projects / linux-block.git / blame_incremental
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (non-incremental) | history | HEAD
ext4: Clean up ext4_get_blocks() so it does not depend on bh_result->b_state
[linux-block.git] / fs / ext4 / super.c
... / ...
CommitLineData
1/*
2 * linux/fs/ext4/super.c
3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
23#include <linux/vmalloc.h>
24#include <linux/jbd2.h>
25#include <linux/slab.h>
26#include <linux/init.h>
27#include <linux/blkdev.h>
28#include <linux/parser.h>
29#include <linux/smp_lock.h>
30#include <linux/buffer_head.h>
31#include <linux/exportfs.h>
32#include <linux/vfs.h>
33#include <linux/random.h>
34#include <linux/mount.h>
35#include <linux/namei.h>
36#include <linux/quotaops.h>
37#include <linux/seq_file.h>
38#include <linux/proc_fs.h>
39#include <linux/ctype.h>
40#include <linux/marker.h>
41#include <linux/log2.h>
42#include <linux/crc16.h>
43#include <asm/uaccess.h>
44
45#include "ext4.h"
46#include "ext4_jbd2.h"
47#include "xattr.h"
48#include "acl.h"
49
50static int default_mb_history_length = 1000;
51
52module_param_named(default_mb_history_length, default_mb_history_length,
53 int, 0644);
54MODULE_PARM_DESC(default_mb_history_length,
55 "Default number of entries saved for mb_history");
56
57struct proc_dir_entry *ext4_proc_root;
58static struct kset *ext4_kset;
59
60static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
61 unsigned long journal_devnum);
62static int ext4_commit_super(struct super_block *sb, int sync);
63static void ext4_mark_recovery_complete(struct super_block *sb,
64 struct ext4_super_block *es);
65static void ext4_clear_journal_err(struct super_block *sb,
66 struct ext4_super_block *es);
67static int ext4_sync_fs(struct super_block *sb, int wait);
68static const char *ext4_decode_error(struct super_block *sb, int errno,
69 char nbuf[16]);
70static int ext4_remount(struct super_block *sb, int *flags, char *data);
71static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
72static int ext4_unfreeze(struct super_block *sb);
73static void ext4_write_super(struct super_block *sb);
74static int ext4_freeze(struct super_block *sb);
75
76
77ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
78 struct ext4_group_desc *bg)
79{
80 return le32_to_cpu(bg->bg_block_bitmap_lo) |
81 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
82 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
83}
84
85ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
86 struct ext4_group_desc *bg)
87{
88 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
89 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
90 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
91}
92
93ext4_fsblk_t ext4_inode_table(struct super_block *sb,
94 struct ext4_group_desc *bg)
95{
96 return le32_to_cpu(bg->bg_inode_table_lo) |
97 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
98 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
99}
100
101__u32 ext4_free_blks_count(struct super_block *sb,
102 struct ext4_group_desc *bg)
103{
104 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
105 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
106 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
107}
108
109__u32 ext4_free_inodes_count(struct super_block *sb,
110 struct ext4_group_desc *bg)
111{
112 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
113 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
114 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
115}
116
117__u32 ext4_used_dirs_count(struct super_block *sb,
118 struct ext4_group_desc *bg)
119{
120 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
121 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
122 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
123}
124
125__u32 ext4_itable_unused_count(struct super_block *sb,
126 struct ext4_group_desc *bg)
127{
128 return le16_to_cpu(bg->bg_itable_unused_lo) |
129 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
130 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
131}
132
133void ext4_block_bitmap_set(struct super_block *sb,
134 struct ext4_group_desc *bg, ext4_fsblk_t blk)
135{
136 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
137 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
138 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
139}
140
141void ext4_inode_bitmap_set(struct super_block *sb,
142 struct ext4_group_desc *bg, ext4_fsblk_t blk)
143{
144 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
145 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
146 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
147}
148
149void ext4_inode_table_set(struct super_block *sb,
150 struct ext4_group_desc *bg, ext4_fsblk_t blk)
151{
152 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
153 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
154 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
155}
156
157void ext4_free_blks_set(struct super_block *sb,
158 struct ext4_group_desc *bg, __u32 count)
159{
160 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
161 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
162 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
163}
164
165void ext4_free_inodes_set(struct super_block *sb,
166 struct ext4_group_desc *bg, __u32 count)
167{
168 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
169 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
170 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
171}
172
173void ext4_used_dirs_set(struct super_block *sb,
174 struct ext4_group_desc *bg, __u32 count)
175{
176 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
177 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
178 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
179}
180
181void ext4_itable_unused_set(struct super_block *sb,
182 struct ext4_group_desc *bg, __u32 count)
183{
184 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
185 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
186 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
187}
188
189/*
190 * Wrappers for jbd2_journal_start/end.
191 *
192 * The only special thing we need to do here is to make sure that all
193 * journal_end calls result in the superblock being marked dirty, so
194 * that sync() will call the filesystem's write_super callback if
195 * appropriate.
196 */
197handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
198{
199 journal_t *journal;
200
201 if (sb->s_flags & MS_RDONLY)
202 return ERR_PTR(-EROFS);
203
204 /* Special case here: if the journal has aborted behind our
205 * backs (eg. EIO in the commit thread), then we still need to
206 * take the FS itself readonly cleanly. */
207 journal = EXT4_SB(sb)->s_journal;
208 if (journal) {
209 if (is_journal_aborted(journal)) {
210 ext4_abort(sb, __func__,
211 "Detected aborted journal");
212 return ERR_PTR(-EROFS);
213 }
214 return jbd2_journal_start(journal, nblocks);
215 }
216 /*
217 * We're not journaling, return the appropriate indication.
218 */
219 current->journal_info = EXT4_NOJOURNAL_HANDLE;
220 return current->journal_info;
221}
222
223/*
224 * The only special thing we need to do here is to make sure that all
225 * jbd2_journal_stop calls result in the superblock being marked dirty, so
226 * that sync() will call the filesystem's write_super callback if
227 * appropriate.
228 */
229int __ext4_journal_stop(const char *where, handle_t *handle)
230{
231 struct super_block *sb;
232 int err;
233 int rc;
234
235 if (!ext4_handle_valid(handle)) {
236 /*
237 * Do this here since we don't call jbd2_journal_stop() in
238 * no-journal mode.
239 */
240 current->journal_info = NULL;
241 return 0;
242 }
243 sb = handle->h_transaction->t_journal->j_private;
244 err = handle->h_err;
245 rc = jbd2_journal_stop(handle);
246
247 if (!err)
248 err = rc;
249 if (err)
250 __ext4_std_error(sb, where, err);
251 return err;
252}
253
254void ext4_journal_abort_handle(const char *caller, const char *err_fn,
255 struct buffer_head *bh, handle_t *handle, int err)
256{
257 char nbuf[16];
258 const char *errstr = ext4_decode_error(NULL, err, nbuf);
259
260 BUG_ON(!ext4_handle_valid(handle));
261
262 if (bh)
263 BUFFER_TRACE(bh, "abort");
264
265 if (!handle->h_err)
266 handle->h_err = err;
267
268 if (is_handle_aborted(handle))
269 return;
270
271 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
272 caller, errstr, err_fn);
273
274 jbd2_journal_abort_handle(handle);
275}
276
277/* Deal with the reporting of failure conditions on a filesystem such as
278 * inconsistencies detected or read IO failures.
279 *
280 * On ext2, we can store the error state of the filesystem in the
281 * superblock. That is not possible on ext4, because we may have other
282 * write ordering constraints on the superblock which prevent us from
283 * writing it out straight away; and given that the journal is about to
284 * be aborted, we can't rely on the current, or future, transactions to
285 * write out the superblock safely.
286 *
287 * We'll just use the jbd2_journal_abort() error code to record an error in
288 * the journal instead. On recovery, the journal will compain about
289 * that error until we've noted it down and cleared it.
290 */
291
292static void ext4_handle_error(struct super_block *sb)
293{
294 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
295
296 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
297 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
298
299 if (sb->s_flags & MS_RDONLY)
300 return;
301
302 if (!test_opt(sb, ERRORS_CONT)) {
303 journal_t *journal = EXT4_SB(sb)->s_journal;
304
305 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
306 if (journal)
307 jbd2_journal_abort(journal, -EIO);
308 }
309 if (test_opt(sb, ERRORS_RO)) {
310 printk(KERN_CRIT "Remounting filesystem read-only\n");
311 sb->s_flags |= MS_RDONLY;
312 }
313 ext4_commit_super(sb, 1);
314 if (test_opt(sb, ERRORS_PANIC))
315 panic("EXT4-fs (device %s): panic forced after error\n",
316 sb->s_id);
317}
318
319void ext4_error(struct super_block *sb, const char *function,
320 const char *fmt, ...)
321{
322 va_list args;
323
324 va_start(args, fmt);
325 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
326 vprintk(fmt, args);
327 printk("\n");
328 va_end(args);
329
330 ext4_handle_error(sb);
331}
332
333static const char *ext4_decode_error(struct super_block *sb, int errno,
334 char nbuf[16])
335{
336 char *errstr = NULL;
337
338 switch (errno) {
339 case -EIO:
340 errstr = "IO failure";
341 break;
342 case -ENOMEM:
343 errstr = "Out of memory";
344 break;
345 case -EROFS:
346 if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
347 errstr = "Journal has aborted";
348 else
349 errstr = "Readonly filesystem";
350 break;
351 default:
352 /* If the caller passed in an extra buffer for unknown
353 * errors, textualise them now. Else we just return
354 * NULL. */
355 if (nbuf) {
356 /* Check for truncated error codes... */
357 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
358 errstr = nbuf;
359 }
360 break;
361 }
362
363 return errstr;
364}
365
366/* __ext4_std_error decodes expected errors from journaling functions
367 * automatically and invokes the appropriate error response. */
368
369void __ext4_std_error(struct super_block *sb, const char *function, int errno)
370{
371 char nbuf[16];
372 const char *errstr;
373
374 /* Special case: if the error is EROFS, and we're not already
375 * inside a transaction, then there's really no point in logging
376 * an error. */
377 if (errno == -EROFS && journal_current_handle() == NULL &&
378 (sb->s_flags & MS_RDONLY))
379 return;
380
381 errstr = ext4_decode_error(sb, errno, nbuf);
382 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
383 sb->s_id, function, errstr);
384
385 ext4_handle_error(sb);
386}
387
388/*
389 * ext4_abort is a much stronger failure handler than ext4_error. The
390 * abort function may be used to deal with unrecoverable failures such
391 * as journal IO errors or ENOMEM at a critical moment in log management.
392 *
393 * We unconditionally force the filesystem into an ABORT|READONLY state,
394 * unless the error response on the fs has been set to panic in which
395 * case we take the easy way out and panic immediately.
396 */
397
398void ext4_abort(struct super_block *sb, const char *function,
399 const char *fmt, ...)
400{
401 va_list args;
402
403 printk(KERN_CRIT "ext4_abort called.\n");
404
405 va_start(args, fmt);
406 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
407 vprintk(fmt, args);
408 printk("\n");
409 va_end(args);
410
411 if (test_opt(sb, ERRORS_PANIC))
412 panic("EXT4-fs panic from previous error\n");
413
414 if (sb->s_flags & MS_RDONLY)
415 return;
416
417 printk(KERN_CRIT "Remounting filesystem read-only\n");
418 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
419 sb->s_flags |= MS_RDONLY;
420 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
421 if (EXT4_SB(sb)->s_journal)
422 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
423}
424
425void ext4_warning(struct super_block *sb, const char *function,
426 const char *fmt, ...)
427{
428 va_list args;
429
430 va_start(args, fmt);
431 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
432 sb->s_id, function);
433 vprintk(fmt, args);
434 printk("\n");
435 va_end(args);
436}
437
438void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp,
439 const char *function, const char *fmt, ...)
440__releases(bitlock)
441__acquires(bitlock)
442{
443 va_list args;
444 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
445
446 va_start(args, fmt);
447 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
448 vprintk(fmt, args);
449 printk("\n");
450 va_end(args);
451
452 if (test_opt(sb, ERRORS_CONT)) {
453 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
454 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
455 ext4_commit_super(sb, 0);
456 return;
457 }
458 ext4_unlock_group(sb, grp);
459 ext4_handle_error(sb);
460 /*
461 * We only get here in the ERRORS_RO case; relocking the group
462 * may be dangerous, but nothing bad will happen since the
463 * filesystem will have already been marked read/only and the
464 * journal has been aborted. We return 1 as a hint to callers
465 * who might what to use the return value from
466 * ext4_grp_locked_error() to distinguish beween the
467 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
468 * aggressively from the ext4 function in question, with a
469 * more appropriate error code.
470 */
471 ext4_lock_group(sb, grp);
472 return;
473}
474
475
476void ext4_update_dynamic_rev(struct super_block *sb)
477{
478 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
479
480 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
481 return;
482
483 ext4_warning(sb, __func__,
484 "updating to rev %d because of new feature flag, "
485 "running e2fsck is recommended",
486 EXT4_DYNAMIC_REV);
487
488 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
489 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
490 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
491 /* leave es->s_feature_*compat flags alone */
492 /* es->s_uuid will be set by e2fsck if empty */
493
494 /*
495 * The rest of the superblock fields should be zero, and if not it
496 * means they are likely already in use, so leave them alone. We
497 * can leave it up to e2fsck to clean up any inconsistencies there.
498 */
499}
500
501/*
502 * Open the external journal device
503 */
504static struct block_device *ext4_blkdev_get(dev_t dev)
505{
506 struct block_device *bdev;
507 char b[BDEVNAME_SIZE];
508
509 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
510 if (IS_ERR(bdev))
511 goto fail;
512 return bdev;
513
514fail:
515 printk(KERN_ERR "EXT4-fs: failed to open journal device %s: %ld\n",
516 __bdevname(dev, b), PTR_ERR(bdev));
517 return NULL;
518}
519
520/*
521 * Release the journal device
522 */
523static int ext4_blkdev_put(struct block_device *bdev)
524{
525 bd_release(bdev);
526 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
527}
528
529static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
530{
531 struct block_device *bdev;
532 int ret = -ENODEV;
533
534 bdev = sbi->journal_bdev;
535 if (bdev) {
536 ret = ext4_blkdev_put(bdev);
537 sbi->journal_bdev = NULL;
538 }
539 return ret;
540}
541
542static inline struct inode *orphan_list_entry(struct list_head *l)
543{
544 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
545}
546
547static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
548{
549 struct list_head *l;
550
551 printk(KERN_ERR "sb orphan head is %d\n",
552 le32_to_cpu(sbi->s_es->s_last_orphan));
553
554 printk(KERN_ERR "sb_info orphan list:\n");
555 list_for_each(l, &sbi->s_orphan) {
556 struct inode *inode = orphan_list_entry(l);
557 printk(KERN_ERR " "
558 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
559 inode->i_sb->s_id, inode->i_ino, inode,
560 inode->i_mode, inode->i_nlink,
561 NEXT_ORPHAN(inode));
562 }
563}
564
565static void ext4_put_super(struct super_block *sb)
566{
567 struct ext4_sb_info *sbi = EXT4_SB(sb);
568 struct ext4_super_block *es = sbi->s_es;
569 int i, err;
570
571 ext4_mb_release(sb);
572 ext4_ext_release(sb);
573 ext4_xattr_put_super(sb);
574 if (sbi->s_journal) {
575 err = jbd2_journal_destroy(sbi->s_journal);
576 sbi->s_journal = NULL;
577 if (err < 0)
578 ext4_abort(sb, __func__,
579 "Couldn't clean up the journal");
580 }
581 if (!(sb->s_flags & MS_RDONLY)) {
582 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
583 es->s_state = cpu_to_le16(sbi->s_mount_state);
584 ext4_commit_super(sb, 1);
585 }
586 if (sbi->s_proc) {
587 remove_proc_entry(sb->s_id, ext4_proc_root);
588 }
589 kobject_del(&sbi->s_kobj);
590
591 for (i = 0; i < sbi->s_gdb_count; i++)
592 brelse(sbi->s_group_desc[i]);
593 kfree(sbi->s_group_desc);
594 if (is_vmalloc_addr(sbi->s_flex_groups))
595 vfree(sbi->s_flex_groups);
596 else
597 kfree(sbi->s_flex_groups);
598 percpu_counter_destroy(&sbi->s_freeblocks_counter);
599 percpu_counter_destroy(&sbi->s_freeinodes_counter);
600 percpu_counter_destroy(&sbi->s_dirs_counter);
601 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
602 brelse(sbi->s_sbh);
603#ifdef CONFIG_QUOTA
604 for (i = 0; i < MAXQUOTAS; i++)
605 kfree(sbi->s_qf_names[i]);
606#endif
607
608 /* Debugging code just in case the in-memory inode orphan list
609 * isn't empty. The on-disk one can be non-empty if we've
610 * detected an error and taken the fs readonly, but the
611 * in-memory list had better be clean by this point. */
612 if (!list_empty(&sbi->s_orphan))
613 dump_orphan_list(sb, sbi);
614 J_ASSERT(list_empty(&sbi->s_orphan));
615
616 invalidate_bdev(sb->s_bdev);
617 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
618 /*
619 * Invalidate the journal device's buffers. We don't want them
620 * floating about in memory - the physical journal device may
621 * hotswapped, and it breaks the `ro-after' testing code.
622 */
623 sync_blockdev(sbi->journal_bdev);
624 invalidate_bdev(sbi->journal_bdev);
625 ext4_blkdev_remove(sbi);
626 }
627 sb->s_fs_info = NULL;
628 /*
629 * Now that we are completely done shutting down the
630 * superblock, we need to actually destroy the kobject.
631 */
632 unlock_kernel();
633 unlock_super(sb);
634 kobject_put(&sbi->s_kobj);
635 wait_for_completion(&sbi->s_kobj_unregister);
636 lock_super(sb);
637 lock_kernel();
638 kfree(sbi->s_blockgroup_lock);
639 kfree(sbi);
640 return;
641}
642
643static struct kmem_cache *ext4_inode_cachep;
644
645/*
646 * Called inside transaction, so use GFP_NOFS
647 */
648static struct inode *ext4_alloc_inode(struct super_block *sb)
649{
650 struct ext4_inode_info *ei;
651
652 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
653 if (!ei)
654 return NULL;
655#ifdef CONFIG_EXT4_FS_POSIX_ACL
656 ei->i_acl = EXT4_ACL_NOT_CACHED;
657 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
658#endif
659 ei->vfs_inode.i_version = 1;
660 ei->vfs_inode.i_data.writeback_index = 0;
661 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
662 INIT_LIST_HEAD(&ei->i_prealloc_list);
663 spin_lock_init(&ei->i_prealloc_lock);
664 /*
665 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
666 * therefore it can be null here. Don't check it, just initialize
667 * jinode.
668 */
669 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
670 ei->i_reserved_data_blocks = 0;
671 ei->i_reserved_meta_blocks = 0;
672 ei->i_allocated_meta_blocks = 0;
673 ei->i_delalloc_reserved_flag = 0;
674 spin_lock_init(&(ei->i_block_reservation_lock));
675 return &ei->vfs_inode;
676}
677
678static void ext4_destroy_inode(struct inode *inode)
679{
680 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
681 printk("EXT4 Inode %p: orphan list check failed!\n",
682 EXT4_I(inode));
683 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
684 EXT4_I(inode), sizeof(struct ext4_inode_info),
685 true);
686 dump_stack();
687 }
688 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
689}
690
691static void init_once(void *foo)
692{
693 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
694
695 INIT_LIST_HEAD(&ei->i_orphan);
696#ifdef CONFIG_EXT4_FS_XATTR
697 init_rwsem(&ei->xattr_sem);
698#endif
699 init_rwsem(&ei->i_data_sem);
700 inode_init_once(&ei->vfs_inode);
701}
702
703static int init_inodecache(void)
704{
705 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
706 sizeof(struct ext4_inode_info),
707 0, (SLAB_RECLAIM_ACCOUNT|
708 SLAB_MEM_SPREAD),
709 init_once);
710 if (ext4_inode_cachep == NULL)
711 return -ENOMEM;
712 return 0;
713}
714
715static void destroy_inodecache(void)
716{
717 kmem_cache_destroy(ext4_inode_cachep);
718}
719
720static void ext4_clear_inode(struct inode *inode)
721{
722#ifdef CONFIG_EXT4_FS_POSIX_ACL
723 if (EXT4_I(inode)->i_acl &&
724 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
725 posix_acl_release(EXT4_I(inode)->i_acl);
726 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
727 }
728 if (EXT4_I(inode)->i_default_acl &&
729 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
730 posix_acl_release(EXT4_I(inode)->i_default_acl);
731 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
732 }
733#endif
734 ext4_discard_preallocations(inode);
735 if (EXT4_JOURNAL(inode))
736 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
737 &EXT4_I(inode)->jinode);
738}
739
740static inline void ext4_show_quota_options(struct seq_file *seq,
741 struct super_block *sb)
742{
743#if defined(CONFIG_QUOTA)
744 struct ext4_sb_info *sbi = EXT4_SB(sb);
745
746 if (sbi->s_jquota_fmt)
747 seq_printf(seq, ",jqfmt=%s",
748 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
749
750 if (sbi->s_qf_names[USRQUOTA])
751 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
752
753 if (sbi->s_qf_names[GRPQUOTA])
754 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
755
756 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
757 seq_puts(seq, ",usrquota");
758
759 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
760 seq_puts(seq, ",grpquota");
761#endif
762}
763
764/*
765 * Show an option if
766 * - it's set to a non-default value OR
767 * - if the per-sb default is different from the global default
768 */
769static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
770{
771 int def_errors;
772 unsigned long def_mount_opts;
773 struct super_block *sb = vfs->mnt_sb;
774 struct ext4_sb_info *sbi = EXT4_SB(sb);
775 struct ext4_super_block *es = sbi->s_es;
776
777 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
778 def_errors = le16_to_cpu(es->s_errors);
779
780 if (sbi->s_sb_block != 1)
781 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
782 if (test_opt(sb, MINIX_DF))
783 seq_puts(seq, ",minixdf");
784 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
785 seq_puts(seq, ",grpid");
786 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
787 seq_puts(seq, ",nogrpid");
788 if (sbi->s_resuid != EXT4_DEF_RESUID ||
789 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
790 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
791 }
792 if (sbi->s_resgid != EXT4_DEF_RESGID ||
793 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
794 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
795 }
796 if (test_opt(sb, ERRORS_RO)) {
797 if (def_errors == EXT4_ERRORS_PANIC ||
798 def_errors == EXT4_ERRORS_CONTINUE) {
799 seq_puts(seq, ",errors=remount-ro");
800 }
801 }
802 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
803 seq_puts(seq, ",errors=continue");
804 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
805 seq_puts(seq, ",errors=panic");
806 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
807 seq_puts(seq, ",nouid32");
808 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
809 seq_puts(seq, ",debug");
810 if (test_opt(sb, OLDALLOC))
811 seq_puts(seq, ",oldalloc");
812#ifdef CONFIG_EXT4_FS_XATTR
813 if (test_opt(sb, XATTR_USER) &&
814 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
815 seq_puts(seq, ",user_xattr");
816 if (!test_opt(sb, XATTR_USER) &&
817 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
818 seq_puts(seq, ",nouser_xattr");
819 }
820#endif
821#ifdef CONFIG_EXT4_FS_POSIX_ACL
822 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
823 seq_puts(seq, ",acl");
824 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
825 seq_puts(seq, ",noacl");
826#endif
827 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
828 seq_printf(seq, ",commit=%u",
829 (unsigned) (sbi->s_commit_interval / HZ));
830 }
831 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
832 seq_printf(seq, ",min_batch_time=%u",
833 (unsigned) sbi->s_min_batch_time);
834 }
835 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
836 seq_printf(seq, ",max_batch_time=%u",
837 (unsigned) sbi->s_min_batch_time);
838 }
839
840 /*
841 * We're changing the default of barrier mount option, so
842 * let's always display its mount state so it's clear what its
843 * status is.
844 */
845 seq_puts(seq, ",barrier=");
846 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
847 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
848 seq_puts(seq, ",journal_async_commit");
849 if (test_opt(sb, NOBH))
850 seq_puts(seq, ",nobh");
851 if (test_opt(sb, I_VERSION))
852 seq_puts(seq, ",i_version");
853 if (!test_opt(sb, DELALLOC))
854 seq_puts(seq, ",nodelalloc");
855
856
857 if (sbi->s_stripe)
858 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
859 /*
860 * journal mode get enabled in different ways
861 * So just print the value even if we didn't specify it
862 */
863 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
864 seq_puts(seq, ",data=journal");
865 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
866 seq_puts(seq, ",data=ordered");
867 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
868 seq_puts(seq, ",data=writeback");
869
870 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
871 seq_printf(seq, ",inode_readahead_blks=%u",
872 sbi->s_inode_readahead_blks);
873
874 if (test_opt(sb, DATA_ERR_ABORT))
875 seq_puts(seq, ",data_err=abort");
876
877 if (test_opt(sb, NO_AUTO_DA_ALLOC))
878 seq_puts(seq, ",noauto_da_alloc");
879
880 ext4_show_quota_options(seq, sb);
881 return 0;
882}
883
884
885static struct inode *ext4_nfs_get_inode(struct super_block *sb,
886 u64 ino, u32 generation)
887{
888 struct inode *inode;
889
890 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
891 return ERR_PTR(-ESTALE);
892 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
893 return ERR_PTR(-ESTALE);
894
895 /* iget isn't really right if the inode is currently unallocated!!
896 *
897 * ext4_read_inode will return a bad_inode if the inode had been
898 * deleted, so we should be safe.
899 *
900 * Currently we don't know the generation for parent directory, so
901 * a generation of 0 means "accept any"
902 */
903 inode = ext4_iget(sb, ino);
904 if (IS_ERR(inode))
905 return ERR_CAST(inode);
906 if (generation && inode->i_generation != generation) {
907 iput(inode);
908 return ERR_PTR(-ESTALE);
909 }
910
911 return inode;
912}
913
914static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
915 int fh_len, int fh_type)
916{
917 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
918 ext4_nfs_get_inode);
919}
920
921static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
922 int fh_len, int fh_type)
923{
924 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
925 ext4_nfs_get_inode);
926}
927
928/*
929 * Try to release metadata pages (indirect blocks, directories) which are
930 * mapped via the block device. Since these pages could have journal heads
931 * which would prevent try_to_free_buffers() from freeing them, we must use
932 * jbd2 layer's try_to_free_buffers() function to release them.
933 */
934static int bdev_try_to_free_page(struct super_block *sb, struct page *page, gfp_t wait)
935{
936 journal_t *journal = EXT4_SB(sb)->s_journal;
937
938 WARN_ON(PageChecked(page));
939 if (!page_has_buffers(page))
940 return 0;
941 if (journal)
942 return jbd2_journal_try_to_free_buffers(journal, page,
943 wait & ~__GFP_WAIT);
944 return try_to_free_buffers(page);
945}
946
947#ifdef CONFIG_QUOTA
948#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
949#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
950
951static int ext4_write_dquot(struct dquot *dquot);
952static int ext4_acquire_dquot(struct dquot *dquot);
953static int ext4_release_dquot(struct dquot *dquot);
954static int ext4_mark_dquot_dirty(struct dquot *dquot);
955static int ext4_write_info(struct super_block *sb, int type);
956static int ext4_quota_on(struct super_block *sb, int type, int format_id,
957 char *path, int remount);
958static int ext4_quota_on_mount(struct super_block *sb, int type);
959static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
960 size_t len, loff_t off);
961static ssize_t ext4_quota_write(struct super_block *sb, int type,
962 const char *data, size_t len, loff_t off);
963
964static struct dquot_operations ext4_quota_operations = {
965 .initialize = dquot_initialize,
966 .drop = dquot_drop,
967 .alloc_space = dquot_alloc_space,
968 .reserve_space = dquot_reserve_space,
969 .claim_space = dquot_claim_space,
970 .release_rsv = dquot_release_reserved_space,
971 .get_reserved_space = ext4_get_reserved_space,
972 .alloc_inode = dquot_alloc_inode,
973 .free_space = dquot_free_space,
974 .free_inode = dquot_free_inode,
975 .transfer = dquot_transfer,
976 .write_dquot = ext4_write_dquot,
977 .acquire_dquot = ext4_acquire_dquot,
978 .release_dquot = ext4_release_dquot,
979 .mark_dirty = ext4_mark_dquot_dirty,
980 .write_info = ext4_write_info,
981 .alloc_dquot = dquot_alloc,
982 .destroy_dquot = dquot_destroy,
983};
984
985static struct quotactl_ops ext4_qctl_operations = {
986 .quota_on = ext4_quota_on,
987 .quota_off = vfs_quota_off,
988 .quota_sync = vfs_quota_sync,
989 .get_info = vfs_get_dqinfo,
990 .set_info = vfs_set_dqinfo,
991 .get_dqblk = vfs_get_dqblk,
992 .set_dqblk = vfs_set_dqblk
993};
994#endif
995
996static const struct super_operations ext4_sops = {
997 .alloc_inode = ext4_alloc_inode,
998 .destroy_inode = ext4_destroy_inode,
999 .write_inode = ext4_write_inode,
1000 .dirty_inode = ext4_dirty_inode,
1001 .delete_inode = ext4_delete_inode,
1002 .put_super = ext4_put_super,
1003 .sync_fs = ext4_sync_fs,
1004 .freeze_fs = ext4_freeze,
1005 .unfreeze_fs = ext4_unfreeze,
1006 .statfs = ext4_statfs,
1007 .remount_fs = ext4_remount,
1008 .clear_inode = ext4_clear_inode,
1009 .show_options = ext4_show_options,
1010#ifdef CONFIG_QUOTA
1011 .quota_read = ext4_quota_read,
1012 .quota_write = ext4_quota_write,
1013#endif
1014 .bdev_try_to_free_page = bdev_try_to_free_page,
1015};
1016
1017static const struct super_operations ext4_nojournal_sops = {
1018 .alloc_inode = ext4_alloc_inode,
1019 .destroy_inode = ext4_destroy_inode,
1020 .write_inode = ext4_write_inode,
1021 .dirty_inode = ext4_dirty_inode,
1022 .delete_inode = ext4_delete_inode,
1023 .write_super = ext4_write_super,
1024 .put_super = ext4_put_super,
1025 .statfs = ext4_statfs,
1026 .remount_fs = ext4_remount,
1027 .clear_inode = ext4_clear_inode,
1028 .show_options = ext4_show_options,
1029#ifdef CONFIG_QUOTA
1030 .quota_read = ext4_quota_read,
1031 .quota_write = ext4_quota_write,
1032#endif
1033 .bdev_try_to_free_page = bdev_try_to_free_page,
1034};
1035
1036static const struct export_operations ext4_export_ops = {
1037 .fh_to_dentry = ext4_fh_to_dentry,
1038 .fh_to_parent = ext4_fh_to_parent,
1039 .get_parent = ext4_get_parent,
1040};
1041
1042enum {
1043 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1044 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1045 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1046 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1047 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1048 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1049 Opt_journal_update, Opt_journal_dev,
1050 Opt_journal_checksum, Opt_journal_async_commit,
1051 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1052 Opt_data_err_abort, Opt_data_err_ignore, Opt_mb_history_length,
1053 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1054 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
1055 Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err, Opt_resize,
1056 Opt_usrquota, Opt_grpquota, Opt_i_version,
1057 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1058 Opt_inode_readahead_blks, Opt_journal_ioprio
1059};
1060
1061static const match_table_t tokens = {
1062 {Opt_bsd_df, "bsddf"},
1063 {Opt_minix_df, "minixdf"},
1064 {Opt_grpid, "grpid"},
1065 {Opt_grpid, "bsdgroups"},
1066 {Opt_nogrpid, "nogrpid"},
1067 {Opt_nogrpid, "sysvgroups"},
1068 {Opt_resgid, "resgid=%u"},
1069 {Opt_resuid, "resuid=%u"},
1070 {Opt_sb, "sb=%u"},
1071 {Opt_err_cont, "errors=continue"},
1072 {Opt_err_panic, "errors=panic"},
1073 {Opt_err_ro, "errors=remount-ro"},
1074 {Opt_nouid32, "nouid32"},
1075 {Opt_debug, "debug"},
1076 {Opt_oldalloc, "oldalloc"},
1077 {Opt_orlov, "orlov"},
1078 {Opt_user_xattr, "user_xattr"},
1079 {Opt_nouser_xattr, "nouser_xattr"},
1080 {Opt_acl, "acl"},
1081 {Opt_noacl, "noacl"},
1082 {Opt_noload, "noload"},
1083 {Opt_nobh, "nobh"},
1084 {Opt_bh, "bh"},
1085 {Opt_commit, "commit=%u"},
1086 {Opt_min_batch_time, "min_batch_time=%u"},
1087 {Opt_max_batch_time, "max_batch_time=%u"},
1088 {Opt_journal_update, "journal=update"},
1089 {Opt_journal_dev, "journal_dev=%u"},
1090 {Opt_journal_checksum, "journal_checksum"},
1091 {Opt_journal_async_commit, "journal_async_commit"},
1092 {Opt_abort, "abort"},
1093 {Opt_data_journal, "data=journal"},
1094 {Opt_data_ordered, "data=ordered"},
1095 {Opt_data_writeback, "data=writeback"},
1096 {Opt_data_err_abort, "data_err=abort"},
1097 {Opt_data_err_ignore, "data_err=ignore"},
1098 {Opt_mb_history_length, "mb_history_length=%u"},
1099 {Opt_offusrjquota, "usrjquota="},
1100 {Opt_usrjquota, "usrjquota=%s"},
1101 {Opt_offgrpjquota, "grpjquota="},
1102 {Opt_grpjquota, "grpjquota=%s"},
1103 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1104 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1105 {Opt_grpquota, "grpquota"},
1106 {Opt_noquota, "noquota"},
1107 {Opt_quota, "quota"},
1108 {Opt_usrquota, "usrquota"},
1109 {Opt_barrier, "barrier=%u"},
1110 {Opt_barrier, "barrier"},
1111 {Opt_nobarrier, "nobarrier"},
1112 {Opt_i_version, "i_version"},
1113 {Opt_stripe, "stripe=%u"},
1114 {Opt_resize, "resize"},
1115 {Opt_delalloc, "delalloc"},
1116 {Opt_nodelalloc, "nodelalloc"},
1117 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1118 {Opt_journal_ioprio, "journal_ioprio=%u"},
1119 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1120 {Opt_auto_da_alloc, "auto_da_alloc"},
1121 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1122 {Opt_err, NULL},
1123};
1124
1125static ext4_fsblk_t get_sb_block(void **data)
1126{
1127 ext4_fsblk_t sb_block;
1128 char *options = (char *) *data;
1129
1130 if (!options || strncmp(options, "sb=", 3) != 0)
1131 return 1; /* Default location */
1132 options += 3;
1133 /*todo: use simple_strtoll with >32bit ext4 */
1134 sb_block = simple_strtoul(options, &options, 0);
1135 if (*options && *options != ',') {
1136 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1137 (char *) *data);
1138 return 1;
1139 }
1140 if (*options == ',')
1141 options++;
1142 *data = (void *) options;
1143 return sb_block;
1144}
1145
1146#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1147
1148static int parse_options(char *options, struct super_block *sb,
1149 unsigned long *journal_devnum,
1150 unsigned int *journal_ioprio,
1151 ext4_fsblk_t *n_blocks_count, int is_remount)
1152{
1153 struct ext4_sb_info *sbi = EXT4_SB(sb);
1154 char *p;
1155 substring_t args[MAX_OPT_ARGS];
1156 int data_opt = 0;
1157 int option;
1158#ifdef CONFIG_QUOTA
1159 int qtype, qfmt;
1160 char *qname;
1161#endif
1162
1163 if (!options)
1164 return 1;
1165
1166 while ((p = strsep(&options, ",")) != NULL) {
1167 int token;
1168 if (!*p)
1169 continue;
1170
1171 token = match_token(p, tokens, args);
1172 switch (token) {
1173 case Opt_bsd_df:
1174 clear_opt(sbi->s_mount_opt, MINIX_DF);
1175 break;
1176 case Opt_minix_df:
1177 set_opt(sbi->s_mount_opt, MINIX_DF);
1178 break;
1179 case Opt_grpid:
1180 set_opt(sbi->s_mount_opt, GRPID);
1181 break;
1182 case Opt_nogrpid:
1183 clear_opt(sbi->s_mount_opt, GRPID);
1184 break;
1185 case Opt_resuid:
1186 if (match_int(&args[0], &option))
1187 return 0;
1188 sbi->s_resuid = option;
1189 break;
1190 case Opt_resgid:
1191 if (match_int(&args[0], &option))
1192 return 0;
1193 sbi->s_resgid = option;
1194 break;
1195 case Opt_sb:
1196 /* handled by get_sb_block() instead of here */
1197 /* *sb_block = match_int(&args[0]); */
1198 break;
1199 case Opt_err_panic:
1200 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1201 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1202 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1203 break;
1204 case Opt_err_ro:
1205 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1206 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1207 set_opt(sbi->s_mount_opt, ERRORS_RO);
1208 break;
1209 case Opt_err_cont:
1210 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1211 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1212 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1213 break;
1214 case Opt_nouid32:
1215 set_opt(sbi->s_mount_opt, NO_UID32);
1216 break;
1217 case Opt_debug:
1218 set_opt(sbi->s_mount_opt, DEBUG);
1219 break;
1220 case Opt_oldalloc:
1221 set_opt(sbi->s_mount_opt, OLDALLOC);
1222 break;
1223 case Opt_orlov:
1224 clear_opt(sbi->s_mount_opt, OLDALLOC);
1225 break;
1226#ifdef CONFIG_EXT4_FS_XATTR
1227 case Opt_user_xattr:
1228 set_opt(sbi->s_mount_opt, XATTR_USER);
1229 break;
1230 case Opt_nouser_xattr:
1231 clear_opt(sbi->s_mount_opt, XATTR_USER);
1232 break;
1233#else
1234 case Opt_user_xattr:
1235 case Opt_nouser_xattr:
1236 printk(KERN_ERR "EXT4 (no)user_xattr options "
1237 "not supported\n");
1238 break;
1239#endif
1240#ifdef CONFIG_EXT4_FS_POSIX_ACL
1241 case Opt_acl:
1242 set_opt(sbi->s_mount_opt, POSIX_ACL);
1243 break;
1244 case Opt_noacl:
1245 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1246 break;
1247#else
1248 case Opt_acl:
1249 case Opt_noacl:
1250 printk(KERN_ERR "EXT4 (no)acl options "
1251 "not supported\n");
1252 break;
1253#endif
1254 case Opt_journal_update:
1255 /* @@@ FIXME */
1256 /* Eventually we will want to be able to create
1257 a journal file here. For now, only allow the
1258 user to specify an existing inode to be the
1259 journal file. */
1260 if (is_remount) {
1261 printk(KERN_ERR "EXT4-fs: cannot specify "
1262 "journal on remount\n");
1263 return 0;
1264 }
1265 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1266 break;
1267 case Opt_journal_dev:
1268 if (is_remount) {
1269 printk(KERN_ERR "EXT4-fs: cannot specify "
1270 "journal on remount\n");
1271 return 0;
1272 }
1273 if (match_int(&args[0], &option))
1274 return 0;
1275 *journal_devnum = option;
1276 break;
1277 case Opt_journal_checksum:
1278 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1279 break;
1280 case Opt_journal_async_commit:
1281 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1282 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1283 break;
1284 case Opt_noload:
1285 set_opt(sbi->s_mount_opt, NOLOAD);
1286 break;
1287 case Opt_commit:
1288 if (match_int(&args[0], &option))
1289 return 0;
1290 if (option < 0)
1291 return 0;
1292 if (option == 0)
1293 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1294 sbi->s_commit_interval = HZ * option;
1295 break;
1296 case Opt_max_batch_time:
1297 if (match_int(&args[0], &option))
1298 return 0;
1299 if (option < 0)
1300 return 0;
1301 if (option == 0)
1302 option = EXT4_DEF_MAX_BATCH_TIME;
1303 sbi->s_max_batch_time = option;
1304 break;
1305 case Opt_min_batch_time:
1306 if (match_int(&args[0], &option))
1307 return 0;
1308 if (option < 0)
1309 return 0;
1310 sbi->s_min_batch_time = option;
1311 break;
1312 case Opt_data_journal:
1313 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1314 goto datacheck;
1315 case Opt_data_ordered:
1316 data_opt = EXT4_MOUNT_ORDERED_DATA;
1317 goto datacheck;
1318 case Opt_data_writeback:
1319 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1320 datacheck:
1321 if (is_remount) {
1322 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1323 != data_opt) {
1324 printk(KERN_ERR
1325 "EXT4-fs: cannot change data "
1326 "mode on remount\n");
1327 return 0;
1328 }
1329 } else {
1330 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1331 sbi->s_mount_opt |= data_opt;
1332 }
1333 break;
1334 case Opt_data_err_abort:
1335 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1336 break;
1337 case Opt_data_err_ignore:
1338 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1339 break;
1340 case Opt_mb_history_length:
1341 if (match_int(&args[0], &option))
1342 return 0;
1343 if (option < 0)
1344 return 0;
1345 sbi->s_mb_history_max = option;
1346 break;
1347#ifdef CONFIG_QUOTA
1348 case Opt_usrjquota:
1349 qtype = USRQUOTA;
1350 goto set_qf_name;
1351 case Opt_grpjquota:
1352 qtype = GRPQUOTA;
1353set_qf_name:
1354 if (sb_any_quota_loaded(sb) &&
1355 !sbi->s_qf_names[qtype]) {
1356 printk(KERN_ERR
1357 "EXT4-fs: Cannot change journaled "
1358 "quota options when quota turned on.\n");
1359 return 0;
1360 }
1361 qname = match_strdup(&args[0]);
1362 if (!qname) {
1363 printk(KERN_ERR
1364 "EXT4-fs: not enough memory for "
1365 "storing quotafile name.\n");
1366 return 0;
1367 }
1368 if (sbi->s_qf_names[qtype] &&
1369 strcmp(sbi->s_qf_names[qtype], qname)) {
1370 printk(KERN_ERR
1371 "EXT4-fs: %s quota file already "
1372 "specified.\n", QTYPE2NAME(qtype));
1373 kfree(qname);
1374 return 0;
1375 }
1376 sbi->s_qf_names[qtype] = qname;
1377 if (strchr(sbi->s_qf_names[qtype], '/')) {
1378 printk(KERN_ERR
1379 "EXT4-fs: quotafile must be on "
1380 "filesystem root.\n");
1381 kfree(sbi->s_qf_names[qtype]);
1382 sbi->s_qf_names[qtype] = NULL;
1383 return 0;
1384 }
1385 set_opt(sbi->s_mount_opt, QUOTA);
1386 break;
1387 case Opt_offusrjquota:
1388 qtype = USRQUOTA;
1389 goto clear_qf_name;
1390 case Opt_offgrpjquota:
1391 qtype = GRPQUOTA;
1392clear_qf_name:
1393 if (sb_any_quota_loaded(sb) &&
1394 sbi->s_qf_names[qtype]) {
1395 printk(KERN_ERR "EXT4-fs: Cannot change "
1396 "journaled quota options when "
1397 "quota turned on.\n");
1398 return 0;
1399 }
1400 /*
1401 * The space will be released later when all options
1402 * are confirmed to be correct
1403 */
1404 sbi->s_qf_names[qtype] = NULL;
1405 break;
1406 case Opt_jqfmt_vfsold:
1407 qfmt = QFMT_VFS_OLD;
1408 goto set_qf_format;
1409 case Opt_jqfmt_vfsv0:
1410 qfmt = QFMT_VFS_V0;
1411set_qf_format:
1412 if (sb_any_quota_loaded(sb) &&
1413 sbi->s_jquota_fmt != qfmt) {
1414 printk(KERN_ERR "EXT4-fs: Cannot change "
1415 "journaled quota options when "
1416 "quota turned on.\n");
1417 return 0;
1418 }
1419 sbi->s_jquota_fmt = qfmt;
1420 break;
1421 case Opt_quota:
1422 case Opt_usrquota:
1423 set_opt(sbi->s_mount_opt, QUOTA);
1424 set_opt(sbi->s_mount_opt, USRQUOTA);
1425 break;
1426 case Opt_grpquota:
1427 set_opt(sbi->s_mount_opt, QUOTA);
1428 set_opt(sbi->s_mount_opt, GRPQUOTA);
1429 break;
1430 case Opt_noquota:
1431 if (sb_any_quota_loaded(sb)) {
1432 printk(KERN_ERR "EXT4-fs: Cannot change quota "
1433 "options when quota turned on.\n");
1434 return 0;
1435 }
1436 clear_opt(sbi->s_mount_opt, QUOTA);
1437 clear_opt(sbi->s_mount_opt, USRQUOTA);
1438 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1439 break;
1440#else
1441 case Opt_quota:
1442 case Opt_usrquota:
1443 case Opt_grpquota:
1444 printk(KERN_ERR
1445 "EXT4-fs: quota options not supported.\n");
1446 break;
1447 case Opt_usrjquota:
1448 case Opt_grpjquota:
1449 case Opt_offusrjquota:
1450 case Opt_offgrpjquota:
1451 case Opt_jqfmt_vfsold:
1452 case Opt_jqfmt_vfsv0:
1453 printk(KERN_ERR
1454 "EXT4-fs: journaled quota options not "
1455 "supported.\n");
1456 break;
1457 case Opt_noquota:
1458 break;
1459#endif
1460 case Opt_abort:
1461 set_opt(sbi->s_mount_opt, ABORT);
1462 break;
1463 case Opt_nobarrier:
1464 clear_opt(sbi->s_mount_opt, BARRIER);
1465 break;
1466 case Opt_barrier:
1467 if (match_int(&args[0], &option)) {
1468 set_opt(sbi->s_mount_opt, BARRIER);
1469 break;
1470 }
1471 if (option)
1472 set_opt(sbi->s_mount_opt, BARRIER);
1473 else
1474 clear_opt(sbi->s_mount_opt, BARRIER);
1475 break;
1476 case Opt_ignore:
1477 break;
1478 case Opt_resize:
1479 if (!is_remount) {
1480 printk("EXT4-fs: resize option only available "
1481 "for remount\n");
1482 return 0;
1483 }
1484 if (match_int(&args[0], &option) != 0)
1485 return 0;
1486 *n_blocks_count = option;
1487 break;
1488 case Opt_nobh:
1489 set_opt(sbi->s_mount_opt, NOBH);
1490 break;
1491 case Opt_bh:
1492 clear_opt(sbi->s_mount_opt, NOBH);
1493 break;
1494 case Opt_i_version:
1495 set_opt(sbi->s_mount_opt, I_VERSION);
1496 sb->s_flags |= MS_I_VERSION;
1497 break;
1498 case Opt_nodelalloc:
1499 clear_opt(sbi->s_mount_opt, DELALLOC);
1500 break;
1501 case Opt_stripe:
1502 if (match_int(&args[0], &option))
1503 return 0;
1504 if (option < 0)
1505 return 0;
1506 sbi->s_stripe = option;
1507 break;
1508 case Opt_delalloc:
1509 set_opt(sbi->s_mount_opt, DELALLOC);
1510 break;
1511 case Opt_inode_readahead_blks:
1512 if (match_int(&args[0], &option))
1513 return 0;
1514 if (option < 0 || option > (1 << 30))
1515 return 0;
1516 if (!is_power_of_2(option)) {
1517 printk(KERN_ERR "EXT4-fs: inode_readahead_blks"
1518 " must be a power of 2\n");
1519 return 0;
1520 }
1521 sbi->s_inode_readahead_blks = option;
1522 break;
1523 case Opt_journal_ioprio:
1524 if (match_int(&args[0], &option))
1525 return 0;
1526 if (option < 0 || option > 7)
1527 break;
1528 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1529 option);
1530 break;
1531 case Opt_noauto_da_alloc:
1532 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1533 break;
1534 case Opt_auto_da_alloc:
1535 if (match_int(&args[0], &option)) {
1536 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1537 break;
1538 }
1539 if (option)
1540 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1541 else
1542 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1543 break;
1544 default:
1545 printk(KERN_ERR
1546 "EXT4-fs: Unrecognized mount option \"%s\" "
1547 "or missing value\n", p);
1548 return 0;
1549 }
1550 }
1551#ifdef CONFIG_QUOTA
1552 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1553 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1554 sbi->s_qf_names[USRQUOTA])
1555 clear_opt(sbi->s_mount_opt, USRQUOTA);
1556
1557 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1558 sbi->s_qf_names[GRPQUOTA])
1559 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1560
1561 if ((sbi->s_qf_names[USRQUOTA] &&
1562 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1563 (sbi->s_qf_names[GRPQUOTA] &&
1564 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1565 printk(KERN_ERR "EXT4-fs: old and new quota "
1566 "format mixing.\n");
1567 return 0;
1568 }
1569
1570 if (!sbi->s_jquota_fmt) {
1571 printk(KERN_ERR "EXT4-fs: journaled quota format "
1572 "not specified.\n");
1573 return 0;
1574 }
1575 } else {
1576 if (sbi->s_jquota_fmt) {
1577 printk(KERN_ERR "EXT4-fs: journaled quota format "
1578 "specified with no journaling "
1579 "enabled.\n");
1580 return 0;
1581 }
1582 }
1583#endif
1584 return 1;
1585}
1586
1587static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1588 int read_only)
1589{
1590 struct ext4_sb_info *sbi = EXT4_SB(sb);
1591 int res = 0;
1592
1593 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1594 printk(KERN_ERR "EXT4-fs warning: revision level too high, "
1595 "forcing read-only mode\n");
1596 res = MS_RDONLY;
1597 }
1598 if (read_only)
1599 return res;
1600 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1601 printk(KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1602 "running e2fsck is recommended\n");
1603 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1604 printk(KERN_WARNING
1605 "EXT4-fs warning: mounting fs with errors, "
1606 "running e2fsck is recommended\n");
1607 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1608 le16_to_cpu(es->s_mnt_count) >=
1609 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1610 printk(KERN_WARNING
1611 "EXT4-fs warning: maximal mount count reached, "
1612 "running e2fsck is recommended\n");
1613 else if (le32_to_cpu(es->s_checkinterval) &&
1614 (le32_to_cpu(es->s_lastcheck) +
1615 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1616 printk(KERN_WARNING
1617 "EXT4-fs warning: checktime reached, "
1618 "running e2fsck is recommended\n");
1619 if (!sbi->s_journal)
1620 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1621 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1622 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1623 le16_add_cpu(&es->s_mnt_count, 1);
1624 es->s_mtime = cpu_to_le32(get_seconds());
1625 ext4_update_dynamic_rev(sb);
1626 if (sbi->s_journal)
1627 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1628
1629 ext4_commit_super(sb, 1);
1630 if (test_opt(sb, DEBUG))
1631 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1632 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1633 sb->s_blocksize,
1634 sbi->s_groups_count,
1635 EXT4_BLOCKS_PER_GROUP(sb),
1636 EXT4_INODES_PER_GROUP(sb),
1637 sbi->s_mount_opt);
1638
1639 if (EXT4_SB(sb)->s_journal) {
1640 printk(KERN_INFO "EXT4 FS on %s, %s journal on %s\n",
1641 sb->s_id, EXT4_SB(sb)->s_journal->j_inode ? "internal" :
1642 "external", EXT4_SB(sb)->s_journal->j_devname);
1643 } else {
1644 printk(KERN_INFO "EXT4 FS on %s, no journal\n", sb->s_id);
1645 }
1646 return res;
1647}
1648
1649static int ext4_fill_flex_info(struct super_block *sb)
1650{
1651 struct ext4_sb_info *sbi = EXT4_SB(sb);
1652 struct ext4_group_desc *gdp = NULL;
1653 struct buffer_head *bh;
1654 ext4_group_t flex_group_count;
1655 ext4_group_t flex_group;
1656 int groups_per_flex = 0;
1657 size_t size;
1658 int i;
1659
1660 if (!sbi->s_es->s_log_groups_per_flex) {
1661 sbi->s_log_groups_per_flex = 0;
1662 return 1;
1663 }
1664
1665 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1666 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1667
1668 /* We allocate both existing and potentially added groups */
1669 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1670 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1671 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1672 size = flex_group_count * sizeof(struct flex_groups);
1673 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1674 if (sbi->s_flex_groups == NULL) {
1675 sbi->s_flex_groups = vmalloc(size);
1676 if (sbi->s_flex_groups)
1677 memset(sbi->s_flex_groups, 0, size);
1678 }
1679 if (sbi->s_flex_groups == NULL) {
1680 printk(KERN_ERR "EXT4-fs: not enough memory for "
1681 "%u flex groups\n", flex_group_count);
1682 goto failed;
1683 }
1684
1685 for (i = 0; i < sbi->s_groups_count; i++) {
1686 gdp = ext4_get_group_desc(sb, i, &bh);
1687
1688 flex_group = ext4_flex_group(sbi, i);
1689 atomic_set(&sbi->s_flex_groups[flex_group].free_inodes,
1690 ext4_free_inodes_count(sb, gdp));
1691 atomic_set(&sbi->s_flex_groups[flex_group].free_blocks,
1692 ext4_free_blks_count(sb, gdp));
1693 atomic_set(&sbi->s_flex_groups[flex_group].used_dirs,
1694 ext4_used_dirs_count(sb, gdp));
1695 }
1696
1697 return 1;
1698failed:
1699 return 0;
1700}
1701
1702__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1703 struct ext4_group_desc *gdp)
1704{
1705 __u16 crc = 0;
1706
1707 if (sbi->s_es->s_feature_ro_compat &
1708 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1709 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1710 __le32 le_group = cpu_to_le32(block_group);
1711
1712 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1713 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1714 crc = crc16(crc, (__u8 *)gdp, offset);
1715 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1716 /* for checksum of struct ext4_group_desc do the rest...*/
1717 if ((sbi->s_es->s_feature_incompat &
1718 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1719 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1720 crc = crc16(crc, (__u8 *)gdp + offset,
1721 le16_to_cpu(sbi->s_es->s_desc_size) -
1722 offset);
1723 }
1724
1725 return cpu_to_le16(crc);
1726}
1727
1728int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1729 struct ext4_group_desc *gdp)
1730{
1731 if ((sbi->s_es->s_feature_ro_compat &
1732 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1733 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1734 return 0;
1735
1736 return 1;
1737}
1738
1739/* Called at mount-time, super-block is locked */
1740static int ext4_check_descriptors(struct super_block *sb)
1741{
1742 struct ext4_sb_info *sbi = EXT4_SB(sb);
1743 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1744 ext4_fsblk_t last_block;
1745 ext4_fsblk_t block_bitmap;
1746 ext4_fsblk_t inode_bitmap;
1747 ext4_fsblk_t inode_table;
1748 int flexbg_flag = 0;
1749 ext4_group_t i;
1750
1751 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1752 flexbg_flag = 1;
1753
1754 ext4_debug("Checking group descriptors");
1755
1756 for (i = 0; i < sbi->s_groups_count; i++) {
1757 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1758
1759 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1760 last_block = ext4_blocks_count(sbi->s_es) - 1;
1761 else
1762 last_block = first_block +
1763 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1764
1765 block_bitmap = ext4_block_bitmap(sb, gdp);
1766 if (block_bitmap < first_block || block_bitmap > last_block) {
1767 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1768 "Block bitmap for group %u not in group "
1769 "(block %llu)!\n", i, block_bitmap);
1770 return 0;
1771 }
1772 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1773 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1774 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1775 "Inode bitmap for group %u not in group "
1776 "(block %llu)!\n", i, inode_bitmap);
1777 return 0;
1778 }
1779 inode_table = ext4_inode_table(sb, gdp);
1780 if (inode_table < first_block ||
1781 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1782 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1783 "Inode table for group %u not in group "
1784 "(block %llu)!\n", i, inode_table);
1785 return 0;
1786 }
1787 ext4_lock_group(sb, i);
1788 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1789 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1790 "Checksum for group %u failed (%u!=%u)\n",
1791 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1792 gdp)), le16_to_cpu(gdp->bg_checksum));
1793 if (!(sb->s_flags & MS_RDONLY)) {
1794 ext4_unlock_group(sb, i);
1795 return 0;
1796 }
1797 }
1798 ext4_unlock_group(sb, i);
1799 if (!flexbg_flag)
1800 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1801 }
1802
1803 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1804 sbi->s_es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
1805 return 1;
1806}
1807
1808/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1809 * the superblock) which were deleted from all directories, but held open by
1810 * a process at the time of a crash. We walk the list and try to delete these
1811 * inodes at recovery time (only with a read-write filesystem).
1812 *
1813 * In order to keep the orphan inode chain consistent during traversal (in
1814 * case of crash during recovery), we link each inode into the superblock
1815 * orphan list_head and handle it the same way as an inode deletion during
1816 * normal operation (which journals the operations for us).
1817 *
1818 * We only do an iget() and an iput() on each inode, which is very safe if we
1819 * accidentally point at an in-use or already deleted inode. The worst that
1820 * can happen in this case is that we get a "bit already cleared" message from
1821 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1822 * e2fsck was run on this filesystem, and it must have already done the orphan
1823 * inode cleanup for us, so we can safely abort without any further action.
1824 */
1825static void ext4_orphan_cleanup(struct super_block *sb,
1826 struct ext4_super_block *es)
1827{
1828 unsigned int s_flags = sb->s_flags;
1829 int nr_orphans = 0, nr_truncates = 0;
1830#ifdef CONFIG_QUOTA
1831 int i;
1832#endif
1833 if (!es->s_last_orphan) {
1834 jbd_debug(4, "no orphan inodes to clean up\n");
1835 return;
1836 }
1837
1838 if (bdev_read_only(sb->s_bdev)) {
1839 printk(KERN_ERR "EXT4-fs: write access "
1840 "unavailable, skipping orphan cleanup.\n");
1841 return;
1842 }
1843
1844 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1845 if (es->s_last_orphan)
1846 jbd_debug(1, "Errors on filesystem, "
1847 "clearing orphan list.\n");
1848 es->s_last_orphan = 0;
1849 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1850 return;
1851 }
1852
1853 if (s_flags & MS_RDONLY) {
1854 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
1855 sb->s_id);
1856 sb->s_flags &= ~MS_RDONLY;
1857 }
1858#ifdef CONFIG_QUOTA
1859 /* Needed for iput() to work correctly and not trash data */
1860 sb->s_flags |= MS_ACTIVE;
1861 /* Turn on quotas so that they are updated correctly */
1862 for (i = 0; i < MAXQUOTAS; i++) {
1863 if (EXT4_SB(sb)->s_qf_names[i]) {
1864 int ret = ext4_quota_on_mount(sb, i);
1865 if (ret < 0)
1866 printk(KERN_ERR
1867 "EXT4-fs: Cannot turn on journaled "
1868 "quota: error %d\n", ret);
1869 }
1870 }
1871#endif
1872
1873 while (es->s_last_orphan) {
1874 struct inode *inode;
1875
1876 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1877 if (IS_ERR(inode)) {
1878 es->s_last_orphan = 0;
1879 break;
1880 }
1881
1882 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1883 vfs_dq_init(inode);
1884 if (inode->i_nlink) {
1885 printk(KERN_DEBUG
1886 "%s: truncating inode %lu to %lld bytes\n",
1887 __func__, inode->i_ino, inode->i_size);
1888 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1889 inode->i_ino, inode->i_size);
1890 ext4_truncate(inode);
1891 nr_truncates++;
1892 } else {
1893 printk(KERN_DEBUG
1894 "%s: deleting unreferenced inode %lu\n",
1895 __func__, inode->i_ino);
1896 jbd_debug(2, "deleting unreferenced inode %lu\n",
1897 inode->i_ino);
1898 nr_orphans++;
1899 }
1900 iput(inode); /* The delete magic happens here! */
1901 }
1902
1903#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1904
1905 if (nr_orphans)
1906 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
1907 sb->s_id, PLURAL(nr_orphans));
1908 if (nr_truncates)
1909 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
1910 sb->s_id, PLURAL(nr_truncates));
1911#ifdef CONFIG_QUOTA
1912 /* Turn quotas off */
1913 for (i = 0; i < MAXQUOTAS; i++) {
1914 if (sb_dqopt(sb)->files[i])
1915 vfs_quota_off(sb, i, 0);
1916 }
1917#endif
1918 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1919}
1920/*
1921 * Maximal extent format file size.
1922 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1923 * extent format containers, within a sector_t, and within i_blocks
1924 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1925 * so that won't be a limiting factor.
1926 *
1927 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1928 */
1929static loff_t ext4_max_size(int blkbits, int has_huge_files)
1930{
1931 loff_t res;
1932 loff_t upper_limit = MAX_LFS_FILESIZE;
1933
1934 /* small i_blocks in vfs inode? */
1935 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1936 /*
1937 * CONFIG_LBD is not enabled implies the inode
1938 * i_block represent total blocks in 512 bytes
1939 * 32 == size of vfs inode i_blocks * 8
1940 */
1941 upper_limit = (1LL << 32) - 1;
1942
1943 /* total blocks in file system block size */
1944 upper_limit >>= (blkbits - 9);
1945 upper_limit <<= blkbits;
1946 }
1947
1948 /* 32-bit extent-start container, ee_block */
1949 res = 1LL << 32;
1950 res <<= blkbits;
1951 res -= 1;
1952
1953 /* Sanity check against vm- & vfs- imposed limits */
1954 if (res > upper_limit)
1955 res = upper_limit;
1956
1957 return res;
1958}
1959
1960/*
1961 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1962 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1963 * We need to be 1 filesystem block less than the 2^48 sector limit.
1964 */
1965static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
1966{
1967 loff_t res = EXT4_NDIR_BLOCKS;
1968 int meta_blocks;
1969 loff_t upper_limit;
1970 /* This is calculated to be the largest file size for a
1971 * dense, bitmapped file such that the total number of
1972 * sectors in the file, including data and all indirect blocks,
1973 * does not exceed 2^48 -1
1974 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1975 * total number of 512 bytes blocks of the file
1976 */
1977
1978 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1979 /*
1980 * !has_huge_files or CONFIG_LBD is not enabled
1981 * implies the inode i_block represent total blocks in
1982 * 512 bytes 32 == size of vfs inode i_blocks * 8
1983 */
1984 upper_limit = (1LL << 32) - 1;
1985
1986 /* total blocks in file system block size */
1987 upper_limit >>= (bits - 9);
1988
1989 } else {
1990 /*
1991 * We use 48 bit ext4_inode i_blocks
1992 * With EXT4_HUGE_FILE_FL set the i_blocks
1993 * represent total number of blocks in
1994 * file system block size
1995 */
1996 upper_limit = (1LL << 48) - 1;
1997
1998 }
1999
2000 /* indirect blocks */
2001 meta_blocks = 1;
2002 /* double indirect blocks */
2003 meta_blocks += 1 + (1LL << (bits-2));
2004 /* tripple indirect blocks */
2005 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2006
2007 upper_limit -= meta_blocks;
2008 upper_limit <<= bits;
2009
2010 res += 1LL << (bits-2);
2011 res += 1LL << (2*(bits-2));
2012 res += 1LL << (3*(bits-2));
2013 res <<= bits;
2014 if (res > upper_limit)
2015 res = upper_limit;
2016
2017 if (res > MAX_LFS_FILESIZE)
2018 res = MAX_LFS_FILESIZE;
2019
2020 return res;
2021}
2022
2023static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2024 ext4_fsblk_t logical_sb_block, int nr)
2025{
2026 struct ext4_sb_info *sbi = EXT4_SB(sb);
2027 ext4_group_t bg, first_meta_bg;
2028 int has_super = 0;
2029
2030 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2031
2032 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2033 nr < first_meta_bg)
2034 return logical_sb_block + nr + 1;
2035 bg = sbi->s_desc_per_block * nr;
2036 if (ext4_bg_has_super(sb, bg))
2037 has_super = 1;
2038 return (has_super + ext4_group_first_block_no(sb, bg));
2039}
2040
2041/**
2042 * ext4_get_stripe_size: Get the stripe size.
2043 * @sbi: In memory super block info
2044 *
2045 * If we have specified it via mount option, then
2046 * use the mount option value. If the value specified at mount time is
2047 * greater than the blocks per group use the super block value.
2048 * If the super block value is greater than blocks per group return 0.
2049 * Allocator needs it be less than blocks per group.
2050 *
2051 */
2052static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2053{
2054 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2055 unsigned long stripe_width =
2056 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2057
2058 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2059 return sbi->s_stripe;
2060
2061 if (stripe_width <= sbi->s_blocks_per_group)
2062 return stripe_width;
2063
2064 if (stride <= sbi->s_blocks_per_group)
2065 return stride;
2066
2067 return 0;
2068}
2069
2070/* sysfs supprt */
2071
2072struct ext4_attr {
2073 struct attribute attr;
2074 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2075 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2076 const char *, size_t);
2077 int offset;
2078};
2079
2080static int parse_strtoul(const char *buf,
2081 unsigned long max, unsigned long *value)
2082{
2083 char *endp;
2084
2085 while (*buf && isspace(*buf))
2086 buf++;
2087 *value = simple_strtoul(buf, &endp, 0);
2088 while (*endp && isspace(*endp))
2089 endp++;
2090 if (*endp || *value > max)
2091 return -EINVAL;
2092
2093 return 0;
2094}
2095
2096static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2097 struct ext4_sb_info *sbi,
2098 char *buf)
2099{
2100 return snprintf(buf, PAGE_SIZE, "%llu\n",
2101 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2102}
2103
2104static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2105 struct ext4_sb_info *sbi, char *buf)
2106{
2107 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2108
2109 return snprintf(buf, PAGE_SIZE, "%lu\n",
2110 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2111 sbi->s_sectors_written_start) >> 1);
2112}
2113
2114static ssize_t lifetime_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, "%llu\n",
2120 sbi->s_kbytes_written +
2121 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2122 EXT4_SB(sb)->s_sectors_written_start) >> 1));
2123}
2124
2125static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2126 struct ext4_sb_info *sbi,
2127 const char *buf, size_t count)
2128{
2129 unsigned long t;
2130
2131 if (parse_strtoul(buf, 0x40000000, &t))
2132 return -EINVAL;
2133
2134 if (!is_power_of_2(t))
2135 return -EINVAL;
2136
2137 sbi->s_inode_readahead_blks = t;
2138 return count;
2139}
2140
2141static ssize_t sbi_ui_show(struct ext4_attr *a,
2142 struct ext4_sb_info *sbi, char *buf)
2143{
2144 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2145
2146 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2147}
2148
2149static ssize_t sbi_ui_store(struct ext4_attr *a,
2150 struct ext4_sb_info *sbi,
2151 const char *buf, size_t count)
2152{
2153 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2154 unsigned long t;
2155
2156 if (parse_strtoul(buf, 0xffffffff, &t))
2157 return -EINVAL;
2158 *ui = t;
2159 return count;
2160}
2161
2162#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2163static struct ext4_attr ext4_attr_##_name = { \
2164 .attr = {.name = __stringify(_name), .mode = _mode }, \
2165 .show = _show, \
2166 .store = _store, \
2167 .offset = offsetof(struct ext4_sb_info, _elname), \
2168}
2169#define EXT4_ATTR(name, mode, show, store) \
2170static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2171
2172#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2173#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2174#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2175 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2176#define ATTR_LIST(name) &ext4_attr_##name.attr
2177
2178EXT4_RO_ATTR(delayed_allocation_blocks);
2179EXT4_RO_ATTR(session_write_kbytes);
2180EXT4_RO_ATTR(lifetime_write_kbytes);
2181EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2182 inode_readahead_blks_store, s_inode_readahead_blks);
2183EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2184EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2185EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2186EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2187EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2188EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2189
2190static struct attribute *ext4_attrs[] = {
2191 ATTR_LIST(delayed_allocation_blocks),
2192 ATTR_LIST(session_write_kbytes),
2193 ATTR_LIST(lifetime_write_kbytes),
2194 ATTR_LIST(inode_readahead_blks),
2195 ATTR_LIST(mb_stats),
2196 ATTR_LIST(mb_max_to_scan),
2197 ATTR_LIST(mb_min_to_scan),
2198 ATTR_LIST(mb_order2_req),
2199 ATTR_LIST(mb_stream_req),
2200 ATTR_LIST(mb_group_prealloc),
2201 NULL,
2202};
2203
2204static ssize_t ext4_attr_show(struct kobject *kobj,
2205 struct attribute *attr, char *buf)
2206{
2207 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2208 s_kobj);
2209 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2210
2211 return a->show ? a->show(a, sbi, buf) : 0;
2212}
2213
2214static ssize_t ext4_attr_store(struct kobject *kobj,
2215 struct attribute *attr,
2216 const char *buf, size_t len)
2217{
2218 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2219 s_kobj);
2220 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2221
2222 return a->store ? a->store(a, sbi, buf, len) : 0;
2223}
2224
2225static void ext4_sb_release(struct kobject *kobj)
2226{
2227 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2228 s_kobj);
2229 complete(&sbi->s_kobj_unregister);
2230}
2231
2232
2233static struct sysfs_ops ext4_attr_ops = {
2234 .show = ext4_attr_show,
2235 .store = ext4_attr_store,
2236};
2237
2238static struct kobj_type ext4_ktype = {
2239 .default_attrs = ext4_attrs,
2240 .sysfs_ops = &ext4_attr_ops,
2241 .release = ext4_sb_release,
2242};
2243
2244static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2245 __releases(kernel_lock)
2246 __acquires(kernel_lock)
2247
2248{
2249 struct buffer_head *bh;
2250 struct ext4_super_block *es = NULL;
2251 struct ext4_sb_info *sbi;
2252 ext4_fsblk_t block;
2253 ext4_fsblk_t sb_block = get_sb_block(&data);
2254 ext4_fsblk_t logical_sb_block;
2255 unsigned long offset = 0;
2256 unsigned long journal_devnum = 0;
2257 unsigned long def_mount_opts;
2258 struct inode *root;
2259 char *cp;
2260 const char *descr;
2261 int ret = -EINVAL;
2262 int blocksize;
2263 unsigned int db_count;
2264 unsigned int i;
2265 int needs_recovery, has_huge_files;
2266 int features;
2267 __u64 blocks_count;
2268 int err;
2269 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2270
2271 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2272 if (!sbi)
2273 return -ENOMEM;
2274
2275 sbi->s_blockgroup_lock =
2276 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2277 if (!sbi->s_blockgroup_lock) {
2278 kfree(sbi);
2279 return -ENOMEM;
2280 }
2281 sb->s_fs_info = sbi;
2282 sbi->s_mount_opt = 0;
2283 sbi->s_resuid = EXT4_DEF_RESUID;
2284 sbi->s_resgid = EXT4_DEF_RESGID;
2285 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2286 sbi->s_sb_block = sb_block;
2287 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2288 sectors[1]);
2289
2290 unlock_kernel();
2291
2292 /* Cleanup superblock name */
2293 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2294 *cp = '!';
2295
2296 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2297 if (!blocksize) {
2298 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
2299 goto out_fail;
2300 }
2301
2302 /*
2303 * The ext4 superblock will not be buffer aligned for other than 1kB
2304 * block sizes. We need to calculate the offset from buffer start.
2305 */
2306 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2307 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2308 offset = do_div(logical_sb_block, blocksize);
2309 } else {
2310 logical_sb_block = sb_block;
2311 }
2312
2313 if (!(bh = sb_bread(sb, logical_sb_block))) {
2314 printk(KERN_ERR "EXT4-fs: unable to read superblock\n");
2315 goto out_fail;
2316 }
2317 /*
2318 * Note: s_es must be initialized as soon as possible because
2319 * some ext4 macro-instructions depend on its value
2320 */
2321 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2322 sbi->s_es = es;
2323 sb->s_magic = le16_to_cpu(es->s_magic);
2324 if (sb->s_magic != EXT4_SUPER_MAGIC)
2325 goto cantfind_ext4;
2326 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2327
2328 /* Set defaults before we parse the mount options */
2329 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2330 if (def_mount_opts & EXT4_DEFM_DEBUG)
2331 set_opt(sbi->s_mount_opt, DEBUG);
2332 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
2333 set_opt(sbi->s_mount_opt, GRPID);
2334 if (def_mount_opts & EXT4_DEFM_UID16)
2335 set_opt(sbi->s_mount_opt, NO_UID32);
2336#ifdef CONFIG_EXT4_FS_XATTR
2337 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2338 set_opt(sbi->s_mount_opt, XATTR_USER);
2339#endif
2340#ifdef CONFIG_EXT4_FS_POSIX_ACL
2341 if (def_mount_opts & EXT4_DEFM_ACL)
2342 set_opt(sbi->s_mount_opt, POSIX_ACL);
2343#endif
2344 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2345 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2346 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2347 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2348 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2349 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2350
2351 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2352 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2353 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2354 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2355 else
2356 set_opt(sbi->s_mount_opt, ERRORS_RO);
2357
2358 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2359 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2360 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2361 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2362 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2363 sbi->s_mb_history_max = default_mb_history_length;
2364
2365 set_opt(sbi->s_mount_opt, BARRIER);
2366
2367 /*
2368 * enable delayed allocation by default
2369 * Use -o nodelalloc to turn it off
2370 */
2371 set_opt(sbi->s_mount_opt, DELALLOC);
2372
2373
2374 if (!parse_options((char *) data, sb, &journal_devnum,
2375 &journal_ioprio, NULL, 0))
2376 goto failed_mount;
2377
2378 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2379 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2380
2381 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2382 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2383 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2384 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2385 printk(KERN_WARNING
2386 "EXT4-fs warning: feature flags set on rev 0 fs, "
2387 "running e2fsck is recommended\n");
2388
2389 /*
2390 * Check feature flags regardless of the revision level, since we
2391 * previously didn't change the revision level when setting the flags,
2392 * so there is a chance incompat flags are set on a rev 0 filesystem.
2393 */
2394 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
2395 if (features) {
2396 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
2397 "unsupported optional features (%x).\n", sb->s_id,
2398 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2399 ~EXT4_FEATURE_INCOMPAT_SUPP));
2400 goto failed_mount;
2401 }
2402 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
2403 if (!(sb->s_flags & MS_RDONLY) && features) {
2404 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
2405 "unsupported optional features (%x).\n", sb->s_id,
2406 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2407 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2408 goto failed_mount;
2409 }
2410 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2411 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2412 if (has_huge_files) {
2413 /*
2414 * Large file size enabled file system can only be
2415 * mount if kernel is build with CONFIG_LBD
2416 */
2417 if (sizeof(root->i_blocks) < sizeof(u64) &&
2418 !(sb->s_flags & MS_RDONLY)) {
2419 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2420 "files cannot be mounted read-write "
2421 "without CONFIG_LBD.\n", sb->s_id);
2422 goto failed_mount;
2423 }
2424 }
2425 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2426
2427 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2428 blocksize > EXT4_MAX_BLOCK_SIZE) {
2429 printk(KERN_ERR
2430 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
2431 blocksize, sb->s_id);
2432 goto failed_mount;
2433 }
2434
2435 if (sb->s_blocksize != blocksize) {
2436
2437 /* Validate the filesystem blocksize */
2438 if (!sb_set_blocksize(sb, blocksize)) {
2439 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2440 blocksize);
2441 goto failed_mount;
2442 }
2443
2444 brelse(bh);
2445 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2446 offset = do_div(logical_sb_block, blocksize);
2447 bh = sb_bread(sb, logical_sb_block);
2448 if (!bh) {
2449 printk(KERN_ERR
2450 "EXT4-fs: Can't read superblock on 2nd try.\n");
2451 goto failed_mount;
2452 }
2453 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2454 sbi->s_es = es;
2455 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2456 printk(KERN_ERR
2457 "EXT4-fs: Magic mismatch, very weird !\n");
2458 goto failed_mount;
2459 }
2460 }
2461
2462 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2463 has_huge_files);
2464 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2465
2466 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2467 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2468 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2469 } else {
2470 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2471 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2472 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2473 (!is_power_of_2(sbi->s_inode_size)) ||
2474 (sbi->s_inode_size > blocksize)) {
2475 printk(KERN_ERR
2476 "EXT4-fs: unsupported inode size: %d\n",
2477 sbi->s_inode_size);
2478 goto failed_mount;
2479 }
2480 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2481 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2482 }
2483 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2484 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2485 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2486 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2487 !is_power_of_2(sbi->s_desc_size)) {
2488 printk(KERN_ERR
2489 "EXT4-fs: unsupported descriptor size %lu\n",
2490 sbi->s_desc_size);
2491 goto failed_mount;
2492 }
2493 } else
2494 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2495 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2496 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2497 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2498 goto cantfind_ext4;
2499 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2500 if (sbi->s_inodes_per_block == 0)
2501 goto cantfind_ext4;
2502 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2503 sbi->s_inodes_per_block;
2504 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2505 sbi->s_sbh = bh;
2506 sbi->s_mount_state = le16_to_cpu(es->s_state);
2507 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2508 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2509 for (i = 0; i < 4; i++)
2510 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2511 sbi->s_def_hash_version = es->s_def_hash_version;
2512 i = le32_to_cpu(es->s_flags);
2513 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2514 sbi->s_hash_unsigned = 3;
2515 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2516#ifdef __CHAR_UNSIGNED__
2517 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2518 sbi->s_hash_unsigned = 3;
2519#else
2520 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2521#endif
2522 sb->s_dirt = 1;
2523 }
2524
2525 if (sbi->s_blocks_per_group > blocksize * 8) {
2526 printk(KERN_ERR
2527 "EXT4-fs: #blocks per group too big: %lu\n",
2528 sbi->s_blocks_per_group);
2529 goto failed_mount;
2530 }
2531 if (sbi->s_inodes_per_group > blocksize * 8) {
2532 printk(KERN_ERR
2533 "EXT4-fs: #inodes per group too big: %lu\n",
2534 sbi->s_inodes_per_group);
2535 goto failed_mount;
2536 }
2537
2538 if (ext4_blocks_count(es) >
2539 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
2540 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
2541 " too large to mount safely\n", sb->s_id);
2542 if (sizeof(sector_t) < 8)
2543 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
2544 "enabled\n");
2545 goto failed_mount;
2546 }
2547
2548 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2549 goto cantfind_ext4;
2550
2551 /* check blocks count against device size */
2552 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2553 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2554 printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu "
2555 "exceeds size of device (%llu blocks)\n",
2556 ext4_blocks_count(es), blocks_count);
2557 goto failed_mount;
2558 }
2559
2560 /*
2561 * It makes no sense for the first data block to be beyond the end
2562 * of the filesystem.
2563 */
2564 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2565 printk(KERN_WARNING "EXT4-fs: bad geometry: first data"
2566 "block %u is beyond end of filesystem (%llu)\n",
2567 le32_to_cpu(es->s_first_data_block),
2568 ext4_blocks_count(es));
2569 goto failed_mount;
2570 }
2571 blocks_count = (ext4_blocks_count(es) -
2572 le32_to_cpu(es->s_first_data_block) +
2573 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2574 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2575 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2576 printk(KERN_WARNING "EXT4-fs: groups count too large: %u "
2577 "(block count %llu, first data block %u, "
2578 "blocks per group %lu)\n", sbi->s_groups_count,
2579 ext4_blocks_count(es),
2580 le32_to_cpu(es->s_first_data_block),
2581 EXT4_BLOCKS_PER_GROUP(sb));
2582 goto failed_mount;
2583 }
2584 sbi->s_groups_count = blocks_count;
2585 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2586 EXT4_DESC_PER_BLOCK(sb);
2587 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2588 GFP_KERNEL);
2589 if (sbi->s_group_desc == NULL) {
2590 printk(KERN_ERR "EXT4-fs: not enough memory\n");
2591 goto failed_mount;
2592 }
2593
2594#ifdef CONFIG_PROC_FS
2595 if (ext4_proc_root)
2596 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2597#endif
2598
2599 bgl_lock_init(sbi->s_blockgroup_lock);
2600
2601 for (i = 0; i < db_count; i++) {
2602 block = descriptor_loc(sb, logical_sb_block, i);
2603 sbi->s_group_desc[i] = sb_bread(sb, block);
2604 if (!sbi->s_group_desc[i]) {
2605 printk(KERN_ERR "EXT4-fs: "
2606 "can't read group descriptor %d\n", i);
2607 db_count = i;
2608 goto failed_mount2;
2609 }
2610 }
2611 if (!ext4_check_descriptors(sb)) {
2612 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
2613 goto failed_mount2;
2614 }
2615 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2616 if (!ext4_fill_flex_info(sb)) {
2617 printk(KERN_ERR
2618 "EXT4-fs: unable to initialize "
2619 "flex_bg meta info!\n");
2620 goto failed_mount2;
2621 }
2622
2623 sbi->s_gdb_count = db_count;
2624 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2625 spin_lock_init(&sbi->s_next_gen_lock);
2626
2627 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2628 ext4_count_free_blocks(sb));
2629 if (!err) {
2630 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2631 ext4_count_free_inodes(sb));
2632 }
2633 if (!err) {
2634 err = percpu_counter_init(&sbi->s_dirs_counter,
2635 ext4_count_dirs(sb));
2636 }
2637 if (!err) {
2638 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2639 }
2640 if (err) {
2641 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2642 goto failed_mount3;
2643 }
2644
2645 sbi->s_stripe = ext4_get_stripe_size(sbi);
2646
2647 /*
2648 * set up enough so that it can read an inode
2649 */
2650 if (!test_opt(sb, NOLOAD) &&
2651 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2652 sb->s_op = &ext4_sops;
2653 else
2654 sb->s_op = &ext4_nojournal_sops;
2655 sb->s_export_op = &ext4_export_ops;
2656 sb->s_xattr = ext4_xattr_handlers;
2657#ifdef CONFIG_QUOTA
2658 sb->s_qcop = &ext4_qctl_operations;
2659 sb->dq_op = &ext4_quota_operations;
2660#endif
2661 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2662 mutex_init(&sbi->s_orphan_lock);
2663 mutex_init(&sbi->s_resize_lock);
2664
2665 sb->s_root = NULL;
2666
2667 needs_recovery = (es->s_last_orphan != 0 ||
2668 EXT4_HAS_INCOMPAT_FEATURE(sb,
2669 EXT4_FEATURE_INCOMPAT_RECOVER));
2670
2671 /*
2672 * The first inode we look at is the journal inode. Don't try
2673 * root first: it may be modified in the journal!
2674 */
2675 if (!test_opt(sb, NOLOAD) &&
2676 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2677 if (ext4_load_journal(sb, es, journal_devnum))
2678 goto failed_mount3;
2679 if (!(sb->s_flags & MS_RDONLY) &&
2680 EXT4_SB(sb)->s_journal->j_failed_commit) {
2681 printk(KERN_CRIT "EXT4-fs error (device %s): "
2682 "ext4_fill_super: Journal transaction "
2683 "%u is corrupt\n", sb->s_id,
2684 EXT4_SB(sb)->s_journal->j_failed_commit);
2685 if (test_opt(sb, ERRORS_RO)) {
2686 printk(KERN_CRIT
2687 "Mounting filesystem read-only\n");
2688 sb->s_flags |= MS_RDONLY;
2689 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2690 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2691 }
2692 if (test_opt(sb, ERRORS_PANIC)) {
2693 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2694 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2695 ext4_commit_super(sb, 1);
2696 goto failed_mount4;
2697 }
2698 }
2699 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2700 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2701 printk(KERN_ERR "EXT4-fs: required journal recovery "
2702 "suppressed and not mounted read-only\n");
2703 goto failed_mount4;
2704 } else {
2705 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2706 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2707 sbi->s_journal = NULL;
2708 needs_recovery = 0;
2709 goto no_journal;
2710 }
2711
2712 if (ext4_blocks_count(es) > 0xffffffffULL &&
2713 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2714 JBD2_FEATURE_INCOMPAT_64BIT)) {
2715 printk(KERN_ERR "EXT4-fs: Failed to set 64-bit journal feature\n");
2716 goto failed_mount4;
2717 }
2718
2719 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2720 jbd2_journal_set_features(sbi->s_journal,
2721 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2722 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2723 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2724 jbd2_journal_set_features(sbi->s_journal,
2725 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2726 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2727 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2728 } else {
2729 jbd2_journal_clear_features(sbi->s_journal,
2730 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2731 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2732 }
2733
2734 /* We have now updated the journal if required, so we can
2735 * validate the data journaling mode. */
2736 switch (test_opt(sb, DATA_FLAGS)) {
2737 case 0:
2738 /* No mode set, assume a default based on the journal
2739 * capabilities: ORDERED_DATA if the journal can
2740 * cope, else JOURNAL_DATA
2741 */
2742 if (jbd2_journal_check_available_features
2743 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2744 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2745 else
2746 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2747 break;
2748
2749 case EXT4_MOUNT_ORDERED_DATA:
2750 case EXT4_MOUNT_WRITEBACK_DATA:
2751 if (!jbd2_journal_check_available_features
2752 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2753 printk(KERN_ERR "EXT4-fs: Journal does not support "
2754 "requested data journaling mode\n");
2755 goto failed_mount4;
2756 }
2757 default:
2758 break;
2759 }
2760 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2761
2762no_journal:
2763
2764 if (test_opt(sb, NOBH)) {
2765 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2766 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
2767 "its supported only with writeback mode\n");
2768 clear_opt(sbi->s_mount_opt, NOBH);
2769 }
2770 }
2771 /*
2772 * The jbd2_journal_load will have done any necessary log recovery,
2773 * so we can safely mount the rest of the filesystem now.
2774 */
2775
2776 root = ext4_iget(sb, EXT4_ROOT_INO);
2777 if (IS_ERR(root)) {
2778 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
2779 ret = PTR_ERR(root);
2780 goto failed_mount4;
2781 }
2782 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2783 iput(root);
2784 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
2785 goto failed_mount4;
2786 }
2787 sb->s_root = d_alloc_root(root);
2788 if (!sb->s_root) {
2789 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2790 iput(root);
2791 ret = -ENOMEM;
2792 goto failed_mount4;
2793 }
2794
2795 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2796
2797 /* determine the minimum size of new large inodes, if present */
2798 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2799 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2800 EXT4_GOOD_OLD_INODE_SIZE;
2801 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2802 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2803 if (sbi->s_want_extra_isize <
2804 le16_to_cpu(es->s_want_extra_isize))
2805 sbi->s_want_extra_isize =
2806 le16_to_cpu(es->s_want_extra_isize);
2807 if (sbi->s_want_extra_isize <
2808 le16_to_cpu(es->s_min_extra_isize))
2809 sbi->s_want_extra_isize =
2810 le16_to_cpu(es->s_min_extra_isize);
2811 }
2812 }
2813 /* Check if enough inode space is available */
2814 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2815 sbi->s_inode_size) {
2816 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2817 EXT4_GOOD_OLD_INODE_SIZE;
2818 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2819 "available.\n");
2820 }
2821
2822 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2823 printk(KERN_WARNING "EXT4-fs: Ignoring delalloc option - "
2824 "requested data journaling mode\n");
2825 clear_opt(sbi->s_mount_opt, DELALLOC);
2826 } else if (test_opt(sb, DELALLOC))
2827 printk(KERN_INFO "EXT4-fs: delayed allocation enabled\n");
2828
2829 ext4_ext_init(sb);
2830 err = ext4_mb_init(sb, needs_recovery);
2831 if (err) {
2832 printk(KERN_ERR "EXT4-fs: failed to initalize mballoc (%d)\n",
2833 err);
2834 goto failed_mount4;
2835 }
2836
2837 sbi->s_kobj.kset = ext4_kset;
2838 init_completion(&sbi->s_kobj_unregister);
2839 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2840 "%s", sb->s_id);
2841 if (err) {
2842 ext4_mb_release(sb);
2843 ext4_ext_release(sb);
2844 goto failed_mount4;
2845 };
2846
2847 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2848 ext4_orphan_cleanup(sb, es);
2849 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2850 if (needs_recovery) {
2851 printk(KERN_INFO "EXT4-fs: recovery complete.\n");
2852 ext4_mark_recovery_complete(sb, es);
2853 }
2854 if (EXT4_SB(sb)->s_journal) {
2855 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2856 descr = " journalled data mode";
2857 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2858 descr = " ordered data mode";
2859 else
2860 descr = " writeback data mode";
2861 } else
2862 descr = "out journal";
2863
2864 printk(KERN_INFO "EXT4-fs: mounted filesystem %s with%s\n",
2865 sb->s_id, descr);
2866
2867 lock_kernel();
2868 return 0;
2869
2870cantfind_ext4:
2871 if (!silent)
2872 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
2873 sb->s_id);
2874 goto failed_mount;
2875
2876failed_mount4:
2877 printk(KERN_ERR "EXT4-fs (device %s): mount failed\n", sb->s_id);
2878 if (sbi->s_journal) {
2879 jbd2_journal_destroy(sbi->s_journal);
2880 sbi->s_journal = NULL;
2881 }
2882failed_mount3:
2883 if (sbi->s_flex_groups) {
2884 if (is_vmalloc_addr(sbi->s_flex_groups))
2885 vfree(sbi->s_flex_groups);
2886 else
2887 kfree(sbi->s_flex_groups);
2888 }
2889 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2890 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2891 percpu_counter_destroy(&sbi->s_dirs_counter);
2892 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2893failed_mount2:
2894 for (i = 0; i < db_count; i++)
2895 brelse(sbi->s_group_desc[i]);
2896 kfree(sbi->s_group_desc);
2897failed_mount:
2898 if (sbi->s_proc) {
2899 remove_proc_entry(sb->s_id, ext4_proc_root);
2900 }
2901#ifdef CONFIG_QUOTA
2902 for (i = 0; i < MAXQUOTAS; i++)
2903 kfree(sbi->s_qf_names[i]);
2904#endif
2905 ext4_blkdev_remove(sbi);
2906 brelse(bh);
2907out_fail:
2908 sb->s_fs_info = NULL;
2909 kfree(sbi);
2910 lock_kernel();
2911 return ret;
2912}
2913
2914/*
2915 * Setup any per-fs journal parameters now. We'll do this both on
2916 * initial mount, once the journal has been initialised but before we've
2917 * done any recovery; and again on any subsequent remount.
2918 */
2919static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2920{
2921 struct ext4_sb_info *sbi = EXT4_SB(sb);
2922
2923 journal->j_commit_interval = sbi->s_commit_interval;
2924 journal->j_min_batch_time = sbi->s_min_batch_time;
2925 journal->j_max_batch_time = sbi->s_max_batch_time;
2926
2927 spin_lock(&journal->j_state_lock);
2928 if (test_opt(sb, BARRIER))
2929 journal->j_flags |= JBD2_BARRIER;
2930 else
2931 journal->j_flags &= ~JBD2_BARRIER;
2932 if (test_opt(sb, DATA_ERR_ABORT))
2933 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2934 else
2935 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
2936 spin_unlock(&journal->j_state_lock);
2937}
2938
2939static journal_t *ext4_get_journal(struct super_block *sb,
2940 unsigned int journal_inum)
2941{
2942 struct inode *journal_inode;
2943 journal_t *journal;
2944
2945 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2946
2947 /* First, test for the existence of a valid inode on disk. Bad
2948 * things happen if we iget() an unused inode, as the subsequent
2949 * iput() will try to delete it. */
2950
2951 journal_inode = ext4_iget(sb, journal_inum);
2952 if (IS_ERR(journal_inode)) {
2953 printk(KERN_ERR "EXT4-fs: no journal found.\n");
2954 return NULL;
2955 }
2956 if (!journal_inode->i_nlink) {
2957 make_bad_inode(journal_inode);
2958 iput(journal_inode);
2959 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
2960 return NULL;
2961 }
2962
2963 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
2964 journal_inode, journal_inode->i_size);
2965 if (!S_ISREG(journal_inode->i_mode)) {
2966 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
2967 iput(journal_inode);
2968 return NULL;
2969 }
2970
2971 journal = jbd2_journal_init_inode(journal_inode);
2972 if (!journal) {
2973 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
2974 iput(journal_inode);
2975 return NULL;
2976 }
2977 journal->j_private = sb;
2978 ext4_init_journal_params(sb, journal);
2979 return journal;
2980}
2981
2982static journal_t *ext4_get_dev_journal(struct super_block *sb,
2983 dev_t j_dev)
2984{
2985 struct buffer_head *bh;
2986 journal_t *journal;
2987 ext4_fsblk_t start;
2988 ext4_fsblk_t len;
2989 int hblock, blocksize;
2990 ext4_fsblk_t sb_block;
2991 unsigned long offset;
2992 struct ext4_super_block *es;
2993 struct block_device *bdev;
2994
2995 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2996
2997 bdev = ext4_blkdev_get(j_dev);
2998 if (bdev == NULL)
2999 return NULL;
3000
3001 if (bd_claim(bdev, sb)) {
3002 printk(KERN_ERR
3003 "EXT4-fs: failed to claim external journal device.\n");
3004 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3005 return NULL;
3006 }
3007
3008 blocksize = sb->s_blocksize;
3009 hblock = bdev_hardsect_size(bdev);
3010 if (blocksize < hblock) {
3011 printk(KERN_ERR
3012 "EXT4-fs: blocksize too small for journal device.\n");
3013 goto out_bdev;
3014 }
3015
3016 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3017 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3018 set_blocksize(bdev, blocksize);
3019 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3020 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
3021 "external journal\n");
3022 goto out_bdev;
3023 }
3024
3025 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3026 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3027 !(le32_to_cpu(es->s_feature_incompat) &
3028 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3029 printk(KERN_ERR "EXT4-fs: external journal has "
3030 "bad superblock\n");
3031 brelse(bh);
3032 goto out_bdev;
3033 }
3034
3035 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3036 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
3037 brelse(bh);
3038 goto out_bdev;
3039 }
3040
3041 len = ext4_blocks_count(es);
3042 start = sb_block + 1;
3043 brelse(bh); /* we're done with the superblock */
3044
3045 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3046 start, len, blocksize);
3047 if (!journal) {
3048 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
3049 goto out_bdev;
3050 }
3051 journal->j_private = sb;
3052 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3053 wait_on_buffer(journal->j_sb_buffer);
3054 if (!buffer_uptodate(journal->j_sb_buffer)) {
3055 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
3056 goto out_journal;
3057 }
3058 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3059 printk(KERN_ERR "EXT4-fs: External journal has more than one "
3060 "user (unsupported) - %d\n",
3061 be32_to_cpu(journal->j_superblock->s_nr_users));
3062 goto out_journal;
3063 }
3064 EXT4_SB(sb)->journal_bdev = bdev;
3065 ext4_init_journal_params(sb, journal);
3066 return journal;
3067out_journal:
3068 jbd2_journal_destroy(journal);
3069out_bdev:
3070 ext4_blkdev_put(bdev);
3071 return NULL;
3072}
3073
3074static int ext4_load_journal(struct super_block *sb,
3075 struct ext4_super_block *es,
3076 unsigned long journal_devnum)
3077{
3078 journal_t *journal;
3079 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3080 dev_t journal_dev;
3081 int err = 0;
3082 int really_read_only;
3083
3084 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3085
3086 if (journal_devnum &&
3087 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3088 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
3089 "numbers have changed\n");
3090 journal_dev = new_decode_dev(journal_devnum);
3091 } else
3092 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3093
3094 really_read_only = bdev_read_only(sb->s_bdev);
3095
3096 /*
3097 * Are we loading a blank journal or performing recovery after a
3098 * crash? For recovery, we need to check in advance whether we
3099 * can get read-write access to the device.
3100 */
3101
3102 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3103 if (sb->s_flags & MS_RDONLY) {
3104 printk(KERN_INFO "EXT4-fs: INFO: recovery "
3105 "required on readonly filesystem.\n");
3106 if (really_read_only) {
3107 printk(KERN_ERR "EXT4-fs: write access "
3108 "unavailable, cannot proceed.\n");
3109 return -EROFS;
3110 }
3111 printk(KERN_INFO "EXT4-fs: write access will "
3112 "be enabled during recovery.\n");
3113 }
3114 }
3115
3116 if (journal_inum && journal_dev) {
3117 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
3118 "and inode journals!\n");
3119 return -EINVAL;
3120 }
3121
3122 if (journal_inum) {
3123 if (!(journal = ext4_get_journal(sb, journal_inum)))
3124 return -EINVAL;
3125 } else {
3126 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3127 return -EINVAL;
3128 }
3129
3130 if (journal->j_flags & JBD2_BARRIER)
3131 printk(KERN_INFO "EXT4-fs: barriers enabled\n");
3132 else
3133 printk(KERN_INFO "EXT4-fs: barriers disabled\n");
3134
3135 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3136 err = jbd2_journal_update_format(journal);
3137 if (err) {
3138 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
3139 jbd2_journal_destroy(journal);
3140 return err;
3141 }
3142 }
3143
3144 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3145 err = jbd2_journal_wipe(journal, !really_read_only);
3146 if (!err)
3147 err = jbd2_journal_load(journal);
3148
3149 if (err) {
3150 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
3151 jbd2_journal_destroy(journal);
3152 return err;
3153 }
3154
3155 EXT4_SB(sb)->s_journal = journal;
3156 ext4_clear_journal_err(sb, es);
3157
3158 if (journal_devnum &&
3159 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3160 es->s_journal_dev = cpu_to_le32(journal_devnum);
3161
3162 /* Make sure we flush the recovery flag to disk. */
3163 ext4_commit_super(sb, 1);
3164 }
3165
3166 return 0;
3167}
3168
3169static int ext4_commit_super(struct super_block *sb, int sync)
3170{
3171 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3172 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3173 int error = 0;
3174
3175 if (!sbh)
3176 return error;
3177 if (buffer_write_io_error(sbh)) {
3178 /*
3179 * Oh, dear. A previous attempt to write the
3180 * superblock failed. This could happen because the
3181 * USB device was yanked out. Or it could happen to
3182 * be a transient write error and maybe the block will
3183 * be remapped. Nothing we can do but to retry the
3184 * write and hope for the best.
3185 */
3186 printk(KERN_ERR "EXT4-fs: previous I/O error to "
3187 "superblock detected for %s.\n", sb->s_id);
3188 clear_buffer_write_io_error(sbh);
3189 set_buffer_uptodate(sbh);
3190 }
3191 es->s_wtime = cpu_to_le32(get_seconds());
3192 es->s_kbytes_written =
3193 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3194 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3195 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3196 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3197 &EXT4_SB(sb)->s_freeblocks_counter));
3198 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3199 &EXT4_SB(sb)->s_freeinodes_counter));
3200 sb->s_dirt = 0;
3201 BUFFER_TRACE(sbh, "marking dirty");
3202 mark_buffer_dirty(sbh);
3203 if (sync) {
3204 error = sync_dirty_buffer(sbh);
3205 if (error)
3206 return error;
3207
3208 error = buffer_write_io_error(sbh);
3209 if (error) {
3210 printk(KERN_ERR "EXT4-fs: I/O error while writing "
3211 "superblock for %s.\n", sb->s_id);
3212 clear_buffer_write_io_error(sbh);
3213 set_buffer_uptodate(sbh);
3214 }
3215 }
3216 return error;
3217}
3218
3219
3220/*
3221 * Have we just finished recovery? If so, and if we are mounting (or
3222 * remounting) the filesystem readonly, then we will end up with a
3223 * consistent fs on disk. Record that fact.
3224 */
3225static void ext4_mark_recovery_complete(struct super_block *sb,
3226 struct ext4_super_block *es)
3227{
3228 journal_t *journal = EXT4_SB(sb)->s_journal;
3229
3230 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3231 BUG_ON(journal != NULL);
3232 return;
3233 }
3234 jbd2_journal_lock_updates(journal);
3235 if (jbd2_journal_flush(journal) < 0)
3236 goto out;
3237
3238 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3239 sb->s_flags & MS_RDONLY) {
3240 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3241 ext4_commit_super(sb, 1);
3242 }
3243
3244out:
3245 jbd2_journal_unlock_updates(journal);
3246}
3247
3248/*
3249 * If we are mounting (or read-write remounting) a filesystem whose journal
3250 * has recorded an error from a previous lifetime, move that error to the
3251 * main filesystem now.
3252 */
3253static void ext4_clear_journal_err(struct super_block *sb,
3254 struct ext4_super_block *es)
3255{
3256 journal_t *journal;
3257 int j_errno;
3258 const char *errstr;
3259
3260 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3261
3262 journal = EXT4_SB(sb)->s_journal;
3263
3264 /*
3265 * Now check for any error status which may have been recorded in the
3266 * journal by a prior ext4_error() or ext4_abort()
3267 */
3268
3269 j_errno = jbd2_journal_errno(journal);
3270 if (j_errno) {
3271 char nbuf[16];
3272
3273 errstr = ext4_decode_error(sb, j_errno, nbuf);
3274 ext4_warning(sb, __func__, "Filesystem error recorded "
3275 "from previous mount: %s", errstr);
3276 ext4_warning(sb, __func__, "Marking fs in need of "
3277 "filesystem check.");
3278
3279 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3280 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3281 ext4_commit_super(sb, 1);
3282
3283 jbd2_journal_clear_err(journal);
3284 }
3285}
3286
3287/*
3288 * Force the running and committing transactions to commit,
3289 * and wait on the commit.
3290 */
3291int ext4_force_commit(struct super_block *sb)
3292{
3293 journal_t *journal;
3294 int ret = 0;
3295
3296 if (sb->s_flags & MS_RDONLY)
3297 return 0;
3298
3299 journal = EXT4_SB(sb)->s_journal;
3300 if (journal)
3301 ret = ext4_journal_force_commit(journal);
3302
3303 return ret;
3304}
3305
3306static void ext4_write_super(struct super_block *sb)
3307{
3308 ext4_commit_super(sb, 1);
3309}
3310
3311static int ext4_sync_fs(struct super_block *sb, int wait)
3312{
3313 int ret = 0;
3314 tid_t target;
3315
3316 trace_mark(ext4_sync_fs, "dev %s wait %d", sb->s_id, wait);
3317 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
3318 if (wait)
3319 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
3320 }
3321 return ret;
3322}
3323
3324/*
3325 * LVM calls this function before a (read-only) snapshot is created. This
3326 * gives us a chance to flush the journal completely and mark the fs clean.
3327 */
3328static int ext4_freeze(struct super_block *sb)
3329{
3330 int error = 0;
3331 journal_t *journal;
3332
3333 if (sb->s_flags & MS_RDONLY)
3334 return 0;
3335
3336 journal = EXT4_SB(sb)->s_journal;
3337
3338 /* Now we set up the journal barrier. */
3339 jbd2_journal_lock_updates(journal);
3340
3341 /*
3342 * Don't clear the needs_recovery flag if we failed to flush
3343 * the journal.
3344 */
3345 error = jbd2_journal_flush(journal);
3346 if (error < 0) {
3347 out:
3348 jbd2_journal_unlock_updates(journal);
3349 return error;
3350 }
3351
3352 /* Journal blocked and flushed, clear needs_recovery flag. */
3353 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3354 error = ext4_commit_super(sb, 1);
3355 if (error)
3356 goto out;
3357 return 0;
3358}
3359
3360/*
3361 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3362 * flag here, even though the filesystem is not technically dirty yet.
3363 */
3364static int ext4_unfreeze(struct super_block *sb)
3365{
3366 if (sb->s_flags & MS_RDONLY)
3367 return 0;
3368
3369 lock_super(sb);
3370 /* Reset the needs_recovery flag before the fs is unlocked. */
3371 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3372 ext4_commit_super(sb, 1);
3373 unlock_super(sb);
3374 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3375 return 0;
3376}
3377
3378static int ext4_remount(struct super_block *sb, int *flags, char *data)
3379{
3380 struct ext4_super_block *es;
3381 struct ext4_sb_info *sbi = EXT4_SB(sb);
3382 ext4_fsblk_t n_blocks_count = 0;
3383 unsigned long old_sb_flags;
3384 struct ext4_mount_options old_opts;
3385 ext4_group_t g;
3386 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3387 int err;
3388#ifdef CONFIG_QUOTA
3389 int i;
3390#endif
3391
3392 /* Store the original options */
3393 old_sb_flags = sb->s_flags;
3394 old_opts.s_mount_opt = sbi->s_mount_opt;
3395 old_opts.s_resuid = sbi->s_resuid;
3396 old_opts.s_resgid = sbi->s_resgid;
3397 old_opts.s_commit_interval = sbi->s_commit_interval;
3398 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3399 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3400#ifdef CONFIG_QUOTA
3401 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3402 for (i = 0; i < MAXQUOTAS; i++)
3403 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3404#endif
3405 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3406 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3407
3408 /*
3409 * Allow the "check" option to be passed as a remount option.
3410 */
3411 if (!parse_options(data, sb, NULL, &journal_ioprio,
3412 &n_blocks_count, 1)) {
3413 err = -EINVAL;
3414 goto restore_opts;
3415 }
3416
3417 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
3418 ext4_abort(sb, __func__, "Abort forced by user");
3419
3420 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3421 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3422
3423 es = sbi->s_es;
3424
3425 if (sbi->s_journal) {
3426 ext4_init_journal_params(sb, sbi->s_journal);
3427 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3428 }
3429
3430 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3431 n_blocks_count > ext4_blocks_count(es)) {
3432 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
3433 err = -EROFS;
3434 goto restore_opts;
3435 }
3436
3437 if (*flags & MS_RDONLY) {
3438 /*
3439 * First of all, the unconditional stuff we have to do
3440 * to disable replay of the journal when we next remount
3441 */
3442 sb->s_flags |= MS_RDONLY;
3443
3444 /*
3445 * OK, test if we are remounting a valid rw partition
3446 * readonly, and if so set the rdonly flag and then
3447 * mark the partition as valid again.
3448 */
3449 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3450 (sbi->s_mount_state & EXT4_VALID_FS))
3451 es->s_state = cpu_to_le16(sbi->s_mount_state);
3452
3453 if (sbi->s_journal)
3454 ext4_mark_recovery_complete(sb, es);
3455 } else {
3456 int ret;
3457 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3458 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
3459 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3460 "remount RDWR because of unsupported "
3461 "optional features (%x).\n", sb->s_id,
3462 (le32_to_cpu(sbi->s_es->s_feature_ro_compat) &
3463 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3464 err = -EROFS;
3465 goto restore_opts;
3466 }
3467
3468 /*
3469 * Make sure the group descriptor checksums
3470 * are sane. If they aren't, refuse to
3471 * remount r/w.
3472 */
3473 for (g = 0; g < sbi->s_groups_count; g++) {
3474 struct ext4_group_desc *gdp =
3475 ext4_get_group_desc(sb, g, NULL);
3476
3477 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3478 printk(KERN_ERR
3479 "EXT4-fs: ext4_remount: "
3480 "Checksum for group %u failed (%u!=%u)\n",
3481 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3482 le16_to_cpu(gdp->bg_checksum));
3483 err = -EINVAL;
3484 goto restore_opts;
3485 }
3486 }
3487
3488 /*
3489 * If we have an unprocessed orphan list hanging
3490 * around from a previously readonly bdev mount,
3491 * require a full umount/remount for now.
3492 */
3493 if (es->s_last_orphan) {
3494 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3495 "remount RDWR because of unprocessed "
3496 "orphan inode list. Please "
3497 "umount/remount instead.\n",
3498 sb->s_id);
3499 err = -EINVAL;
3500 goto restore_opts;
3501 }
3502
3503 /*
3504 * Mounting a RDONLY partition read-write, so reread
3505 * and store the current valid flag. (It may have
3506 * been changed by e2fsck since we originally mounted
3507 * the partition.)
3508 */
3509 if (sbi->s_journal)
3510 ext4_clear_journal_err(sb, es);
3511 sbi->s_mount_state = le16_to_cpu(es->s_state);
3512 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3513 goto restore_opts;
3514 if (!ext4_setup_super(sb, es, 0))
3515 sb->s_flags &= ~MS_RDONLY;
3516 }
3517 }
3518 if (sbi->s_journal == NULL)
3519 ext4_commit_super(sb, 1);
3520
3521#ifdef CONFIG_QUOTA
3522 /* Release old quota file names */
3523 for (i = 0; i < MAXQUOTAS; i++)
3524 if (old_opts.s_qf_names[i] &&
3525 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3526 kfree(old_opts.s_qf_names[i]);
3527#endif
3528 return 0;
3529restore_opts:
3530 sb->s_flags = old_sb_flags;
3531 sbi->s_mount_opt = old_opts.s_mount_opt;
3532 sbi->s_resuid = old_opts.s_resuid;
3533 sbi->s_resgid = old_opts.s_resgid;
3534 sbi->s_commit_interval = old_opts.s_commit_interval;
3535 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3536 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3537#ifdef CONFIG_QUOTA
3538 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3539 for (i = 0; i < MAXQUOTAS; i++) {
3540 if (sbi->s_qf_names[i] &&
3541 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3542 kfree(sbi->s_qf_names[i]);
3543 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3544 }
3545#endif
3546 return err;
3547}
3548
3549static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3550{
3551 struct super_block *sb = dentry->d_sb;
3552 struct ext4_sb_info *sbi = EXT4_SB(sb);
3553 struct ext4_super_block *es = sbi->s_es;
3554 u64 fsid;
3555
3556 if (test_opt(sb, MINIX_DF)) {
3557 sbi->s_overhead_last = 0;
3558 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3559 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3560 ext4_fsblk_t overhead = 0;
3561
3562 /*
3563 * Compute the overhead (FS structures). This is constant
3564 * for a given filesystem unless the number of block groups
3565 * changes so we cache the previous value until it does.
3566 */
3567
3568 /*
3569 * All of the blocks before first_data_block are
3570 * overhead
3571 */
3572 overhead = le32_to_cpu(es->s_first_data_block);
3573
3574 /*
3575 * Add the overhead attributed to the superblock and
3576 * block group descriptors. If the sparse superblocks
3577 * feature is turned on, then not all groups have this.
3578 */
3579 for (i = 0; i < ngroups; i++) {
3580 overhead += ext4_bg_has_super(sb, i) +
3581 ext4_bg_num_gdb(sb, i);
3582 cond_resched();
3583 }
3584
3585 /*
3586 * Every block group has an inode bitmap, a block
3587 * bitmap, and an inode table.
3588 */
3589 overhead += ngroups * (2 + sbi->s_itb_per_group);
3590 sbi->s_overhead_last = overhead;
3591 smp_wmb();
3592 sbi->s_blocks_last = ext4_blocks_count(es);
3593 }
3594
3595 buf->f_type = EXT4_SUPER_MAGIC;
3596 buf->f_bsize = sb->s_blocksize;
3597 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3598 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3599 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3600 ext4_free_blocks_count_set(es, buf->f_bfree);
3601 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3602 if (buf->f_bfree < ext4_r_blocks_count(es))
3603 buf->f_bavail = 0;
3604 buf->f_files = le32_to_cpu(es->s_inodes_count);
3605 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3606 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
3607 buf->f_namelen = EXT4_NAME_LEN;
3608 fsid = le64_to_cpup((void *)es->s_uuid) ^
3609 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3610 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3611 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3612 return 0;
3613}
3614
3615/* Helper function for writing quotas on sync - we need to start transaction before quota file
3616 * is locked for write. Otherwise the are possible deadlocks:
3617 * Process 1 Process 2
3618 * ext4_create() quota_sync()
3619 * jbd2_journal_start() write_dquot()
3620 * vfs_dq_init() down(dqio_mutex)
3621 * down(dqio_mutex) jbd2_journal_start()
3622 *
3623 */
3624
3625#ifdef CONFIG_QUOTA
3626
3627static inline struct inode *dquot_to_inode(struct dquot *dquot)
3628{
3629 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3630}
3631
3632static int ext4_write_dquot(struct dquot *dquot)
3633{
3634 int ret, err;
3635 handle_t *handle;
3636 struct inode *inode;
3637
3638 inode = dquot_to_inode(dquot);
3639 handle = ext4_journal_start(inode,
3640 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3641 if (IS_ERR(handle))
3642 return PTR_ERR(handle);
3643 ret = dquot_commit(dquot);
3644 err = ext4_journal_stop(handle);
3645 if (!ret)
3646 ret = err;
3647 return ret;
3648}
3649
3650static int ext4_acquire_dquot(struct dquot *dquot)
3651{
3652 int ret, err;
3653 handle_t *handle;
3654
3655 handle = ext4_journal_start(dquot_to_inode(dquot),
3656 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3657 if (IS_ERR(handle))
3658 return PTR_ERR(handle);
3659 ret = dquot_acquire(dquot);
3660 err = ext4_journal_stop(handle);
3661 if (!ret)
3662 ret = err;
3663 return ret;
3664}
3665
3666static int ext4_release_dquot(struct dquot *dquot)
3667{
3668 int ret, err;
3669 handle_t *handle;
3670
3671 handle = ext4_journal_start(dquot_to_inode(dquot),
3672 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3673 if (IS_ERR(handle)) {
3674 /* Release dquot anyway to avoid endless cycle in dqput() */
3675 dquot_release(dquot);
3676 return PTR_ERR(handle);
3677 }
3678 ret = dquot_release(dquot);
3679 err = ext4_journal_stop(handle);
3680 if (!ret)
3681 ret = err;
3682 return ret;
3683}
3684
3685static int ext4_mark_dquot_dirty(struct dquot *dquot)
3686{
3687 /* Are we journaling quotas? */
3688 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3689 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3690 dquot_mark_dquot_dirty(dquot);
3691 return ext4_write_dquot(dquot);
3692 } else {
3693 return dquot_mark_dquot_dirty(dquot);
3694 }
3695}
3696
3697static int ext4_write_info(struct super_block *sb, int type)
3698{
3699 int ret, err;
3700 handle_t *handle;
3701
3702 /* Data block + inode block */
3703 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3704 if (IS_ERR(handle))
3705 return PTR_ERR(handle);
3706 ret = dquot_commit_info(sb, type);
3707 err = ext4_journal_stop(handle);
3708 if (!ret)
3709 ret = err;
3710 return ret;
3711}
3712
3713/*
3714 * Turn on quotas during mount time - we need to find
3715 * the quota file and such...
3716 */
3717static int ext4_quota_on_mount(struct super_block *sb, int type)
3718{
3719 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3720 EXT4_SB(sb)->s_jquota_fmt, type);
3721}
3722
3723/*
3724 * Standard function to be called on quota_on
3725 */
3726static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3727 char *name, int remount)
3728{
3729 int err;
3730 struct path path;
3731
3732 if (!test_opt(sb, QUOTA))
3733 return -EINVAL;
3734 /* When remounting, no checks are needed and in fact, name is NULL */
3735 if (remount)
3736 return vfs_quota_on(sb, type, format_id, name, remount);
3737
3738 err = kern_path(name, LOOKUP_FOLLOW, &path);
3739 if (err)
3740 return err;
3741
3742 /* Quotafile not on the same filesystem? */
3743 if (path.mnt->mnt_sb != sb) {
3744 path_put(&path);
3745 return -EXDEV;
3746 }
3747 /* Journaling quota? */
3748 if (EXT4_SB(sb)->s_qf_names[type]) {
3749 /* Quotafile not in fs root? */
3750 if (path.dentry->d_parent != sb->s_root)
3751 printk(KERN_WARNING
3752 "EXT4-fs: Quota file not on filesystem root. "
3753 "Journaled quota will not work.\n");
3754 }
3755
3756 /*
3757 * When we journal data on quota file, we have to flush journal to see
3758 * all updates to the file when we bypass pagecache...
3759 */
3760 if (EXT4_SB(sb)->s_journal &&
3761 ext4_should_journal_data(path.dentry->d_inode)) {
3762 /*
3763 * We don't need to lock updates but journal_flush() could
3764 * otherwise be livelocked...
3765 */
3766 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3767 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3768 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3769 if (err) {
3770 path_put(&path);
3771 return err;
3772 }
3773 }
3774
3775 err = vfs_quota_on_path(sb, type, format_id, &path);
3776 path_put(&path);
3777 return err;
3778}
3779
3780/* Read data from quotafile - avoid pagecache and such because we cannot afford
3781 * acquiring the locks... As quota files are never truncated and quota code
3782 * itself serializes the operations (and noone else should touch the files)
3783 * we don't have to be afraid of races */
3784static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3785 size_t len, loff_t off)
3786{
3787 struct inode *inode = sb_dqopt(sb)->files[type];
3788 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3789 int err = 0;
3790 int offset = off & (sb->s_blocksize - 1);
3791 int tocopy;
3792 size_t toread;
3793 struct buffer_head *bh;
3794 loff_t i_size = i_size_read(inode);
3795
3796 if (off > i_size)
3797 return 0;
3798 if (off+len > i_size)
3799 len = i_size-off;
3800 toread = len;
3801 while (toread > 0) {
3802 tocopy = sb->s_blocksize - offset < toread ?
3803 sb->s_blocksize - offset : toread;
3804 bh = ext4_bread(NULL, inode, blk, 0, &err);
3805 if (err)
3806 return err;
3807 if (!bh) /* A hole? */
3808 memset(data, 0, tocopy);
3809 else
3810 memcpy(data, bh->b_data+offset, tocopy);
3811 brelse(bh);
3812 offset = 0;
3813 toread -= tocopy;
3814 data += tocopy;
3815 blk++;
3816 }
3817 return len;
3818}
3819
3820/* Write to quotafile (we know the transaction is already started and has
3821 * enough credits) */
3822static ssize_t ext4_quota_write(struct super_block *sb, int type,
3823 const char *data, size_t len, loff_t off)
3824{
3825 struct inode *inode = sb_dqopt(sb)->files[type];
3826 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3827 int err = 0;
3828 int offset = off & (sb->s_blocksize - 1);
3829 int tocopy;
3830 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3831 size_t towrite = len;
3832 struct buffer_head *bh;
3833 handle_t *handle = journal_current_handle();
3834
3835 if (EXT4_SB(sb)->s_journal && !handle) {
3836 printk(KERN_WARNING "EXT4-fs: Quota write (off=%llu, len=%llu)"
3837 " cancelled because transaction is not started.\n",
3838 (unsigned long long)off, (unsigned long long)len);
3839 return -EIO;
3840 }
3841 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3842 while (towrite > 0) {
3843 tocopy = sb->s_blocksize - offset < towrite ?
3844 sb->s_blocksize - offset : towrite;
3845 bh = ext4_bread(handle, inode, blk, 1, &err);
3846 if (!bh)
3847 goto out;
3848 if (journal_quota) {
3849 err = ext4_journal_get_write_access(handle, bh);
3850 if (err) {
3851 brelse(bh);
3852 goto out;
3853 }
3854 }
3855 lock_buffer(bh);
3856 memcpy(bh->b_data+offset, data, tocopy);
3857 flush_dcache_page(bh->b_page);
3858 unlock_buffer(bh);
3859 if (journal_quota)
3860 err = ext4_handle_dirty_metadata(handle, NULL, bh);
3861 else {
3862 /* Always do at least ordered writes for quotas */
3863 err = ext4_jbd2_file_inode(handle, inode);
3864 mark_buffer_dirty(bh);
3865 }
3866 brelse(bh);
3867 if (err)
3868 goto out;
3869 offset = 0;
3870 towrite -= tocopy;
3871 data += tocopy;
3872 blk++;
3873 }
3874out:
3875 if (len == towrite) {
3876 mutex_unlock(&inode->i_mutex);
3877 return err;
3878 }
3879 if (inode->i_size < off+len-towrite) {
3880 i_size_write(inode, off+len-towrite);
3881 EXT4_I(inode)->i_disksize = inode->i_size;
3882 }
3883 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3884 ext4_mark_inode_dirty(handle, inode);
3885 mutex_unlock(&inode->i_mutex);
3886 return len - towrite;
3887}
3888
3889#endif
3890
3891static int ext4_get_sb(struct file_system_type *fs_type,
3892 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3893{
3894 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3895}
3896
3897static struct file_system_type ext4_fs_type = {
3898 .owner = THIS_MODULE,
3899 .name = "ext4",
3900 .get_sb = ext4_get_sb,
3901 .kill_sb = kill_block_super,
3902 .fs_flags = FS_REQUIRES_DEV,
3903};
3904
3905#ifdef CONFIG_EXT4DEV_COMPAT
3906static int ext4dev_get_sb(struct file_system_type *fs_type,
3907 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3908{
3909 printk(KERN_WARNING "EXT4-fs: Update your userspace programs "
3910 "to mount using ext4\n");
3911 printk(KERN_WARNING "EXT4-fs: ext4dev backwards compatibility "
3912 "will go away by 2.6.31\n");
3913 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3914}
3915
3916static struct file_system_type ext4dev_fs_type = {
3917 .owner = THIS_MODULE,
3918 .name = "ext4dev",
3919 .get_sb = ext4dev_get_sb,
3920 .kill_sb = kill_block_super,
3921 .fs_flags = FS_REQUIRES_DEV,
3922};
3923MODULE_ALIAS("ext4dev");
3924#endif
3925
3926static int __init init_ext4_fs(void)
3927{
3928 int err;
3929
3930 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
3931 if (!ext4_kset)
3932 return -ENOMEM;
3933 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
3934 err = init_ext4_mballoc();
3935 if (err)
3936 return err;
3937
3938 err = init_ext4_xattr();
3939 if (err)
3940 goto out2;
3941 err = init_inodecache();
3942 if (err)
3943 goto out1;
3944 err = register_filesystem(&ext4_fs_type);
3945 if (err)
3946 goto out;
3947#ifdef CONFIG_EXT4DEV_COMPAT
3948 err = register_filesystem(&ext4dev_fs_type);
3949 if (err) {
3950 unregister_filesystem(&ext4_fs_type);
3951 goto out;
3952 }
3953#endif
3954 return 0;
3955out:
3956 destroy_inodecache();
3957out1:
3958 exit_ext4_xattr();
3959out2:
3960 exit_ext4_mballoc();
3961 return err;
3962}
3963
3964static void __exit exit_ext4_fs(void)
3965{
3966 unregister_filesystem(&ext4_fs_type);
3967#ifdef CONFIG_EXT4DEV_COMPAT
3968 unregister_filesystem(&ext4dev_fs_type);
3969#endif
3970 destroy_inodecache();
3971 exit_ext4_xattr();
3972 exit_ext4_mballoc();
3973 remove_proc_entry("fs/ext4", NULL);
3974 kset_unregister(ext4_kset);
3975}
3976
3977MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3978MODULE_DESCRIPTION("Fourth Extended Filesystem");
3979MODULE_LICENSE("GPL");
3980module_init(init_ext4_fs)
3981module_exit(exit_ext4_fs)
Linux 6.x block layer and io_uring tree(s)