?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
System call wrappers part 05
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / fs / ext4 / super.c
blob7726e8e0e3ef8612564674633bfc2bc882454e0b
1 /*
2 * linux/fs/ext4/super.c
3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19 #include <linux/module.h>
20 #include <linux/string.h>
21 #include <linux/fs.h>
22 #include <linux/time.h>
23 #include <linux/jbd2.h>
24 #include <linux/slab.h>
25 #include <linux/init.h>
26 #include <linux/blkdev.h>
27 #include <linux/parser.h>
28 #include <linux/smp_lock.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/log2.h>
38 #include <linux/crc16.h>
39 #include <asm/uaccess.h>
40
41 #include "ext4.h"
42 #include "ext4_jbd2.h"
43 #include "xattr.h"
44 #include "acl.h"
45 #include "namei.h"
46 #include "group.h"
47
48 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
49 unsigned long journal_devnum);
50 static int ext4_create_journal(struct super_block *, struct ext4_super_block *,
51 unsigned int);
52 static void ext4_commit_super(struct super_block *sb,
53 struct ext4_super_block *es, int sync);
54 static void ext4_mark_recovery_complete(struct super_block *sb,
55 struct ext4_super_block *es);
56 static void ext4_clear_journal_err(struct super_block *sb,
57 struct ext4_super_block *es);
58 static int ext4_sync_fs(struct super_block *sb, int wait);
59 static const char *ext4_decode_error(struct super_block *sb, int errno,
60 char nbuf[16]);
61 static int ext4_remount(struct super_block *sb, int *flags, char *data);
62 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
63 static void ext4_unlockfs(struct super_block *sb);
64 static void ext4_write_super(struct super_block *sb);
65 static void ext4_write_super_lockfs(struct super_block *sb);
66
67
68 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
69 struct ext4_group_desc *bg)
70 {
71 return le32_to_cpu(bg->bg_block_bitmap_lo) |
72 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
73 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
74 }
75
76 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
77 struct ext4_group_desc *bg)
78 {
79 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
80 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
81 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
82 }
83
84 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
85 struct ext4_group_desc *bg)
86 {
87 return le32_to_cpu(bg->bg_inode_table_lo) |
88 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
89 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
90 }
91
92 void ext4_block_bitmap_set(struct super_block *sb,
93 struct ext4_group_desc *bg, ext4_fsblk_t blk)
94 {
95 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
96 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
97 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
98 }
99
100 void ext4_inode_bitmap_set(struct super_block *sb,
101 struct ext4_group_desc *bg, ext4_fsblk_t blk)
102 {
103 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
104 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
105 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
106 }
107
108 void ext4_inode_table_set(struct super_block *sb,
109 struct ext4_group_desc *bg, ext4_fsblk_t blk)
110 {
111 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
112 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
113 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
114 }
115
116 /*
117 * Wrappers for jbd2_journal_start/end.
118 *
119 * The only special thing we need to do here is to make sure that all
120 * journal_end calls result in the superblock being marked dirty, so
121 * that sync() will call the filesystem's write_super callback if
122 * appropriate.
123 */
124 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
125 {
126 journal_t *journal;
127
128 if (sb->s_flags & MS_RDONLY)
129 return ERR_PTR(-EROFS);
130
131 /* Special case here: if the journal has aborted behind our
132 * backs (eg. EIO in the commit thread), then we still need to
133 * take the FS itself readonly cleanly. */
134 journal = EXT4_SB(sb)->s_journal;
135 if (is_journal_aborted(journal)) {
136 ext4_abort(sb, __func__,
137 "Detected aborted journal");
138 return ERR_PTR(-EROFS);
139 }
140
141 return jbd2_journal_start(journal, nblocks);
142 }
143
144 /*
145 * The only special thing we need to do here is to make sure that all
146 * jbd2_journal_stop calls result in the superblock being marked dirty, so
147 * that sync() will call the filesystem's write_super callback if
148 * appropriate.
149 */
150 int __ext4_journal_stop(const char *where, handle_t *handle)
151 {
152 struct super_block *sb;
153 int err;
154 int rc;
155
156 sb = handle->h_transaction->t_journal->j_private;
157 err = handle->h_err;
158 rc = jbd2_journal_stop(handle);
159
160 if (!err)
161 err = rc;
162 if (err)
163 __ext4_std_error(sb, where, err);
164 return err;
165 }
166
167 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
168 struct buffer_head *bh, handle_t *handle, int err)
169 {
170 char nbuf[16];
171 const char *errstr = ext4_decode_error(NULL, err, nbuf);
172
173 if (bh)
174 BUFFER_TRACE(bh, "abort");
175
176 if (!handle->h_err)
177 handle->h_err = err;
178
179 if (is_handle_aborted(handle))
180 return;
181
182 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
183 caller, errstr, err_fn);
184
185 jbd2_journal_abort_handle(handle);
186 }
187
188 /* Deal with the reporting of failure conditions on a filesystem such as
189 * inconsistencies detected or read IO failures.
190 *
191 * On ext2, we can store the error state of the filesystem in the
192 * superblock. That is not possible on ext4, because we may have other
193 * write ordering constraints on the superblock which prevent us from
194 * writing it out straight away; and given that the journal is about to
195 * be aborted, we can't rely on the current, or future, transactions to
196 * write out the superblock safely.
197 *
198 * We'll just use the jbd2_journal_abort() error code to record an error in
199 * the journal instead. On recovery, the journal will compain about
200 * that error until we've noted it down and cleared it.
201 */
202
203 static void ext4_handle_error(struct super_block *sb)
204 {
205 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
206
207 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
208 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
209
210 if (sb->s_flags & MS_RDONLY)
211 return;
212
213 if (!test_opt(sb, ERRORS_CONT)) {
214 journal_t *journal = EXT4_SB(sb)->s_journal;
215
216 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
217 if (journal)
218 jbd2_journal_abort(journal, -EIO);
219 }
220 if (test_opt(sb, ERRORS_RO)) {
221 printk(KERN_CRIT "Remounting filesystem read-only\n");
222 sb->s_flags |= MS_RDONLY;
223 }
224 ext4_commit_super(sb, es, 1);
225 if (test_opt(sb, ERRORS_PANIC))
226 panic("EXT4-fs (device %s): panic forced after error\n",
227 sb->s_id);
228 }
229
230 void ext4_error(struct super_block *sb, const char *function,
231 const char *fmt, ...)
232 {
233 va_list args;
234
235 va_start(args, fmt);
236 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
237 vprintk(fmt, args);
238 printk("\n");
239 va_end(args);
240
241 ext4_handle_error(sb);
242 }
243
244 static const char *ext4_decode_error(struct super_block *sb, int errno,
245 char nbuf[16])
246 {
247 char *errstr = NULL;
248
249 switch (errno) {
250 case -EIO:
251 errstr = "IO failure";
252 break;
253 case -ENOMEM:
254 errstr = "Out of memory";
255 break;
256 case -EROFS:
257 if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
258 errstr = "Journal has aborted";
259 else
260 errstr = "Readonly filesystem";
261 break;
262 default:
263 /* If the caller passed in an extra buffer for unknown
264 * errors, textualise them now. Else we just return
265 * NULL. */
266 if (nbuf) {
267 /* Check for truncated error codes... */
268 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
269 errstr = nbuf;
270 }
271 break;
272 }
273
274 return errstr;
275 }
276
277 /* __ext4_std_error decodes expected errors from journaling functions
278 * automatically and invokes the appropriate error response. */
279
280 void __ext4_std_error(struct super_block *sb, const char *function, int errno)
281 {
282 char nbuf[16];
283 const char *errstr;
284
285 /* Special case: if the error is EROFS, and we're not already
286 * inside a transaction, then there's really no point in logging
287 * an error. */
288 if (errno == -EROFS && journal_current_handle() == NULL &&
289 (sb->s_flags & MS_RDONLY))
290 return;
291
292 errstr = ext4_decode_error(sb, errno, nbuf);
293 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
294 sb->s_id, function, errstr);
295
296 ext4_handle_error(sb);
297 }
298
299 /*
300 * ext4_abort is a much stronger failure handler than ext4_error. The
301 * abort function may be used to deal with unrecoverable failures such
302 * as journal IO errors or ENOMEM at a critical moment in log management.
303 *
304 * We unconditionally force the filesystem into an ABORT|READONLY state,
305 * unless the error response on the fs has been set to panic in which
306 * case we take the easy way out and panic immediately.
307 */
308
309 void ext4_abort(struct super_block *sb, const char *function,
310 const char *fmt, ...)
311 {
312 va_list args;
313
314 printk(KERN_CRIT "ext4_abort called.\n");
315
316 va_start(args, fmt);
317 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
318 vprintk(fmt, args);
319 printk("\n");
320 va_end(args);
321
322 if (test_opt(sb, ERRORS_PANIC))
323 panic("EXT4-fs panic from previous error\n");
324
325 if (sb->s_flags & MS_RDONLY)
326 return;
327
328 printk(KERN_CRIT "Remounting filesystem read-only\n");
329 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
330 sb->s_flags |= MS_RDONLY;
331 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
332 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
333 }
334
335 void ext4_warning(struct super_block *sb, const char *function,
336 const char *fmt, ...)
337 {
338 va_list args;
339
340 va_start(args, fmt);
341 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
342 sb->s_id, function);
343 vprintk(fmt, args);
344 printk("\n");
345 va_end(args);
346 }
347
348 void ext4_update_dynamic_rev(struct super_block *sb)
349 {
350 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
351
352 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
353 return;
354
355 ext4_warning(sb, __func__,
356 "updating to rev %d because of new feature flag, "
357 "running e2fsck is recommended",
358 EXT4_DYNAMIC_REV);
359
360 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
361 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
362 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
363 /* leave es->s_feature_*compat flags alone */
364 /* es->s_uuid will be set by e2fsck if empty */
365
366 /*
367 * The rest of the superblock fields should be zero, and if not it
368 * means they are likely already in use, so leave them alone. We
369 * can leave it up to e2fsck to clean up any inconsistencies there.
370 */
371 }
372
373 int ext4_update_compat_feature(handle_t *handle,
374 struct super_block *sb, __u32 compat)
375 {
376 int err = 0;
377 if (!EXT4_HAS_COMPAT_FEATURE(sb, compat)) {
378 err = ext4_journal_get_write_access(handle,
379 EXT4_SB(sb)->s_sbh);
380 if (err)
381 return err;
382 EXT4_SET_COMPAT_FEATURE(sb, compat);
383 sb->s_dirt = 1;
384 handle->h_sync = 1;
385 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
386 "call ext4_journal_dirty_met adata");
387 err = ext4_journal_dirty_metadata(handle,
388 EXT4_SB(sb)->s_sbh);
389 }
390 return err;
391 }
392
393 int ext4_update_rocompat_feature(handle_t *handle,
394 struct super_block *sb, __u32 rocompat)
395 {
396 int err = 0;
397 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, rocompat)) {
398 err = ext4_journal_get_write_access(handle,
399 EXT4_SB(sb)->s_sbh);
400 if (err)
401 return err;
402 EXT4_SET_RO_COMPAT_FEATURE(sb, rocompat);
403 sb->s_dirt = 1;
404 handle->h_sync = 1;
405 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
406 "call ext4_journal_dirty_met adata");
407 err = ext4_journal_dirty_metadata(handle,
408 EXT4_SB(sb)->s_sbh);
409 }
410 return err;
411 }
412
413 int ext4_update_incompat_feature(handle_t *handle,
414 struct super_block *sb, __u32 incompat)
415 {
416 int err = 0;
417 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, incompat)) {
418 err = ext4_journal_get_write_access(handle,
419 EXT4_SB(sb)->s_sbh);
420 if (err)
421 return err;
422 EXT4_SET_INCOMPAT_FEATURE(sb, incompat);
423 sb->s_dirt = 1;
424 handle->h_sync = 1;
425 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
426 "call ext4_journal_dirty_met adata");
427 err = ext4_journal_dirty_metadata(handle,
428 EXT4_SB(sb)->s_sbh);
429 }
430 return err;
431 }
432
433 /*
434 * Open the external journal device
435 */
436 static struct block_device *ext4_blkdev_get(dev_t dev)
437 {
438 struct block_device *bdev;
439 char b[BDEVNAME_SIZE];
440
441 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
442 if (IS_ERR(bdev))
443 goto fail;
444 return bdev;
445
446 fail:
447 printk(KERN_ERR "EXT4: failed to open journal device %s: %ld\n",
448 __bdevname(dev, b), PTR_ERR(bdev));
449 return NULL;
450 }
451
452 /*
453 * Release the journal device
454 */
455 static int ext4_blkdev_put(struct block_device *bdev)
456 {
457 bd_release(bdev);
458 return blkdev_put(bdev);
459 }
460
461 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
462 {
463 struct block_device *bdev;
464 int ret = -ENODEV;
465
466 bdev = sbi->journal_bdev;
467 if (bdev) {
468 ret = ext4_blkdev_put(bdev);
469 sbi->journal_bdev = NULL;
470 }
471 return ret;
472 }
473
474 static inline struct inode *orphan_list_entry(struct list_head *l)
475 {
476 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
477 }
478
479 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
480 {
481 struct list_head *l;
482
483 printk(KERN_ERR "sb orphan head is %d\n",
484 le32_to_cpu(sbi->s_es->s_last_orphan));
485
486 printk(KERN_ERR "sb_info orphan list:\n");
487 list_for_each(l, &sbi->s_orphan) {
488 struct inode *inode = orphan_list_entry(l);
489 printk(KERN_ERR " "
490 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
491 inode->i_sb->s_id, inode->i_ino, inode,
492 inode->i_mode, inode->i_nlink,
493 NEXT_ORPHAN(inode));
494 }
495 }
496
497 static void ext4_put_super(struct super_block *sb)
498 {
499 struct ext4_sb_info *sbi = EXT4_SB(sb);
500 struct ext4_super_block *es = sbi->s_es;
501 int i;
502
503 ext4_mb_release(sb);
504 ext4_ext_release(sb);
505 ext4_xattr_put_super(sb);
506 jbd2_journal_destroy(sbi->s_journal);
507 sbi->s_journal = NULL;
508 if (!(sb->s_flags & MS_RDONLY)) {
509 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
510 es->s_state = cpu_to_le16(sbi->s_mount_state);
511 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
512 mark_buffer_dirty(sbi->s_sbh);
513 ext4_commit_super(sb, es, 1);
514 }
515
516 for (i = 0; i < sbi->s_gdb_count; i++)
517 brelse(sbi->s_group_desc[i]);
518 kfree(sbi->s_group_desc);
519 kfree(sbi->s_flex_groups);
520 percpu_counter_destroy(&sbi->s_freeblocks_counter);
521 percpu_counter_destroy(&sbi->s_freeinodes_counter);
522 percpu_counter_destroy(&sbi->s_dirs_counter);
523 brelse(sbi->s_sbh);
524 #ifdef CONFIG_QUOTA
525 for (i = 0; i < MAXQUOTAS; i++)
526 kfree(sbi->s_qf_names[i]);
527 #endif
528
529 /* Debugging code just in case the in-memory inode orphan list
530 * isn't empty. The on-disk one can be non-empty if we've
531 * detected an error and taken the fs readonly, but the
532 * in-memory list had better be clean by this point. */
533 if (!list_empty(&sbi->s_orphan))
534 dump_orphan_list(sb, sbi);
535 J_ASSERT(list_empty(&sbi->s_orphan));
536
537 invalidate_bdev(sb->s_bdev);
538 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
539 /*
540 * Invalidate the journal device's buffers. We don't want them
541 * floating about in memory - the physical journal device may
542 * hotswapped, and it breaks the `ro-after' testing code.
543 */
544 sync_blockdev(sbi->journal_bdev);
545 invalidate_bdev(sbi->journal_bdev);
546 ext4_blkdev_remove(sbi);
547 }
548 sb->s_fs_info = NULL;
549 kfree(sbi);
550 return;
551 }
552
553 static struct kmem_cache *ext4_inode_cachep;
554
555 /*
556 * Called inside transaction, so use GFP_NOFS
557 */
558 static struct inode *ext4_alloc_inode(struct super_block *sb)
559 {
560 struct ext4_inode_info *ei;
561
562 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
563 if (!ei)
564 return NULL;
565 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
566 ei->i_acl = EXT4_ACL_NOT_CACHED;
567 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
568 #endif
569 ei->i_block_alloc_info = NULL;
570 ei->vfs_inode.i_version = 1;
571 ei->vfs_inode.i_data.writeback_index = 0;
572 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
573 INIT_LIST_HEAD(&ei->i_prealloc_list);
574 spin_lock_init(&ei->i_prealloc_lock);
575 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
576 ei->i_reserved_data_blocks = 0;
577 ei->i_reserved_meta_blocks = 0;
578 ei->i_allocated_meta_blocks = 0;
579 ei->i_delalloc_reserved_flag = 0;
580 spin_lock_init(&(ei->i_block_reservation_lock));
581 return &ei->vfs_inode;
582 }
583
584 static void ext4_destroy_inode(struct inode *inode)
585 {
586 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
587 printk("EXT4 Inode %p: orphan list check failed!\n",
588 EXT4_I(inode));
589 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
590 EXT4_I(inode), sizeof(struct ext4_inode_info),
591 true);
592 dump_stack();
593 }
594 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
595 }
596
597 static void init_once(void *foo)
598 {
599 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
600
601 INIT_LIST_HEAD(&ei->i_orphan);
602 #ifdef CONFIG_EXT4DEV_FS_XATTR
603 init_rwsem(&ei->xattr_sem);
604 #endif
605 init_rwsem(&ei->i_data_sem);
606 inode_init_once(&ei->vfs_inode);
607 }
608
609 static int init_inodecache(void)
610 {
611 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
612 sizeof(struct ext4_inode_info),
613 0, (SLAB_RECLAIM_ACCOUNT|
614 SLAB_MEM_SPREAD),
615 init_once);
616 if (ext4_inode_cachep == NULL)
617 return -ENOMEM;
618 return 0;
619 }
620
621 static void destroy_inodecache(void)
622 {
623 kmem_cache_destroy(ext4_inode_cachep);
624 }
625
626 static void ext4_clear_inode(struct inode *inode)
627 {
628 struct ext4_block_alloc_info *rsv = EXT4_I(inode)->i_block_alloc_info;
629 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
630 if (EXT4_I(inode)->i_acl &&
631 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
632 posix_acl_release(EXT4_I(inode)->i_acl);
633 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
634 }
635 if (EXT4_I(inode)->i_default_acl &&
636 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
637 posix_acl_release(EXT4_I(inode)->i_default_acl);
638 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
639 }
640 #endif
641 ext4_discard_reservation(inode);
642 EXT4_I(inode)->i_block_alloc_info = NULL;
643 if (unlikely(rsv))
644 kfree(rsv);
645 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
646 &EXT4_I(inode)->jinode);
647 }
648
649 static inline void ext4_show_quota_options(struct seq_file *seq,
650 struct super_block *sb)
651 {
652 #if defined(CONFIG_QUOTA)
653 struct ext4_sb_info *sbi = EXT4_SB(sb);
654
655 if (sbi->s_jquota_fmt)
656 seq_printf(seq, ",jqfmt=%s",
657 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
658
659 if (sbi->s_qf_names[USRQUOTA])
660 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
661
662 if (sbi->s_qf_names[GRPQUOTA])
663 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
664
665 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
666 seq_puts(seq, ",usrquota");
667
668 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
669 seq_puts(seq, ",grpquota");
670 #endif
671 }
672
673 /*
674 * Show an option if
675 * - it's set to a non-default value OR
676 * - if the per-sb default is different from the global default
677 */
678 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
679 {
680 int def_errors;
681 unsigned long def_mount_opts;
682 struct super_block *sb = vfs->mnt_sb;
683 struct ext4_sb_info *sbi = EXT4_SB(sb);
684 struct ext4_super_block *es = sbi->s_es;
685
686 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
687 def_errors = le16_to_cpu(es->s_errors);
688
689 if (sbi->s_sb_block != 1)
690 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
691 if (test_opt(sb, MINIX_DF))
692 seq_puts(seq, ",minixdf");
693 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
694 seq_puts(seq, ",grpid");
695 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
696 seq_puts(seq, ",nogrpid");
697 if (sbi->s_resuid != EXT4_DEF_RESUID ||
698 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
699 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
700 }
701 if (sbi->s_resgid != EXT4_DEF_RESGID ||
702 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
703 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
704 }
705 if (test_opt(sb, ERRORS_RO)) {
706 if (def_errors == EXT4_ERRORS_PANIC ||
707 def_errors == EXT4_ERRORS_CONTINUE) {
708 seq_puts(seq, ",errors=remount-ro");
709 }
710 }
711 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
712 seq_puts(seq, ",errors=continue");
713 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
714 seq_puts(seq, ",errors=panic");
715 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
716 seq_puts(seq, ",nouid32");
717 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
718 seq_puts(seq, ",debug");
719 if (test_opt(sb, OLDALLOC))
720 seq_puts(seq, ",oldalloc");
721 #ifdef CONFIG_EXT4DEV_FS_XATTR
722 if (test_opt(sb, XATTR_USER) &&
723 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
724 seq_puts(seq, ",user_xattr");
725 if (!test_opt(sb, XATTR_USER) &&
726 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
727 seq_puts(seq, ",nouser_xattr");
728 }
729 #endif
730 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
731 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
732 seq_puts(seq, ",acl");
733 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
734 seq_puts(seq, ",noacl");
735 #endif
736 if (!test_opt(sb, RESERVATION))
737 seq_puts(seq, ",noreservation");
738 if (sbi->s_commit_interval) {
739 seq_printf(seq, ",commit=%u",
740 (unsigned) (sbi->s_commit_interval / HZ));
741 }
742 /*
743 * We're changing the default of barrier mount option, so
744 * let's always display its mount state so it's clear what its
745 * status is.
746 */
747 seq_puts(seq, ",barrier=");
748 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
749 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
750 seq_puts(seq, ",journal_async_commit");
751 if (test_opt(sb, NOBH))
752 seq_puts(seq, ",nobh");
753 if (!test_opt(sb, EXTENTS))
754 seq_puts(seq, ",noextents");
755 if (!test_opt(sb, MBALLOC))
756 seq_puts(seq, ",nomballoc");
757 if (test_opt(sb, I_VERSION))
758 seq_puts(seq, ",i_version");
759 if (!test_opt(sb, DELALLOC))
760 seq_puts(seq, ",nodelalloc");
761
762
763 if (sbi->s_stripe)
764 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
765 /*
766 * journal mode get enabled in different ways
767 * So just print the value even if we didn't specify it
768 */
769 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
770 seq_puts(seq, ",data=journal");
771 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
772 seq_puts(seq, ",data=ordered");
773 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
774 seq_puts(seq, ",data=writeback");
775
776 ext4_show_quota_options(seq, sb);
777 return 0;
778 }
779
780
781 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
782 u64 ino, u32 generation)
783 {
784 struct inode *inode;
785
786 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
787 return ERR_PTR(-ESTALE);
788 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
789 return ERR_PTR(-ESTALE);
790
791 /* iget isn't really right if the inode is currently unallocated!!
792 *
793 * ext4_read_inode will return a bad_inode if the inode had been
794 * deleted, so we should be safe.
795 *
796 * Currently we don't know the generation for parent directory, so
797 * a generation of 0 means "accept any"
798 */
799 inode = ext4_iget(sb, ino);
800 if (IS_ERR(inode))
801 return ERR_CAST(inode);
802 if (generation && inode->i_generation != generation) {
803 iput(inode);
804 return ERR_PTR(-ESTALE);
805 }
806
807 return inode;
808 }
809
810 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
811 int fh_len, int fh_type)
812 {
813 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
814 ext4_nfs_get_inode);
815 }
816
817 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
818 int fh_len, int fh_type)
819 {
820 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
821 ext4_nfs_get_inode);
822 }
823
824 #ifdef CONFIG_QUOTA
825 #define QTYPE2NAME(t) ((t) == USRQUOTA?"user":"group")
826 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
827
828 static int ext4_dquot_initialize(struct inode *inode, int type);
829 static int ext4_dquot_drop(struct inode *inode);
830 static int ext4_write_dquot(struct dquot *dquot);
831 static int ext4_acquire_dquot(struct dquot *dquot);
832 static int ext4_release_dquot(struct dquot *dquot);
833 static int ext4_mark_dquot_dirty(struct dquot *dquot);
834 static int ext4_write_info(struct super_block *sb, int type);
835 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
836 char *path, int remount);
837 static int ext4_quota_on_mount(struct super_block *sb, int type);
838 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
839 size_t len, loff_t off);
840 static ssize_t ext4_quota_write(struct super_block *sb, int type,
841 const char *data, size_t len, loff_t off);
842
843 static struct dquot_operations ext4_quota_operations = {
844 .initialize = ext4_dquot_initialize,
845 .drop = ext4_dquot_drop,
846 .alloc_space = dquot_alloc_space,
847 .alloc_inode = dquot_alloc_inode,
848 .free_space = dquot_free_space,
849 .free_inode = dquot_free_inode,
850 .transfer = dquot_transfer,
851 .write_dquot = ext4_write_dquot,
852 .acquire_dquot = ext4_acquire_dquot,
853 .release_dquot = ext4_release_dquot,
854 .mark_dirty = ext4_mark_dquot_dirty,
855 .write_info = ext4_write_info
856 };
857
858 static struct quotactl_ops ext4_qctl_operations = {
859 .quota_on = ext4_quota_on,
860 .quota_off = vfs_quota_off,
861 .quota_sync = vfs_quota_sync,
862 .get_info = vfs_get_dqinfo,
863 .set_info = vfs_set_dqinfo,
864 .get_dqblk = vfs_get_dqblk,
865 .set_dqblk = vfs_set_dqblk
866 };
867 #endif
868
869 static const struct super_operations ext4_sops = {
870 .alloc_inode = ext4_alloc_inode,
871 .destroy_inode = ext4_destroy_inode,
872 .write_inode = ext4_write_inode,
873 .dirty_inode = ext4_dirty_inode,
874 .delete_inode = ext4_delete_inode,
875 .put_super = ext4_put_super,
876 .write_super = ext4_write_super,
877 .sync_fs = ext4_sync_fs,
878 .write_super_lockfs = ext4_write_super_lockfs,
879 .unlockfs = ext4_unlockfs,
880 .statfs = ext4_statfs,
881 .remount_fs = ext4_remount,
882 .clear_inode = ext4_clear_inode,
883 .show_options = ext4_show_options,
884 #ifdef CONFIG_QUOTA
885 .quota_read = ext4_quota_read,
886 .quota_write = ext4_quota_write,
887 #endif
888 };
889
890 static const struct export_operations ext4_export_ops = {
891 .fh_to_dentry = ext4_fh_to_dentry,
892 .fh_to_parent = ext4_fh_to_parent,
893 .get_parent = ext4_get_parent,
894 };
895
896 enum {
897 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
898 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
899 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
900 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
901 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
902 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
903 Opt_journal_checksum, Opt_journal_async_commit,
904 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
905 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
906 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
907 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
908 Opt_grpquota, Opt_extents, Opt_noextents, Opt_i_version,
909 Opt_mballoc, Opt_nomballoc, Opt_stripe, Opt_delalloc, Opt_nodelalloc,
910 };
911
912 static match_table_t tokens = {
913 {Opt_bsd_df, "bsddf"},
914 {Opt_minix_df, "minixdf"},
915 {Opt_grpid, "grpid"},
916 {Opt_grpid, "bsdgroups"},
917 {Opt_nogrpid, "nogrpid"},
918 {Opt_nogrpid, "sysvgroups"},
919 {Opt_resgid, "resgid=%u"},
920 {Opt_resuid, "resuid=%u"},
921 {Opt_sb, "sb=%u"},
922 {Opt_err_cont, "errors=continue"},
923 {Opt_err_panic, "errors=panic"},
924 {Opt_err_ro, "errors=remount-ro"},
925 {Opt_nouid32, "nouid32"},
926 {Opt_nocheck, "nocheck"},
927 {Opt_nocheck, "check=none"},
928 {Opt_debug, "debug"},
929 {Opt_oldalloc, "oldalloc"},
930 {Opt_orlov, "orlov"},
931 {Opt_user_xattr, "user_xattr"},
932 {Opt_nouser_xattr, "nouser_xattr"},
933 {Opt_acl, "acl"},
934 {Opt_noacl, "noacl"},
935 {Opt_reservation, "reservation"},
936 {Opt_noreservation, "noreservation"},
937 {Opt_noload, "noload"},
938 {Opt_nobh, "nobh"},
939 {Opt_bh, "bh"},
940 {Opt_commit, "commit=%u"},
941 {Opt_journal_update, "journal=update"},
942 {Opt_journal_inum, "journal=%u"},
943 {Opt_journal_dev, "journal_dev=%u"},
944 {Opt_journal_checksum, "journal_checksum"},
945 {Opt_journal_async_commit, "journal_async_commit"},
946 {Opt_abort, "abort"},
947 {Opt_data_journal, "data=journal"},
948 {Opt_data_ordered, "data=ordered"},
949 {Opt_data_writeback, "data=writeback"},
950 {Opt_offusrjquota, "usrjquota="},
951 {Opt_usrjquota, "usrjquota=%s"},
952 {Opt_offgrpjquota, "grpjquota="},
953 {Opt_grpjquota, "grpjquota=%s"},
954 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
955 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
956 {Opt_grpquota, "grpquota"},
957 {Opt_noquota, "noquota"},
958 {Opt_quota, "quota"},
959 {Opt_usrquota, "usrquota"},
960 {Opt_barrier, "barrier=%u"},
961 {Opt_extents, "extents"},
962 {Opt_noextents, "noextents"},
963 {Opt_i_version, "i_version"},
964 {Opt_mballoc, "mballoc"},
965 {Opt_nomballoc, "nomballoc"},
966 {Opt_stripe, "stripe=%u"},
967 {Opt_resize, "resize"},
968 {Opt_delalloc, "delalloc"},
969 {Opt_nodelalloc, "nodelalloc"},
970 {Opt_err, NULL},
971 };
972
973 static ext4_fsblk_t get_sb_block(void **data)
974 {
975 ext4_fsblk_t sb_block;
976 char *options = (char *) *data;
977
978 if (!options || strncmp(options, "sb=", 3) != 0)
979 return 1; /* Default location */
980 options += 3;
981 /*todo: use simple_strtoll with >32bit ext4 */
982 sb_block = simple_strtoul(options, &options, 0);
983 if (*options && *options != ',') {
984 printk("EXT4-fs: Invalid sb specification: %s\n",
985 (char *) *data);
986 return 1;
987 }
988 if (*options == ',')
989 options++;
990 *data = (void *) options;
991 return sb_block;
992 }
993
994 static int parse_options(char *options, struct super_block *sb,
995 unsigned int *inum, unsigned long *journal_devnum,
996 ext4_fsblk_t *n_blocks_count, int is_remount)
997 {
998 struct ext4_sb_info *sbi = EXT4_SB(sb);
999 char *p;
1000 substring_t args[MAX_OPT_ARGS];
1001 int data_opt = 0;
1002 int option;
1003 #ifdef CONFIG_QUOTA
1004 int qtype, qfmt;
1005 char *qname;
1006 #endif
1007 ext4_fsblk_t last_block;
1008
1009 if (!options)
1010 return 1;
1011
1012 while ((p = strsep(&options, ",")) != NULL) {
1013 int token;
1014 if (!*p)
1015 continue;
1016
1017 token = match_token(p, tokens, args);
1018 switch (token) {
1019 case Opt_bsd_df:
1020 clear_opt(sbi->s_mount_opt, MINIX_DF);
1021 break;
1022 case Opt_minix_df:
1023 set_opt(sbi->s_mount_opt, MINIX_DF);
1024 break;
1025 case Opt_grpid:
1026 set_opt(sbi->s_mount_opt, GRPID);
1027 break;
1028 case Opt_nogrpid:
1029 clear_opt(sbi->s_mount_opt, GRPID);
1030 break;
1031 case Opt_resuid:
1032 if (match_int(&args[0], &option))
1033 return 0;
1034 sbi->s_resuid = option;
1035 break;
1036 case Opt_resgid:
1037 if (match_int(&args[0], &option))
1038 return 0;
1039 sbi->s_resgid = option;
1040 break;
1041 case Opt_sb:
1042 /* handled by get_sb_block() instead of here */
1043 /* *sb_block = match_int(&args[0]); */
1044 break;
1045 case Opt_err_panic:
1046 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1047 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1048 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1049 break;
1050 case Opt_err_ro:
1051 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1052 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1053 set_opt(sbi->s_mount_opt, ERRORS_RO);
1054 break;
1055 case Opt_err_cont:
1056 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1057 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1058 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1059 break;
1060 case Opt_nouid32:
1061 set_opt(sbi->s_mount_opt, NO_UID32);
1062 break;
1063 case Opt_nocheck:
1064 clear_opt(sbi->s_mount_opt, CHECK);
1065 break;
1066 case Opt_debug:
1067 set_opt(sbi->s_mount_opt, DEBUG);
1068 break;
1069 case Opt_oldalloc:
1070 set_opt(sbi->s_mount_opt, OLDALLOC);
1071 break;
1072 case Opt_orlov:
1073 clear_opt(sbi->s_mount_opt, OLDALLOC);
1074 break;
1075 #ifdef CONFIG_EXT4DEV_FS_XATTR
1076 case Opt_user_xattr:
1077 set_opt(sbi->s_mount_opt, XATTR_USER);
1078 break;
1079 case Opt_nouser_xattr:
1080 clear_opt(sbi->s_mount_opt, XATTR_USER);
1081 break;
1082 #else
1083 case Opt_user_xattr:
1084 case Opt_nouser_xattr:
1085 printk("EXT4 (no)user_xattr options not supported\n");
1086 break;
1087 #endif
1088 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
1089 case Opt_acl:
1090 set_opt(sbi->s_mount_opt, POSIX_ACL);
1091 break;
1092 case Opt_noacl:
1093 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1094 break;
1095 #else
1096 case Opt_acl:
1097 case Opt_noacl:
1098 printk("EXT4 (no)acl options not supported\n");
1099 break;
1100 #endif
1101 case Opt_reservation:
1102 set_opt(sbi->s_mount_opt, RESERVATION);
1103 break;
1104 case Opt_noreservation:
1105 clear_opt(sbi->s_mount_opt, RESERVATION);
1106 break;
1107 case Opt_journal_update:
1108 /* @@@ FIXME */
1109 /* Eventually we will want to be able to create
1110 a journal file here. For now, only allow the
1111 user to specify an existing inode to be the
1112 journal file. */
1113 if (is_remount) {
1114 printk(KERN_ERR "EXT4-fs: cannot specify "
1115 "journal on remount\n");
1116 return 0;
1117 }
1118 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1119 break;
1120 case Opt_journal_inum:
1121 if (is_remount) {
1122 printk(KERN_ERR "EXT4-fs: cannot specify "
1123 "journal on remount\n");
1124 return 0;
1125 }
1126 if (match_int(&args[0], &option))
1127 return 0;
1128 *inum = option;
1129 break;
1130 case Opt_journal_dev:
1131 if (is_remount) {
1132 printk(KERN_ERR "EXT4-fs: cannot specify "
1133 "journal on remount\n");
1134 return 0;
1135 }
1136 if (match_int(&args[0], &option))
1137 return 0;
1138 *journal_devnum = option;
1139 break;
1140 case Opt_journal_checksum:
1141 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1142 break;
1143 case Opt_journal_async_commit:
1144 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1145 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1146 break;
1147 case Opt_noload:
1148 set_opt(sbi->s_mount_opt, NOLOAD);
1149 break;
1150 case Opt_commit:
1151 if (match_int(&args[0], &option))
1152 return 0;
1153 if (option < 0)
1154 return 0;
1155 if (option == 0)
1156 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1157 sbi->s_commit_interval = HZ * option;
1158 break;
1159 case Opt_data_journal:
1160 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1161 goto datacheck;
1162 case Opt_data_ordered:
1163 data_opt = EXT4_MOUNT_ORDERED_DATA;
1164 goto datacheck;
1165 case Opt_data_writeback:
1166 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1167 datacheck:
1168 if (is_remount) {
1169 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1170 != data_opt) {
1171 printk(KERN_ERR
1172 "EXT4-fs: cannot change data "
1173 "mode on remount\n");
1174 return 0;
1175 }
1176 } else {
1177 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1178 sbi->s_mount_opt |= data_opt;
1179 }
1180 break;
1181 #ifdef CONFIG_QUOTA
1182 case Opt_usrjquota:
1183 qtype = USRQUOTA;
1184 goto set_qf_name;
1185 case Opt_grpjquota:
1186 qtype = GRPQUOTA;
1187 set_qf_name:
1188 if ((sb_any_quota_enabled(sb) ||
1189 sb_any_quota_suspended(sb)) &&
1190 !sbi->s_qf_names[qtype]) {
1191 printk(KERN_ERR
1192 "EXT4-fs: Cannot change journaled "
1193 "quota options when quota turned on.\n");
1194 return 0;
1195 }
1196 qname = match_strdup(&args[0]);
1197 if (!qname) {
1198 printk(KERN_ERR
1199 "EXT4-fs: not enough memory for "
1200 "storing quotafile name.\n");
1201 return 0;
1202 }
1203 if (sbi->s_qf_names[qtype] &&
1204 strcmp(sbi->s_qf_names[qtype], qname)) {
1205 printk(KERN_ERR
1206 "EXT4-fs: %s quota file already "
1207 "specified.\n", QTYPE2NAME(qtype));
1208 kfree(qname);
1209 return 0;
1210 }
1211 sbi->s_qf_names[qtype] = qname;
1212 if (strchr(sbi->s_qf_names[qtype], '/')) {
1213 printk(KERN_ERR
1214 "EXT4-fs: quotafile must be on "
1215 "filesystem root.\n");
1216 kfree(sbi->s_qf_names[qtype]);
1217 sbi->s_qf_names[qtype] = NULL;
1218 return 0;
1219 }
1220 set_opt(sbi->s_mount_opt, QUOTA);
1221 break;
1222 case Opt_offusrjquota:
1223 qtype = USRQUOTA;
1224 goto clear_qf_name;
1225 case Opt_offgrpjquota:
1226 qtype = GRPQUOTA;
1227 clear_qf_name:
1228 if ((sb_any_quota_enabled(sb) ||
1229 sb_any_quota_suspended(sb)) &&
1230 sbi->s_qf_names[qtype]) {
1231 printk(KERN_ERR "EXT4-fs: Cannot change "
1232 "journaled quota options when "
1233 "quota turned on.\n");
1234 return 0;
1235 }
1236 /*
1237 * The space will be released later when all options
1238 * are confirmed to be correct
1239 */
1240 sbi->s_qf_names[qtype] = NULL;
1241 break;
1242 case Opt_jqfmt_vfsold:
1243 qfmt = QFMT_VFS_OLD;
1244 goto set_qf_format;
1245 case Opt_jqfmt_vfsv0:
1246 qfmt = QFMT_VFS_V0;
1247 set_qf_format:
1248 if ((sb_any_quota_enabled(sb) ||
1249 sb_any_quota_suspended(sb)) &&
1250 sbi->s_jquota_fmt != qfmt) {
1251 printk(KERN_ERR "EXT4-fs: Cannot change "
1252 "journaled quota options when "
1253 "quota turned on.\n");
1254 return 0;
1255 }
1256 sbi->s_jquota_fmt = qfmt;
1257 break;
1258 case Opt_quota:
1259 case Opt_usrquota:
1260 set_opt(sbi->s_mount_opt, QUOTA);
1261 set_opt(sbi->s_mount_opt, USRQUOTA);
1262 break;
1263 case Opt_grpquota:
1264 set_opt(sbi->s_mount_opt, QUOTA);
1265 set_opt(sbi->s_mount_opt, GRPQUOTA);
1266 break;
1267 case Opt_noquota:
1268 if (sb_any_quota_enabled(sb)) {
1269 printk(KERN_ERR "EXT4-fs: Cannot change quota "
1270 "options when quota turned on.\n");
1271 return 0;
1272 }
1273 clear_opt(sbi->s_mount_opt, QUOTA);
1274 clear_opt(sbi->s_mount_opt, USRQUOTA);
1275 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1276 break;
1277 #else
1278 case Opt_quota:
1279 case Opt_usrquota:
1280 case Opt_grpquota:
1281 printk(KERN_ERR
1282 "EXT4-fs: quota options not supported.\n");
1283 break;
1284 case Opt_usrjquota:
1285 case Opt_grpjquota:
1286 case Opt_offusrjquota:
1287 case Opt_offgrpjquota:
1288 case Opt_jqfmt_vfsold:
1289 case Opt_jqfmt_vfsv0:
1290 printk(KERN_ERR
1291 "EXT4-fs: journaled quota options not "
1292 "supported.\n");
1293 break;
1294 case Opt_noquota:
1295 break;
1296 #endif
1297 case Opt_abort:
1298 set_opt(sbi->s_mount_opt, ABORT);
1299 break;
1300 case Opt_barrier:
1301 if (match_int(&args[0], &option))
1302 return 0;
1303 if (option)
1304 set_opt(sbi->s_mount_opt, BARRIER);
1305 else
1306 clear_opt(sbi->s_mount_opt, BARRIER);
1307 break;
1308 case Opt_ignore:
1309 break;
1310 case Opt_resize:
1311 if (!is_remount) {
1312 printk("EXT4-fs: resize option only available "
1313 "for remount\n");
1314 return 0;
1315 }
1316 if (match_int(&args[0], &option) != 0)
1317 return 0;
1318 *n_blocks_count = option;
1319 break;
1320 case Opt_nobh:
1321 set_opt(sbi->s_mount_opt, NOBH);
1322 break;
1323 case Opt_bh:
1324 clear_opt(sbi->s_mount_opt, NOBH);
1325 break;
1326 case Opt_extents:
1327 if (!EXT4_HAS_INCOMPAT_FEATURE(sb,
1328 EXT4_FEATURE_INCOMPAT_EXTENTS)) {
1329 ext4_warning(sb, __func__,
1330 "extents feature not enabled "
1331 "on this filesystem, use tune2fs\n");
1332 return 0;
1333 }
1334 set_opt(sbi->s_mount_opt, EXTENTS);
1335 break;
1336 case Opt_noextents:
1337 /*
1338 * When e2fsprogs support resizing an already existing
1339 * ext3 file system to greater than 2**32 we need to
1340 * add support to block allocator to handle growing
1341 * already existing block mapped inode so that blocks
1342 * allocated for them fall within 2**32
1343 */
1344 last_block = ext4_blocks_count(sbi->s_es) - 1;
1345 if (last_block > 0xffffffffULL) {
1346 printk(KERN_ERR "EXT4-fs: Filesystem too "
1347 "large to mount with "
1348 "-o noextents options\n");
1349 return 0;
1350 }
1351 clear_opt(sbi->s_mount_opt, EXTENTS);
1352 break;
1353 case Opt_i_version:
1354 set_opt(sbi->s_mount_opt, I_VERSION);
1355 sb->s_flags |= MS_I_VERSION;
1356 break;
1357 case Opt_nodelalloc:
1358 clear_opt(sbi->s_mount_opt, DELALLOC);
1359 break;
1360 case Opt_mballoc:
1361 set_opt(sbi->s_mount_opt, MBALLOC);
1362 break;
1363 case Opt_nomballoc:
1364 clear_opt(sbi->s_mount_opt, MBALLOC);
1365 break;
1366 case Opt_stripe:
1367 if (match_int(&args[0], &option))
1368 return 0;
1369 if (option < 0)
1370 return 0;
1371 sbi->s_stripe = option;
1372 break;
1373 case Opt_delalloc:
1374 set_opt(sbi->s_mount_opt, DELALLOC);
1375 break;
1376 default:
1377 printk(KERN_ERR
1378 "EXT4-fs: Unrecognized mount option \"%s\" "
1379 "or missing value\n", p);
1380 return 0;
1381 }
1382 }
1383 #ifdef CONFIG_QUOTA
1384 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1385 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1386 sbi->s_qf_names[USRQUOTA])
1387 clear_opt(sbi->s_mount_opt, USRQUOTA);
1388
1389 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1390 sbi->s_qf_names[GRPQUOTA])
1391 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1392
1393 if ((sbi->s_qf_names[USRQUOTA] &&
1394 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1395 (sbi->s_qf_names[GRPQUOTA] &&
1396 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1397 printk(KERN_ERR "EXT4-fs: old and new quota "
1398 "format mixing.\n");
1399 return 0;
1400 }
1401
1402 if (!sbi->s_jquota_fmt) {
1403 printk(KERN_ERR "EXT4-fs: journaled quota format "
1404 "not specified.\n");
1405 return 0;
1406 }
1407 } else {
1408 if (sbi->s_jquota_fmt) {
1409 printk(KERN_ERR "EXT4-fs: journaled quota format "
1410 "specified with no journaling "
1411 "enabled.\n");
1412 return 0;
1413 }
1414 }
1415 #endif
1416 return 1;
1417 }
1418
1419 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1420 int read_only)
1421 {
1422 struct ext4_sb_info *sbi = EXT4_SB(sb);
1423 int res = 0;
1424
1425 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1426 printk(KERN_ERR "EXT4-fs warning: revision level too high, "
1427 "forcing read-only mode\n");
1428 res = MS_RDONLY;
1429 }
1430 if (read_only)
1431 return res;
1432 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1433 printk(KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1434 "running e2fsck is recommended\n");
1435 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1436 printk(KERN_WARNING
1437 "EXT4-fs warning: mounting fs with errors, "
1438 "running e2fsck is recommended\n");
1439 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1440 le16_to_cpu(es->s_mnt_count) >=
1441 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1442 printk(KERN_WARNING
1443 "EXT4-fs warning: maximal mount count reached, "
1444 "running e2fsck is recommended\n");
1445 else if (le32_to_cpu(es->s_checkinterval) &&
1446 (le32_to_cpu(es->s_lastcheck) +
1447 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1448 printk(KERN_WARNING
1449 "EXT4-fs warning: checktime reached, "
1450 "running e2fsck is recommended\n");
1451 #if 0
1452 /* @@@ We _will_ want to clear the valid bit if we find
1453 * inconsistencies, to force a fsck at reboot. But for
1454 * a plain journaled filesystem we can keep it set as
1455 * valid forever! :)
1456 */
1457 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1458 #endif
1459 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1460 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1461 le16_add_cpu(&es->s_mnt_count, 1);
1462 es->s_mtime = cpu_to_le32(get_seconds());
1463 ext4_update_dynamic_rev(sb);
1464 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1465
1466 ext4_commit_super(sb, es, 1);
1467 if (test_opt(sb, DEBUG))
1468 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
1469 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1470 sb->s_blocksize,
1471 sbi->s_groups_count,
1472 EXT4_BLOCKS_PER_GROUP(sb),
1473 EXT4_INODES_PER_GROUP(sb),
1474 sbi->s_mount_opt);
1475
1476 printk(KERN_INFO "EXT4 FS on %s, ", sb->s_id);
1477 if (EXT4_SB(sb)->s_journal->j_inode == NULL) {
1478 char b[BDEVNAME_SIZE];
1479
1480 printk("external journal on %s\n",
1481 bdevname(EXT4_SB(sb)->s_journal->j_dev, b));
1482 } else {
1483 printk("internal journal\n");
1484 }
1485 return res;
1486 }
1487
1488 static int ext4_fill_flex_info(struct super_block *sb)
1489 {
1490 struct ext4_sb_info *sbi = EXT4_SB(sb);
1491 struct ext4_group_desc *gdp = NULL;
1492 struct buffer_head *bh;
1493 ext4_group_t flex_group_count;
1494 ext4_group_t flex_group;
1495 int groups_per_flex = 0;
1496 __u64 block_bitmap = 0;
1497 int i;
1498
1499 if (!sbi->s_es->s_log_groups_per_flex) {
1500 sbi->s_log_groups_per_flex = 0;
1501 return 1;
1502 }
1503
1504 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1505 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1506
1507 /* We allocate both existing and potentially added groups */
1508 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1509 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1510 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1511 sbi->s_flex_groups = kzalloc(flex_group_count *
1512 sizeof(struct flex_groups), GFP_KERNEL);
1513 if (sbi->s_flex_groups == NULL) {
1514 printk(KERN_ERR "EXT4-fs: not enough memory for "
1515 "%lu flex groups\n", flex_group_count);
1516 goto failed;
1517 }
1518
1519 gdp = ext4_get_group_desc(sb, 1, &bh);
1520 block_bitmap = ext4_block_bitmap(sb, gdp) - 1;
1521
1522 for (i = 0; i < sbi->s_groups_count; i++) {
1523 gdp = ext4_get_group_desc(sb, i, &bh);
1524
1525 flex_group = ext4_flex_group(sbi, i);
1526 sbi->s_flex_groups[flex_group].free_inodes +=
1527 le16_to_cpu(gdp->bg_free_inodes_count);
1528 sbi->s_flex_groups[flex_group].free_blocks +=
1529 le16_to_cpu(gdp->bg_free_blocks_count);
1530 }
1531
1532 return 1;
1533 failed:
1534 return 0;
1535 }
1536
1537 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1538 struct ext4_group_desc *gdp)
1539 {
1540 __u16 crc = 0;
1541
1542 if (sbi->s_es->s_feature_ro_compat &
1543 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1544 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1545 __le32 le_group = cpu_to_le32(block_group);
1546
1547 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1548 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1549 crc = crc16(crc, (__u8 *)gdp, offset);
1550 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1551 /* for checksum of struct ext4_group_desc do the rest...*/
1552 if ((sbi->s_es->s_feature_incompat &
1553 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1554 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1555 crc = crc16(crc, (__u8 *)gdp + offset,
1556 le16_to_cpu(sbi->s_es->s_desc_size) -
1557 offset);
1558 }
1559
1560 return cpu_to_le16(crc);
1561 }
1562
1563 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1564 struct ext4_group_desc *gdp)
1565 {
1566 if ((sbi->s_es->s_feature_ro_compat &
1567 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1568 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1569 return 0;
1570
1571 return 1;
1572 }
1573
1574 /* Called at mount-time, super-block is locked */
1575 static int ext4_check_descriptors(struct super_block *sb)
1576 {
1577 struct ext4_sb_info *sbi = EXT4_SB(sb);
1578 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1579 ext4_fsblk_t last_block;
1580 ext4_fsblk_t block_bitmap;
1581 ext4_fsblk_t inode_bitmap;
1582 ext4_fsblk_t inode_table;
1583 int flexbg_flag = 0;
1584 ext4_group_t i;
1585
1586 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1587 flexbg_flag = 1;
1588
1589 ext4_debug ("Checking group descriptors");
1590
1591 for (i = 0; i < sbi->s_groups_count; i++) {
1592 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1593
1594 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1595 last_block = ext4_blocks_count(sbi->s_es) - 1;
1596 else
1597 last_block = first_block +
1598 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1599
1600 block_bitmap = ext4_block_bitmap(sb, gdp);
1601 if (block_bitmap < first_block || block_bitmap > last_block) {
1602 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1603 "Block bitmap for group %lu not in group "
1604 "(block %llu)!", i, block_bitmap);
1605 return 0;
1606 }
1607 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1608 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1609 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1610 "Inode bitmap for group %lu not in group "
1611 "(block %llu)!", i, inode_bitmap);
1612 return 0;
1613 }
1614 inode_table = ext4_inode_table(sb, gdp);
1615 if (inode_table < first_block ||
1616 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1617 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1618 "Inode table for group %lu not in group "
1619 "(block %llu)!", i, inode_table);
1620 return 0;
1621 }
1622 spin_lock(sb_bgl_lock(sbi, i));
1623 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1624 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1625 "Checksum for group %lu failed (%u!=%u)\n",
1626 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1627 gdp)), le16_to_cpu(gdp->bg_checksum));
1628 if (!(sb->s_flags & MS_RDONLY)) {
1629 spin_unlock(sb_bgl_lock(sbi, i));
1630 return 0;
1631 }
1632 }
1633 spin_unlock(sb_bgl_lock(sbi, i));
1634 if (!flexbg_flag)
1635 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1636 }
1637
1638 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1639 sbi->s_es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
1640 return 1;
1641 }
1642
1643 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1644 * the superblock) which were deleted from all directories, but held open by
1645 * a process at the time of a crash. We walk the list and try to delete these
1646 * inodes at recovery time (only with a read-write filesystem).
1647 *
1648 * In order to keep the orphan inode chain consistent during traversal (in
1649 * case of crash during recovery), we link each inode into the superblock
1650 * orphan list_head and handle it the same way as an inode deletion during
1651 * normal operation (which journals the operations for us).
1652 *
1653 * We only do an iget() and an iput() on each inode, which is very safe if we
1654 * accidentally point at an in-use or already deleted inode. The worst that
1655 * can happen in this case is that we get a "bit already cleared" message from
1656 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1657 * e2fsck was run on this filesystem, and it must have already done the orphan
1658 * inode cleanup for us, so we can safely abort without any further action.
1659 */
1660 static void ext4_orphan_cleanup(struct super_block *sb,
1661 struct ext4_super_block *es)
1662 {
1663 unsigned int s_flags = sb->s_flags;
1664 int nr_orphans = 0, nr_truncates = 0;
1665 #ifdef CONFIG_QUOTA
1666 int i;
1667 #endif
1668 if (!es->s_last_orphan) {
1669 jbd_debug(4, "no orphan inodes to clean up\n");
1670 return;
1671 }
1672
1673 if (bdev_read_only(sb->s_bdev)) {
1674 printk(KERN_ERR "EXT4-fs: write access "
1675 "unavailable, skipping orphan cleanup.\n");
1676 return;
1677 }
1678
1679 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1680 if (es->s_last_orphan)
1681 jbd_debug(1, "Errors on filesystem, "
1682 "clearing orphan list.\n");
1683 es->s_last_orphan = 0;
1684 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1685 return;
1686 }
1687
1688 if (s_flags & MS_RDONLY) {
1689 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
1690 sb->s_id);
1691 sb->s_flags &= ~MS_RDONLY;
1692 }
1693 #ifdef CONFIG_QUOTA
1694 /* Needed for iput() to work correctly and not trash data */
1695 sb->s_flags |= MS_ACTIVE;
1696 /* Turn on quotas so that they are updated correctly */
1697 for (i = 0; i < MAXQUOTAS; i++) {
1698 if (EXT4_SB(sb)->s_qf_names[i]) {
1699 int ret = ext4_quota_on_mount(sb, i);
1700 if (ret < 0)
1701 printk(KERN_ERR
1702 "EXT4-fs: Cannot turn on journaled "
1703 "quota: error %d\n", ret);
1704 }
1705 }
1706 #endif
1707
1708 while (es->s_last_orphan) {
1709 struct inode *inode;
1710
1711 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1712 if (IS_ERR(inode)) {
1713 es->s_last_orphan = 0;
1714 break;
1715 }
1716
1717 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1718 DQUOT_INIT(inode);
1719 if (inode->i_nlink) {
1720 printk(KERN_DEBUG
1721 "%s: truncating inode %lu to %Ld bytes\n",
1722 __func__, inode->i_ino, inode->i_size);
1723 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1724 inode->i_ino, inode->i_size);
1725 ext4_truncate(inode);
1726 nr_truncates++;
1727 } else {
1728 printk(KERN_DEBUG
1729 "%s: deleting unreferenced inode %lu\n",
1730 __func__, inode->i_ino);
1731 jbd_debug(2, "deleting unreferenced inode %lu\n",
1732 inode->i_ino);
1733 nr_orphans++;
1734 }
1735 iput(inode); /* The delete magic happens here! */
1736 }
1737
1738 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1739
1740 if (nr_orphans)
1741 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
1742 sb->s_id, PLURAL(nr_orphans));
1743 if (nr_truncates)
1744 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
1745 sb->s_id, PLURAL(nr_truncates));
1746 #ifdef CONFIG_QUOTA
1747 /* Turn quotas off */
1748 for (i = 0; i < MAXQUOTAS; i++) {
1749 if (sb_dqopt(sb)->files[i])
1750 vfs_quota_off(sb, i, 0);
1751 }
1752 #endif
1753 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1754 }
1755 /*
1756 * Maximal extent format file size.
1757 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1758 * extent format containers, within a sector_t, and within i_blocks
1759 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1760 * so that won't be a limiting factor.
1761 *
1762 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1763 */
1764 static loff_t ext4_max_size(int blkbits)
1765 {
1766 loff_t res;
1767 loff_t upper_limit = MAX_LFS_FILESIZE;
1768
1769 /* small i_blocks in vfs inode? */
1770 if (sizeof(blkcnt_t) < sizeof(u64)) {
1771 /*
1772 * CONFIG_LSF is not enabled implies the inode
1773 * i_block represent total blocks in 512 bytes
1774 * 32 == size of vfs inode i_blocks * 8
1775 */
1776 upper_limit = (1LL << 32) - 1;
1777
1778 /* total blocks in file system block size */
1779 upper_limit >>= (blkbits - 9);
1780 upper_limit <<= blkbits;
1781 }
1782
1783 /* 32-bit extent-start container, ee_block */
1784 res = 1LL << 32;
1785 res <<= blkbits;
1786 res -= 1;
1787
1788 /* Sanity check against vm- & vfs- imposed limits */
1789 if (res > upper_limit)
1790 res = upper_limit;
1791
1792 return res;
1793 }
1794
1795 /*
1796 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1797 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1798 * We need to be 1 filesystem block less than the 2^48 sector limit.
1799 */
1800 static loff_t ext4_max_bitmap_size(int bits)
1801 {
1802 loff_t res = EXT4_NDIR_BLOCKS;
1803 int meta_blocks;
1804 loff_t upper_limit;
1805 /* This is calculated to be the largest file size for a
1806 * dense, bitmapped file such that the total number of
1807 * sectors in the file, including data and all indirect blocks,
1808 * does not exceed 2^48 -1
1809 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1810 * total number of 512 bytes blocks of the file
1811 */
1812
1813 if (sizeof(blkcnt_t) < sizeof(u64)) {
1814 /*
1815 * CONFIG_LSF is not enabled implies the inode
1816 * i_block represent total blocks in 512 bytes
1817 * 32 == size of vfs inode i_blocks * 8
1818 */
1819 upper_limit = (1LL << 32) - 1;
1820
1821 /* total blocks in file system block size */
1822 upper_limit >>= (bits - 9);
1823
1824 } else {
1825 /*
1826 * We use 48 bit ext4_inode i_blocks
1827 * With EXT4_HUGE_FILE_FL set the i_blocks
1828 * represent total number of blocks in
1829 * file system block size
1830 */
1831 upper_limit = (1LL << 48) - 1;
1832
1833 }
1834
1835 /* indirect blocks */
1836 meta_blocks = 1;
1837 /* double indirect blocks */
1838 meta_blocks += 1 + (1LL << (bits-2));
1839 /* tripple indirect blocks */
1840 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1841
1842 upper_limit -= meta_blocks;
1843 upper_limit <<= bits;
1844
1845 res += 1LL << (bits-2);
1846 res += 1LL << (2*(bits-2));
1847 res += 1LL << (3*(bits-2));
1848 res <<= bits;
1849 if (res > upper_limit)
1850 res = upper_limit;
1851
1852 if (res > MAX_LFS_FILESIZE)
1853 res = MAX_LFS_FILESIZE;
1854
1855 return res;
1856 }
1857
1858 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
1859 ext4_fsblk_t logical_sb_block, int nr)
1860 {
1861 struct ext4_sb_info *sbi = EXT4_SB(sb);
1862 ext4_group_t bg, first_meta_bg;
1863 int has_super = 0;
1864
1865 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1866
1867 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
1868 nr < first_meta_bg)
1869 return logical_sb_block + nr + 1;
1870 bg = sbi->s_desc_per_block * nr;
1871 if (ext4_bg_has_super(sb, bg))
1872 has_super = 1;
1873 return (has_super + ext4_group_first_block_no(sb, bg));
1874 }
1875
1876 /**
1877 * ext4_get_stripe_size: Get the stripe size.
1878 * @sbi: In memory super block info
1879 *
1880 * If we have specified it via mount option, then
1881 * use the mount option value. If the value specified at mount time is
1882 * greater than the blocks per group use the super block value.
1883 * If the super block value is greater than blocks per group return 0.
1884 * Allocator needs it be less than blocks per group.
1885 *
1886 */
1887 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
1888 {
1889 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
1890 unsigned long stripe_width =
1891 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
1892
1893 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
1894 return sbi->s_stripe;
1895
1896 if (stripe_width <= sbi->s_blocks_per_group)
1897 return stripe_width;
1898
1899 if (stride <= sbi->s_blocks_per_group)
1900 return stride;
1901
1902 return 0;
1903 }
1904
1905 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
1906 __releases(kernel_lock)
1907 __acquires(kernel_lock)
1908
1909 {
1910 struct buffer_head *bh;
1911 struct ext4_super_block *es = NULL;
1912 struct ext4_sb_info *sbi;
1913 ext4_fsblk_t block;
1914 ext4_fsblk_t sb_block = get_sb_block(&data);
1915 ext4_fsblk_t logical_sb_block;
1916 unsigned long offset = 0;
1917 unsigned int journal_inum = 0;
1918 unsigned long journal_devnum = 0;
1919 unsigned long def_mount_opts;
1920 struct inode *root;
1921 int ret = -EINVAL;
1922 int blocksize;
1923 int db_count;
1924 int i;
1925 int needs_recovery;
1926 __le32 features;
1927 __u64 blocks_count;
1928 int err;
1929
1930 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1931 if (!sbi)
1932 return -ENOMEM;
1933 sb->s_fs_info = sbi;
1934 sbi->s_mount_opt = 0;
1935 sbi->s_resuid = EXT4_DEF_RESUID;
1936 sbi->s_resgid = EXT4_DEF_RESGID;
1937 sbi->s_sb_block = sb_block;
1938
1939 unlock_kernel();
1940
1941 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
1942 if (!blocksize) {
1943 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
1944 goto out_fail;
1945 }
1946
1947 /*
1948 * The ext4 superblock will not be buffer aligned for other than 1kB
1949 * block sizes. We need to calculate the offset from buffer start.
1950 */
1951 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
1952 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1953 offset = do_div(logical_sb_block, blocksize);
1954 } else {
1955 logical_sb_block = sb_block;
1956 }
1957
1958 if (!(bh = sb_bread(sb, logical_sb_block))) {
1959 printk(KERN_ERR "EXT4-fs: unable to read superblock\n");
1960 goto out_fail;
1961 }
1962 /*
1963 * Note: s_es must be initialized as soon as possible because
1964 * some ext4 macro-instructions depend on its value
1965 */
1966 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
1967 sbi->s_es = es;
1968 sb->s_magic = le16_to_cpu(es->s_magic);
1969 if (sb->s_magic != EXT4_SUPER_MAGIC)
1970 goto cantfind_ext4;
1971
1972 /* Set defaults before we parse the mount options */
1973 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1974 if (def_mount_opts & EXT4_DEFM_DEBUG)
1975 set_opt(sbi->s_mount_opt, DEBUG);
1976 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
1977 set_opt(sbi->s_mount_opt, GRPID);
1978 if (def_mount_opts & EXT4_DEFM_UID16)
1979 set_opt(sbi->s_mount_opt, NO_UID32);
1980 #ifdef CONFIG_EXT4DEV_FS_XATTR
1981 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
1982 set_opt(sbi->s_mount_opt, XATTR_USER);
1983 #endif
1984 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
1985 if (def_mount_opts & EXT4_DEFM_ACL)
1986 set_opt(sbi->s_mount_opt, POSIX_ACL);
1987 #endif
1988 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
1989 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
1990 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
1991 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
1992 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
1993 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
1994
1995 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
1996 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1997 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
1998 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1999 else
2000 set_opt(sbi->s_mount_opt, ERRORS_RO);
2001
2002 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2003 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2004
2005 set_opt(sbi->s_mount_opt, RESERVATION);
2006 set_opt(sbi->s_mount_opt, BARRIER);
2007
2008 /*
2009 * turn on extents feature by default in ext4 filesystem
2010 * only if feature flag already set by mkfs or tune2fs.
2011 * Use -o noextents to turn it off
2012 */
2013 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
2014 set_opt(sbi->s_mount_opt, EXTENTS);
2015 else
2016 ext4_warning(sb, __func__,
2017 "extents feature not enabled on this filesystem, "
2018 "use tune2fs.\n");
2019 /*
2020 * turn on mballoc code by default in ext4 filesystem
2021 * Use -o nomballoc to turn it off
2022 */
2023 set_opt(sbi->s_mount_opt, MBALLOC);
2024
2025 /*
2026 * enable delayed allocation by default
2027 * Use -o nodelalloc to turn it off
2028 */
2029 set_opt(sbi->s_mount_opt, DELALLOC);
2030
2031
2032 if (!parse_options((char *) data, sb, &journal_inum, &journal_devnum,
2033 NULL, 0))
2034 goto failed_mount;
2035
2036 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2037 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2038
2039 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2040 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2041 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2042 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2043 printk(KERN_WARNING
2044 "EXT4-fs warning: feature flags set on rev 0 fs, "
2045 "running e2fsck is recommended\n");
2046
2047 /*
2048 * Since ext4 is still considered development code, we require
2049 * that the TEST_FILESYS flag in s->flags be set.
2050 */
2051 if (!(le32_to_cpu(es->s_flags) & EXT2_FLAGS_TEST_FILESYS)) {
2052 printk(KERN_WARNING "EXT4-fs: %s: not marked "
2053 "OK to use with test code.\n", sb->s_id);
2054 goto failed_mount;
2055 }
2056
2057 /*
2058 * Check feature flags regardless of the revision level, since we
2059 * previously didn't change the revision level when setting the flags,
2060 * so there is a chance incompat flags are set on a rev 0 filesystem.
2061 */
2062 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
2063 if (features) {
2064 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
2065 "unsupported optional features (%x).\n",
2066 sb->s_id, le32_to_cpu(features));
2067 goto failed_mount;
2068 }
2069 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
2070 if (!(sb->s_flags & MS_RDONLY) && features) {
2071 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
2072 "unsupported optional features (%x).\n",
2073 sb->s_id, le32_to_cpu(features));
2074 goto failed_mount;
2075 }
2076 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2077 /*
2078 * Large file size enabled file system can only be
2079 * mount if kernel is build with CONFIG_LSF
2080 */
2081 if (sizeof(root->i_blocks) < sizeof(u64) &&
2082 !(sb->s_flags & MS_RDONLY)) {
2083 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2084 "files cannot be mounted read-write "
2085 "without CONFIG_LSF.\n", sb->s_id);
2086 goto failed_mount;
2087 }
2088 }
2089 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2090
2091 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2092 blocksize > EXT4_MAX_BLOCK_SIZE) {
2093 printk(KERN_ERR
2094 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
2095 blocksize, sb->s_id);
2096 goto failed_mount;
2097 }
2098
2099 if (sb->s_blocksize != blocksize) {
2100
2101 /* Validate the filesystem blocksize */
2102 if (!sb_set_blocksize(sb, blocksize)) {
2103 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2104 blocksize);
2105 goto failed_mount;
2106 }
2107
2108 brelse(bh);
2109 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2110 offset = do_div(logical_sb_block, blocksize);
2111 bh = sb_bread(sb, logical_sb_block);
2112 if (!bh) {
2113 printk(KERN_ERR
2114 "EXT4-fs: Can't read superblock on 2nd try.\n");
2115 goto failed_mount;
2116 }
2117 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2118 sbi->s_es = es;
2119 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2120 printk(KERN_ERR
2121 "EXT4-fs: Magic mismatch, very weird !\n");
2122 goto failed_mount;
2123 }
2124 }
2125
2126 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits);
2127 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits);
2128
2129 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2130 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2131 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2132 } else {
2133 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2134 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2135 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2136 (!is_power_of_2(sbi->s_inode_size)) ||
2137 (sbi->s_inode_size > blocksize)) {
2138 printk(KERN_ERR
2139 "EXT4-fs: unsupported inode size: %d\n",
2140 sbi->s_inode_size);
2141 goto failed_mount;
2142 }
2143 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2144 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2145 }
2146 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2147 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2148 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2149 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2150 !is_power_of_2(sbi->s_desc_size)) {
2151 printk(KERN_ERR
2152 "EXT4-fs: unsupported descriptor size %lu\n",
2153 sbi->s_desc_size);
2154 goto failed_mount;
2155 }
2156 } else
2157 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2158 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2159 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2160 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2161 goto cantfind_ext4;
2162 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2163 if (sbi->s_inodes_per_block == 0)
2164 goto cantfind_ext4;
2165 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2166 sbi->s_inodes_per_block;
2167 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2168 sbi->s_sbh = bh;
2169 sbi->s_mount_state = le16_to_cpu(es->s_state);
2170 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2171 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2172 for (i = 0; i < 4; i++)
2173 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2174 sbi->s_def_hash_version = es->s_def_hash_version;
2175
2176 if (sbi->s_blocks_per_group > blocksize * 8) {
2177 printk(KERN_ERR
2178 "EXT4-fs: #blocks per group too big: %lu\n",
2179 sbi->s_blocks_per_group);
2180 goto failed_mount;
2181 }
2182 if (sbi->s_inodes_per_group > blocksize * 8) {
2183 printk(KERN_ERR
2184 "EXT4-fs: #inodes per group too big: %lu\n",
2185 sbi->s_inodes_per_group);
2186 goto failed_mount;
2187 }
2188
2189 if (ext4_blocks_count(es) >
2190 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
2191 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
2192 " too large to mount safely\n", sb->s_id);
2193 if (sizeof(sector_t) < 8)
2194 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
2195 "enabled\n");
2196 goto failed_mount;
2197 }
2198
2199 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2200 goto cantfind_ext4;
2201
2202 /* ensure blocks_count calculation below doesn't sign-extend */
2203 if (ext4_blocks_count(es) + EXT4_BLOCKS_PER_GROUP(sb) <
2204 le32_to_cpu(es->s_first_data_block) + 1) {
2205 printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu, "
2206 "first data block %u, blocks per group %lu\n",
2207 ext4_blocks_count(es),
2208 le32_to_cpu(es->s_first_data_block),
2209 EXT4_BLOCKS_PER_GROUP(sb));
2210 goto failed_mount;
2211 }
2212 blocks_count = (ext4_blocks_count(es) -
2213 le32_to_cpu(es->s_first_data_block) +
2214 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2215 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2216 sbi->s_groups_count = blocks_count;
2217 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2218 EXT4_DESC_PER_BLOCK(sb);
2219 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2220 GFP_KERNEL);
2221 if (sbi->s_group_desc == NULL) {
2222 printk(KERN_ERR "EXT4-fs: not enough memory\n");
2223 goto failed_mount;
2224 }
2225
2226 bgl_lock_init(&sbi->s_blockgroup_lock);
2227
2228 for (i = 0; i < db_count; i++) {
2229 block = descriptor_loc(sb, logical_sb_block, i);
2230 sbi->s_group_desc[i] = sb_bread(sb, block);
2231 if (!sbi->s_group_desc[i]) {
2232 printk(KERN_ERR "EXT4-fs: "
2233 "can't read group descriptor %d\n", i);
2234 db_count = i;
2235 goto failed_mount2;
2236 }
2237 }
2238 if (!ext4_check_descriptors(sb)) {
2239 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
2240 goto failed_mount2;
2241 }
2242 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2243 if (!ext4_fill_flex_info(sb)) {
2244 printk(KERN_ERR
2245 "EXT4-fs: unable to initialize "
2246 "flex_bg meta info!\n");
2247 goto failed_mount2;
2248 }
2249
2250 sbi->s_gdb_count = db_count;
2251 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2252 spin_lock_init(&sbi->s_next_gen_lock);
2253
2254 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2255 ext4_count_free_blocks(sb));
2256 if (!err) {
2257 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2258 ext4_count_free_inodes(sb));
2259 }
2260 if (!err) {
2261 err = percpu_counter_init(&sbi->s_dirs_counter,
2262 ext4_count_dirs(sb));
2263 }
2264 if (err) {
2265 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2266 goto failed_mount3;
2267 }
2268
2269 /* per fileystem reservation list head & lock */
2270 spin_lock_init(&sbi->s_rsv_window_lock);
2271 sbi->s_rsv_window_root = RB_ROOT;
2272 /* Add a single, static dummy reservation to the start of the
2273 * reservation window list --- it gives us a placeholder for
2274 * append-at-start-of-list which makes the allocation logic
2275 * _much_ simpler. */
2276 sbi->s_rsv_window_head.rsv_start = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
2277 sbi->s_rsv_window_head.rsv_end = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
2278 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
2279 sbi->s_rsv_window_head.rsv_goal_size = 0;
2280 ext4_rsv_window_add(sb, &sbi->s_rsv_window_head);
2281
2282 sbi->s_stripe = ext4_get_stripe_size(sbi);
2283
2284 /*
2285 * set up enough so that it can read an inode
2286 */
2287 sb->s_op = &ext4_sops;
2288 sb->s_export_op = &ext4_export_ops;
2289 sb->s_xattr = ext4_xattr_handlers;
2290 #ifdef CONFIG_QUOTA
2291 sb->s_qcop = &ext4_qctl_operations;
2292 sb->dq_op = &ext4_quota_operations;
2293 #endif
2294 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2295
2296 sb->s_root = NULL;
2297
2298 needs_recovery = (es->s_last_orphan != 0 ||
2299 EXT4_HAS_INCOMPAT_FEATURE(sb,
2300 EXT4_FEATURE_INCOMPAT_RECOVER));
2301
2302 /*
2303 * The first inode we look at is the journal inode. Don't try
2304 * root first: it may be modified in the journal!
2305 */
2306 if (!test_opt(sb, NOLOAD) &&
2307 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2308 if (ext4_load_journal(sb, es, journal_devnum))
2309 goto failed_mount3;
2310 if (!(sb->s_flags & MS_RDONLY) &&
2311 EXT4_SB(sb)->s_journal->j_failed_commit) {
2312 printk(KERN_CRIT "EXT4-fs error (device %s): "
2313 "ext4_fill_super: Journal transaction "
2314 "%u is corrupt\n", sb->s_id,
2315 EXT4_SB(sb)->s_journal->j_failed_commit);
2316 if (test_opt(sb, ERRORS_RO)) {
2317 printk(KERN_CRIT
2318 "Mounting filesystem read-only\n");
2319 sb->s_flags |= MS_RDONLY;
2320 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2321 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2322 }
2323 if (test_opt(sb, ERRORS_PANIC)) {
2324 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2325 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2326 ext4_commit_super(sb, es, 1);
2327 printk(KERN_CRIT
2328 "EXT4-fs (device %s): mount failed\n",
2329 sb->s_id);
2330 goto failed_mount4;
2331 }
2332 }
2333 } else if (journal_inum) {
2334 if (ext4_create_journal(sb, es, journal_inum))
2335 goto failed_mount3;
2336 } else {
2337 if (!silent)
2338 printk(KERN_ERR
2339 "ext4: No journal on filesystem on %s\n",
2340 sb->s_id);
2341 goto failed_mount3;
2342 }
2343
2344 if (ext4_blocks_count(es) > 0xffffffffULL &&
2345 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2346 JBD2_FEATURE_INCOMPAT_64BIT)) {
2347 printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
2348 goto failed_mount4;
2349 }
2350
2351 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2352 jbd2_journal_set_features(sbi->s_journal,
2353 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2354 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2355 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2356 jbd2_journal_set_features(sbi->s_journal,
2357 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2358 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2359 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2360 } else {
2361 jbd2_journal_clear_features(sbi->s_journal,
2362 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2363 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2364 }
2365
2366 /* We have now updated the journal if required, so we can
2367 * validate the data journaling mode. */
2368 switch (test_opt(sb, DATA_FLAGS)) {
2369 case 0:
2370 /* No mode set, assume a default based on the journal
2371 * capabilities: ORDERED_DATA if the journal can
2372 * cope, else JOURNAL_DATA
2373 */
2374 if (jbd2_journal_check_available_features
2375 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2376 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2377 else
2378 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2379 break;
2380
2381 case EXT4_MOUNT_ORDERED_DATA:
2382 case EXT4_MOUNT_WRITEBACK_DATA:
2383 if (!jbd2_journal_check_available_features
2384 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2385 printk(KERN_ERR "EXT4-fs: Journal does not support "
2386 "requested data journaling mode\n");
2387 goto failed_mount4;
2388 }
2389 default:
2390 break;
2391 }
2392
2393 if (test_opt(sb, NOBH)) {
2394 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2395 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
2396 "its supported only with writeback mode\n");
2397 clear_opt(sbi->s_mount_opt, NOBH);
2398 }
2399 }
2400 /*
2401 * The jbd2_journal_load will have done any necessary log recovery,
2402 * so we can safely mount the rest of the filesystem now.
2403 */
2404
2405 root = ext4_iget(sb, EXT4_ROOT_INO);
2406 if (IS_ERR(root)) {
2407 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
2408 ret = PTR_ERR(root);
2409 goto failed_mount4;
2410 }
2411 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2412 iput(root);
2413 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
2414 goto failed_mount4;
2415 }
2416 sb->s_root = d_alloc_root(root);
2417 if (!sb->s_root) {
2418 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2419 iput(root);
2420 ret = -ENOMEM;
2421 goto failed_mount4;
2422 }
2423
2424 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2425
2426 /* determine the minimum size of new large inodes, if present */
2427 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2428 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2429 EXT4_GOOD_OLD_INODE_SIZE;
2430 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2431 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2432 if (sbi->s_want_extra_isize <
2433 le16_to_cpu(es->s_want_extra_isize))
2434 sbi->s_want_extra_isize =
2435 le16_to_cpu(es->s_want_extra_isize);
2436 if (sbi->s_want_extra_isize <
2437 le16_to_cpu(es->s_min_extra_isize))
2438 sbi->s_want_extra_isize =
2439 le16_to_cpu(es->s_min_extra_isize);
2440 }
2441 }
2442 /* Check if enough inode space is available */
2443 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2444 sbi->s_inode_size) {
2445 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2446 EXT4_GOOD_OLD_INODE_SIZE;
2447 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2448 "available.\n");
2449 }
2450
2451 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2452 printk(KERN_WARNING "EXT4-fs: Ignoring delalloc option - "
2453 "requested data journaling mode\n");
2454 clear_opt(sbi->s_mount_opt, DELALLOC);
2455 } else if (test_opt(sb, DELALLOC))
2456 printk(KERN_INFO "EXT4-fs: delayed allocation enabled\n");
2457
2458 ext4_ext_init(sb);
2459 err = ext4_mb_init(sb, needs_recovery);
2460 if (err) {
2461 printk(KERN_ERR "EXT4-fs: failed to initalize mballoc (%d)\n",
2462 err);
2463 goto failed_mount4;
2464 }
2465
2466 /*
2467 * akpm: core read_super() calls in here with the superblock locked.
2468 * That deadlocks, because orphan cleanup needs to lock the superblock
2469 * in numerous places. Here we just pop the lock - it's relatively
2470 * harmless, because we are now ready to accept write_super() requests,
2471 * and aviro says that's the only reason for hanging onto the
2472 * superblock lock.
2473 */
2474 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2475 ext4_orphan_cleanup(sb, es);
2476 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2477 if (needs_recovery)
2478 printk(KERN_INFO "EXT4-fs: recovery complete.\n");
2479 ext4_mark_recovery_complete(sb, es);
2480 printk(KERN_INFO "EXT4-fs: mounted filesystem with %s data mode.\n",
2481 test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ? "journal":
2482 test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA ? "ordered":
2483 "writeback");
2484
2485 lock_kernel();
2486 return 0;
2487
2488 cantfind_ext4:
2489 if (!silent)
2490 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
2491 sb->s_id);
2492 goto failed_mount;
2493
2494 failed_mount4:
2495 jbd2_journal_destroy(sbi->s_journal);
2496 sbi->s_journal = NULL;
2497 failed_mount3:
2498 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2499 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2500 percpu_counter_destroy(&sbi->s_dirs_counter);
2501 failed_mount2:
2502 for (i = 0; i < db_count; i++)
2503 brelse(sbi->s_group_desc[i]);
2504 kfree(sbi->s_group_desc);
2505 failed_mount:
2506 #ifdef CONFIG_QUOTA
2507 for (i = 0; i < MAXQUOTAS; i++)
2508 kfree(sbi->s_qf_names[i]);
2509 #endif
2510 ext4_blkdev_remove(sbi);
2511 brelse(bh);
2512 out_fail:
2513 sb->s_fs_info = NULL;
2514 kfree(sbi);
2515 lock_kernel();
2516 return ret;
2517 }
2518
2519 /*
2520 * Setup any per-fs journal parameters now. We'll do this both on
2521 * initial mount, once the journal has been initialised but before we've
2522 * done any recovery; and again on any subsequent remount.
2523 */
2524 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2525 {
2526 struct ext4_sb_info *sbi = EXT4_SB(sb);
2527
2528 if (sbi->s_commit_interval)
2529 journal->j_commit_interval = sbi->s_commit_interval;
2530 /* We could also set up an ext4-specific default for the commit
2531 * interval here, but for now we'll just fall back to the jbd
2532 * default. */
2533
2534 spin_lock(&journal->j_state_lock);
2535 if (test_opt(sb, BARRIER))
2536 journal->j_flags |= JBD2_BARRIER;
2537 else
2538 journal->j_flags &= ~JBD2_BARRIER;
2539 spin_unlock(&journal->j_state_lock);
2540 }
2541
2542 static journal_t *ext4_get_journal(struct super_block *sb,
2543 unsigned int journal_inum)
2544 {
2545 struct inode *journal_inode;
2546 journal_t *journal;
2547
2548 /* First, test for the existence of a valid inode on disk. Bad
2549 * things happen if we iget() an unused inode, as the subsequent
2550 * iput() will try to delete it. */
2551
2552 journal_inode = ext4_iget(sb, journal_inum);
2553 if (IS_ERR(journal_inode)) {
2554 printk(KERN_ERR "EXT4-fs: no journal found.\n");
2555 return NULL;
2556 }
2557 if (!journal_inode->i_nlink) {
2558 make_bad_inode(journal_inode);
2559 iput(journal_inode);
2560 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
2561 return NULL;
2562 }
2563
2564 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2565 journal_inode, journal_inode->i_size);
2566 if (!S_ISREG(journal_inode->i_mode)) {
2567 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
2568 iput(journal_inode);
2569 return NULL;
2570 }
2571
2572 journal = jbd2_journal_init_inode(journal_inode);
2573 if (!journal) {
2574 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
2575 iput(journal_inode);
2576 return NULL;
2577 }
2578 journal->j_private = sb;
2579 ext4_init_journal_params(sb, journal);
2580 return journal;
2581 }
2582
2583 static journal_t *ext4_get_dev_journal(struct super_block *sb,
2584 dev_t j_dev)
2585 {
2586 struct buffer_head *bh;
2587 journal_t *journal;
2588 ext4_fsblk_t start;
2589 ext4_fsblk_t len;
2590 int hblock, blocksize;
2591 ext4_fsblk_t sb_block;
2592 unsigned long offset;
2593 struct ext4_super_block *es;
2594 struct block_device *bdev;
2595
2596 bdev = ext4_blkdev_get(j_dev);
2597 if (bdev == NULL)
2598 return NULL;
2599
2600 if (bd_claim(bdev, sb)) {
2601 printk(KERN_ERR
2602 "EXT4: failed to claim external journal device.\n");
2603 blkdev_put(bdev);
2604 return NULL;
2605 }
2606
2607 blocksize = sb->s_blocksize;
2608 hblock = bdev_hardsect_size(bdev);
2609 if (blocksize < hblock) {
2610 printk(KERN_ERR
2611 "EXT4-fs: blocksize too small for journal device.\n");
2612 goto out_bdev;
2613 }
2614
2615 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2616 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
2617 set_blocksize(bdev, blocksize);
2618 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2619 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
2620 "external journal\n");
2621 goto out_bdev;
2622 }
2623
2624 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2625 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
2626 !(le32_to_cpu(es->s_feature_incompat) &
2627 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2628 printk(KERN_ERR "EXT4-fs: external journal has "
2629 "bad superblock\n");
2630 brelse(bh);
2631 goto out_bdev;
2632 }
2633
2634 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2635 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
2636 brelse(bh);
2637 goto out_bdev;
2638 }
2639
2640 len = ext4_blocks_count(es);
2641 start = sb_block + 1;
2642 brelse(bh); /* we're done with the superblock */
2643
2644 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
2645 start, len, blocksize);
2646 if (!journal) {
2647 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
2648 goto out_bdev;
2649 }
2650 journal->j_private = sb;
2651 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2652 wait_on_buffer(journal->j_sb_buffer);
2653 if (!buffer_uptodate(journal->j_sb_buffer)) {
2654 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
2655 goto out_journal;
2656 }
2657 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2658 printk(KERN_ERR "EXT4-fs: External journal has more than one "
2659 "user (unsupported) - %d\n",
2660 be32_to_cpu(journal->j_superblock->s_nr_users));
2661 goto out_journal;
2662 }
2663 EXT4_SB(sb)->journal_bdev = bdev;
2664 ext4_init_journal_params(sb, journal);
2665 return journal;
2666 out_journal:
2667 jbd2_journal_destroy(journal);
2668 out_bdev:
2669 ext4_blkdev_put(bdev);
2670 return NULL;
2671 }
2672
2673 static int ext4_load_journal(struct super_block *sb,
2674 struct ext4_super_block *es,
2675 unsigned long journal_devnum)
2676 {
2677 journal_t *journal;
2678 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2679 dev_t journal_dev;
2680 int err = 0;
2681 int really_read_only;
2682
2683 if (journal_devnum &&
2684 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2685 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
2686 "numbers have changed\n");
2687 journal_dev = new_decode_dev(journal_devnum);
2688 } else
2689 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2690
2691 really_read_only = bdev_read_only(sb->s_bdev);
2692
2693 /*
2694 * Are we loading a blank journal or performing recovery after a
2695 * crash? For recovery, we need to check in advance whether we
2696 * can get read-write access to the device.
2697 */
2698
2699 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2700 if (sb->s_flags & MS_RDONLY) {
2701 printk(KERN_INFO "EXT4-fs: INFO: recovery "
2702 "required on readonly filesystem.\n");
2703 if (really_read_only) {
2704 printk(KERN_ERR "EXT4-fs: write access "
2705 "unavailable, cannot proceed.\n");
2706 return -EROFS;
2707 }
2708 printk(KERN_INFO "EXT4-fs: write access will "
2709 "be enabled during recovery.\n");
2710 }
2711 }
2712
2713 if (journal_inum && journal_dev) {
2714 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
2715 "and inode journals!\n");
2716 return -EINVAL;
2717 }
2718
2719 if (journal_inum) {
2720 if (!(journal = ext4_get_journal(sb, journal_inum)))
2721 return -EINVAL;
2722 } else {
2723 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
2724 return -EINVAL;
2725 }
2726
2727 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2728 err = jbd2_journal_update_format(journal);
2729 if (err) {
2730 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
2731 jbd2_journal_destroy(journal);
2732 return err;
2733 }
2734 }
2735
2736 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
2737 err = jbd2_journal_wipe(journal, !really_read_only);
2738 if (!err)
2739 err = jbd2_journal_load(journal);
2740
2741 if (err) {
2742 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
2743 jbd2_journal_destroy(journal);
2744 return err;
2745 }
2746
2747 EXT4_SB(sb)->s_journal = journal;
2748 ext4_clear_journal_err(sb, es);
2749
2750 if (journal_devnum &&
2751 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2752 es->s_journal_dev = cpu_to_le32(journal_devnum);
2753 sb->s_dirt = 1;
2754
2755 /* Make sure we flush the recovery flag to disk. */
2756 ext4_commit_super(sb, es, 1);
2757 }
2758
2759 return 0;
2760 }
2761
2762 static int ext4_create_journal(struct super_block *sb,
2763 struct ext4_super_block *es,
2764 unsigned int journal_inum)
2765 {
2766 journal_t *journal;
2767 int err;
2768
2769 if (sb->s_flags & MS_RDONLY) {
2770 printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
2771 "create journal.\n");
2772 return -EROFS;
2773 }
2774
2775 journal = ext4_get_journal(sb, journal_inum);
2776 if (!journal)
2777 return -EINVAL;
2778
2779 printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
2780 journal_inum);
2781
2782 err = jbd2_journal_create(journal);
2783 if (err) {
2784 printk(KERN_ERR "EXT4-fs: error creating journal.\n");
2785 jbd2_journal_destroy(journal);
2786 return -EIO;
2787 }
2788
2789 EXT4_SB(sb)->s_journal = journal;
2790
2791 ext4_update_dynamic_rev(sb);
2792 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2793 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
2794
2795 es->s_journal_inum = cpu_to_le32(journal_inum);
2796 sb->s_dirt = 1;
2797
2798 /* Make sure we flush the recovery flag to disk. */
2799 ext4_commit_super(sb, es, 1);
2800
2801 return 0;
2802 }
2803
2804 static void ext4_commit_super(struct super_block *sb,
2805 struct ext4_super_block *es, int sync)
2806 {
2807 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
2808
2809 if (!sbh)
2810 return;
2811 if (buffer_write_io_error(sbh)) {
2812 /*
2813 * Oh, dear. A previous attempt to write the
2814 * superblock failed. This could happen because the
2815 * USB device was yanked out. Or it could happen to
2816 * be a transient write error and maybe the block will
2817 * be remapped. Nothing we can do but to retry the
2818 * write and hope for the best.
2819 */
2820 printk(KERN_ERR "ext4: previous I/O error to "
2821 "superblock detected for %s.\n", sb->s_id);
2822 clear_buffer_write_io_error(sbh);
2823 set_buffer_uptodate(sbh);
2824 }
2825 es->s_wtime = cpu_to_le32(get_seconds());
2826 ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb));
2827 es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
2828 BUFFER_TRACE(sbh, "marking dirty");
2829 mark_buffer_dirty(sbh);
2830 if (sync) {
2831 sync_dirty_buffer(sbh);
2832 if (buffer_write_io_error(sbh)) {
2833 printk(KERN_ERR "ext4: I/O error while writing "
2834 "superblock for %s.\n", sb->s_id);
2835 clear_buffer_write_io_error(sbh);
2836 set_buffer_uptodate(sbh);
2837 }
2838 }
2839 }
2840
2841
2842 /*
2843 * Have we just finished recovery? If so, and if we are mounting (or
2844 * remounting) the filesystem readonly, then we will end up with a
2845 * consistent fs on disk. Record that fact.
2846 */
2847 static void ext4_mark_recovery_complete(struct super_block *sb,
2848 struct ext4_super_block *es)
2849 {
2850 journal_t *journal = EXT4_SB(sb)->s_journal;
2851
2852 jbd2_journal_lock_updates(journal);
2853 jbd2_journal_flush(journal);
2854 lock_super(sb);
2855 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
2856 sb->s_flags & MS_RDONLY) {
2857 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2858 sb->s_dirt = 0;
2859 ext4_commit_super(sb, es, 1);
2860 }
2861 unlock_super(sb);
2862 jbd2_journal_unlock_updates(journal);
2863 }
2864
2865 /*
2866 * If we are mounting (or read-write remounting) a filesystem whose journal
2867 * has recorded an error from a previous lifetime, move that error to the
2868 * main filesystem now.
2869 */
2870 static void ext4_clear_journal_err(struct super_block *sb,
2871 struct ext4_super_block *es)
2872 {
2873 journal_t *journal;
2874 int j_errno;
2875 const char *errstr;
2876
2877 journal = EXT4_SB(sb)->s_journal;
2878
2879 /*
2880 * Now check for any error status which may have been recorded in the
2881 * journal by a prior ext4_error() or ext4_abort()
2882 */
2883
2884 j_errno = jbd2_journal_errno(journal);
2885 if (j_errno) {
2886 char nbuf[16];
2887
2888 errstr = ext4_decode_error(sb, j_errno, nbuf);
2889 ext4_warning(sb, __func__, "Filesystem error recorded "
2890 "from previous mount: %s", errstr);
2891 ext4_warning(sb, __func__, "Marking fs in need of "
2892 "filesystem check.");
2893
2894 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2895 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2896 ext4_commit_super(sb, es, 1);
2897
2898 jbd2_journal_clear_err(journal);
2899 }
2900 }
2901
2902 /*
2903 * Force the running and committing transactions to commit,
2904 * and wait on the commit.
2905 */
2906 int ext4_force_commit(struct super_block *sb)
2907 {
2908 journal_t *journal;
2909 int ret;
2910
2911 if (sb->s_flags & MS_RDONLY)
2912 return 0;
2913
2914 journal = EXT4_SB(sb)->s_journal;
2915 sb->s_dirt = 0;
2916 ret = ext4_journal_force_commit(journal);
2917 return ret;
2918 }
2919
2920 /*
2921 * Ext4 always journals updates to the superblock itself, so we don't
2922 * have to propagate any other updates to the superblock on disk at this
2923 * point. (We can probably nuke this function altogether, and remove
2924 * any mention to sb->s_dirt in all of fs/ext4; eventual cleanup...)
2925 */
2926 static void ext4_write_super(struct super_block *sb)
2927 {
2928 if (mutex_trylock(&sb->s_lock) != 0)
2929 BUG();
2930 sb->s_dirt = 0;
2931 }
2932
2933 static int ext4_sync_fs(struct super_block *sb, int wait)
2934 {
2935 int ret = 0;
2936
2937 sb->s_dirt = 0;
2938 if (wait)
2939 ret = ext4_force_commit(sb);
2940 else
2941 jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, NULL);
2942 return ret;
2943 }
2944
2945 /*
2946 * LVM calls this function before a (read-only) snapshot is created. This
2947 * gives us a chance to flush the journal completely and mark the fs clean.
2948 */
2949 static void ext4_write_super_lockfs(struct super_block *sb)
2950 {
2951 sb->s_dirt = 0;
2952
2953 if (!(sb->s_flags & MS_RDONLY)) {
2954 journal_t *journal = EXT4_SB(sb)->s_journal;
2955
2956 /* Now we set up the journal barrier. */
2957 jbd2_journal_lock_updates(journal);
2958 jbd2_journal_flush(journal);
2959
2960 /* Journal blocked and flushed, clear needs_recovery flag. */
2961 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2962 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2963 }
2964 }
2965
2966 /*
2967 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2968 * flag here, even though the filesystem is not technically dirty yet.
2969 */
2970 static void ext4_unlockfs(struct super_block *sb)
2971 {
2972 if (!(sb->s_flags & MS_RDONLY)) {
2973 lock_super(sb);
2974 /* Reser the needs_recovery flag before the fs is unlocked. */
2975 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2976 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2977 unlock_super(sb);
2978 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
2979 }
2980 }
2981
2982 static int ext4_remount(struct super_block *sb, int *flags, char *data)
2983 {
2984 struct ext4_super_block *es;
2985 struct ext4_sb_info *sbi = EXT4_SB(sb);
2986 ext4_fsblk_t n_blocks_count = 0;
2987 unsigned long old_sb_flags;
2988 struct ext4_mount_options old_opts;
2989 ext4_group_t g;
2990 int err;
2991 #ifdef CONFIG_QUOTA
2992 int i;
2993 #endif
2994
2995 /* Store the original options */
2996 old_sb_flags = sb->s_flags;
2997 old_opts.s_mount_opt = sbi->s_mount_opt;
2998 old_opts.s_resuid = sbi->s_resuid;
2999 old_opts.s_resgid = sbi->s_resgid;
3000 old_opts.s_commit_interval = sbi->s_commit_interval;
3001 #ifdef CONFIG_QUOTA
3002 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3003 for (i = 0; i < MAXQUOTAS; i++)
3004 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3005 #endif
3006
3007 /*
3008 * Allow the "check" option to be passed as a remount option.
3009 */
3010 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
3011 err = -EINVAL;
3012 goto restore_opts;
3013 }
3014
3015 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
3016 ext4_abort(sb, __func__, "Abort forced by user");
3017
3018 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3019 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3020
3021 es = sbi->s_es;
3022
3023 ext4_init_journal_params(sb, sbi->s_journal);
3024
3025 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3026 n_blocks_count > ext4_blocks_count(es)) {
3027 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
3028 err = -EROFS;
3029 goto restore_opts;
3030 }
3031
3032 if (*flags & MS_RDONLY) {
3033 /*
3034 * First of all, the unconditional stuff we have to do
3035 * to disable replay of the journal when we next remount
3036 */
3037 sb->s_flags |= MS_RDONLY;
3038
3039 /*
3040 * OK, test if we are remounting a valid rw partition
3041 * readonly, and if so set the rdonly flag and then
3042 * mark the partition as valid again.
3043 */
3044 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3045 (sbi->s_mount_state & EXT4_VALID_FS))
3046 es->s_state = cpu_to_le16(sbi->s_mount_state);
3047
3048 /*
3049 * We have to unlock super so that we can wait for
3050 * transactions.
3051 */
3052 unlock_super(sb);
3053 ext4_mark_recovery_complete(sb, es);
3054 lock_super(sb);
3055 } else {
3056 __le32 ret;
3057 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3058 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
3059 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3060 "remount RDWR because of unsupported "
3061 "optional features (%x).\n",
3062 sb->s_id, le32_to_cpu(ret));
3063 err = -EROFS;
3064 goto restore_opts;
3065 }
3066
3067 /*
3068 * Make sure the group descriptor checksums
3069 * are sane. If they aren't, refuse to
3070 * remount r/w.
3071 */
3072 for (g = 0; g < sbi->s_groups_count; g++) {
3073 struct ext4_group_desc *gdp =
3074 ext4_get_group_desc(sb, g, NULL);
3075
3076 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3077 printk(KERN_ERR
3078 "EXT4-fs: ext4_remount: "
3079 "Checksum for group %lu failed (%u!=%u)\n",
3080 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3081 le16_to_cpu(gdp->bg_checksum));
3082 err = -EINVAL;
3083 goto restore_opts;
3084 }
3085 }
3086
3087 /*
3088 * If we have an unprocessed orphan list hanging
3089 * around from a previously readonly bdev mount,
3090 * require a full umount/remount for now.
3091 */
3092 if (es->s_last_orphan) {
3093 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3094 "remount RDWR because of unprocessed "
3095 "orphan inode list. Please "
3096 "umount/remount instead.\n",
3097 sb->s_id);
3098 err = -EINVAL;
3099 goto restore_opts;
3100 }
3101
3102 /*
3103 * Mounting a RDONLY partition read-write, so reread
3104 * and store the current valid flag. (It may have
3105 * been changed by e2fsck since we originally mounted
3106 * the partition.)
3107 */
3108 ext4_clear_journal_err(sb, es);
3109 sbi->s_mount_state = le16_to_cpu(es->s_state);
3110 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3111 goto restore_opts;
3112 if (!ext4_setup_super(sb, es, 0))
3113 sb->s_flags &= ~MS_RDONLY;
3114 }
3115 }
3116 #ifdef CONFIG_QUOTA
3117 /* Release old quota file names */
3118 for (i = 0; i < MAXQUOTAS; i++)
3119 if (old_opts.s_qf_names[i] &&
3120 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3121 kfree(old_opts.s_qf_names[i]);
3122 #endif
3123 return 0;
3124 restore_opts:
3125 sb->s_flags = old_sb_flags;
3126 sbi->s_mount_opt = old_opts.s_mount_opt;
3127 sbi->s_resuid = old_opts.s_resuid;
3128 sbi->s_resgid = old_opts.s_resgid;
3129 sbi->s_commit_interval = old_opts.s_commit_interval;
3130 #ifdef CONFIG_QUOTA
3131 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3132 for (i = 0; i < MAXQUOTAS; i++) {
3133 if (sbi->s_qf_names[i] &&
3134 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3135 kfree(sbi->s_qf_names[i]);
3136 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3137 }
3138 #endif
3139 return err;
3140 }
3141
3142 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3143 {
3144 struct super_block *sb = dentry->d_sb;
3145 struct ext4_sb_info *sbi = EXT4_SB(sb);
3146 struct ext4_super_block *es = sbi->s_es;
3147 u64 fsid;
3148
3149 if (test_opt(sb, MINIX_DF)) {
3150 sbi->s_overhead_last = 0;
3151 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3152 ext4_group_t ngroups = sbi->s_groups_count, i;
3153 ext4_fsblk_t overhead = 0;
3154 smp_rmb();
3155
3156 /*
3157 * Compute the overhead (FS structures). This is constant
3158 * for a given filesystem unless the number of block groups
3159 * changes so we cache the previous value until it does.
3160 */
3161
3162 /*
3163 * All of the blocks before first_data_block are
3164 * overhead
3165 */
3166 overhead = le32_to_cpu(es->s_first_data_block);
3167
3168 /*
3169 * Add the overhead attributed to the superblock and
3170 * block group descriptors. If the sparse superblocks
3171 * feature is turned on, then not all groups have this.
3172 */
3173 for (i = 0; i < ngroups; i++) {
3174 overhead += ext4_bg_has_super(sb, i) +
3175 ext4_bg_num_gdb(sb, i);
3176 cond_resched();
3177 }
3178
3179 /*
3180 * Every block group has an inode bitmap, a block
3181 * bitmap, and an inode table.
3182 */
3183 overhead += ngroups * (2 + sbi->s_itb_per_group);
3184 sbi->s_overhead_last = overhead;
3185 smp_wmb();
3186 sbi->s_blocks_last = ext4_blocks_count(es);
3187 }
3188
3189 buf->f_type = EXT4_SUPER_MAGIC;
3190 buf->f_bsize = sb->s_blocksize;
3191 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3192 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
3193 ext4_free_blocks_count_set(es, buf->f_bfree);
3194 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3195 if (buf->f_bfree < ext4_r_blocks_count(es))
3196 buf->f_bavail = 0;
3197 buf->f_files = le32_to_cpu(es->s_inodes_count);
3198 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3199 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
3200 buf->f_namelen = EXT4_NAME_LEN;
3201 fsid = le64_to_cpup((void *)es->s_uuid) ^
3202 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3203 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3204 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3205 return 0;
3206 }
3207
3208 /* Helper function for writing quotas on sync - we need to start transaction before quota file
3209 * is locked for write. Otherwise the are possible deadlocks:
3210 * Process 1 Process 2
3211 * ext4_create() quota_sync()
3212 * jbd2_journal_start() write_dquot()
3213 * DQUOT_INIT() down(dqio_mutex)
3214 * down(dqio_mutex) jbd2_journal_start()
3215 *
3216 */
3217
3218 #ifdef CONFIG_QUOTA
3219
3220 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3221 {
3222 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3223 }
3224
3225 static int ext4_dquot_initialize(struct inode *inode, int type)
3226 {
3227 handle_t *handle;
3228 int ret, err;
3229
3230 /* We may create quota structure so we need to reserve enough blocks */
3231 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
3232 if (IS_ERR(handle))
3233 return PTR_ERR(handle);
3234 ret = dquot_initialize(inode, type);
3235 err = ext4_journal_stop(handle);
3236 if (!ret)
3237 ret = err;
3238 return ret;
3239 }
3240
3241 static int ext4_dquot_drop(struct inode *inode)
3242 {
3243 handle_t *handle;
3244 int ret, err;
3245
3246 /* We may delete quota structure so we need to reserve enough blocks */
3247 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
3248 if (IS_ERR(handle)) {
3249 /*
3250 * We call dquot_drop() anyway to at least release references
3251 * to quota structures so that umount does not hang.
3252 */
3253 dquot_drop(inode);
3254 return PTR_ERR(handle);
3255 }
3256 ret = dquot_drop(inode);
3257 err = ext4_journal_stop(handle);
3258 if (!ret)
3259 ret = err;
3260 return ret;
3261 }
3262
3263 static int ext4_write_dquot(struct dquot *dquot)
3264 {
3265 int ret, err;
3266 handle_t *handle;
3267 struct inode *inode;
3268
3269 inode = dquot_to_inode(dquot);
3270 handle = ext4_journal_start(inode,
3271 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3272 if (IS_ERR(handle))
3273 return PTR_ERR(handle);
3274 ret = dquot_commit(dquot);
3275 err = ext4_journal_stop(handle);
3276 if (!ret)
3277 ret = err;
3278 return ret;
3279 }
3280
3281 static int ext4_acquire_dquot(struct dquot *dquot)
3282 {
3283 int ret, err;
3284 handle_t *handle;
3285
3286 handle = ext4_journal_start(dquot_to_inode(dquot),
3287 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3288 if (IS_ERR(handle))
3289 return PTR_ERR(handle);
3290 ret = dquot_acquire(dquot);
3291 err = ext4_journal_stop(handle);
3292 if (!ret)
3293 ret = err;
3294 return ret;
3295 }
3296
3297 static int ext4_release_dquot(struct dquot *dquot)
3298 {
3299 int ret, err;
3300 handle_t *handle;
3301
3302 handle = ext4_journal_start(dquot_to_inode(dquot),
3303 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3304 if (IS_ERR(handle)) {
3305 /* Release dquot anyway to avoid endless cycle in dqput() */
3306 dquot_release(dquot);
3307 return PTR_ERR(handle);
3308 }
3309 ret = dquot_release(dquot);
3310 err = ext4_journal_stop(handle);
3311 if (!ret)
3312 ret = err;
3313 return ret;
3314 }
3315
3316 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3317 {
3318 /* Are we journaling quotas? */
3319 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3320 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3321 dquot_mark_dquot_dirty(dquot);
3322 return ext4_write_dquot(dquot);
3323 } else {
3324 return dquot_mark_dquot_dirty(dquot);
3325 }
3326 }
3327
3328 static int ext4_write_info(struct super_block *sb, int type)
3329 {
3330 int ret, err;
3331 handle_t *handle;
3332
3333 /* Data block + inode block */
3334 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3335 if (IS_ERR(handle))
3336 return PTR_ERR(handle);
3337 ret = dquot_commit_info(sb, type);
3338 err = ext4_journal_stop(handle);
3339 if (!ret)
3340 ret = err;
3341 return ret;
3342 }
3343
3344 /*
3345 * Turn on quotas during mount time - we need to find
3346 * the quota file and such...
3347 */
3348 static int ext4_quota_on_mount(struct super_block *sb, int type)
3349 {
3350 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3351 EXT4_SB(sb)->s_jquota_fmt, type);
3352 }
3353
3354 /*
3355 * Standard function to be called on quota_on
3356 */
3357 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3358 char *path, int remount)
3359 {
3360 int err;
3361 struct nameidata nd;
3362
3363 if (!test_opt(sb, QUOTA))
3364 return -EINVAL;
3365 /* When remounting, no checks are needed and in fact, path is NULL */
3366 if (remount)
3367 return vfs_quota_on(sb, type, format_id, path, remount);
3368
3369 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
3370 if (err)
3371 return err;
3372
3373 /* Quotafile not on the same filesystem? */
3374 if (nd.path.mnt->mnt_sb != sb) {
3375 path_put(&nd.path);
3376 return -EXDEV;
3377 }
3378 /* Journaling quota? */
3379 if (EXT4_SB(sb)->s_qf_names[type]) {
3380 /* Quotafile not in fs root? */
3381 if (nd.path.dentry->d_parent->d_inode != sb->s_root->d_inode)
3382 printk(KERN_WARNING
3383 "EXT4-fs: Quota file not on filesystem root. "
3384 "Journaled quota will not work.\n");
3385 }
3386
3387 /*
3388 * When we journal data on quota file, we have to flush journal to see
3389 * all updates to the file when we bypass pagecache...
3390 */
3391 if (ext4_should_journal_data(nd.path.dentry->d_inode)) {
3392 /*
3393 * We don't need to lock updates but journal_flush() could
3394 * otherwise be livelocked...
3395 */
3396 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3397 jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3398 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3399 }
3400
3401 err = vfs_quota_on_path(sb, type, format_id, &nd.path);
3402 path_put(&nd.path);
3403 return err;
3404 }
3405
3406 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3407 * acquiring the locks... As quota files are never truncated and quota code
3408 * itself serializes the operations (and noone else should touch the files)
3409 * we don't have to be afraid of races */
3410 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3411 size_t len, loff_t off)
3412 {
3413 struct inode *inode = sb_dqopt(sb)->files[type];
3414 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3415 int err = 0;
3416 int offset = off & (sb->s_blocksize - 1);
3417 int tocopy;
3418 size_t toread;
3419 struct buffer_head *bh;
3420 loff_t i_size = i_size_read(inode);
3421
3422 if (off > i_size)
3423 return 0;
3424 if (off+len > i_size)
3425 len = i_size-off;
3426 toread = len;
3427 while (toread > 0) {
3428 tocopy = sb->s_blocksize - offset < toread ?
3429 sb->s_blocksize - offset : toread;
3430 bh = ext4_bread(NULL, inode, blk, 0, &err);
3431 if (err)
3432 return err;
3433 if (!bh) /* A hole? */
3434 memset(data, 0, tocopy);
3435 else
3436 memcpy(data, bh->b_data+offset, tocopy);
3437 brelse(bh);
3438 offset = 0;
3439 toread -= tocopy;
3440 data += tocopy;
3441 blk++;
3442 }
3443 return len;
3444 }
3445
3446 /* Write to quotafile (we know the transaction is already started and has
3447 * enough credits) */
3448 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3449 const char *data, size_t len, loff_t off)
3450 {
3451 struct inode *inode = sb_dqopt(sb)->files[type];
3452 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3453 int err = 0;
3454 int offset = off & (sb->s_blocksize - 1);
3455 int tocopy;
3456 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3457 size_t towrite = len;
3458 struct buffer_head *bh;
3459 handle_t *handle = journal_current_handle();
3460
3461 if (!handle) {
3462 printk(KERN_WARNING "EXT4-fs: Quota write (off=%Lu, len=%Lu)"
3463 " cancelled because transaction is not started.\n",
3464 (unsigned long long)off, (unsigned long long)len);
3465 return -EIO;
3466 }
3467 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3468 while (towrite > 0) {
3469 tocopy = sb->s_blocksize - offset < towrite ?
3470 sb->s_blocksize - offset : towrite;
3471 bh = ext4_bread(handle, inode, blk, 1, &err);
3472 if (!bh)
3473 goto out;
3474 if (journal_quota) {
3475 err = ext4_journal_get_write_access(handle, bh);
3476 if (err) {
3477 brelse(bh);
3478 goto out;
3479 }
3480 }
3481 lock_buffer(bh);
3482 memcpy(bh->b_data+offset, data, tocopy);
3483 flush_dcache_page(bh->b_page);
3484 unlock_buffer(bh);
3485 if (journal_quota)
3486 err = ext4_journal_dirty_metadata(handle, bh);
3487 else {
3488 /* Always do at least ordered writes for quotas */
3489 err = ext4_jbd2_file_inode(handle, inode);
3490 mark_buffer_dirty(bh);
3491 }
3492 brelse(bh);
3493 if (err)
3494 goto out;
3495 offset = 0;
3496 towrite -= tocopy;
3497 data += tocopy;
3498 blk++;
3499 }
3500 out:
3501 if (len == towrite) {
3502 mutex_unlock(&inode->i_mutex);
3503 return err;
3504 }
3505 if (inode->i_size < off+len-towrite) {
3506 i_size_write(inode, off+len-towrite);
3507 EXT4_I(inode)->i_disksize = inode->i_size;
3508 }
3509 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3510 ext4_mark_inode_dirty(handle, inode);
3511 mutex_unlock(&inode->i_mutex);
3512 return len - towrite;
3513 }
3514
3515 #endif
3516
3517 static int ext4_get_sb(struct file_system_type *fs_type,
3518 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3519 {
3520 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3521 }
3522
3523 static struct file_system_type ext4dev_fs_type = {
3524 .owner = THIS_MODULE,
3525 .name = "ext4dev",
3526 .get_sb = ext4_get_sb,
3527 .kill_sb = kill_block_super,
3528 .fs_flags = FS_REQUIRES_DEV,
3529 };
3530
3531 static int __init init_ext4_fs(void)
3532 {
3533 int err;
3534
3535 err = init_ext4_mballoc();
3536 if (err)
3537 return err;
3538
3539 err = init_ext4_xattr();
3540 if (err)
3541 goto out2;
3542 err = init_inodecache();
3543 if (err)
3544 goto out1;
3545 err = register_filesystem(&ext4dev_fs_type);
3546 if (err)
3547 goto out;
3548 return 0;
3549 out:
3550 destroy_inodecache();
3551 out1:
3552 exit_ext4_xattr();
3553 out2:
3554 exit_ext4_mballoc();
3555 return err;
3556 }
3557
3558 static void __exit exit_ext4_fs(void)
3559 {
3560 unregister_filesystem(&ext4dev_fs_type);
3561 destroy_inodecache();
3562 exit_ext4_xattr();
3563 exit_ext4_mballoc();
3564 }
3565
3566 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3567 MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
3568 MODULE_LICENSE("GPL");
3569 module_init(init_ext4_fs)
3570 module_exit(exit_ext4_fs)
Linux 2.6 ibm-acpi/thinkpad-acpi driver git tree