 else
2972 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
2973 spin_unlock(&journal->j_state_lock);
2974 }
2975
2976 static journal_t *ext4_get_journal(struct super_block *sb,
2977 unsigned int journal_inum)
2978 {
2979 struct inode *journal_inode;
2980 journal_t *journal;
2981
2982 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2983
2984 /* First, test for the existence of a valid inode on disk. Bad
2985 * things happen if we iget() an unused inode, as the subsequent
2986 * iput() will try to delete it. */
2987
2988 journal_inode = ext4_iget(sb, journal_inum);
2989 if (IS_ERR(journal_inode)) {
2990 ext4_msg(sb, KERN_ERR, "no journal found");
2991 return NULL;
2992 }
2993 if (!journal_inode->i_nlink) {
2994 make_bad_inode(journal_inode);
2995 iput(journal_inode);
2996 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
2997 return NULL;
2998 }
2999
3000 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3001 journal_inode, journal_inode->i_size);
3002 if (!S_ISREG(journal_inode->i_mode)) {
3003 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3004 iput(journal_inode);
3005 return NULL;
3006 }
3007
3008 journal = jbd2_journal_init_inode(journal_inode);
3009 if (!journal) {
3010 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3011 iput(journal_inode);
3012 return NULL;
3013 }
3014 journal->j_private = sb;
3015 ext4_init_journal_params(sb, journal);
3016 return journal;
3017 }
3018
3019 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3020 dev_t j_dev)
3021 {
3022 struct buffer_head *bh;
3023 journal_t *journal;
3024 ext4_fsblk_t start;
3025 ext4_fsblk_t len;
3026 int hblock, blocksize;
3027 ext4_fsblk_t sb_block;
3028 unsigned long offset;
3029 struct ext4_super_block *es;
3030 struct block_device *bdev;
3031
3032 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3033
3034 bdev = ext4_blkdev_get(j_dev, sb);
3035 if (bdev == NULL)
3036 return NULL;
3037
3038 if (bd_claim(bdev, sb)) {
3039 ext4_msg(sb, KERN_ERR,
3040 "failed to claim external journal device");
3041 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3042 return NULL;
3043 }
3044
3045 blocksize = sb->s_blocksize;
3046 hblock = bdev_logical_block_size(bdev);
3047 if (blocksize < hblock) {
3048 ext4_msg(sb, KERN_ERR,
3049 "blocksize too small for journal device");
3050 goto out_bdev;
3051 }
3052
3053 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3054 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3055 set_blocksize(bdev, blocksize);
3056 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3057 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3058 "external journal");
3059 goto out_bdev;
3060 }
3061
3062 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3063 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3064 !(le32_to_cpu(es->s_feature_incompat) &
3065 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3066 ext4_msg(sb, KERN_ERR, "external journal has "
3067 "bad superblock");
3068 brelse(bh);
3069 goto out_bdev;
3070 }
3071
3072 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3073 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3074 brelse(bh);
3075 goto out_bdev;
3076 }
3077
3078 len = ext4_blocks_count(es);
3079 start = sb_block + 1;
3080 brelse(bh); /* we're done with the superblock */
3081
3082 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3083 start, len, blocksize);
3084 if (!journal) {
3085 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3086 goto out_bdev;
3087 }
3088 journal->j_private = sb;
3089 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3090 wait_on_buffer(journal->j_sb_buffer);
3091 if (!buffer_uptodate(journal->j_sb_buffer)) {
3092 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3093 goto out_journal;
3094 }
3095 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3096 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3097 "user (unsupported) - %d",
3098 be32_to_cpu(journal->j_superblock->s_nr_users));
3099 goto out_journal;
3100 }
3101 EXT4_SB(sb)->journal_bdev = bdev;
3102 ext4_init_journal_params(sb, journal);
3103 return journal;
3104
3105 out_journal:
3106 jbd2_journal_destroy(journal);
3107 out_bdev:
3108 ext4_blkdev_put(bdev);
3109 return NULL;
3110 }
3111
3112 static int ext4_load_journal(struct super_block *sb,
3113 struct ext4_super_block *es,
3114 unsigned long journal_devnum)
3115 {
3116 journal_t *journal;
3117 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3118 dev_t journal_dev;
3119 int err = 0;
3120 int really_read_only;
3121
3122 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3123
3124 if (journal_devnum &&
3125 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3126 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3127 "numbers have changed");
3128 journal_dev = new_decode_dev(journal_devnum);
3129 } else
3130 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3131
3132 really_read_only = bdev_read_only(sb->s_bdev);
3133
3134 /*
3135 * Are we loading a blank journal or performing recovery after a
3136 * crash? For recovery, we need to check in advance whether we
3137 * can get read-write access to the device.
3138 */
3139 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3140 if (sb->s_flags & MS_RDONLY) {
3141 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3142 "required on readonly filesystem");
3143 if (really_read_only) {
3144 ext4_msg(sb, KERN_ERR, "write access "
3145 "unavailable, cannot proceed");
3146 return -EROFS;
3147 }
3148 ext4_msg(sb, KERN_INFO, "write access will "
3149 "be enabled during recovery");
3150 }
3151 }
3152
3153 if (journal_inum && journal_dev) {
3154 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3155 "and inode journals!");
3156 return -EINVAL;
3157 }
3158
3159 if (journal_inum) {
3160 if (!(journal = ext4_get_journal(sb, journal_inum)))
3161 return -EINVAL;
3162 } else {
3163 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3164 return -EINVAL;
3165 }
3166
3167 if (journal->j_flags & JBD2_BARRIER)
3168 ext4_msg(sb, KERN_INFO, "barriers enabled");
3169 else
3170 ext4_msg(sb, KERN_INFO, "barriers disabled");
3171
3172 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3173 err = jbd2_journal_update_format(journal);
3174 if (err) {
3175 ext4_msg(sb, KERN_ERR, "error updating journal");
3176 jbd2_journal_destroy(journal);
3177 return err;
3178 }
3179 }
3180
3181 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3182 err = jbd2_journal_wipe(journal, !really_read_only);
3183 if (!err)
3184 err = jbd2_journal_load(journal);
3185
3186 if (err) {
3187 ext4_msg(sb, KERN_ERR, "error loading journal");
3188 jbd2_journal_destroy(journal);
3189 return err;
3190 }
3191
3192 EXT4_SB(sb)->s_journal = journal;
3193 ext4_clear_journal_err(sb, es);
3194
3195 if (journal_devnum &&
3196 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3197 es->s_journal_dev = cpu_to_le32(journal_devnum);
3198
3199 /* Make sure we flush the recovery flag to disk. */
3200 ext4_commit_super(sb, 1);
3201 }
3202
3203 return 0;
3204 }
3205
3206 static int ext4_commit_super(struct super_block *sb, int sync)
3207 {
3208 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3209 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3210 int error = 0;
3211
3212 if (!sbh)
3213 return error;
3214 if (buffer_write_io_error(sbh)) {
3215 /*
3216 * Oh, dear. A previous attempt to write the
3217 * superblock failed. This could happen because the
3218 * USB device was yanked out. Or it could happen to
3219 * be a transient write error and maybe the block will
3220 * be remapped. Nothing we can do but to retry the
3221 * write and hope for the best.
3222 */
3223 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3224 "superblock detected");
3225 clear_buffer_write_io_error(sbh);
3226 set_buffer_uptodate(sbh);
3227 }
3228 /*
3229 * If the file system is mounted read-only, don't update the
3230 * superblock write time. This avoids updating the superblock
3231 * write time when we are mounting the root file system
3232 * read/only but we need to replay the journal; at that point,
3233 * for people who are east of GMT and who make their clock
3234 * tick in localtime for Windows bug-for-bug compatibility,
3235 * the clock is set in the future, and this will cause e2fsck
3236 * to complain and force a full file system check.
3237 */
3238 if (!(sb->s_flags & MS_RDONLY))
3239 es->s_wtime = cpu_to_le32(get_seconds());
3240 es->s_kbytes_written =
3241 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3242 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3243 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3244 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3245 &EXT4_SB(sb)->s_freeblocks_counter));
3246 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3247 &EXT4_SB(sb)->s_freeinodes_counter));
3248 sb->s_dirt = 0;
3249 BUFFER_TRACE(sbh, "marking dirty");
3250 mark_buffer_dirty(sbh);
3251 if (sync) {
3252 error = sync_dirty_buffer(sbh);
3253 if (error)
3254 return error;
3255
3256 error = buffer_write_io_error(sbh);
3257 if (error) {
3258 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3259 "superblock");
3260 clear_buffer_write_io_error(sbh);
3261 set_buffer_uptodate(sbh);
3262 }
3263 }
3264 return error;
3265 }
3266
3267 /*
3268 * Have we just finished recovery? If so, and if we are mounting (or
3269 * remounting) the filesystem readonly, then we will end up with a
3270 * consistent fs on disk. Record that fact.
3271 */
3272 static void ext4_mark_recovery_complete(struct super_block *sb,
3273 struct ext4_super_block *es)
3274 {
3275 journal_t *journal = EXT4_SB(sb)->s_journal;
3276
3277 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3278 BUG_ON(journal != NULL);
3279 return;
3280 }
3281 jbd2_journal_lock_updates(journal);
3282 if (jbd2_journal_flush(journal) < 0)
3283 goto out;
3284
3285 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3286 sb->s_flags & MS_RDONLY) {
3287 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3288 ext4_commit_super(sb, 1);
3289 }
3290
3291 out:
3292 jbd2_journal_unlock_updates(journal);
3293 }
3294
3295 /*
3296 * If we are mounting (or read-write remounting) a filesystem whose journal
3297 * has recorded an error from a previous lifetime, move that error to the
3298 * main filesystem now.
3299 */
3300 static void ext4_clear_journal_err(struct super_block *sb,
3301 struct ext4_super_block *es)
3302 {
3303 journal_t *journal;
3304 int j_errno;
3305 const char *errstr;
3306
3307 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3308
3309 journal = EXT4_SB(sb)->s_journal;
3310
3311 /*
3312 * Now check for any error status which may have been recorded in the
3313 * journal by a prior ext4_error() or ext4_abort()
3314 */
3315
3316 j_errno = jbd2_journal_errno(journal);
3317 if (j_errno) {
3318 char nbuf[16];
3319
3320 errstr = ext4_decode_error(sb, j_errno, nbuf);
3321 ext4_warning(sb, __func__, "Filesystem error recorded "
3322 "from previous mount: %s", errstr);
3323 ext4_warning(sb, __func__, "Marking fs in need of "
3324 "filesystem check.");
3325
3326 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3327 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3328 ext4_commit_super(sb, 1);
3329
3330 jbd2_journal_clear_err(journal);
3331 }
3332 }
3333
3334 /*
3335 * Force the running and committing transactions to commit,
3336 * and wait on the commit.
3337 */
3338 int ext4_force_commit(struct super_block *sb)
3339 {
3340 journal_t *journal;
3341 int ret = 0;
3342
3343 if (sb->s_flags & MS_RDONLY)
3344 return 0;
3345
3346 journal = EXT4_SB(sb)->s_journal;
3347 if (journal)
3348 ret = ext4_journal_force_commit(journal);
3349
3350 return ret;
3351 }
3352
3353 static void ext4_write_super(struct super_block *sb)
3354 {
3355 lock_super(sb);
3356 ext4_commit_super(sb, 1);
3357 unlock_super(sb);
3358 }
3359
3360 static int ext4_sync_fs(struct super_block *sb, int wait)
3361 {
3362 int ret = 0;
3363 tid_t target;
3364
3365 trace_ext4_sync_fs(sb, wait);
3366 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
3367 if (wait)
3368 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
3369 }
3370 return ret;
3371 }
3372
3373 /*
3374 * LVM calls this function before a (read-only) snapshot is created. This
3375 * gives us a chance to flush the journal completely and mark the fs clean.
3376 */
3377 static int ext4_freeze(struct super_block *sb)
3378 {
3379 int error = 0;
3380 journal_t *journal;
3381
3382 if (sb->s_flags & MS_RDONLY)
3383 return 0;
3384
3385 journal = EXT4_SB(sb)->s_journal;
3386
3387 /* Now we set up the journal barrier. */
3388 jbd2_journal_lock_updates(journal);
3389
3390 /*
3391 * Don't clear the needs_recovery flag if we failed to flush
3392 * the journal.
3393 */
3394 error = jbd2_journal_flush(journal);
3395 if (error < 0) {
3396 out:
3397 jbd2_journal_unlock_updates(journal);
3398 return error;
3399 }
3400
3401 /* Journal blocked and flushed, clear needs_recovery flag. */
3402 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3403 error = ext4_commit_super(sb, 1);
3404 if (error)
3405 goto out;
3406 return 0;
3407 }
3408
3409 /*
3410 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3411 * flag here, even though the filesystem is not technically dirty yet.
3412 */
3413 static int ext4_unfreeze(struct super_block *sb)
3414 {
3415 if (sb->s_flags & MS_RDONLY)
3416 return 0;
3417
3418 lock_super(sb);
3419 /* Reset the needs_recovery flag before the fs is unlocked. */
3420 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3421 ext4_commit_super(sb, 1);
3422 unlock_super(sb);
3423 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3424 return 0;
3425 }
3426
3427 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3428 {
3429 struct ext4_super_block *es;
3430 struct ext4_sb_info *sbi = EXT4_SB(sb);
3431 ext4_fsblk_t n_blocks_count = 0;
3432 unsigned long old_sb_flags;
3433 struct ext4_mount_options old_opts;
3434 ext4_group_t g;
3435 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3436 int err;
3437 #ifdef CONFIG_QUOTA
3438 int i;
3439 #endif
3440
3441 lock_kernel();
3442
3443 /* Store the original options */
3444 lock_super(sb);
3445 old_sb_flags = sb->s_flags;
3446 old_opts.s_mount_opt = sbi->s_mount_opt;
3447 old_opts.s_resuid = sbi->s_resuid;
3448 old_opts.s_resgid = sbi->s_resgid;
3449 old_opts.s_commit_interval = sbi->s_commit_interval;
3450 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3451 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3452 #ifdef CONFIG_QUOTA
3453 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3454 for (i = 0; i < MAXQUOTAS; i++)
3455 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3456 #endif
3457 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3458 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3459
3460 /*
3461 * Allow the "check" option to be passed as a remount option.
3462 */
3463 if (!parse_options(data, sb, NULL, &journal_ioprio,
3464 &n_blocks_count, 1)) {
3465 err = -EINVAL;
3466 goto restore_opts;
3467 }
3468
3469 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
3470 ext4_abort(sb, __func__, "Abort forced by user");
3471
3472 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3473 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3474
3475 es = sbi->s_es;
3476
3477 if (sbi->s_journal) {
3478 ext4_init_journal_params(sb, sbi->s_journal);
3479 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3480 }
3481
3482 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3483 n_blocks_count > ext4_blocks_count(es)) {
3484 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
3485 err = -EROFS;
3486 goto restore_opts;
3487 }
3488
3489 if (*flags & MS_RDONLY) {
3490 /*
3491 * First of all, the unconditional stuff we have to do
3492 * to disable replay of the journal when we next remount
3493 */
3494 sb->s_flags |= MS_RDONLY;
3495
3496 /*
3497 * OK, test if we are remounting a valid rw partition
3498 * readonly, and if so set the rdonly flag and then
3499 * mark the partition as valid again.
3500 */
3501 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3502 (sbi->s_mount_state & EXT4_VALID_FS))
3503 es->s_state = cpu_to_le16(sbi->s_mount_state);
3504
3505 if (sbi->s_journal)
3506 ext4_mark_recovery_complete(sb, es);
3507 } else {
3508 /* Make sure we can mount this feature set readwrite */
3509 if (!ext4_feature_set_ok(sb, 0)) {
3510 err = -EROFS;
3511 goto restore_opts;
3512 }
3513 /*
3514 * Make sure the group descriptor checksums
3515 * are sane. If they aren't, refuse to remount r/w.
3516 */
3517 for (g = 0; g < sbi->s_groups_count; g++) {
3518 struct ext4_group_desc *gdp =
3519 ext4_get_group_desc(sb, g, NULL);
3520
3521 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3522 ext4_msg(sb, KERN_ERR,
3523 "ext4_remount: Checksum for group %u failed (%u!=%u)",
3524 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3525 le16_to_cpu(gdp->bg_checksum));
3526 err = -EINVAL;
3527 goto restore_opts;
3528 }
3529 }
3530
3531 /*
3532 * If we have an unprocessed orphan list hanging
3533 * around from a previously readonly bdev mount,
3534 * require a full umount/remount for now.
3535 */
3536 if (es->s_last_orphan) {
3537 ext4_msg(sb, KERN_WARNING, "Couldn't "
3538 "remount RDWR because of unprocessed "
3539 "orphan inode list. Please "
3540 "umount/remount instead");
3541 err = -EINVAL;
3542 goto restore_opts;
3543 }
3544
3545 /*
3546 * Mounting a RDONLY partition read-write, so reread
3547 * and store the current valid flag. (It may have
3548 * been changed by e2fsck since we originally mounted
3549 * the partition.)
3550 */
3551 if (sbi->s_journal)
3552 ext4_clear_journal_err(sb, es);
3553 sbi->s_mount_state = le16_to_cpu(es->s_state);
3554 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3555 goto restore_opts;
3556 if (!ext4_setup_super(sb, es, 0))
3557 sb->s_flags &= ~MS_RDONLY;
3558 }
3559 }
3560 ext4_setup_system_zone(sb);
3561 if (sbi->s_journal == NULL)
3562 ext4_commit_super(sb, 1);
3563
3564 #ifdef CONFIG_QUOTA
3565 /* Release old quota file names */
3566 for (i = 0; i < MAXQUOTAS; i++)
3567 if (old_opts.s_qf_names[i] &&
3568 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3569 kfree(old_opts.s_qf_names[i]);
3570 #endif
3571 unlock_super(sb);
3572 unlock_kernel();
3573 return 0;
3574
3575 restore_opts:
3576 sb->s_flags = old_sb_flags;
3577 sbi->s_mount_opt = old_opts.s_mount_opt;
3578 sbi->s_resuid = old_opts.s_resuid;
3579 sbi->s_resgid = old_opts.s_resgid;
3580 sbi->s_commit_interval = old_opts.s_commit_interval;
3581 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3582 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3583 #ifdef CONFIG_QUOTA
3584 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3585 for (i = 0; i < MAXQUOTAS; i++) {
3586 if (sbi->s_qf_names[i] &&
3587 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3588 kfree(sbi->s_qf_names[i]);
3589 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3590 }
3591 #endif
3592 unlock_super(sb);
3593 unlock_kernel();
3594 return err;
3595 }
3596
3597 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3598 {
3599 struct super_block *sb = dentry->d_sb;
3600 struct ext4_sb_info *sbi = EXT4_SB(sb);
3601 struct ext4_super_block *es = sbi->s_es;
3602 u64 fsid;
3603
3604 if (test_opt(sb, MINIX_DF)) {
3605 sbi->s_overhead_last = 0;
3606 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3607 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3608 ext4_fsblk_t overhead = 0;
3609
3610 /*
3611 * Compute the overhead (FS structures). This is constant
3612 * for a given filesystem unless the number of block groups
3613 * changes so we cache the previous value until it does.
3614 */
3615
3616 /*
3617 * All of the blocks before first_data_block are
3618 * overhead
3619 */
3620 overhead = le32_to_cpu(es->s_first_data_block);
3621
3622 /*
3623 * Add the overhead attributed to the superblock and
3624 * block group descriptors. If the sparse superblocks
3625 * feature is turned on, then not all groups have this.
3626 */
3627 for (i = 0; i < ngroups; i++) {
3628 overhead += ext4_bg_has_super(sb, i) +
3629 ext4_bg_num_gdb(sb, i);
3630 cond_resched();
3631 }
3632
3633 /*
3634 * Every block group has an inode bitmap, a block
3635 * bitmap, and an inode table.
3636 */
3637 overhead += ngroups * (2 + sbi->s_itb_per_group);
3638 sbi->s_overhead_last = overhead;
3639 smp_wmb();
3640 sbi->s_blocks_last = ext4_blocks_count(es);
3641 }
3642
3643 buf->f_type = EXT4_SUPER_MAGIC;
3644 buf->f_bsize = sb->s_blocksize;
3645 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3646 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3647 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3648 ext4_free_blocks_count_set(es, buf->f_bfree);
3649 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3650 if (buf->f_bfree < ext4_r_blocks_count(es))
3651 buf->f_bavail = 0;
3652 buf->f_files = le32_to_cpu(es->s_inodes_count);
3653 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3654 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
3655 buf->f_namelen = EXT4_NAME_LEN;
3656 fsid = le64_to_cpup((void *)es->s_uuid) ^
3657 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3658 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3659 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3660
3661 return 0;
3662 }
3663
3664 /* Helper function for writing quotas on sync - we need to start transaction
3665 * before quota file is locked for write. Otherwise the are possible deadlocks:
3666 * Process 1 Process 2
3667 * ext4_create() quota_sync()
3668 * jbd2_journal_start() write_dquot()
3669 * vfs_dq_init() down(dqio_mutex)
3670 * down(dqio_mutex) jbd2_journal_start()
3671 *
3672 */
3673
3674 #ifdef CONFIG_QUOTA
3675
3676 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3677 {
3678 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3679 }
3680
3681 static int ext4_write_dquot(struct dquot *dquot)
3682 {
3683 int ret, err;
3684 handle_t *handle;
3685 struct inode *inode;
3686
3687 inode = dquot_to_inode(dquot);
3688 handle = ext4_journal_start(inode,
3689 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3690 if (IS_ERR(handle))
3691 return PTR_ERR(handle);
3692 ret = dquot_commit(dquot);
3693 err = ext4_journal_stop(handle);
3694 if (!ret)
3695 ret = err;
3696 return ret;
3697 }
3698
3699 static int ext4_acquire_dquot(struct dquot *dquot)
3700 {
3701 int ret, err;
3702 handle_t *handle;
3703
3704 handle = ext4_journal_start(dquot_to_inode(dquot),
3705 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3706 if (IS_ERR(handle))
3707 return PTR_ERR(handle);
3708 ret = dquot_acquire(dquot);
3709 err = ext4_journal_stop(handle);
3710 if (!ret)
3711 ret = err;
3712 return ret;
3713 }
3714
3715 static int ext4_release_dquot(struct dquot *dquot)
3716 {
3717 int ret, err;
3718 handle_t *handle;
3719
3720 handle = ext4_journal_start(dquot_to_inode(dquot),
3721 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3722 if (IS_ERR(handle)) {
3723 /* Release dquot anyway to avoid endless cycle in dqput() */
3724 dquot_release(dquot);
3725 return PTR_ERR(handle);
3726 }
3727 ret = dquot_release(dquot);
3728 err = ext4_journal_stop(handle);
3729 if (!ret)
3730 ret = err;
3731 return ret;
3732 }
3733
3734 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3735 {
3736 /* Are we journaling quotas? */
3737 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3738 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3739 dquot_mark_dquot_dirty(dquot);
3740 return ext4_write_dquot(dquot);
3741 } else {
3742 return dquot_mark_dquot_dirty(dquot);
3743 }
3744 }
3745
3746 static int ext4_write_info(struct super_block *sb, int type)
3747 {
3748 int ret, err;
3749 handle_t *handle;
3750
3751 /* Data block + inode block */
3752 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3753 if (IS_ERR(handle))
3754 return PTR_ERR(handle);
3755 ret = dquot_commit_info(sb, type);
3756 err = ext4_journal_stop(handle);
3757 if (!ret)
3758 ret = err;
3759 return ret;
3760 }
3761
3762 /*
3763 * Turn on quotas during mount time - we need to find
3764 * the quota file and such...
3765 */
3766 static int ext4_quota_on_mount(struct super_block *sb, int type)
3767 {
3768 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3769 EXT4_SB(sb)->s_jquota_fmt, type);
3770 }
3771
3772 /*
3773 * Standard function to be called on quota_on
3774 */
3775 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3776 char *name, int remount)
3777 {
3778 int err;
3779 struct path path;
3780
3781 if (!test_opt(sb, QUOTA))
3782 return -EINVAL;
3783 /* When remounting, no checks are needed and in fact, name is NULL */
3784 if (remount)
3785 return vfs_quota_on(sb, type, format_id, name, remount);
3786
3787 err = kern_path(name, LOOKUP_FOLLOW, &path);
3788 if (err)
3789 return err;
3790
3791 /* Quotafile not on the same filesystem? */
3792 if (path.mnt->mnt_sb != sb) {
3793 path_put(&path);
3794 return -EXDEV;
3795 }
3796 /* Journaling quota? */
3797 if (EXT4_SB(sb)->s_qf_names[type]) {
3798 /* Quotafile not in fs root? */
3799 if (path.dentry->d_parent != sb->s_root)
3800 ext4_msg(sb, KERN_WARNING,
3801 "Quota file not on filesystem root. "
3802 "Journaled quota will not work");
3803 }
3804
3805 /*
3806 * When we journal data on quota file, we have to flush journal to see
3807 * all updates to the file when we bypass pagecache...
3808 */
3809 if (EXT4_SB(sb)->s_journal &&
3810 ext4_should_journal_data(path.dentry->d_inode)) {
3811 /*
3812 * We don't need to lock updates but journal_flush() could
3813 * otherwise be livelocked...
3814 */
3815 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3816 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3817 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3818 if (err) {
3819 path_put(&path);
3820 return err;
3821 }
3822 }
3823
3824 err = vfs_quota_on_path(sb, type, format_id, &path);
3825 path_put(&path);
3826 return err;
3827 }
3828
3829 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3830 * acquiring the locks... As quota files are never truncated and quota code
3831 * itself serializes the operations (and noone else should touch the files)
3832 * we don't have to be afraid of races */
3833 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3834 size_t len, loff_t off)
3835 {
3836 struct inode *inode = sb_dqopt(sb)->files[type];
3837 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3838 int err = 0;
3839 int offset = off & (sb->s_blocksize - 1);
3840 int tocopy;
3841 size_t toread;
3842 struct buffer_head *bh;
3843 loff_t i_size = i_size_read(inode);
3844
3845 if (off > i_size)
3846 return 0;
3847 if (off+len > i_size)
3848 len = i_size-off;
3849 toread = len;
3850 while (toread > 0) {
3851 tocopy = sb->s_blocksize - offset < toread ?
3852 sb->s_blocksize - offset : toread;
3853 bh = ext4_bread(NULL, inode, blk, 0, &err);
3854 if (err)
3855 return err;
3856 if (!bh) /* A hole? */
3857 memset(data, 0, tocopy);
3858 else
3859 memcpy(data, bh->b_data+offset, tocopy);
3860 brelse(bh);
3861 offset = 0;
3862 toread -= tocopy;
3863 data += tocopy;
3864 blk++;
3865 }
3866 return len;
3867 }
3868
3869 /* Write to quotafile (we know the transaction is already started and has
3870 * enough credits) */
3871 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3872 const char *data, size_t len, loff_t off)
3873 {
3874 struct inode *inode = sb_dqopt(sb)->files[type];
3875 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3876 int err = 0;
3877 int offset = off & (sb->s_blocksize - 1);
3878 int tocopy;
3879 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3880 size_t towrite = len;
3881 struct buffer_head *bh;
3882 handle_t *handle = journal_current_handle();
3883
3884 if (EXT4_SB(sb)->s_journal && !handle) {
3885 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
3886 " cancelled because transaction is not started",
3887 (unsigned long long)off, (unsigned long long)len);
3888 return -EIO;
3889 }
3890 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3891 while (towrite > 0) {
3892 tocopy = sb->s_blocksize - offset < towrite ?
3893 sb->s_blocksize - offset : towrite;
3894 bh = ext4_bread(handle, inode, blk, 1, &err);
3895 if (!bh)
3896 goto out;
3897 if (journal_quota) {
3898 err = ext4_journal_get_write_access(handle, bh);
3899 if (err) {
3900 brelse(bh);
3901 goto out;
3902 }
3903 }
3904 lock_buffer(bh);
3905 memcpy(bh->b_data+offset, data, tocopy);
3906 flush_dcache_page(bh->b_page);
3907 unlock_buffer(bh);
3908 if (journal_quota)
3909 err = ext4_handle_dirty_metadata(handle, NULL, bh);
3910 else {
3911 /* Always do at least ordered writes for quotas */
3912 err = ext4_jbd2_file_inode(handle, inode);
3913 mark_buffer_dirty(bh);
3914 }
3915 brelse(bh);
3916 if (err)
3917 goto out;
3918 offset = 0;
3919 towrite -= tocopy;
3920 data += tocopy;
3921 blk++;
3922 }
3923 out:
3924 if (len == towrite) {
3925 mutex_unlock(&inode->i_mutex);
3926 return err;
3927 }
3928 if (inode->i_size < off+len-towrite) {
3929 i_size_write(inode, off+len-towrite);
3930 EXT4_I(inode)->i_disksize = inode->i_size;
3931 }
3932 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3933 ext4_mark_inode_dirty(handle, inode);
3934 mutex_unlock(&inode->i_mutex);
3935 return len - towrite;
3936 }
3937
3938 #endif
3939
3940 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
3941 const char *dev_name, void *data, struct vfsmount *mnt)
3942 {
3943 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
3944 }
3945
3946 static struct file_system_type ext4_fs_type = {
3947 .owner = THIS_MODULE,
3948 .name = "ext4",
3949 .get_sb = ext4_get_sb,
3950 .kill_sb = kill_block_super,
3951 .fs_flags = FS_REQUIRES_DEV,
3952 };
3953
3954 #ifdef CONFIG_EXT4DEV_COMPAT
3955 static int ext4dev_get_sb(struct file_system_type *fs_type, int flags,
3956 const char *dev_name, void *data,struct vfsmount *mnt)
3957 {
3958 printk(KERN_WARNING "EXT4-fs (%s): Update your userspace programs "
3959 "to mount using ext4\n", dev_name);
3960 printk(KERN_WARNING "EXT4-fs (%s): ext4dev backwards compatibility "
3961 "will go away by 2.6.31\n", dev_name);
3962 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
3963 }
3964
3965 static struct file_system_type ext4dev_fs_type = {
3966 .owner = THIS_MODULE,
3967 .name = "ext4dev",
3968 .get_sb = ext4dev_get_sb,
3969 .kill_sb = kill_block_super,
3970 .fs_flags = FS_REQUIRES_DEV,
3971 };
3972 MODULE_ALIAS("ext4dev");
3973 #endif
3974
3975 static int __init init_ext4_fs(void)
3976 {
3977 int err;
3978
3979 err = init_ext4_system_zone();
3980 if (err)
3981 return err;
3982 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
3983 if (!ext4_kset)
3984 goto out4;
3985 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
3986 err = init_ext4_mballoc();
3987 if (err)
3988 goto out3;
3989
3990 err = init_ext4_xattr();
3991 if (err)
3992 goto out2;
3993 err = init_inodecache();
3994 if (err)
3995 goto out1;
3996 err = register_filesystem(&ext4_fs_type);
3997 if (err)
3998 goto out;
3999 #ifdef CONFIG_EXT4DEV_COMPAT
4000 err = register_filesystem(&ext4dev_fs_type);
4001 if (err) {
4002 unregister_filesystem(&ext4_fs_type);
4003 goto out;
4004 }
4005 #endif
4006 return 0;
4007 out:
4008 destroy_inodecache();
4009 out1:
4010 exit_ext4_xattr();
4011 out2:
4012 exit_ext4_mballoc();
4013 out3:
4014 remove_proc_entry("fs/ext4", NULL);
4015 kset_unregister(ext4_kset);
4016 out4:
4017 exit_ext4_system_zone();
4018 return err;
4019 }
4020
4021 static void __exit exit_ext4_fs(void)
4022 {
4023 unregister_filesystem(&ext4_fs_type);
4024 #ifdef CONFIG_EXT4DEV_COMPAT
4025 unregister_filesystem(&ext4dev_fs_type);
4026 #endif
4027 destroy_inodecache();
4028 exit_ext4_xattr();
4029 exit_ext4_mballoc();
4030 remove_proc_entry("fs/ext4", NULL);
4031 kset_unregister(ext4_kset);
4032 exit_ext4_system_zone();
4033 }
4034
4035 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4036 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4037 MODULE_LICENSE("GPL");
4038 module_init(init_ext4_fs)
4039 module_exit(exit_ext4_fs)
Linux 2.6 ibm-acpi/thinkpad-acpi driver git tree