2 * inode.c - NTFS kernel inode handling.
4 * Copyright (c) 2001-2014 Anton Altaparmakov and Tuxera Inc.
6 * This program/include file is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License as published
8 * by the Free Software Foundation; either version 2 of the License, or
9 * (at your option) any later version.
11 * This program/include file is distributed in the hope that it will be
12 * useful, but WITHOUT ANY WARRANTY; without even the implied warranty
13 * of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 * GNU General Public License for more details.
16 * You should have received a copy of the GNU General Public License
17 * along with this program (in the main directory of the Linux-NTFS
18 * distribution in the file COPYING); if not, write to the Free Software
19 * Foundation,Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
22 #include <linux/buffer_head.h>
25 #include <linux/mount.h>
26 #include <linux/mutex.h>
27 #include <linux/pagemap.h>
28 #include <linux/quotaops.h>
29 #include <linux/slab.h>
30 #include <linux/log2.h>
31 #include <linux/aio.h>
46 * ntfs_test_inode - compare two (possibly fake) inodes for equality
47 * @vi: vfs inode which to test
48 * @na: ntfs attribute which is being tested with
50 * Compare the ntfs attribute embedded in the ntfs specific part of the vfs
51 * inode @vi for equality with the ntfs attribute @na.
53 * If searching for the normal file/directory inode, set @na->type to AT_UNUSED.
54 * @na->name and @na->name_len are then ignored.
56 * Return 1 if the attributes match and 0 if not.
58 * NOTE: This function runs with the inode_hash_lock spin lock held so it is not
61 int ntfs_test_inode(struct inode
*vi
, ntfs_attr
*na
)
65 if (vi
->i_ino
!= na
->mft_no
)
68 /* If !NInoAttr(ni), @vi is a normal file or directory inode. */
69 if (likely(!NInoAttr(ni
))) {
70 /* If not looking for a normal inode this is a mismatch. */
71 if (unlikely(na
->type
!= AT_UNUSED
))
74 /* A fake inode describing an attribute. */
75 if (ni
->type
!= na
->type
)
77 if (ni
->name_len
!= na
->name_len
)
79 if (na
->name_len
&& memcmp(ni
->name
, na
->name
,
80 na
->name_len
* sizeof(ntfschar
)))
88 * ntfs_init_locked_inode - initialize an inode
89 * @vi: vfs inode to initialize
90 * @na: ntfs attribute which to initialize @vi to
92 * Initialize the vfs inode @vi with the values from the ntfs attribute @na in
93 * order to enable ntfs_test_inode() to do its work.
95 * If initializing the normal file/directory inode, set @na->type to AT_UNUSED.
96 * In that case, @na->name and @na->name_len should be set to NULL and 0,
97 * respectively. Although that is not strictly necessary as
98 * ntfs_read_locked_inode() will fill them in later.
100 * Return 0 on success and -errno on error.
102 * NOTE: This function runs with the inode->i_lock spin lock held so it is not
103 * allowed to sleep. (Hence the GFP_ATOMIC allocation.)
105 static int ntfs_init_locked_inode(struct inode
*vi
, ntfs_attr
*na
)
107 ntfs_inode
*ni
= NTFS_I(vi
);
109 vi
->i_ino
= na
->mft_no
;
112 if (na
->type
== AT_INDEX_ALLOCATION
)
113 NInoSetMstProtected(ni
);
116 ni
->name_len
= na
->name_len
;
118 /* If initializing a normal inode, we are done. */
119 if (likely(na
->type
== AT_UNUSED
)) {
121 BUG_ON(na
->name_len
);
125 /* It is a fake inode. */
129 * We have I30 global constant as an optimization as it is the name
130 * in >99.9% of named attributes! The other <0.1% incur a GFP_ATOMIC
131 * allocation but that is ok. And most attributes are unnamed anyway,
132 * thus the fraction of named attributes with name != I30 is actually
135 if (na
->name_len
&& na
->name
!= I30
) {
139 i
= na
->name_len
* sizeof(ntfschar
);
140 ni
->name
= kmalloc(i
+ sizeof(ntfschar
), GFP_ATOMIC
);
143 memcpy(ni
->name
, na
->name
, i
);
144 ni
->name
[na
->name_len
] = 0;
149 typedef int (*set_t
)(struct inode
*, void *);
150 static int ntfs_read_locked_inode(struct inode
*vi
);
151 static int ntfs_read_locked_attr_inode(struct inode
*base_vi
, struct inode
*vi
);
152 static int ntfs_read_locked_index_inode(struct inode
*base_vi
,
156 * ntfs_iget - obtain a struct inode corresponding to a specific normal inode
157 * @sb: super block of mounted volume
158 * @mft_no: mft record number / inode number to obtain
160 * Obtain the struct inode corresponding to a specific normal inode (i.e. a
161 * file or directory).
163 * If the inode is in the cache, it is just returned with an increased
164 * reference count. Otherwise, a new struct inode is allocated and initialized,
165 * and finally ntfs_read_locked_inode() is called to read in the inode and
166 * fill in the remainder of the inode structure.
168 * Return the struct inode on success. Check the return value with IS_ERR() and
169 * if true, the function failed and the error code is obtained from PTR_ERR().
171 struct inode
*ntfs_iget(struct super_block
*sb
, unsigned long mft_no
)
182 vi
= iget5_locked(sb
, mft_no
, (test_t
)ntfs_test_inode
,
183 (set_t
)ntfs_init_locked_inode
, &na
);
185 return ERR_PTR(-ENOMEM
);
189 /* If this is a freshly allocated inode, need to read it now. */
190 if (vi
->i_state
& I_NEW
) {
191 err
= ntfs_read_locked_inode(vi
);
192 unlock_new_inode(vi
);
195 * There is no point in keeping bad inodes around if the failure was
196 * due to ENOMEM. We want to be able to retry again later.
198 if (unlikely(err
== -ENOMEM
)) {
206 * ntfs_attr_iget - obtain a struct inode corresponding to an attribute
207 * @base_vi: vfs base inode containing the attribute
208 * @type: attribute type
209 * @name: Unicode name of the attribute (NULL if unnamed)
210 * @name_len: length of @name in Unicode characters (0 if unnamed)
212 * Obtain the (fake) struct inode corresponding to the attribute specified by
213 * @type, @name, and @name_len, which is present in the base mft record
214 * specified by the vfs inode @base_vi.
216 * If the attribute inode is in the cache, it is just returned with an
217 * increased reference count. Otherwise, a new struct inode is allocated and
218 * initialized, and finally ntfs_read_locked_attr_inode() is called to read the
219 * attribute and fill in the inode structure.
221 * Note, for index allocation attributes, you need to use ntfs_index_iget()
222 * instead of ntfs_attr_iget() as working with indices is a lot more complex.
224 * Return the struct inode of the attribute inode on success. Check the return
225 * value with IS_ERR() and if true, the function failed and the error code is
226 * obtained from PTR_ERR().
228 struct inode
*ntfs_attr_iget(struct inode
*base_vi
, ATTR_TYPE type
,
229 ntfschar
*name
, u32 name_len
)
235 /* Make sure no one calls ntfs_attr_iget() for indices. */
236 BUG_ON(type
== AT_INDEX_ALLOCATION
);
238 na
.mft_no
= base_vi
->i_ino
;
241 na
.name_len
= name_len
;
243 vi
= iget5_locked(base_vi
->i_sb
, na
.mft_no
, (test_t
)ntfs_test_inode
,
244 (set_t
)ntfs_init_locked_inode
, &na
);
246 return ERR_PTR(-ENOMEM
);
250 /* If this is a freshly allocated inode, need to read it now. */
251 if (vi
->i_state
& I_NEW
) {
252 err
= ntfs_read_locked_attr_inode(base_vi
, vi
);
253 unlock_new_inode(vi
);
256 * There is no point in keeping bad attribute inodes around. This also
257 * simplifies things in that we never need to check for bad attribute
268 * ntfs_index_iget - obtain a struct inode corresponding to an index
269 * @base_vi: vfs base inode containing the index related attributes
270 * @name: Unicode name of the index
271 * @name_len: length of @name in Unicode characters
273 * Obtain the (fake) struct inode corresponding to the index specified by @name
274 * and @name_len, which is present in the base mft record specified by the vfs
277 * If the index inode is in the cache, it is just returned with an increased
278 * reference count. Otherwise, a new struct inode is allocated and
279 * initialized, and finally ntfs_read_locked_index_inode() is called to read
280 * the index related attributes and fill in the inode structure.
282 * Return the struct inode of the index inode on success. Check the return
283 * value with IS_ERR() and if true, the function failed and the error code is
284 * obtained from PTR_ERR().
286 struct inode
*ntfs_index_iget(struct inode
*base_vi
, ntfschar
*name
,
293 na
.mft_no
= base_vi
->i_ino
;
294 na
.type
= AT_INDEX_ALLOCATION
;
296 na
.name_len
= name_len
;
298 vi
= iget5_locked(base_vi
->i_sb
, na
.mft_no
, (test_t
)ntfs_test_inode
,
299 (set_t
)ntfs_init_locked_inode
, &na
);
301 return ERR_PTR(-ENOMEM
);
305 /* If this is a freshly allocated inode, need to read it now. */
306 if (vi
->i_state
& I_NEW
) {
307 err
= ntfs_read_locked_index_inode(base_vi
, vi
);
308 unlock_new_inode(vi
);
311 * There is no point in keeping bad index inodes around. This also
312 * simplifies things in that we never need to check for bad index
322 struct inode
*ntfs_alloc_big_inode(struct super_block
*sb
)
326 ntfs_debug("Entering.");
327 ni
= kmem_cache_alloc(ntfs_big_inode_cache
, GFP_NOFS
);
328 if (likely(ni
!= NULL
)) {
332 ntfs_error(sb
, "Allocation of NTFS big inode structure failed.");
336 static void ntfs_i_callback(struct rcu_head
*head
)
338 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
339 kmem_cache_free(ntfs_big_inode_cache
, NTFS_I(inode
));
342 void ntfs_destroy_big_inode(struct inode
*inode
)
344 ntfs_inode
*ni
= NTFS_I(inode
);
346 ntfs_debug("Entering.");
348 if (!atomic_dec_and_test(&ni
->count
))
350 call_rcu(&inode
->i_rcu
, ntfs_i_callback
);
353 static inline ntfs_inode
*ntfs_alloc_extent_inode(void)
357 ntfs_debug("Entering.");
358 ni
= kmem_cache_alloc(ntfs_inode_cache
, GFP_NOFS
);
359 if (likely(ni
!= NULL
)) {
363 ntfs_error(NULL
, "Allocation of NTFS inode structure failed.");
367 static void ntfs_destroy_extent_inode(ntfs_inode
*ni
)
369 ntfs_debug("Entering.");
371 if (!atomic_dec_and_test(&ni
->count
))
373 kmem_cache_free(ntfs_inode_cache
, ni
);
377 * The attribute runlist lock has separate locking rules from the
378 * normal runlist lock, so split the two lock-classes:
380 static struct lock_class_key attr_list_rl_lock_class
;
383 * __ntfs_init_inode - initialize ntfs specific part of an inode
384 * @sb: super block of mounted volume
385 * @ni: freshly allocated ntfs inode which to initialize
387 * Initialize an ntfs inode to defaults.
389 * NOTE: ni->mft_no, ni->state, ni->type, ni->name, and ni->name_len are left
390 * untouched. Make sure to initialize them elsewhere.
392 * Return zero on success and -ENOMEM on error.
394 void __ntfs_init_inode(struct super_block
*sb
, ntfs_inode
*ni
)
396 ntfs_debug("Entering.");
397 rwlock_init(&ni
->size_lock
);
398 ni
->initialized_size
= ni
->allocated_size
= 0;
400 atomic_set(&ni
->count
, 1);
401 ni
->vol
= NTFS_SB(sb
);
402 ntfs_init_runlist(&ni
->runlist
);
403 mutex_init(&ni
->mrec_lock
);
406 ni
->attr_list_size
= 0;
407 ni
->attr_list
= NULL
;
408 ntfs_init_runlist(&ni
->attr_list_rl
);
409 lockdep_set_class(&ni
->attr_list_rl
.lock
,
410 &attr_list_rl_lock_class
);
411 ni
->itype
.index
.block_size
= 0;
412 ni
->itype
.index
.vcn_size
= 0;
413 ni
->itype
.index
.collation_rule
= 0;
414 ni
->itype
.index
.block_size_bits
= 0;
415 ni
->itype
.index
.vcn_size_bits
= 0;
416 mutex_init(&ni
->extent_lock
);
418 ni
->ext
.base_ntfs_ino
= NULL
;
422 * Extent inodes get MFT-mapped in a nested way, while the base inode
423 * is still mapped. Teach this nesting to the lock validator by creating
424 * a separate class for nested inode's mrec_lock's:
426 static struct lock_class_key extent_inode_mrec_lock_key
;
428 inline ntfs_inode
*ntfs_new_extent_inode(struct super_block
*sb
,
429 unsigned long mft_no
)
431 ntfs_inode
*ni
= ntfs_alloc_extent_inode();
433 ntfs_debug("Entering.");
434 if (likely(ni
!= NULL
)) {
435 __ntfs_init_inode(sb
, ni
);
436 lockdep_set_class(&ni
->mrec_lock
, &extent_inode_mrec_lock_key
);
438 ni
->type
= AT_UNUSED
;
446 * ntfs_is_extended_system_file - check if a file is in the $Extend directory
447 * @ctx: initialized attribute search context
449 * Search all file name attributes in the inode described by the attribute
450 * search context @ctx and check if any of the names are in the $Extend system
454 * 1: file is in $Extend directory
455 * 0: file is not in $Extend directory
456 * -errno: failed to determine if the file is in the $Extend directory
458 static int ntfs_is_extended_system_file(ntfs_attr_search_ctx
*ctx
)
462 /* Restart search. */
463 ntfs_attr_reinit_search_ctx(ctx
);
465 /* Get number of hard links. */
466 nr_links
= le16_to_cpu(ctx
->mrec
->link_count
);
468 /* Loop through all hard links. */
469 while (!(err
= ntfs_attr_lookup(AT_FILE_NAME
, NULL
, 0, 0, 0, NULL
, 0,
471 FILE_NAME_ATTR
*file_name_attr
;
472 ATTR_RECORD
*attr
= ctx
->attr
;
477 * Maximum sanity checking as we are called on an inode that
478 * we suspect might be corrupt.
480 p
= (u8
*)attr
+ le32_to_cpu(attr
->length
);
481 if (p
< (u8
*)ctx
->mrec
|| (u8
*)p
> (u8
*)ctx
->mrec
+
482 le32_to_cpu(ctx
->mrec
->bytes_in_use
)) {
484 ntfs_error(ctx
->ntfs_ino
->vol
->sb
, "Corrupt file name "
485 "attribute. You should run chkdsk.");
488 if (attr
->non_resident
) {
489 ntfs_error(ctx
->ntfs_ino
->vol
->sb
, "Non-resident file "
490 "name. You should run chkdsk.");
494 ntfs_error(ctx
->ntfs_ino
->vol
->sb
, "File name with "
495 "invalid flags. You should run "
499 if (!(attr
->data
.resident
.flags
& RESIDENT_ATTR_IS_INDEXED
)) {
500 ntfs_error(ctx
->ntfs_ino
->vol
->sb
, "Unindexed file "
501 "name. You should run chkdsk.");
504 file_name_attr
= (FILE_NAME_ATTR
*)((u8
*)attr
+
505 le16_to_cpu(attr
->data
.resident
.value_offset
));
506 p2
= (u8
*)attr
+ le32_to_cpu(attr
->data
.resident
.value_length
);
507 if (p2
< (u8
*)attr
|| p2
> p
)
508 goto err_corrupt_attr
;
509 /* This attribute is ok, but is it in the $Extend directory? */
510 if (MREF_LE(file_name_attr
->parent_directory
) == FILE_Extend
)
511 return 1; /* YES, it's an extended system file. */
513 if (unlikely(err
!= -ENOENT
))
515 if (unlikely(nr_links
)) {
516 ntfs_error(ctx
->ntfs_ino
->vol
->sb
, "Inode hard link count "
517 "doesn't match number of name attributes. You "
518 "should run chkdsk.");
521 return 0; /* NO, it is not an extended system file. */
525 * ntfs_read_locked_inode - read an inode from its device
528 * ntfs_read_locked_inode() is called from ntfs_iget() to read the inode
529 * described by @vi into memory from the device.
531 * The only fields in @vi that we need to/can look at when the function is
532 * called are i_sb, pointing to the mounted device's super block, and i_ino,
533 * the number of the inode to load.
535 * ntfs_read_locked_inode() maps, pins and locks the mft record number i_ino
536 * for reading and sets up the necessary @vi fields as well as initializing
539 * Q: What locks are held when the function is called?
540 * A: i_state has I_NEW set, hence the inode is locked, also
541 * i_count is set to 1, so it is not going to go away
542 * i_flags is set to 0 and we have no business touching it. Only an ioctl()
543 * is allowed to write to them. We should of course be honouring them but
544 * we need to do that using the IS_* macros defined in include/linux/fs.h.
545 * In any case ntfs_read_locked_inode() has nothing to do with i_flags.
547 * Return 0 on success and -errno on error. In the error case, the inode will
548 * have had make_bad_inode() executed on it.
550 static int ntfs_read_locked_inode(struct inode
*vi
)
552 ntfs_volume
*vol
= NTFS_SB(vi
->i_sb
);
557 STANDARD_INFORMATION
*si
;
558 ntfs_attr_search_ctx
*ctx
;
561 ntfs_debug("Entering for i_ino 0x%lx.", vi
->i_ino
);
563 /* Setup the generic vfs inode parts now. */
566 * This is for checking whether an inode has changed w.r.t. a file so
567 * that the file can be updated if necessary (compare with f_version).
571 vi
->i_uid
= vol
->uid
;
572 vi
->i_gid
= vol
->gid
;
576 * Initialize the ntfs specific part of @vi special casing
577 * FILE_MFT which we need to do at mount time.
579 if (vi
->i_ino
!= FILE_MFT
)
580 ntfs_init_big_inode(vi
);
583 m
= map_mft_record(ni
);
588 ctx
= ntfs_attr_get_search_ctx(ni
, m
);
594 if (!(m
->flags
& MFT_RECORD_IN_USE
)) {
595 ntfs_error(vi
->i_sb
, "Inode is not in use!");
598 if (m
->base_mft_record
) {
599 ntfs_error(vi
->i_sb
, "Inode is an extent inode!");
603 /* Transfer information from mft record into vfs and ntfs inodes. */
604 vi
->i_generation
= ni
->seq_no
= le16_to_cpu(m
->sequence_number
);
607 * FIXME: Keep in mind that link_count is two for files which have both
608 * a long file name and a short file name as separate entries, so if
609 * we are hiding short file names this will be too high. Either we need
610 * to account for the short file names by subtracting them or we need
611 * to make sure we delete files even though i_nlink is not zero which
612 * might be tricky due to vfs interactions. Need to think about this
613 * some more when implementing the unlink command.
615 set_nlink(vi
, le16_to_cpu(m
->link_count
));
617 * FIXME: Reparse points can have the directory bit set even though
618 * they would be S_IFLNK. Need to deal with this further below when we
619 * implement reparse points / symbolic links but it will do for now.
620 * Also if not a directory, it could be something else, rather than
621 * a regular file. But again, will do for now.
623 /* Everyone gets all permissions. */
624 vi
->i_mode
|= S_IRWXUGO
;
625 /* If read-only, no one gets write permissions. */
627 vi
->i_mode
&= ~S_IWUGO
;
628 if (m
->flags
& MFT_RECORD_IS_DIRECTORY
) {
629 vi
->i_mode
|= S_IFDIR
;
631 * Apply the directory permissions mask set in the mount
634 vi
->i_mode
&= ~vol
->dmask
;
635 /* Things break without this kludge! */
639 vi
->i_mode
|= S_IFREG
;
640 /* Apply the file permissions mask set in the mount options. */
641 vi
->i_mode
&= ~vol
->fmask
;
644 * Find the standard information attribute in the mft record. At this
645 * stage we haven't setup the attribute list stuff yet, so this could
646 * in fact fail if the standard information is in an extent record, but
647 * I don't think this actually ever happens.
649 err
= ntfs_attr_lookup(AT_STANDARD_INFORMATION
, NULL
, 0, 0, 0, NULL
, 0,
652 if (err
== -ENOENT
) {
654 * TODO: We should be performing a hot fix here (if the
655 * recover mount option is set) by creating a new
658 ntfs_error(vi
->i_sb
, "$STANDARD_INFORMATION attribute "
664 /* Get the standard information attribute value. */
665 si
= (STANDARD_INFORMATION
*)((u8
*)a
+
666 le16_to_cpu(a
->data
.resident
.value_offset
));
668 /* Transfer information from the standard information into vi. */
670 * Note: The i_?times do not quite map perfectly onto the NTFS times,
671 * but they are close enough, and in the end it doesn't really matter
675 * mtime is the last change of the data within the file. Not changed
676 * when only metadata is changed, e.g. a rename doesn't affect mtime.
678 vi
->i_mtime
= ntfs2utc(si
->last_data_change_time
);
680 * ctime is the last change of the metadata of the file. This obviously
681 * always changes, when mtime is changed. ctime can be changed on its
682 * own, mtime is then not changed, e.g. when a file is renamed.
684 vi
->i_ctime
= ntfs2utc(si
->last_mft_change_time
);
686 * Last access to the data within the file. Not changed during a rename
687 * for example but changed whenever the file is written to.
689 vi
->i_atime
= ntfs2utc(si
->last_access_time
);
691 /* Find the attribute list attribute if present. */
692 ntfs_attr_reinit_search_ctx(ctx
);
693 err
= ntfs_attr_lookup(AT_ATTRIBUTE_LIST
, NULL
, 0, 0, 0, NULL
, 0, ctx
);
695 if (unlikely(err
!= -ENOENT
)) {
696 ntfs_error(vi
->i_sb
, "Failed to lookup attribute list "
700 } else /* if (!err) */ {
701 if (vi
->i_ino
== FILE_MFT
)
702 goto skip_attr_list_load
;
703 ntfs_debug("Attribute list found in inode 0x%lx.", vi
->i_ino
);
706 if (a
->flags
& ATTR_COMPRESSION_MASK
) {
707 ntfs_error(vi
->i_sb
, "Attribute list attribute is "
711 if (a
->flags
& ATTR_IS_ENCRYPTED
||
712 a
->flags
& ATTR_IS_SPARSE
) {
713 if (a
->non_resident
) {
714 ntfs_error(vi
->i_sb
, "Non-resident attribute "
715 "list attribute is encrypted/"
719 ntfs_warning(vi
->i_sb
, "Resident attribute list "
720 "attribute in inode 0x%lx is marked "
721 "encrypted/sparse which is not true. "
722 "However, Windows allows this and "
723 "chkdsk does not detect or correct it "
724 "so we will just ignore the invalid "
725 "flags and pretend they are not set.",
728 /* Now allocate memory for the attribute list. */
729 ni
->attr_list_size
= (u32
)ntfs_attr_size(a
);
730 ni
->attr_list
= ntfs_malloc_nofs(ni
->attr_list_size
);
731 if (!ni
->attr_list
) {
732 ntfs_error(vi
->i_sb
, "Not enough memory to allocate "
733 "buffer for attribute list.");
737 if (a
->non_resident
) {
738 NInoSetAttrListNonResident(ni
);
739 if (a
->data
.non_resident
.lowest_vcn
) {
740 ntfs_error(vi
->i_sb
, "Attribute list has non "
745 * Setup the runlist. No need for locking as we have
746 * exclusive access to the inode at this time.
748 ni
->attr_list_rl
.rl
= ntfs_mapping_pairs_decompress(vol
,
750 if (IS_ERR(ni
->attr_list_rl
.rl
)) {
751 err
= PTR_ERR(ni
->attr_list_rl
.rl
);
752 ni
->attr_list_rl
.rl
= NULL
;
753 ntfs_error(vi
->i_sb
, "Mapping pairs "
754 "decompression failed.");
757 /* Now load the attribute list. */
758 if ((err
= load_attribute_list(vol
, &ni
->attr_list_rl
,
759 ni
->attr_list
, ni
->attr_list_size
,
760 sle64_to_cpu(a
->data
.non_resident
.
761 initialized_size
)))) {
762 ntfs_error(vi
->i_sb
, "Failed to load "
763 "attribute list attribute.");
766 } else /* if (!a->non_resident) */ {
767 if ((u8
*)a
+ le16_to_cpu(a
->data
.resident
.value_offset
)
769 a
->data
.resident
.value_length
) >
770 (u8
*)ctx
->mrec
+ vol
->mft_record_size
) {
771 ntfs_error(vi
->i_sb
, "Corrupt attribute list "
775 /* Now copy the attribute list. */
776 memcpy(ni
->attr_list
, (u8
*)a
+ le16_to_cpu(
777 a
->data
.resident
.value_offset
),
779 a
->data
.resident
.value_length
));
784 * If an attribute list is present we now have the attribute list value
785 * in ntfs_ino->attr_list and it is ntfs_ino->attr_list_size bytes.
787 if (S_ISDIR(vi
->i_mode
)) {
791 u8
*ir_end
, *index_end
;
793 /* It is a directory, find index root attribute. */
794 ntfs_attr_reinit_search_ctx(ctx
);
795 err
= ntfs_attr_lookup(AT_INDEX_ROOT
, I30
, 4, CASE_SENSITIVE
,
798 if (err
== -ENOENT
) {
799 // FIXME: File is corrupt! Hot-fix with empty
800 // index root attribute if recovery option is
802 ntfs_error(vi
->i_sb
, "$INDEX_ROOT attribute "
808 /* Set up the state. */
809 if (unlikely(a
->non_resident
)) {
810 ntfs_error(vol
->sb
, "$INDEX_ROOT attribute is not "
814 /* Ensure the attribute name is placed before the value. */
815 if (unlikely(a
->name_length
&& (le16_to_cpu(a
->name_offset
) >=
816 le16_to_cpu(a
->data
.resident
.value_offset
)))) {
817 ntfs_error(vol
->sb
, "$INDEX_ROOT attribute name is "
818 "placed after the attribute value.");
822 * Compressed/encrypted index root just means that the newly
823 * created files in that directory should be created compressed/
824 * encrypted. However index root cannot be both compressed and
827 if (a
->flags
& ATTR_COMPRESSION_MASK
)
828 NInoSetCompressed(ni
);
829 if (a
->flags
& ATTR_IS_ENCRYPTED
) {
830 if (a
->flags
& ATTR_COMPRESSION_MASK
) {
831 ntfs_error(vi
->i_sb
, "Found encrypted and "
832 "compressed attribute.");
835 NInoSetEncrypted(ni
);
837 if (a
->flags
& ATTR_IS_SPARSE
)
839 ir
= (INDEX_ROOT
*)((u8
*)a
+
840 le16_to_cpu(a
->data
.resident
.value_offset
));
841 ir_end
= (u8
*)ir
+ le32_to_cpu(a
->data
.resident
.value_length
);
842 if (ir_end
> (u8
*)ctx
->mrec
+ vol
->mft_record_size
) {
843 ntfs_error(vi
->i_sb
, "$INDEX_ROOT attribute is "
847 index_end
= (u8
*)&ir
->index
+
848 le32_to_cpu(ir
->index
.index_length
);
849 if (index_end
> ir_end
) {
850 ntfs_error(vi
->i_sb
, "Directory index is corrupt.");
853 if (ir
->type
!= AT_FILE_NAME
) {
854 ntfs_error(vi
->i_sb
, "Indexed attribute is not "
858 if (ir
->collation_rule
!= COLLATION_FILE_NAME
) {
859 ntfs_error(vi
->i_sb
, "Index collation rule is not "
860 "COLLATION_FILE_NAME.");
863 ni
->itype
.index
.collation_rule
= ir
->collation_rule
;
864 ni
->itype
.index
.block_size
= le32_to_cpu(ir
->index_block_size
);
865 if (ni
->itype
.index
.block_size
&
866 (ni
->itype
.index
.block_size
- 1)) {
867 ntfs_error(vi
->i_sb
, "Index block size (%u) is not a "
869 ni
->itype
.index
.block_size
);
872 if (ni
->itype
.index
.block_size
> PAGE_CACHE_SIZE
) {
873 ntfs_error(vi
->i_sb
, "Index block size (%u) > "
874 "PAGE_CACHE_SIZE (%ld) is not "
876 ni
->itype
.index
.block_size
,
881 if (ni
->itype
.index
.block_size
< NTFS_BLOCK_SIZE
) {
882 ntfs_error(vi
->i_sb
, "Index block size (%u) < "
883 "NTFS_BLOCK_SIZE (%i) is not "
885 ni
->itype
.index
.block_size
,
890 ni
->itype
.index
.block_size_bits
=
891 ffs(ni
->itype
.index
.block_size
) - 1;
892 /* Determine the size of a vcn in the directory index. */
893 if (vol
->cluster_size
<= ni
->itype
.index
.block_size
) {
894 ni
->itype
.index
.vcn_size
= vol
->cluster_size
;
895 ni
->itype
.index
.vcn_size_bits
= vol
->cluster_size_bits
;
897 ni
->itype
.index
.vcn_size
= vol
->sector_size
;
898 ni
->itype
.index
.vcn_size_bits
= vol
->sector_size_bits
;
901 /* Setup the index allocation attribute, even if not present. */
902 NInoSetMstProtected(ni
);
903 ni
->type
= AT_INDEX_ALLOCATION
;
907 if (!(ir
->index
.flags
& LARGE_INDEX
)) {
908 /* No index allocation. */
909 vi
->i_size
= ni
->initialized_size
=
910 ni
->allocated_size
= 0;
911 /* We are done with the mft record, so we release it. */
912 ntfs_attr_put_search_ctx(ctx
);
913 unmap_mft_record(ni
);
916 goto skip_large_dir_stuff
;
917 } /* LARGE_INDEX: Index allocation present. Setup state. */
918 NInoSetIndexAllocPresent(ni
);
919 /* Find index allocation attribute. */
920 ntfs_attr_reinit_search_ctx(ctx
);
921 err
= ntfs_attr_lookup(AT_INDEX_ALLOCATION
, I30
, 4,
922 CASE_SENSITIVE
, 0, NULL
, 0, ctx
);
925 ntfs_error(vi
->i_sb
, "$INDEX_ALLOCATION "
926 "attribute is not present but "
927 "$INDEX_ROOT indicated it is.");
929 ntfs_error(vi
->i_sb
, "Failed to lookup "
935 if (!a
->non_resident
) {
936 ntfs_error(vi
->i_sb
, "$INDEX_ALLOCATION attribute "
941 * Ensure the attribute name is placed before the mapping pairs
944 if (unlikely(a
->name_length
&& (le16_to_cpu(a
->name_offset
) >=
946 a
->data
.non_resident
.mapping_pairs_offset
)))) {
947 ntfs_error(vol
->sb
, "$INDEX_ALLOCATION attribute name "
948 "is placed after the mapping pairs "
952 if (a
->flags
& ATTR_IS_ENCRYPTED
) {
953 ntfs_error(vi
->i_sb
, "$INDEX_ALLOCATION attribute "
957 if (a
->flags
& ATTR_IS_SPARSE
) {
958 ntfs_error(vi
->i_sb
, "$INDEX_ALLOCATION attribute "
962 if (a
->flags
& ATTR_COMPRESSION_MASK
) {
963 ntfs_error(vi
->i_sb
, "$INDEX_ALLOCATION attribute "
967 if (a
->data
.non_resident
.lowest_vcn
) {
968 ntfs_error(vi
->i_sb
, "First extent of "
969 "$INDEX_ALLOCATION attribute has non "
973 vi
->i_size
= sle64_to_cpu(a
->data
.non_resident
.data_size
);
974 ni
->initialized_size
= sle64_to_cpu(
975 a
->data
.non_resident
.initialized_size
);
976 ni
->allocated_size
= sle64_to_cpu(
977 a
->data
.non_resident
.allocated_size
);
979 * We are done with the mft record, so we release it. Otherwise
980 * we would deadlock in ntfs_attr_iget().
982 ntfs_attr_put_search_ctx(ctx
);
983 unmap_mft_record(ni
);
986 /* Get the index bitmap attribute inode. */
987 bvi
= ntfs_attr_iget(vi
, AT_BITMAP
, I30
, 4);
989 ntfs_error(vi
->i_sb
, "Failed to get bitmap attribute.");
994 if (NInoCompressed(bni
) || NInoEncrypted(bni
) ||
996 ntfs_error(vi
->i_sb
, "$BITMAP attribute is compressed "
997 "and/or encrypted and/or sparse.");
998 goto iput_unm_err_out
;
1000 /* Consistency check bitmap size vs. index allocation size. */
1001 bvi_size
= i_size_read(bvi
);
1002 if ((bvi_size
<< 3) < (vi
->i_size
>>
1003 ni
->itype
.index
.block_size_bits
)) {
1004 ntfs_error(vi
->i_sb
, "Index bitmap too small (0x%llx) "
1005 "for index allocation (0x%llx).",
1006 bvi_size
<< 3, vi
->i_size
);
1007 goto iput_unm_err_out
;
1009 /* No longer need the bitmap attribute inode. */
1011 skip_large_dir_stuff
:
1012 /* Setup the operations for this inode. */
1013 vi
->i_op
= &ntfs_dir_inode_ops
;
1014 vi
->i_fop
= &ntfs_dir_ops
;
1015 vi
->i_mapping
->a_ops
= &ntfs_mst_aops
;
1018 ntfs_attr_reinit_search_ctx(ctx
);
1020 /* Setup the data attribute, even if not present. */
1025 /* Find first extent of the unnamed data attribute. */
1026 err
= ntfs_attr_lookup(AT_DATA
, NULL
, 0, 0, 0, NULL
, 0, ctx
);
1027 if (unlikely(err
)) {
1028 vi
->i_size
= ni
->initialized_size
=
1029 ni
->allocated_size
= 0;
1030 if (err
!= -ENOENT
) {
1031 ntfs_error(vi
->i_sb
, "Failed to lookup $DATA "
1036 * FILE_Secure does not have an unnamed $DATA
1037 * attribute, so we special case it here.
1039 if (vi
->i_ino
== FILE_Secure
)
1040 goto no_data_attr_special_case
;
1042 * Most if not all the system files in the $Extend
1043 * system directory do not have unnamed data
1044 * attributes so we need to check if the parent
1045 * directory of the file is FILE_Extend and if it is
1046 * ignore this error. To do this we need to get the
1047 * name of this inode from the mft record as the name
1048 * contains the back reference to the parent directory.
1050 if (ntfs_is_extended_system_file(ctx
) > 0)
1051 goto no_data_attr_special_case
;
1052 // FIXME: File is corrupt! Hot-fix with empty data
1053 // attribute if recovery option is set.
1054 ntfs_error(vi
->i_sb
, "$DATA attribute is missing.");
1058 /* Setup the state. */
1059 if (a
->flags
& (ATTR_COMPRESSION_MASK
| ATTR_IS_SPARSE
)) {
1060 if (a
->flags
& ATTR_COMPRESSION_MASK
) {
1061 NInoSetCompressed(ni
);
1062 if (vol
->cluster_size
> 4096) {
1063 ntfs_error(vi
->i_sb
, "Found "
1064 "compressed data but "
1067 "cluster size (%i) > "
1072 if ((a
->flags
& ATTR_COMPRESSION_MASK
)
1073 != ATTR_IS_COMPRESSED
) {
1074 ntfs_error(vi
->i_sb
, "Found unknown "
1075 "compression method "
1076 "or corrupt file.");
1080 if (a
->flags
& ATTR_IS_SPARSE
)
1083 if (a
->flags
& ATTR_IS_ENCRYPTED
) {
1084 if (NInoCompressed(ni
)) {
1085 ntfs_error(vi
->i_sb
, "Found encrypted and "
1086 "compressed data.");
1089 NInoSetEncrypted(ni
);
1091 if (a
->non_resident
) {
1092 NInoSetNonResident(ni
);
1093 if (NInoCompressed(ni
) || NInoSparse(ni
)) {
1094 if (NInoCompressed(ni
) && a
->data
.non_resident
.
1095 compression_unit
!= 4) {
1096 ntfs_error(vi
->i_sb
, "Found "
1098 "compression unit (%u "
1100 "Cannot handle this.",
1101 a
->data
.non_resident
.
1106 if (a
->data
.non_resident
.compression_unit
) {
1107 ni
->itype
.compressed
.block_size
= 1U <<
1108 (a
->data
.non_resident
.
1110 vol
->cluster_size_bits
);
1111 ni
->itype
.compressed
.block_size_bits
=
1115 ni
->itype
.compressed
.block_clusters
=
1120 ni
->itype
.compressed
.block_size
= 0;
1121 ni
->itype
.compressed
.block_size_bits
=
1123 ni
->itype
.compressed
.block_clusters
=
1126 ni
->itype
.compressed
.size
= sle64_to_cpu(
1127 a
->data
.non_resident
.
1130 if (a
->data
.non_resident
.lowest_vcn
) {
1131 ntfs_error(vi
->i_sb
, "First extent of $DATA "
1132 "attribute has non zero "
1136 vi
->i_size
= sle64_to_cpu(
1137 a
->data
.non_resident
.data_size
);
1138 ni
->initialized_size
= sle64_to_cpu(
1139 a
->data
.non_resident
.initialized_size
);
1140 ni
->allocated_size
= sle64_to_cpu(
1141 a
->data
.non_resident
.allocated_size
);
1142 } else { /* Resident attribute. */
1143 vi
->i_size
= ni
->initialized_size
= le32_to_cpu(
1144 a
->data
.resident
.value_length
);
1145 ni
->allocated_size
= le32_to_cpu(a
->length
) -
1147 a
->data
.resident
.value_offset
);
1148 if (vi
->i_size
> ni
->allocated_size
) {
1149 ntfs_error(vi
->i_sb
, "Resident data attribute "
1150 "is corrupt (size exceeds "
1155 no_data_attr_special_case
:
1156 /* We are done with the mft record, so we release it. */
1157 ntfs_attr_put_search_ctx(ctx
);
1158 unmap_mft_record(ni
);
1161 /* Setup the operations for this inode. */
1162 vi
->i_op
= &ntfs_file_inode_ops
;
1163 vi
->i_fop
= &ntfs_file_ops
;
1164 vi
->i_mapping
->a_ops
= &ntfs_normal_aops
;
1165 if (NInoMstProtected(ni
))
1166 vi
->i_mapping
->a_ops
= &ntfs_mst_aops
;
1167 else if (NInoCompressed(ni
))
1168 vi
->i_mapping
->a_ops
= &ntfs_compressed_aops
;
1171 * The number of 512-byte blocks used on disk (for stat). This is in so
1172 * far inaccurate as it doesn't account for any named streams or other
1173 * special non-resident attributes, but that is how Windows works, too,
1174 * so we are at least consistent with Windows, if not entirely
1175 * consistent with the Linux Way. Doing it the Linux Way would cause a
1176 * significant slowdown as it would involve iterating over all
1177 * attributes in the mft record and adding the allocated/compressed
1178 * sizes of all non-resident attributes present to give us the Linux
1179 * correct size that should go into i_blocks (after division by 512).
1181 if (S_ISREG(vi
->i_mode
) && (NInoCompressed(ni
) || NInoSparse(ni
)))
1182 vi
->i_blocks
= ni
->itype
.compressed
.size
>> 9;
1184 vi
->i_blocks
= ni
->allocated_size
>> 9;
1185 ntfs_debug("Done.");
1193 ntfs_attr_put_search_ctx(ctx
);
1195 unmap_mft_record(ni
);
1197 ntfs_error(vol
->sb
, "Failed with error code %i. Marking corrupt "
1198 "inode 0x%lx as bad. Run chkdsk.", err
, vi
->i_ino
);
1200 if (err
!= -EOPNOTSUPP
&& err
!= -ENOMEM
)
1206 * ntfs_read_locked_attr_inode - read an attribute inode from its base inode
1207 * @base_vi: base inode
1208 * @vi: attribute inode to read
1210 * ntfs_read_locked_attr_inode() is called from ntfs_attr_iget() to read the
1211 * attribute inode described by @vi into memory from the base mft record
1212 * described by @base_ni.
1214 * ntfs_read_locked_attr_inode() maps, pins and locks the base inode for
1215 * reading and looks up the attribute described by @vi before setting up the
1216 * necessary fields in @vi as well as initializing the ntfs inode.
1218 * Q: What locks are held when the function is called?
1219 * A: i_state has I_NEW set, hence the inode is locked, also
1220 * i_count is set to 1, so it is not going to go away
1222 * Return 0 on success and -errno on error. In the error case, the inode will
1223 * have had make_bad_inode() executed on it.
1225 * Note this cannot be called for AT_INDEX_ALLOCATION.
1227 static int ntfs_read_locked_attr_inode(struct inode
*base_vi
, struct inode
*vi
)
1229 ntfs_volume
*vol
= NTFS_SB(vi
->i_sb
);
1230 ntfs_inode
*ni
, *base_ni
;
1233 ntfs_attr_search_ctx
*ctx
;
1236 ntfs_debug("Entering for i_ino 0x%lx.", vi
->i_ino
);
1238 ntfs_init_big_inode(vi
);
1241 base_ni
= NTFS_I(base_vi
);
1243 /* Just mirror the values from the base inode. */
1244 vi
->i_version
= base_vi
->i_version
;
1245 vi
->i_uid
= base_vi
->i_uid
;
1246 vi
->i_gid
= base_vi
->i_gid
;
1247 set_nlink(vi
, base_vi
->i_nlink
);
1248 vi
->i_mtime
= base_vi
->i_mtime
;
1249 vi
->i_ctime
= base_vi
->i_ctime
;
1250 vi
->i_atime
= base_vi
->i_atime
;
1251 vi
->i_generation
= ni
->seq_no
= base_ni
->seq_no
;
1253 /* Set inode type to zero but preserve permissions. */
1254 vi
->i_mode
= base_vi
->i_mode
& ~S_IFMT
;
1256 m
= map_mft_record(base_ni
);
1261 ctx
= ntfs_attr_get_search_ctx(base_ni
, m
);
1266 /* Find the attribute. */
1267 err
= ntfs_attr_lookup(ni
->type
, ni
->name
, ni
->name_len
,
1268 CASE_SENSITIVE
, 0, NULL
, 0, ctx
);
1272 if (a
->flags
& (ATTR_COMPRESSION_MASK
| ATTR_IS_SPARSE
)) {
1273 if (a
->flags
& ATTR_COMPRESSION_MASK
) {
1274 NInoSetCompressed(ni
);
1275 if ((ni
->type
!= AT_DATA
) || (ni
->type
== AT_DATA
&&
1277 ntfs_error(vi
->i_sb
, "Found compressed "
1278 "non-data or named data "
1279 "attribute. Please report "
1280 "you saw this message to "
1281 "linux-ntfs-dev@lists."
1285 if (vol
->cluster_size
> 4096) {
1286 ntfs_error(vi
->i_sb
, "Found compressed "
1287 "attribute but compression is "
1288 "disabled due to cluster size "
1293 if ((a
->flags
& ATTR_COMPRESSION_MASK
) !=
1294 ATTR_IS_COMPRESSED
) {
1295 ntfs_error(vi
->i_sb
, "Found unknown "
1296 "compression method.");
1301 * The compressed/sparse flag set in an index root just means
1302 * to compress all files.
1304 if (NInoMstProtected(ni
) && ni
->type
!= AT_INDEX_ROOT
) {
1305 ntfs_error(vi
->i_sb
, "Found mst protected attribute "
1306 "but the attribute is %s. Please "
1307 "report you saw this message to "
1308 "linux-ntfs-dev@lists.sourceforge.net",
1309 NInoCompressed(ni
) ? "compressed" :
1313 if (a
->flags
& ATTR_IS_SPARSE
)
1316 if (a
->flags
& ATTR_IS_ENCRYPTED
) {
1317 if (NInoCompressed(ni
)) {
1318 ntfs_error(vi
->i_sb
, "Found encrypted and compressed "
1323 * The encryption flag set in an index root just means to
1324 * encrypt all files.
1326 if (NInoMstProtected(ni
) && ni
->type
!= AT_INDEX_ROOT
) {
1327 ntfs_error(vi
->i_sb
, "Found mst protected attribute "
1328 "but the attribute is encrypted. "
1329 "Please report you saw this message "
1330 "to linux-ntfs-dev@lists.sourceforge."
1334 if (ni
->type
!= AT_DATA
) {
1335 ntfs_error(vi
->i_sb
, "Found encrypted non-data "
1339 NInoSetEncrypted(ni
);
1341 if (!a
->non_resident
) {
1342 /* Ensure the attribute name is placed before the value. */
1343 if (unlikely(a
->name_length
&& (le16_to_cpu(a
->name_offset
) >=
1344 le16_to_cpu(a
->data
.resident
.value_offset
)))) {
1345 ntfs_error(vol
->sb
, "Attribute name is placed after "
1346 "the attribute value.");
1349 if (NInoMstProtected(ni
)) {
1350 ntfs_error(vi
->i_sb
, "Found mst protected attribute "
1351 "but the attribute is resident. "
1352 "Please report you saw this message to "
1353 "linux-ntfs-dev@lists.sourceforge.net");
1356 vi
->i_size
= ni
->initialized_size
= le32_to_cpu(
1357 a
->data
.resident
.value_length
);
1358 ni
->allocated_size
= le32_to_cpu(a
->length
) -
1359 le16_to_cpu(a
->data
.resident
.value_offset
);
1360 if (vi
->i_size
> ni
->allocated_size
) {
1361 ntfs_error(vi
->i_sb
, "Resident attribute is corrupt "
1362 "(size exceeds allocation).");
1366 NInoSetNonResident(ni
);
1368 * Ensure the attribute name is placed before the mapping pairs
1371 if (unlikely(a
->name_length
&& (le16_to_cpu(a
->name_offset
) >=
1373 a
->data
.non_resident
.mapping_pairs_offset
)))) {
1374 ntfs_error(vol
->sb
, "Attribute name is placed after "
1375 "the mapping pairs array.");
1378 if (NInoCompressed(ni
) || NInoSparse(ni
)) {
1379 if (NInoCompressed(ni
) && a
->data
.non_resident
.
1380 compression_unit
!= 4) {
1381 ntfs_error(vi
->i_sb
, "Found non-standard "
1382 "compression unit (%u instead "
1383 "of 4). Cannot handle this.",
1384 a
->data
.non_resident
.
1389 if (a
->data
.non_resident
.compression_unit
) {
1390 ni
->itype
.compressed
.block_size
= 1U <<
1391 (a
->data
.non_resident
.
1393 vol
->cluster_size_bits
);
1394 ni
->itype
.compressed
.block_size_bits
=
1395 ffs(ni
->itype
.compressed
.
1397 ni
->itype
.compressed
.block_clusters
= 1U <<
1398 a
->data
.non_resident
.
1401 ni
->itype
.compressed
.block_size
= 0;
1402 ni
->itype
.compressed
.block_size_bits
= 0;
1403 ni
->itype
.compressed
.block_clusters
= 0;
1405 ni
->itype
.compressed
.size
= sle64_to_cpu(
1406 a
->data
.non_resident
.compressed_size
);
1408 if (a
->data
.non_resident
.lowest_vcn
) {
1409 ntfs_error(vi
->i_sb
, "First extent of attribute has "
1410 "non-zero lowest_vcn.");
1413 vi
->i_size
= sle64_to_cpu(a
->data
.non_resident
.data_size
);
1414 ni
->initialized_size
= sle64_to_cpu(
1415 a
->data
.non_resident
.initialized_size
);
1416 ni
->allocated_size
= sle64_to_cpu(
1417 a
->data
.non_resident
.allocated_size
);
1419 vi
->i_mapping
->a_ops
= &ntfs_normal_aops
;
1420 if (NInoMstProtected(ni
))
1421 vi
->i_mapping
->a_ops
= &ntfs_mst_aops
;
1422 else if (NInoCompressed(ni
))
1423 vi
->i_mapping
->a_ops
= &ntfs_compressed_aops
;
1424 if ((NInoCompressed(ni
) || NInoSparse(ni
)) && ni
->type
!= AT_INDEX_ROOT
)
1425 vi
->i_blocks
= ni
->itype
.compressed
.size
>> 9;
1427 vi
->i_blocks
= ni
->allocated_size
>> 9;
1429 * Make sure the base inode does not go away and attach it to the
1433 ni
->ext
.base_ntfs_ino
= base_ni
;
1434 ni
->nr_extents
= -1;
1436 ntfs_attr_put_search_ctx(ctx
);
1437 unmap_mft_record(base_ni
);
1439 ntfs_debug("Done.");
1446 ntfs_attr_put_search_ctx(ctx
);
1447 unmap_mft_record(base_ni
);
1449 ntfs_error(vol
->sb
, "Failed with error code %i while reading attribute "
1450 "inode (mft_no 0x%lx, type 0x%x, name_len %i). "
1451 "Marking corrupt inode and base inode 0x%lx as bad. "
1452 "Run chkdsk.", err
, vi
->i_ino
, ni
->type
, ni
->name_len
,
1461 * ntfs_read_locked_index_inode - read an index inode from its base inode
1462 * @base_vi: base inode
1463 * @vi: index inode to read
1465 * ntfs_read_locked_index_inode() is called from ntfs_index_iget() to read the
1466 * index inode described by @vi into memory from the base mft record described
1469 * ntfs_read_locked_index_inode() maps, pins and locks the base inode for
1470 * reading and looks up the attributes relating to the index described by @vi
1471 * before setting up the necessary fields in @vi as well as initializing the
1474 * Note, index inodes are essentially attribute inodes (NInoAttr() is true)
1475 * with the attribute type set to AT_INDEX_ALLOCATION. Apart from that, they
1476 * are setup like directory inodes since directories are a special case of
1477 * indices ao they need to be treated in much the same way. Most importantly,
1478 * for small indices the index allocation attribute might not actually exist.
1479 * However, the index root attribute always exists but this does not need to
1480 * have an inode associated with it and this is why we define a new inode type
1481 * index. Also, like for directories, we need to have an attribute inode for
1482 * the bitmap attribute corresponding to the index allocation attribute and we
1483 * can store this in the appropriate field of the inode, just like we do for
1484 * normal directory inodes.
1486 * Q: What locks are held when the function is called?
1487 * A: i_state has I_NEW set, hence the inode is locked, also
1488 * i_count is set to 1, so it is not going to go away
1490 * Return 0 on success and -errno on error. In the error case, the inode will
1491 * have had make_bad_inode() executed on it.
1493 static int ntfs_read_locked_index_inode(struct inode
*base_vi
, struct inode
*vi
)
1496 ntfs_volume
*vol
= NTFS_SB(vi
->i_sb
);
1497 ntfs_inode
*ni
, *base_ni
, *bni
;
1501 ntfs_attr_search_ctx
*ctx
;
1503 u8
*ir_end
, *index_end
;
1506 ntfs_debug("Entering for i_ino 0x%lx.", vi
->i_ino
);
1507 ntfs_init_big_inode(vi
);
1509 base_ni
= NTFS_I(base_vi
);
1510 /* Just mirror the values from the base inode. */
1511 vi
->i_version
= base_vi
->i_version
;
1512 vi
->i_uid
= base_vi
->i_uid
;
1513 vi
->i_gid
= base_vi
->i_gid
;
1514 set_nlink(vi
, base_vi
->i_nlink
);
1515 vi
->i_mtime
= base_vi
->i_mtime
;
1516 vi
->i_ctime
= base_vi
->i_ctime
;
1517 vi
->i_atime
= base_vi
->i_atime
;
1518 vi
->i_generation
= ni
->seq_no
= base_ni
->seq_no
;
1519 /* Set inode type to zero but preserve permissions. */
1520 vi
->i_mode
= base_vi
->i_mode
& ~S_IFMT
;
1521 /* Map the mft record for the base inode. */
1522 m
= map_mft_record(base_ni
);
1527 ctx
= ntfs_attr_get_search_ctx(base_ni
, m
);
1532 /* Find the index root attribute. */
1533 err
= ntfs_attr_lookup(AT_INDEX_ROOT
, ni
->name
, ni
->name_len
,
1534 CASE_SENSITIVE
, 0, NULL
, 0, ctx
);
1535 if (unlikely(err
)) {
1537 ntfs_error(vi
->i_sb
, "$INDEX_ROOT attribute is "
1542 /* Set up the state. */
1543 if (unlikely(a
->non_resident
)) {
1544 ntfs_error(vol
->sb
, "$INDEX_ROOT attribute is not resident.");
1547 /* Ensure the attribute name is placed before the value. */
1548 if (unlikely(a
->name_length
&& (le16_to_cpu(a
->name_offset
) >=
1549 le16_to_cpu(a
->data
.resident
.value_offset
)))) {
1550 ntfs_error(vol
->sb
, "$INDEX_ROOT attribute name is placed "
1551 "after the attribute value.");
1555 * Compressed/encrypted/sparse index root is not allowed, except for
1556 * directories of course but those are not dealt with here.
1558 if (a
->flags
& (ATTR_COMPRESSION_MASK
| ATTR_IS_ENCRYPTED
|
1560 ntfs_error(vi
->i_sb
, "Found compressed/encrypted/sparse index "
1564 ir
= (INDEX_ROOT
*)((u8
*)a
+ le16_to_cpu(a
->data
.resident
.value_offset
));
1565 ir_end
= (u8
*)ir
+ le32_to_cpu(a
->data
.resident
.value_length
);
1566 if (ir_end
> (u8
*)ctx
->mrec
+ vol
->mft_record_size
) {
1567 ntfs_error(vi
->i_sb
, "$INDEX_ROOT attribute is corrupt.");
1570 index_end
= (u8
*)&ir
->index
+ le32_to_cpu(ir
->index
.index_length
);
1571 if (index_end
> ir_end
) {
1572 ntfs_error(vi
->i_sb
, "Index is corrupt.");
1576 ntfs_error(vi
->i_sb
, "Index type is not 0 (type is 0x%x).",
1577 le32_to_cpu(ir
->type
));
1580 ni
->itype
.index
.collation_rule
= ir
->collation_rule
;
1581 ntfs_debug("Index collation rule is 0x%x.",
1582 le32_to_cpu(ir
->collation_rule
));
1583 ni
->itype
.index
.block_size
= le32_to_cpu(ir
->index_block_size
);
1584 if (!is_power_of_2(ni
->itype
.index
.block_size
)) {
1585 ntfs_error(vi
->i_sb
, "Index block size (%u) is not a power of "
1586 "two.", ni
->itype
.index
.block_size
);
1589 if (ni
->itype
.index
.block_size
> PAGE_CACHE_SIZE
) {
1590 ntfs_error(vi
->i_sb
, "Index block size (%u) > PAGE_CACHE_SIZE "
1591 "(%ld) is not supported. Sorry.",
1592 ni
->itype
.index
.block_size
, PAGE_CACHE_SIZE
);
1596 if (ni
->itype
.index
.block_size
< NTFS_BLOCK_SIZE
) {
1597 ntfs_error(vi
->i_sb
, "Index block size (%u) < NTFS_BLOCK_SIZE "
1598 "(%i) is not supported. Sorry.",
1599 ni
->itype
.index
.block_size
, NTFS_BLOCK_SIZE
);
1603 ni
->itype
.index
.block_size_bits
= ffs(ni
->itype
.index
.block_size
) - 1;
1604 /* Determine the size of a vcn in the index. */
1605 if (vol
->cluster_size
<= ni
->itype
.index
.block_size
) {
1606 ni
->itype
.index
.vcn_size
= vol
->cluster_size
;
1607 ni
->itype
.index
.vcn_size_bits
= vol
->cluster_size_bits
;
1609 ni
->itype
.index
.vcn_size
= vol
->sector_size
;
1610 ni
->itype
.index
.vcn_size_bits
= vol
->sector_size_bits
;
1612 /* Check for presence of index allocation attribute. */
1613 if (!(ir
->index
.flags
& LARGE_INDEX
)) {
1614 /* No index allocation. */
1615 vi
->i_size
= ni
->initialized_size
= ni
->allocated_size
= 0;
1616 /* We are done with the mft record, so we release it. */
1617 ntfs_attr_put_search_ctx(ctx
);
1618 unmap_mft_record(base_ni
);
1621 goto skip_large_index_stuff
;
1622 } /* LARGE_INDEX: Index allocation present. Setup state. */
1623 NInoSetIndexAllocPresent(ni
);
1624 /* Find index allocation attribute. */
1625 ntfs_attr_reinit_search_ctx(ctx
);
1626 err
= ntfs_attr_lookup(AT_INDEX_ALLOCATION
, ni
->name
, ni
->name_len
,
1627 CASE_SENSITIVE
, 0, NULL
, 0, ctx
);
1628 if (unlikely(err
)) {
1630 ntfs_error(vi
->i_sb
, "$INDEX_ALLOCATION attribute is "
1631 "not present but $INDEX_ROOT "
1632 "indicated it is.");
1634 ntfs_error(vi
->i_sb
, "Failed to lookup "
1635 "$INDEX_ALLOCATION attribute.");
1639 if (!a
->non_resident
) {
1640 ntfs_error(vi
->i_sb
, "$INDEX_ALLOCATION attribute is "
1645 * Ensure the attribute name is placed before the mapping pairs array.
1647 if (unlikely(a
->name_length
&& (le16_to_cpu(a
->name_offset
) >=
1649 a
->data
.non_resident
.mapping_pairs_offset
)))) {
1650 ntfs_error(vol
->sb
, "$INDEX_ALLOCATION attribute name is "
1651 "placed after the mapping pairs array.");
1654 if (a
->flags
& ATTR_IS_ENCRYPTED
) {
1655 ntfs_error(vi
->i_sb
, "$INDEX_ALLOCATION attribute is "
1659 if (a
->flags
& ATTR_IS_SPARSE
) {
1660 ntfs_error(vi
->i_sb
, "$INDEX_ALLOCATION attribute is sparse.");
1663 if (a
->flags
& ATTR_COMPRESSION_MASK
) {
1664 ntfs_error(vi
->i_sb
, "$INDEX_ALLOCATION attribute is "
1668 if (a
->data
.non_resident
.lowest_vcn
) {
1669 ntfs_error(vi
->i_sb
, "First extent of $INDEX_ALLOCATION "
1670 "attribute has non zero lowest_vcn.");
1673 vi
->i_size
= sle64_to_cpu(a
->data
.non_resident
.data_size
);
1674 ni
->initialized_size
= sle64_to_cpu(
1675 a
->data
.non_resident
.initialized_size
);
1676 ni
->allocated_size
= sle64_to_cpu(a
->data
.non_resident
.allocated_size
);
1678 * We are done with the mft record, so we release it. Otherwise
1679 * we would deadlock in ntfs_attr_iget().
1681 ntfs_attr_put_search_ctx(ctx
);
1682 unmap_mft_record(base_ni
);
1685 /* Get the index bitmap attribute inode. */
1686 bvi
= ntfs_attr_iget(base_vi
, AT_BITMAP
, ni
->name
, ni
->name_len
);
1688 ntfs_error(vi
->i_sb
, "Failed to get bitmap attribute.");
1693 if (NInoCompressed(bni
) || NInoEncrypted(bni
) ||
1695 ntfs_error(vi
->i_sb
, "$BITMAP attribute is compressed and/or "
1696 "encrypted and/or sparse.");
1697 goto iput_unm_err_out
;
1699 /* Consistency check bitmap size vs. index allocation size. */
1700 bvi_size
= i_size_read(bvi
);
1701 if ((bvi_size
<< 3) < (vi
->i_size
>> ni
->itype
.index
.block_size_bits
)) {
1702 ntfs_error(vi
->i_sb
, "Index bitmap too small (0x%llx) for "
1703 "index allocation (0x%llx).", bvi_size
<< 3,
1705 goto iput_unm_err_out
;
1708 skip_large_index_stuff
:
1709 /* Setup the operations for this index inode. */
1710 vi
->i_mapping
->a_ops
= &ntfs_mst_aops
;
1711 vi
->i_blocks
= ni
->allocated_size
>> 9;
1713 * Make sure the base inode doesn't go away and attach it to the
1717 ni
->ext
.base_ntfs_ino
= base_ni
;
1718 ni
->nr_extents
= -1;
1720 ntfs_debug("Done.");
1728 ntfs_attr_put_search_ctx(ctx
);
1730 unmap_mft_record(base_ni
);
1732 ntfs_error(vi
->i_sb
, "Failed with error code %i while reading index "
1733 "inode (mft_no 0x%lx, name_len %i.", err
, vi
->i_ino
,
1736 if (err
!= -EOPNOTSUPP
&& err
!= -ENOMEM
)
1742 * The MFT inode has special locking, so teach the lock validator
1743 * about this by splitting off the locking rules of the MFT from
1744 * the locking rules of other inodes. The MFT inode can never be
1745 * accessed from the VFS side (or even internally), only by the
1746 * map_mft functions.
1748 static struct lock_class_key mft_ni_runlist_lock_key
, mft_ni_mrec_lock_key
;
1751 * ntfs_read_inode_mount - special read_inode for mount time use only
1752 * @vi: inode to read
1754 * Read inode FILE_MFT at mount time, only called with super_block lock
1755 * held from within the read_super() code path.
1757 * This function exists because when it is called the page cache for $MFT/$DATA
1758 * is not initialized and hence we cannot get at the contents of mft records
1759 * by calling map_mft_record*().
1761 * Further it needs to cope with the circular references problem, i.e. cannot
1762 * load any attributes other than $ATTRIBUTE_LIST until $DATA is loaded, because
1763 * we do not know where the other extent mft records are yet and again, because
1764 * we cannot call map_mft_record*() yet. Obviously this applies only when an
1765 * attribute list is actually present in $MFT inode.
1767 * We solve these problems by starting with the $DATA attribute before anything
1768 * else and iterating using ntfs_attr_lookup($DATA) over all extents. As each
1769 * extent is found, we ntfs_mapping_pairs_decompress() including the implied
1770 * ntfs_runlists_merge(). Each step of the iteration necessarily provides
1771 * sufficient information for the next step to complete.
1773 * This should work but there are two possible pit falls (see inline comments
1774 * below), but only time will tell if they are real pits or just smoke...
1776 int ntfs_read_inode_mount(struct inode
*vi
)
1778 VCN next_vcn
, last_vcn
, highest_vcn
;
1780 struct super_block
*sb
= vi
->i_sb
;
1781 ntfs_volume
*vol
= NTFS_SB(sb
);
1782 struct buffer_head
*bh
;
1784 MFT_RECORD
*m
= NULL
;
1786 ntfs_attr_search_ctx
*ctx
;
1787 unsigned int i
, nr_blocks
;
1790 ntfs_debug("Entering.");
1792 /* Initialize the ntfs specific part of @vi. */
1793 ntfs_init_big_inode(vi
);
1797 /* Setup the data attribute. It is special as it is mst protected. */
1798 NInoSetNonResident(ni
);
1799 NInoSetMstProtected(ni
);
1800 NInoSetSparseDisabled(ni
);
1805 * This sets up our little cheat allowing us to reuse the async read io
1806 * completion handler for directories.
1808 ni
->itype
.index
.block_size
= vol
->mft_record_size
;
1809 ni
->itype
.index
.block_size_bits
= vol
->mft_record_size_bits
;
1811 /* Very important! Needed to be able to call map_mft_record*(). */
1814 /* Allocate enough memory to read the first mft record. */
1815 if (vol
->mft_record_size
> 64 * 1024) {
1816 ntfs_error(sb
, "Unsupported mft record size %i (max 64kiB).",
1817 vol
->mft_record_size
);
1820 i
= vol
->mft_record_size
;
1821 if (i
< sb
->s_blocksize
)
1822 i
= sb
->s_blocksize
;
1823 m
= (MFT_RECORD
*)ntfs_malloc_nofs(i
);
1825 ntfs_error(sb
, "Failed to allocate buffer for $MFT record 0.");
1829 /* Determine the first block of the $MFT/$DATA attribute. */
1830 block
= vol
->mft_lcn
<< vol
->cluster_size_bits
>>
1831 sb
->s_blocksize_bits
;
1832 nr_blocks
= vol
->mft_record_size
>> sb
->s_blocksize_bits
;
1836 /* Load $MFT/$DATA's first mft record. */
1837 for (i
= 0; i
< nr_blocks
; i
++) {
1838 bh
= sb_bread(sb
, block
++);
1840 ntfs_error(sb
, "Device read failed.");
1843 memcpy((char*)m
+ (i
<< sb
->s_blocksize_bits
), bh
->b_data
,
1848 /* Apply the mst fixups. */
1849 if (post_read_mst_fixup((NTFS_RECORD
*)m
, vol
->mft_record_size
)) {
1850 /* FIXME: Try to use the $MFTMirr now. */
1851 ntfs_error(sb
, "MST fixup failed. $MFT is corrupt.");
1855 /* Need this to sanity check attribute list references to $MFT. */
1856 vi
->i_generation
= ni
->seq_no
= le16_to_cpu(m
->sequence_number
);
1858 /* Provides readpage() and sync_page() for map_mft_record(). */
1859 vi
->i_mapping
->a_ops
= &ntfs_mst_aops
;
1861 ctx
= ntfs_attr_get_search_ctx(ni
, m
);
1867 /* Find the attribute list attribute if present. */
1868 err
= ntfs_attr_lookup(AT_ATTRIBUTE_LIST
, NULL
, 0, 0, 0, NULL
, 0, ctx
);
1870 if (unlikely(err
!= -ENOENT
)) {
1871 ntfs_error(sb
, "Failed to lookup attribute list "
1872 "attribute. You should run chkdsk.");
1875 } else /* if (!err) */ {
1876 ATTR_LIST_ENTRY
*al_entry
, *next_al_entry
;
1878 static const char *es
= " Not allowed. $MFT is corrupt. "
1879 "You should run chkdsk.";
1881 ntfs_debug("Attribute list attribute found in $MFT.");
1882 NInoSetAttrList(ni
);
1884 if (a
->flags
& ATTR_COMPRESSION_MASK
) {
1885 ntfs_error(sb
, "Attribute list attribute is "
1886 "compressed.%s", es
);
1889 if (a
->flags
& ATTR_IS_ENCRYPTED
||
1890 a
->flags
& ATTR_IS_SPARSE
) {
1891 if (a
->non_resident
) {
1892 ntfs_error(sb
, "Non-resident attribute list "
1893 "attribute is encrypted/"
1897 ntfs_warning(sb
, "Resident attribute list attribute "
1898 "in $MFT system file is marked "
1899 "encrypted/sparse which is not true. "
1900 "However, Windows allows this and "
1901 "chkdsk does not detect or correct it "
1902 "so we will just ignore the invalid "
1903 "flags and pretend they are not set.");
1905 /* Now allocate memory for the attribute list. */
1906 ni
->attr_list_size
= (u32
)ntfs_attr_size(a
);
1907 ni
->attr_list
= ntfs_malloc_nofs(ni
->attr_list_size
);
1908 if (!ni
->attr_list
) {
1909 ntfs_error(sb
, "Not enough memory to allocate buffer "
1910 "for attribute list.");
1913 if (a
->non_resident
) {
1914 NInoSetAttrListNonResident(ni
);
1915 if (a
->data
.non_resident
.lowest_vcn
) {
1916 ntfs_error(sb
, "Attribute list has non zero "
1917 "lowest_vcn. $MFT is corrupt. "
1918 "You should run chkdsk.");
1921 /* Setup the runlist. */
1922 ni
->attr_list_rl
.rl
= ntfs_mapping_pairs_decompress(vol
,
1924 if (IS_ERR(ni
->attr_list_rl
.rl
)) {
1925 err
= PTR_ERR(ni
->attr_list_rl
.rl
);
1926 ni
->attr_list_rl
.rl
= NULL
;
1927 ntfs_error(sb
, "Mapping pairs decompression "
1928 "failed with error code %i.",
1932 /* Now load the attribute list. */
1933 if ((err
= load_attribute_list(vol
, &ni
->attr_list_rl
,
1934 ni
->attr_list
, ni
->attr_list_size
,
1935 sle64_to_cpu(a
->data
.
1936 non_resident
.initialized_size
)))) {
1937 ntfs_error(sb
, "Failed to load attribute list "
1938 "attribute with error code %i.",
1942 } else /* if (!ctx.attr->non_resident) */ {
1943 if ((u8
*)a
+ le16_to_cpu(
1944 a
->data
.resident
.value_offset
) +
1946 a
->data
.resident
.value_length
) >
1947 (u8
*)ctx
->mrec
+ vol
->mft_record_size
) {
1948 ntfs_error(sb
, "Corrupt attribute list "
1952 /* Now copy the attribute list. */
1953 memcpy(ni
->attr_list
, (u8
*)a
+ le16_to_cpu(
1954 a
->data
.resident
.value_offset
),
1956 a
->data
.resident
.value_length
));
1958 /* The attribute list is now setup in memory. */
1960 * FIXME: I don't know if this case is actually possible.
1961 * According to logic it is not possible but I have seen too
1962 * many weird things in MS software to rely on logic... Thus we
1963 * perform a manual search and make sure the first $MFT/$DATA
1964 * extent is in the base inode. If it is not we abort with an
1965 * error and if we ever see a report of this error we will need
1966 * to do some magic in order to have the necessary mft record
1967 * loaded and in the right place in the page cache. But
1968 * hopefully logic will prevail and this never happens...
1970 al_entry
= (ATTR_LIST_ENTRY
*)ni
->attr_list
;
1971 al_end
= (u8
*)al_entry
+ ni
->attr_list_size
;
1972 for (;; al_entry
= next_al_entry
) {
1973 /* Out of bounds check. */
1974 if ((u8
*)al_entry
< ni
->attr_list
||
1975 (u8
*)al_entry
> al_end
)
1976 goto em_put_err_out
;
1977 /* Catch the end of the attribute list. */
1978 if ((u8
*)al_entry
== al_end
)
1979 goto em_put_err_out
;
1980 if (!al_entry
->length
)
1981 goto em_put_err_out
;
1982 if ((u8
*)al_entry
+ 6 > al_end
|| (u8
*)al_entry
+
1983 le16_to_cpu(al_entry
->length
) > al_end
)
1984 goto em_put_err_out
;
1985 next_al_entry
= (ATTR_LIST_ENTRY
*)((u8
*)al_entry
+
1986 le16_to_cpu(al_entry
->length
));
1987 if (le32_to_cpu(al_entry
->type
) > le32_to_cpu(AT_DATA
))
1988 goto em_put_err_out
;
1989 if (AT_DATA
!= al_entry
->type
)
1991 /* We want an unnamed attribute. */
1992 if (al_entry
->name_length
)
1993 goto em_put_err_out
;
1994 /* Want the first entry, i.e. lowest_vcn == 0. */
1995 if (al_entry
->lowest_vcn
)
1996 goto em_put_err_out
;
1997 /* First entry has to be in the base mft record. */
1998 if (MREF_LE(al_entry
->mft_reference
) != vi
->i_ino
) {
1999 /* MFT references do not match, logic fails. */
2000 ntfs_error(sb
, "BUG: The first $DATA extent "
2001 "of $MFT is not in the base "
2002 "mft record. Please report "
2003 "you saw this message to "
2004 "linux-ntfs-dev@lists."
2008 /* Sequence numbers must match. */
2009 if (MSEQNO_LE(al_entry
->mft_reference
) !=
2011 goto em_put_err_out
;
2012 /* Got it. All is ok. We can stop now. */
2018 ntfs_attr_reinit_search_ctx(ctx
);
2020 /* Now load all attribute extents. */
2022 next_vcn
= last_vcn
= highest_vcn
= 0;
2023 while (!(err
= ntfs_attr_lookup(AT_DATA
, NULL
, 0, 0, next_vcn
, NULL
, 0,
2025 runlist_element
*nrl
;
2027 /* Cache the current attribute. */
2029 /* $MFT must be non-resident. */
2030 if (!a
->non_resident
) {
2031 ntfs_error(sb
, "$MFT must be non-resident but a "
2032 "resident extent was found. $MFT is "
2033 "corrupt. Run chkdsk.");
2036 /* $MFT must be uncompressed and unencrypted. */
2037 if (a
->flags
& ATTR_COMPRESSION_MASK
||
2038 a
->flags
& ATTR_IS_ENCRYPTED
||
2039 a
->flags
& ATTR_IS_SPARSE
) {
2040 ntfs_error(sb
, "$MFT must be uncompressed, "
2041 "non-sparse, and unencrypted but a "
2042 "compressed/sparse/encrypted extent "
2043 "was found. $MFT is corrupt. Run "
2048 * Decompress the mapping pairs array of this extent and merge
2049 * the result into the existing runlist. No need for locking
2050 * as we have exclusive access to the inode at this time and we
2051 * are a mount in progress task, too.
2053 nrl
= ntfs_mapping_pairs_decompress(vol
, a
, ni
->runlist
.rl
);
2055 ntfs_error(sb
, "ntfs_mapping_pairs_decompress() "
2056 "failed with error code %ld. $MFT is "
2057 "corrupt.", PTR_ERR(nrl
));
2060 ni
->runlist
.rl
= nrl
;
2062 /* Are we in the first extent? */
2064 if (a
->data
.non_resident
.lowest_vcn
) {
2065 ntfs_error(sb
, "First extent of $DATA "
2066 "attribute has non zero "
2067 "lowest_vcn. $MFT is corrupt. "
2068 "You should run chkdsk.");
2071 /* Get the last vcn in the $DATA attribute. */
2072 last_vcn
= sle64_to_cpu(
2073 a
->data
.non_resident
.allocated_size
)
2074 >> vol
->cluster_size_bits
;
2075 /* Fill in the inode size. */
2076 vi
->i_size
= sle64_to_cpu(
2077 a
->data
.non_resident
.data_size
);
2078 ni
->initialized_size
= sle64_to_cpu(
2079 a
->data
.non_resident
.initialized_size
);
2080 ni
->allocated_size
= sle64_to_cpu(
2081 a
->data
.non_resident
.allocated_size
);
2083 * Verify the number of mft records does not exceed
2086 if ((vi
->i_size
>> vol
->mft_record_size_bits
) >=
2088 ntfs_error(sb
, "$MFT is too big! Aborting.");
2092 * We have got the first extent of the runlist for
2093 * $MFT which means it is now relatively safe to call
2094 * the normal ntfs_read_inode() function.
2095 * Complete reading the inode, this will actually
2096 * re-read the mft record for $MFT, this time entering
2097 * it into the page cache with which we complete the
2098 * kick start of the volume. It should be safe to do
2099 * this now as the first extent of $MFT/$DATA is
2100 * already known and we would hope that we don't need
2101 * further extents in order to find the other
2102 * attributes belonging to $MFT. Only time will tell if
2103 * this is really the case. If not we will have to play
2104 * magic at this point, possibly duplicating a lot of
2105 * ntfs_read_inode() at this point. We will need to
2106 * ensure we do enough of its work to be able to call
2107 * ntfs_read_inode() on extents of $MFT/$DATA. But lets
2108 * hope this never happens...
2110 ntfs_read_locked_inode(vi
);
2111 if (is_bad_inode(vi
)) {
2112 ntfs_error(sb
, "ntfs_read_inode() of $MFT "
2113 "failed. BUG or corrupt $MFT. "
2114 "Run chkdsk and if no errors "
2115 "are found, please report you "
2116 "saw this message to "
2117 "linux-ntfs-dev@lists."
2119 ntfs_attr_put_search_ctx(ctx
);
2120 /* Revert to the safe super operations. */
2125 * Re-initialize some specifics about $MFT's inode as
2126 * ntfs_read_inode() will have set up the default ones.
2128 /* Set uid and gid to root. */
2129 vi
->i_uid
= GLOBAL_ROOT_UID
;
2130 vi
->i_gid
= GLOBAL_ROOT_GID
;
2131 /* Regular file. No access for anyone. */
2132 vi
->i_mode
= S_IFREG
;
2133 /* No VFS initiated operations allowed for $MFT. */
2134 vi
->i_op
= &ntfs_empty_inode_ops
;
2135 vi
->i_fop
= &ntfs_empty_file_ops
;
2138 /* Get the lowest vcn for the next extent. */
2139 highest_vcn
= sle64_to_cpu(a
->data
.non_resident
.highest_vcn
);
2140 next_vcn
= highest_vcn
+ 1;
2142 /* Only one extent or error, which we catch below. */
2146 /* Avoid endless loops due to corruption. */
2147 if (next_vcn
< sle64_to_cpu(
2148 a
->data
.non_resident
.lowest_vcn
)) {
2149 ntfs_error(sb
, "$MFT has corrupt attribute list "
2150 "attribute. Run chkdsk.");
2154 if (err
!= -ENOENT
) {
2155 ntfs_error(sb
, "Failed to lookup $MFT/$DATA attribute extent. "
2156 "$MFT is corrupt. Run chkdsk.");
2160 ntfs_error(sb
, "$MFT/$DATA attribute not found. $MFT is "
2161 "corrupt. Run chkdsk.");
2164 if (highest_vcn
&& highest_vcn
!= last_vcn
- 1) {
2165 ntfs_error(sb
, "Failed to load the complete runlist for "
2166 "$MFT/$DATA. Driver bug or corrupt $MFT. "
2168 ntfs_debug("highest_vcn = 0x%llx, last_vcn - 1 = 0x%llx",
2169 (unsigned long long)highest_vcn
,
2170 (unsigned long long)last_vcn
- 1);
2173 ntfs_attr_put_search_ctx(ctx
);
2174 ntfs_debug("Done.");
2178 * Split the locking rules of the MFT inode from the
2179 * locking rules of other inodes:
2181 lockdep_set_class(&ni
->runlist
.lock
, &mft_ni_runlist_lock_key
);
2182 lockdep_set_class(&ni
->mrec_lock
, &mft_ni_mrec_lock_key
);
2187 ntfs_error(sb
, "Couldn't find first extent of $DATA attribute in "
2188 "attribute list. $MFT is corrupt. Run chkdsk.");
2190 ntfs_attr_put_search_ctx(ctx
);
2192 ntfs_error(sb
, "Failed. Marking inode as bad.");
2198 static void __ntfs_clear_inode(ntfs_inode
*ni
)
2200 /* Free all alocated memory. */
2201 down_write(&ni
->runlist
.lock
);
2202 if (ni
->runlist
.rl
) {
2203 ntfs_free(ni
->runlist
.rl
);
2204 ni
->runlist
.rl
= NULL
;
2206 up_write(&ni
->runlist
.lock
);
2208 if (ni
->attr_list
) {
2209 ntfs_free(ni
->attr_list
);
2210 ni
->attr_list
= NULL
;
2213 down_write(&ni
->attr_list_rl
.lock
);
2214 if (ni
->attr_list_rl
.rl
) {
2215 ntfs_free(ni
->attr_list_rl
.rl
);
2216 ni
->attr_list_rl
.rl
= NULL
;
2218 up_write(&ni
->attr_list_rl
.lock
);
2220 if (ni
->name_len
&& ni
->name
!= I30
) {
2227 void ntfs_clear_extent_inode(ntfs_inode
*ni
)
2229 ntfs_debug("Entering for inode 0x%lx.", ni
->mft_no
);
2231 BUG_ON(NInoAttr(ni
));
2232 BUG_ON(ni
->nr_extents
!= -1);
2235 if (NInoDirty(ni
)) {
2236 if (!is_bad_inode(VFS_I(ni
->ext
.base_ntfs_ino
)))
2237 ntfs_error(ni
->vol
->sb
, "Clearing dirty extent inode! "
2238 "Losing data! This is a BUG!!!");
2239 // FIXME: Do something!!!
2241 #endif /* NTFS_RW */
2243 __ntfs_clear_inode(ni
);
2246 ntfs_destroy_extent_inode(ni
);
2250 * ntfs_evict_big_inode - clean up the ntfs specific part of an inode
2251 * @vi: vfs inode pending annihilation
2253 * When the VFS is going to remove an inode from memory, ntfs_clear_big_inode()
2254 * is called, which deallocates all memory belonging to the NTFS specific part
2255 * of the inode and returns.
2257 * If the MFT record is dirty, we commit it before doing anything else.
2259 void ntfs_evict_big_inode(struct inode
*vi
)
2261 ntfs_inode
*ni
= NTFS_I(vi
);
2263 truncate_inode_pages_final(&vi
->i_data
);
2267 if (NInoDirty(ni
)) {
2268 bool was_bad
= (is_bad_inode(vi
));
2270 /* Committing the inode also commits all extent inodes. */
2271 ntfs_commit_inode(vi
);
2273 if (!was_bad
&& (is_bad_inode(vi
) || NInoDirty(ni
))) {
2274 ntfs_error(vi
->i_sb
, "Failed to commit dirty inode "
2275 "0x%lx. Losing data!", vi
->i_ino
);
2276 // FIXME: Do something!!!
2279 #endif /* NTFS_RW */
2281 /* No need to lock at this stage as no one else has a reference. */
2282 if (ni
->nr_extents
> 0) {
2285 for (i
= 0; i
< ni
->nr_extents
; i
++)
2286 ntfs_clear_extent_inode(ni
->ext
.extent_ntfs_inos
[i
]);
2287 kfree(ni
->ext
.extent_ntfs_inos
);
2290 __ntfs_clear_inode(ni
);
2293 /* Release the base inode if we are holding it. */
2294 if (ni
->nr_extents
== -1) {
2295 iput(VFS_I(ni
->ext
.base_ntfs_ino
));
2297 ni
->ext
.base_ntfs_ino
= NULL
;
2304 * ntfs_show_options - show mount options in /proc/mounts
2305 * @sf: seq_file in which to write our mount options
2306 * @root: root of the mounted tree whose mount options to display
2308 * Called by the VFS once for each mounted ntfs volume when someone reads
2309 * /proc/mounts in order to display the NTFS specific mount options of each
2310 * mount. The mount options of fs specified by @root are written to the seq file
2311 * @sf and success is returned.
2313 int ntfs_show_options(struct seq_file
*sf
, struct dentry
*root
)
2315 ntfs_volume
*vol
= NTFS_SB(root
->d_sb
);
2318 seq_printf(sf
, ",uid=%i", from_kuid_munged(&init_user_ns
, vol
->uid
));
2319 seq_printf(sf
, ",gid=%i", from_kgid_munged(&init_user_ns
, vol
->gid
));
2320 if (vol
->fmask
== vol
->dmask
)
2321 seq_printf(sf
, ",umask=0%o", vol
->fmask
);
2323 seq_printf(sf
, ",fmask=0%o", vol
->fmask
);
2324 seq_printf(sf
, ",dmask=0%o", vol
->dmask
);
2326 seq_printf(sf
, ",nls=%s", vol
->nls_map
->charset
);
2327 if (NVolCaseSensitive(vol
))
2328 seq_printf(sf
, ",case_sensitive");
2329 if (NVolShowSystemFiles(vol
))
2330 seq_printf(sf
, ",show_sys_files");
2331 if (!NVolSparseEnabled(vol
))
2332 seq_printf(sf
, ",disable_sparse");
2333 for (i
= 0; on_errors_arr
[i
].val
; i
++) {
2334 if (on_errors_arr
[i
].val
& vol
->on_errors
)
2335 seq_printf(sf
, ",errors=%s", on_errors_arr
[i
].str
);
2337 seq_printf(sf
, ",mft_zone_multiplier=%i", vol
->mft_zone_multiplier
);
2343 static const char *es
= " Leaving inconsistent metadata. Unmount and run "
2347 * ntfs_truncate - called when the i_size of an ntfs inode is changed
2348 * @vi: inode for which the i_size was changed
2350 * We only support i_size changes for normal files at present, i.e. not
2351 * compressed and not encrypted. This is enforced in ntfs_setattr(), see
2354 * The kernel guarantees that @vi is a regular file (S_ISREG() is true) and
2355 * that the change is allowed.
2357 * This implies for us that @vi is a file inode rather than a directory, index,
2358 * or attribute inode as well as that @vi is a base inode.
2360 * Returns 0 on success or -errno on error.
2362 * Called with ->i_mutex held.
2364 int ntfs_truncate(struct inode
*vi
)
2366 s64 new_size
, old_size
, nr_freed
, new_alloc_size
, old_alloc_size
;
2368 unsigned long flags
;
2369 ntfs_inode
*base_ni
, *ni
= NTFS_I(vi
);
2370 ntfs_volume
*vol
= ni
->vol
;
2371 ntfs_attr_search_ctx
*ctx
;
2374 const char *te
= " Leaving file length out of sync with i_size.";
2375 int err
, mp_size
, size_change
, alloc_change
;
2378 ntfs_debug("Entering for inode 0x%lx.", vi
->i_ino
);
2379 BUG_ON(NInoAttr(ni
));
2380 BUG_ON(S_ISDIR(vi
->i_mode
));
2381 BUG_ON(NInoMstProtected(ni
));
2382 BUG_ON(ni
->nr_extents
< 0);
2385 * Lock the runlist for writing and map the mft record to ensure it is
2386 * safe to mess with the attribute runlist and sizes.
2388 down_write(&ni
->runlist
.lock
);
2392 base_ni
= ni
->ext
.base_ntfs_ino
;
2393 m
= map_mft_record(base_ni
);
2396 ntfs_error(vi
->i_sb
, "Failed to map mft record for inode 0x%lx "
2397 "(error code %d).%s", vi
->i_ino
, err
, te
);
2402 ctx
= ntfs_attr_get_search_ctx(base_ni
, m
);
2403 if (unlikely(!ctx
)) {
2404 ntfs_error(vi
->i_sb
, "Failed to allocate a search context for "
2405 "inode 0x%lx (not enough memory).%s",
2410 err
= ntfs_attr_lookup(ni
->type
, ni
->name
, ni
->name_len
,
2411 CASE_SENSITIVE
, 0, NULL
, 0, ctx
);
2412 if (unlikely(err
)) {
2413 if (err
== -ENOENT
) {
2414 ntfs_error(vi
->i_sb
, "Open attribute is missing from "
2415 "mft record. Inode 0x%lx is corrupt. "
2416 "Run chkdsk.%s", vi
->i_ino
, te
);
2419 ntfs_error(vi
->i_sb
, "Failed to lookup attribute in "
2420 "inode 0x%lx (error code %d).%s",
2421 vi
->i_ino
, err
, te
);
2427 * The i_size of the vfs inode is the new size for the attribute value.
2429 new_size
= i_size_read(vi
);
2430 /* The current size of the attribute value is the old size. */
2431 old_size
= ntfs_attr_size(a
);
2432 /* Calculate the new allocated size. */
2433 if (NInoNonResident(ni
))
2434 new_alloc_size
= (new_size
+ vol
->cluster_size
- 1) &
2435 ~(s64
)vol
->cluster_size_mask
;
2437 new_alloc_size
= (new_size
+ 7) & ~7;
2438 /* The current allocated size is the old allocated size. */
2439 read_lock_irqsave(&ni
->size_lock
, flags
);
2440 old_alloc_size
= ni
->allocated_size
;
2441 read_unlock_irqrestore(&ni
->size_lock
, flags
);
2443 * The change in the file size. This will be 0 if no change, >0 if the
2444 * size is growing, and <0 if the size is shrinking.
2447 if (new_size
- old_size
>= 0) {
2449 if (new_size
== old_size
)
2452 /* As above for the allocated size. */
2454 if (new_alloc_size
- old_alloc_size
>= 0) {
2456 if (new_alloc_size
== old_alloc_size
)
2460 * If neither the size nor the allocation are being changed there is
2463 if (!size_change
&& !alloc_change
)
2465 /* If the size is changing, check if new size is allowed in $AttrDef. */
2467 err
= ntfs_attr_size_bounds_check(vol
, ni
->type
, new_size
);
2468 if (unlikely(err
)) {
2469 if (err
== -ERANGE
) {
2470 ntfs_error(vol
->sb
, "Truncate would cause the "
2471 "inode 0x%lx to %simum size "
2472 "for its attribute type "
2473 "(0x%x). Aborting truncate.",
2475 new_size
> old_size
? "exceed "
2476 "the max" : "go under the min",
2477 le32_to_cpu(ni
->type
));
2480 ntfs_error(vol
->sb
, "Inode 0x%lx has unknown "
2481 "attribute type 0x%x. "
2482 "Aborting truncate.",
2484 le32_to_cpu(ni
->type
));
2487 /* Reset the vfs inode size to the old size. */
2488 i_size_write(vi
, old_size
);
2492 if (NInoCompressed(ni
) || NInoEncrypted(ni
)) {
2493 ntfs_warning(vi
->i_sb
, "Changes in inode size are not "
2494 "supported yet for %s files, ignoring.",
2495 NInoCompressed(ni
) ? "compressed" :
2500 if (a
->non_resident
)
2501 goto do_non_resident_truncate
;
2502 BUG_ON(NInoNonResident(ni
));
2503 /* Resize the attribute record to best fit the new attribute size. */
2504 if (new_size
< vol
->mft_record_size
&&
2505 !ntfs_resident_attr_value_resize(m
, a
, new_size
)) {
2506 /* The resize succeeded! */
2507 flush_dcache_mft_record_page(ctx
->ntfs_ino
);
2508 mark_mft_record_dirty(ctx
->ntfs_ino
);
2509 write_lock_irqsave(&ni
->size_lock
, flags
);
2510 /* Update the sizes in the ntfs inode and all is done. */
2511 ni
->allocated_size
= le32_to_cpu(a
->length
) -
2512 le16_to_cpu(a
->data
.resident
.value_offset
);
2514 * Note ntfs_resident_attr_value_resize() has already done any
2515 * necessary data clearing in the attribute record. When the
2516 * file is being shrunk vmtruncate() will already have cleared
2517 * the top part of the last partial page, i.e. since this is
2518 * the resident case this is the page with index 0. However,
2519 * when the file is being expanded, the page cache page data
2520 * between the old data_size, i.e. old_size, and the new_size
2521 * has not been zeroed. Fortunately, we do not need to zero it
2522 * either since on one hand it will either already be zero due
2523 * to both readpage and writepage clearing partial page data
2524 * beyond i_size in which case there is nothing to do or in the
2525 * case of the file being mmap()ped at the same time, POSIX
2526 * specifies that the behaviour is unspecified thus we do not
2527 * have to do anything. This means that in our implementation
2528 * in the rare case that the file is mmap()ped and a write
2529 * occurred into the mmap()ped region just beyond the file size
2530 * and writepage has not yet been called to write out the page
2531 * (which would clear the area beyond the file size) and we now
2532 * extend the file size to incorporate this dirty region
2533 * outside the file size, a write of the page would result in
2534 * this data being written to disk instead of being cleared.
2535 * Given both POSIX and the Linux mmap(2) man page specify that
2536 * this corner case is undefined, we choose to leave it like
2537 * that as this is much simpler for us as we cannot lock the
2538 * relevant page now since we are holding too many ntfs locks
2539 * which would result in a lock reversal deadlock.
2541 ni
->initialized_size
= new_size
;
2542 write_unlock_irqrestore(&ni
->size_lock
, flags
);
2545 /* If the above resize failed, this must be an attribute extension. */
2546 BUG_ON(size_change
< 0);
2548 * We have to drop all the locks so we can call
2549 * ntfs_attr_make_non_resident(). This could be optimised by try-
2550 * locking the first page cache page and only if that fails dropping
2551 * the locks, locking the page, and redoing all the locking and
2552 * lookups. While this would be a huge optimisation, it is not worth
2553 * it as this is definitely a slow code path as it only ever can happen
2554 * once for any given file.
2556 ntfs_attr_put_search_ctx(ctx
);
2557 unmap_mft_record(base_ni
);
2558 up_write(&ni
->runlist
.lock
);
2560 * Not enough space in the mft record, try to make the attribute
2561 * non-resident and if successful restart the truncation process.
2563 err
= ntfs_attr_make_non_resident(ni
, old_size
);
2565 goto retry_truncate
;
2567 * Could not make non-resident. If this is due to this not being
2568 * permitted for this attribute type or there not being enough space,
2569 * try to make other attributes non-resident. Otherwise fail.
2571 if (unlikely(err
!= -EPERM
&& err
!= -ENOSPC
)) {
2572 ntfs_error(vol
->sb
, "Cannot truncate inode 0x%lx, attribute "
2573 "type 0x%x, because the conversion from "
2574 "resident to non-resident attribute failed "
2575 "with error code %i.", vi
->i_ino
,
2576 (unsigned)le32_to_cpu(ni
->type
), err
);
2581 /* TODO: Not implemented from here, abort. */
2583 ntfs_error(vol
->sb
, "Not enough space in the mft record/on "
2584 "disk for the non-resident attribute value. "
2585 "This case is not implemented yet.");
2586 else /* if (err == -EPERM) */
2587 ntfs_error(vol
->sb
, "This attribute type may not be "
2588 "non-resident. This case is not implemented "
2593 // TODO: Attempt to make other attributes non-resident.
2595 goto do_resident_extend
;
2597 * Both the attribute list attribute and the standard information
2598 * attribute must remain in the base inode. Thus, if this is one of
2599 * these attributes, we have to try to move other attributes out into
2600 * extent mft records instead.
2602 if (ni
->type
== AT_ATTRIBUTE_LIST
||
2603 ni
->type
== AT_STANDARD_INFORMATION
) {
2604 // TODO: Attempt to move other attributes into extent mft
2608 goto do_resident_extend
;
2611 // TODO: Attempt to move this attribute to an extent mft record, but
2612 // only if it is not already the only attribute in an mft record in
2613 // which case there would be nothing to gain.
2616 goto do_resident_extend
;
2617 /* There is nothing we can do to make enough space. )-: */
2620 do_non_resident_truncate
:
2621 BUG_ON(!NInoNonResident(ni
));
2622 if (alloc_change
< 0) {
2623 highest_vcn
= sle64_to_cpu(a
->data
.non_resident
.highest_vcn
);
2624 if (highest_vcn
> 0 &&
2625 old_alloc_size
>> vol
->cluster_size_bits
>
2628 * This attribute has multiple extents. Not yet
2631 ntfs_error(vol
->sb
, "Cannot truncate inode 0x%lx, "
2632 "attribute type 0x%x, because the "
2633 "attribute is highly fragmented (it "
2634 "consists of multiple extents) and "
2635 "this case is not implemented yet.",
2637 (unsigned)le32_to_cpu(ni
->type
));
2643 * If the size is shrinking, need to reduce the initialized_size and
2644 * the data_size before reducing the allocation.
2646 if (size_change
< 0) {
2648 * Make the valid size smaller (i_size is already up-to-date).
2650 write_lock_irqsave(&ni
->size_lock
, flags
);
2651 if (new_size
< ni
->initialized_size
) {
2652 ni
->initialized_size
= new_size
;
2653 a
->data
.non_resident
.initialized_size
=
2654 cpu_to_sle64(new_size
);
2656 a
->data
.non_resident
.data_size
= cpu_to_sle64(new_size
);
2657 write_unlock_irqrestore(&ni
->size_lock
, flags
);
2658 flush_dcache_mft_record_page(ctx
->ntfs_ino
);
2659 mark_mft_record_dirty(ctx
->ntfs_ino
);
2660 /* If the allocated size is not changing, we are done. */
2664 * If the size is shrinking it makes no sense for the
2665 * allocation to be growing.
2667 BUG_ON(alloc_change
> 0);
2668 } else /* if (size_change >= 0) */ {
2670 * The file size is growing or staying the same but the
2671 * allocation can be shrinking, growing or staying the same.
2673 if (alloc_change
> 0) {
2675 * We need to extend the allocation and possibly update
2676 * the data size. If we are updating the data size,
2677 * since we are not touching the initialized_size we do
2678 * not need to worry about the actual data on disk.
2679 * And as far as the page cache is concerned, there
2680 * will be no pages beyond the old data size and any
2681 * partial region in the last page between the old and
2682 * new data size (or the end of the page if the new
2683 * data size is outside the page) does not need to be
2684 * modified as explained above for the resident
2685 * attribute truncate case. To do this, we simply drop
2686 * the locks we hold and leave all the work to our
2687 * friendly helper ntfs_attr_extend_allocation().
2689 ntfs_attr_put_search_ctx(ctx
);
2690 unmap_mft_record(base_ni
);
2691 up_write(&ni
->runlist
.lock
);
2692 err
= ntfs_attr_extend_allocation(ni
, new_size
,
2693 size_change
> 0 ? new_size
: -1, -1);
2695 * ntfs_attr_extend_allocation() will have done error
2703 /* alloc_change < 0 */
2704 /* Free the clusters. */
2705 nr_freed
= ntfs_cluster_free(ni
, new_alloc_size
>>
2706 vol
->cluster_size_bits
, -1, ctx
);
2709 if (unlikely(nr_freed
< 0)) {
2710 ntfs_error(vol
->sb
, "Failed to release cluster(s) (error code "
2711 "%lli). Unmount and run chkdsk to recover "
2712 "the lost cluster(s).", (long long)nr_freed
);
2716 /* Truncate the runlist. */
2717 err
= ntfs_rl_truncate_nolock(vol
, &ni
->runlist
,
2718 new_alloc_size
>> vol
->cluster_size_bits
);
2720 * If the runlist truncation failed and/or the search context is no
2721 * longer valid, we cannot resize the attribute record or build the
2722 * mapping pairs array thus we mark the inode bad so that no access to
2723 * the freed clusters can happen.
2725 if (unlikely(err
|| IS_ERR(m
))) {
2726 ntfs_error(vol
->sb
, "Failed to %s (error code %li).%s",
2728 "restore attribute search context" :
2729 "truncate attribute runlist",
2730 IS_ERR(m
) ? PTR_ERR(m
) : err
, es
);
2734 /* Get the size for the shrunk mapping pairs array for the runlist. */
2735 mp_size
= ntfs_get_size_for_mapping_pairs(vol
, ni
->runlist
.rl
, 0, -1);
2736 if (unlikely(mp_size
<= 0)) {
2737 ntfs_error(vol
->sb
, "Cannot shrink allocation of inode 0x%lx, "
2738 "attribute type 0x%x, because determining the "
2739 "size for the mapping pairs failed with error "
2740 "code %i.%s", vi
->i_ino
,
2741 (unsigned)le32_to_cpu(ni
->type
), mp_size
, es
);
2746 * Shrink the attribute record for the new mapping pairs array. Note,
2747 * this cannot fail since we are making the attribute smaller thus by
2748 * definition there is enough space to do so.
2750 attr_len
= le32_to_cpu(a
->length
);
2751 err
= ntfs_attr_record_resize(m
, a
, mp_size
+
2752 le16_to_cpu(a
->data
.non_resident
.mapping_pairs_offset
));
2755 * Generate the mapping pairs array directly into the attribute record.
2757 err
= ntfs_mapping_pairs_build(vol
, (u8
*)a
+
2758 le16_to_cpu(a
->data
.non_resident
.mapping_pairs_offset
),
2759 mp_size
, ni
->runlist
.rl
, 0, -1, NULL
);
2760 if (unlikely(err
)) {
2761 ntfs_error(vol
->sb
, "Cannot shrink allocation of inode 0x%lx, "
2762 "attribute type 0x%x, because building the "
2763 "mapping pairs failed with error code %i.%s",
2764 vi
->i_ino
, (unsigned)le32_to_cpu(ni
->type
),
2769 /* Update the allocated/compressed size as well as the highest vcn. */
2770 a
->data
.non_resident
.highest_vcn
= cpu_to_sle64((new_alloc_size
>>
2771 vol
->cluster_size_bits
) - 1);
2772 write_lock_irqsave(&ni
->size_lock
, flags
);
2773 ni
->allocated_size
= new_alloc_size
;
2774 a
->data
.non_resident
.allocated_size
= cpu_to_sle64(new_alloc_size
);
2775 if (NInoSparse(ni
) || NInoCompressed(ni
)) {
2777 ni
->itype
.compressed
.size
-= nr_freed
<<
2778 vol
->cluster_size_bits
;
2779 BUG_ON(ni
->itype
.compressed
.size
< 0);
2780 a
->data
.non_resident
.compressed_size
= cpu_to_sle64(
2781 ni
->itype
.compressed
.size
);
2782 vi
->i_blocks
= ni
->itype
.compressed
.size
>> 9;
2785 vi
->i_blocks
= new_alloc_size
>> 9;
2786 write_unlock_irqrestore(&ni
->size_lock
, flags
);
2788 * We have shrunk the allocation. If this is a shrinking truncate we
2789 * have already dealt with the initialized_size and the data_size above
2790 * and we are done. If the truncate is only changing the allocation
2791 * and not the data_size, we are also done. If this is an extending
2792 * truncate, need to extend the data_size now which is ensured by the
2793 * fact that @size_change is positive.
2797 * If the size is growing, need to update it now. If it is shrinking,
2798 * we have already updated it above (before the allocation change).
2800 if (size_change
> 0)
2801 a
->data
.non_resident
.data_size
= cpu_to_sle64(new_size
);
2802 /* Ensure the modified mft record is written out. */
2803 flush_dcache_mft_record_page(ctx
->ntfs_ino
);
2804 mark_mft_record_dirty(ctx
->ntfs_ino
);
2806 ntfs_attr_put_search_ctx(ctx
);
2807 unmap_mft_record(base_ni
);
2808 up_write(&ni
->runlist
.lock
);
2810 /* Update the mtime and ctime on the base inode. */
2811 /* normally ->truncate shouldn't update ctime or mtime,
2812 * but ntfs did before so it got a copy & paste version
2813 * of file_update_time. one day someone should fix this
2816 if (!IS_NOCMTIME(VFS_I(base_ni
)) && !IS_RDONLY(VFS_I(base_ni
))) {
2817 struct timespec now
= current_fs_time(VFS_I(base_ni
)->i_sb
);
2820 if (!timespec_equal(&VFS_I(base_ni
)->i_mtime
, &now
) ||
2821 !timespec_equal(&VFS_I(base_ni
)->i_ctime
, &now
))
2823 VFS_I(base_ni
)->i_mtime
= now
;
2824 VFS_I(base_ni
)->i_ctime
= now
;
2827 mark_inode_dirty_sync(VFS_I(base_ni
));
2831 NInoClearTruncateFailed(ni
);
2832 ntfs_debug("Done.");
2838 if (err
!= -ENOMEM
&& err
!= -EOPNOTSUPP
)
2840 if (err
!= -EOPNOTSUPP
)
2841 NInoSetTruncateFailed(ni
);
2842 else if (old_size
>= 0)
2843 i_size_write(vi
, old_size
);
2846 ntfs_attr_put_search_ctx(ctx
);
2848 unmap_mft_record(base_ni
);
2849 up_write(&ni
->runlist
.lock
);
2851 ntfs_debug("Failed. Returning error code %i.", err
);
2854 if (err
!= -ENOMEM
&& err
!= -EOPNOTSUPP
)
2856 if (err
!= -EOPNOTSUPP
)
2857 NInoSetTruncateFailed(ni
);
2859 i_size_write(vi
, old_size
);
2864 * ntfs_truncate_vfs - wrapper for ntfs_truncate() that has no return value
2865 * @vi: inode for which the i_size was changed
2867 * Wrapper for ntfs_truncate() that has no return value.
2869 * See ntfs_truncate() description above for details.
2872 void ntfs_truncate_vfs(struct inode
*vi
) {
2878 * ntfs_setattr - called from notify_change() when an attribute is being changed
2879 * @dentry: dentry whose attributes to change
2880 * @attr: structure describing the attributes and the changes
2882 * We have to trap VFS attempts to truncate the file described by @dentry as
2883 * soon as possible, because we do not implement changes in i_size yet. So we
2884 * abort all i_size changes here.
2886 * We also abort all changes of user, group, and mode as we do not implement
2887 * the NTFS ACLs yet.
2889 * Called with ->i_mutex held.
2891 int ntfs_setattr(struct dentry
*dentry
, struct iattr
*attr
)
2893 struct inode
*vi
= dentry
->d_inode
;
2895 unsigned int ia_valid
= attr
->ia_valid
;
2897 err
= inode_change_ok(vi
, attr
);
2900 /* We do not support NTFS ACLs yet. */
2901 if (ia_valid
& (ATTR_UID
| ATTR_GID
| ATTR_MODE
)) {
2902 ntfs_warning(vi
->i_sb
, "Changes in user/group/mode are not "
2903 "supported yet, ignoring.");
2907 if (ia_valid
& ATTR_SIZE
) {
2908 if (attr
->ia_size
!= i_size_read(vi
)) {
2909 ntfs_inode
*ni
= NTFS_I(vi
);
2911 * FIXME: For now we do not support resizing of
2912 * compressed or encrypted files yet.
2914 if (NInoCompressed(ni
) || NInoEncrypted(ni
)) {
2915 ntfs_warning(vi
->i_sb
, "Changes in inode size "
2916 "are not supported yet for "
2917 "%s files, ignoring.",
2918 NInoCompressed(ni
) ?
2919 "compressed" : "encrypted");
2922 truncate_setsize(vi
, attr
->ia_size
);
2923 ntfs_truncate_vfs(vi
);
2925 if (err
|| ia_valid
== ATTR_SIZE
)
2929 * We skipped the truncate but must still update
2932 ia_valid
|= ATTR_MTIME
| ATTR_CTIME
;
2935 if (ia_valid
& ATTR_ATIME
)
2936 vi
->i_atime
= timespec_trunc(attr
->ia_atime
,
2937 vi
->i_sb
->s_time_gran
);
2938 if (ia_valid
& ATTR_MTIME
)
2939 vi
->i_mtime
= timespec_trunc(attr
->ia_mtime
,
2940 vi
->i_sb
->s_time_gran
);
2941 if (ia_valid
& ATTR_CTIME
)
2942 vi
->i_ctime
= timespec_trunc(attr
->ia_ctime
,
2943 vi
->i_sb
->s_time_gran
);
2944 mark_inode_dirty(vi
);
2950 * ntfs_write_inode - write out a dirty inode
2951 * @vi: inode to write out
2952 * @sync: if true, write out synchronously
2954 * Write out a dirty inode to disk including any extent inodes if present.
2956 * If @sync is true, commit the inode to disk and wait for io completion. This
2957 * is done using write_mft_record().
2959 * If @sync is false, just schedule the write to happen but do not wait for i/o
2960 * completion. In 2.6 kernels, scheduling usually happens just by virtue of
2961 * marking the page (and in this case mft record) dirty but we do not implement
2962 * this yet as write_mft_record() largely ignores the @sync parameter and
2963 * always performs synchronous writes.
2965 * Return 0 on success and -errno on error.
2967 int __ntfs_write_inode(struct inode
*vi
, int sync
)
2970 ntfs_inode
*ni
= NTFS_I(vi
);
2971 ntfs_attr_search_ctx
*ctx
;
2973 STANDARD_INFORMATION
*si
;
2975 bool modified
= false;
2977 ntfs_debug("Entering for %sinode 0x%lx.", NInoAttr(ni
) ? "attr " : "",
2980 * Dirty attribute inodes are written via their real inodes so just
2981 * clean them here. Access time updates are taken care off when the
2982 * real inode is written.
2986 ntfs_debug("Done.");
2989 /* Map, pin, and lock the mft record belonging to the inode. */
2990 m
= map_mft_record(ni
);
2995 /* Update the access times in the standard information attribute. */
2996 ctx
= ntfs_attr_get_search_ctx(ni
, m
);
2997 if (unlikely(!ctx
)) {
3001 err
= ntfs_attr_lookup(AT_STANDARD_INFORMATION
, NULL
, 0,
3002 CASE_SENSITIVE
, 0, NULL
, 0, ctx
);
3003 if (unlikely(err
)) {
3004 ntfs_attr_put_search_ctx(ctx
);
3007 si
= (STANDARD_INFORMATION
*)((u8
*)ctx
->attr
+
3008 le16_to_cpu(ctx
->attr
->data
.resident
.value_offset
));
3009 /* Update the access times if they have changed. */
3010 nt
= utc2ntfs(vi
->i_mtime
);
3011 if (si
->last_data_change_time
!= nt
) {
3012 ntfs_debug("Updating mtime for inode 0x%lx: old = 0x%llx, "
3013 "new = 0x%llx", vi
->i_ino
, (long long)
3014 sle64_to_cpu(si
->last_data_change_time
),
3015 (long long)sle64_to_cpu(nt
));
3016 si
->last_data_change_time
= nt
;
3019 nt
= utc2ntfs(vi
->i_ctime
);
3020 if (si
->last_mft_change_time
!= nt
) {
3021 ntfs_debug("Updating ctime for inode 0x%lx: old = 0x%llx, "
3022 "new = 0x%llx", vi
->i_ino
, (long long)
3023 sle64_to_cpu(si
->last_mft_change_time
),
3024 (long long)sle64_to_cpu(nt
));
3025 si
->last_mft_change_time
= nt
;
3028 nt
= utc2ntfs(vi
->i_atime
);
3029 if (si
->last_access_time
!= nt
) {
3030 ntfs_debug("Updating atime for inode 0x%lx: old = 0x%llx, "
3031 "new = 0x%llx", vi
->i_ino
,
3032 (long long)sle64_to_cpu(si
->last_access_time
),
3033 (long long)sle64_to_cpu(nt
));
3034 si
->last_access_time
= nt
;
3038 * If we just modified the standard information attribute we need to
3039 * mark the mft record it is in dirty. We do this manually so that
3040 * mark_inode_dirty() is not called which would redirty the inode and
3041 * hence result in an infinite loop of trying to write the inode.
3042 * There is no need to mark the base inode nor the base mft record
3043 * dirty, since we are going to write this mft record below in any case
3044 * and the base mft record may actually not have been modified so it
3045 * might not need to be written out.
3046 * NOTE: It is not a problem when the inode for $MFT itself is being
3047 * written out as mark_ntfs_record_dirty() will only set I_DIRTY_PAGES
3048 * on the $MFT inode and hence ntfs_write_inode() will not be
3049 * re-invoked because of it which in turn is ok since the dirtied mft
3050 * record will be cleaned and written out to disk below, i.e. before
3051 * this function returns.
3054 flush_dcache_mft_record_page(ctx
->ntfs_ino
);
3055 if (!NInoTestSetDirty(ctx
->ntfs_ino
))
3056 mark_ntfs_record_dirty(ctx
->ntfs_ino
->page
,
3057 ctx
->ntfs_ino
->page_ofs
);
3059 ntfs_attr_put_search_ctx(ctx
);
3060 /* Now the access times are updated, write the base mft record. */
3062 err
= write_mft_record(ni
, m
, sync
);
3063 /* Write all attached extent mft records. */
3064 mutex_lock(&ni
->extent_lock
);
3065 if (ni
->nr_extents
> 0) {
3066 ntfs_inode
**extent_nis
= ni
->ext
.extent_ntfs_inos
;
3069 ntfs_debug("Writing %i extent inodes.", ni
->nr_extents
);
3070 for (i
= 0; i
< ni
->nr_extents
; i
++) {
3071 ntfs_inode
*tni
= extent_nis
[i
];
3073 if (NInoDirty(tni
)) {
3074 MFT_RECORD
*tm
= map_mft_record(tni
);
3078 if (!err
|| err
== -ENOMEM
)
3082 ret
= write_mft_record(tni
, tm
, sync
);
3083 unmap_mft_record(tni
);
3084 if (unlikely(ret
)) {
3085 if (!err
|| err
== -ENOMEM
)
3091 mutex_unlock(&ni
->extent_lock
);
3092 unmap_mft_record(ni
);
3095 ntfs_debug("Done.");
3098 unmap_mft_record(ni
);
3100 if (err
== -ENOMEM
) {
3101 ntfs_warning(vi
->i_sb
, "Not enough memory to write inode. "
3102 "Marking the inode dirty again, so the VFS "
3104 mark_inode_dirty(vi
);
3106 ntfs_error(vi
->i_sb
, "Failed (error %i): Run chkdsk.", -err
);
3107 NVolSetErrors(ni
->vol
);
3112 #endif /* NTFS_RW */