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