1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * inode.c - NTFS kernel inode handling.
5 * Copyright (c) 2001-2014 Anton Altaparmakov and Tuxera Inc.
8 #include <linux/buffer_head.h>
11 #include <linux/mount.h>
12 #include <linux/mutex.h>
13 #include <linux/pagemap.h>
14 #include <linux/quotaops.h>
15 #include <linux/slab.h>
16 #include <linux/log2.h>
31 * ntfs_test_inode - compare two (possibly fake) inodes for equality
32 * @vi: vfs inode which to test
33 * @data: data which is being tested with
35 * Compare the ntfs attribute embedded in the ntfs specific part of the vfs
36 * inode @vi for equality with the ntfs attribute @data.
38 * If searching for the normal file/directory inode, set @na->type to AT_UNUSED.
39 * @na->name and @na->name_len are then ignored.
41 * Return 1 if the attributes match and 0 if not.
43 * NOTE: This function runs with the inode_hash_lock spin lock held so it is not
46 int ntfs_test_inode(struct inode
*vi
, void *data
)
48 ntfs_attr
*na
= (ntfs_attr
*)data
;
51 if (vi
->i_ino
!= na
->mft_no
)
54 /* If !NInoAttr(ni), @vi is a normal file or directory inode. */
55 if (likely(!NInoAttr(ni
))) {
56 /* If not looking for a normal inode this is a mismatch. */
57 if (unlikely(na
->type
!= AT_UNUSED
))
60 /* A fake inode describing an attribute. */
61 if (ni
->type
!= na
->type
)
63 if (ni
->name_len
!= na
->name_len
)
65 if (na
->name_len
&& memcmp(ni
->name
, na
->name
,
66 na
->name_len
* sizeof(ntfschar
)))
74 * ntfs_init_locked_inode - initialize an inode
75 * @vi: vfs inode to initialize
76 * @data: data which to initialize @vi to
78 * Initialize the vfs inode @vi with the values from the ntfs attribute @data in
79 * order to enable ntfs_test_inode() to do its work.
81 * If initializing the normal file/directory inode, set @na->type to AT_UNUSED.
82 * In that case, @na->name and @na->name_len should be set to NULL and 0,
83 * respectively. Although that is not strictly necessary as
84 * ntfs_read_locked_inode() will fill them in later.
86 * Return 0 on success and -errno on error.
88 * NOTE: This function runs with the inode->i_lock spin lock held so it is not
89 * allowed to sleep. (Hence the GFP_ATOMIC allocation.)
91 static int ntfs_init_locked_inode(struct inode
*vi
, void *data
)
93 ntfs_attr
*na
= (ntfs_attr
*)data
;
94 ntfs_inode
*ni
= NTFS_I(vi
);
96 vi
->i_ino
= na
->mft_no
;
99 if (na
->type
== AT_INDEX_ALLOCATION
)
100 NInoSetMstProtected(ni
);
103 ni
->name_len
= na
->name_len
;
105 /* If initializing a normal inode, we are done. */
106 if (likely(na
->type
== AT_UNUSED
)) {
108 BUG_ON(na
->name_len
);
112 /* It is a fake inode. */
116 * We have I30 global constant as an optimization as it is the name
117 * in >99.9% of named attributes! The other <0.1% incur a GFP_ATOMIC
118 * allocation but that is ok. And most attributes are unnamed anyway,
119 * thus the fraction of named attributes with name != I30 is actually
122 if (na
->name_len
&& na
->name
!= I30
) {
126 i
= na
->name_len
* sizeof(ntfschar
);
127 ni
->name
= kmalloc(i
+ sizeof(ntfschar
), GFP_ATOMIC
);
130 memcpy(ni
->name
, na
->name
, i
);
131 ni
->name
[na
->name_len
] = 0;
136 static int ntfs_read_locked_inode(struct inode
*vi
);
137 static int ntfs_read_locked_attr_inode(struct inode
*base_vi
, struct inode
*vi
);
138 static int ntfs_read_locked_index_inode(struct inode
*base_vi
,
142 * ntfs_iget - obtain a struct inode corresponding to a specific normal inode
143 * @sb: super block of mounted volume
144 * @mft_no: mft record number / inode number to obtain
146 * Obtain the struct inode corresponding to a specific normal inode (i.e. a
147 * file or directory).
149 * If the inode is in the cache, it is just returned with an increased
150 * reference count. Otherwise, a new struct inode is allocated and initialized,
151 * and finally ntfs_read_locked_inode() is called to read in the inode and
152 * fill in the remainder of the inode structure.
154 * Return the struct inode on success. Check the return value with IS_ERR() and
155 * if true, the function failed and the error code is obtained from PTR_ERR().
157 struct inode
*ntfs_iget(struct super_block
*sb
, unsigned long mft_no
)
168 vi
= iget5_locked(sb
, mft_no
, ntfs_test_inode
,
169 ntfs_init_locked_inode
, &na
);
171 return ERR_PTR(-ENOMEM
);
175 /* If this is a freshly allocated inode, need to read it now. */
176 if (vi
->i_state
& I_NEW
) {
177 err
= ntfs_read_locked_inode(vi
);
178 unlock_new_inode(vi
);
181 * There is no point in keeping bad inodes around if the failure was
182 * due to ENOMEM. We want to be able to retry again later.
184 if (unlikely(err
== -ENOMEM
)) {
192 * ntfs_attr_iget - obtain a struct inode corresponding to an attribute
193 * @base_vi: vfs base inode containing the attribute
194 * @type: attribute type
195 * @name: Unicode name of the attribute (NULL if unnamed)
196 * @name_len: length of @name in Unicode characters (0 if unnamed)
198 * Obtain the (fake) struct inode corresponding to the attribute specified by
199 * @type, @name, and @name_len, which is present in the base mft record
200 * specified by the vfs inode @base_vi.
202 * If the attribute inode is in the cache, it is just returned with an
203 * increased reference count. Otherwise, a new struct inode is allocated and
204 * initialized, and finally ntfs_read_locked_attr_inode() is called to read the
205 * attribute and fill in the inode structure.
207 * Note, for index allocation attributes, you need to use ntfs_index_iget()
208 * instead of ntfs_attr_iget() as working with indices is a lot more complex.
210 * Return the struct inode of the attribute inode on success. Check the return
211 * value with IS_ERR() and if true, the function failed and the error code is
212 * obtained from PTR_ERR().
214 struct inode
*ntfs_attr_iget(struct inode
*base_vi
, ATTR_TYPE type
,
215 ntfschar
*name
, u32 name_len
)
221 /* Make sure no one calls ntfs_attr_iget() for indices. */
222 BUG_ON(type
== AT_INDEX_ALLOCATION
);
224 na
.mft_no
= base_vi
->i_ino
;
227 na
.name_len
= name_len
;
229 vi
= iget5_locked(base_vi
->i_sb
, na
.mft_no
, ntfs_test_inode
,
230 ntfs_init_locked_inode
, &na
);
232 return ERR_PTR(-ENOMEM
);
236 /* If this is a freshly allocated inode, need to read it now. */
237 if (vi
->i_state
& I_NEW
) {
238 err
= ntfs_read_locked_attr_inode(base_vi
, vi
);
239 unlock_new_inode(vi
);
242 * There is no point in keeping bad attribute inodes around. This also
243 * simplifies things in that we never need to check for bad attribute
254 * ntfs_index_iget - obtain a struct inode corresponding to an index
255 * @base_vi: vfs base inode containing the index related attributes
256 * @name: Unicode name of the index
257 * @name_len: length of @name in Unicode characters
259 * Obtain the (fake) struct inode corresponding to the index specified by @name
260 * and @name_len, which is present in the base mft record specified by the vfs
263 * If the index inode is in the cache, it is just returned with an increased
264 * reference count. Otherwise, a new struct inode is allocated and
265 * initialized, and finally ntfs_read_locked_index_inode() is called to read
266 * the index related attributes and fill in the inode structure.
268 * Return the struct inode of the index inode on success. Check the return
269 * value with IS_ERR() and if true, the function failed and the error code is
270 * obtained from PTR_ERR().
272 struct inode
*ntfs_index_iget(struct inode
*base_vi
, ntfschar
*name
,
279 na
.mft_no
= base_vi
->i_ino
;
280 na
.type
= AT_INDEX_ALLOCATION
;
282 na
.name_len
= name_len
;
284 vi
= iget5_locked(base_vi
->i_sb
, na
.mft_no
, ntfs_test_inode
,
285 ntfs_init_locked_inode
, &na
);
287 return ERR_PTR(-ENOMEM
);
291 /* If this is a freshly allocated inode, need to read it now. */
292 if (vi
->i_state
& I_NEW
) {
293 err
= ntfs_read_locked_index_inode(base_vi
, vi
);
294 unlock_new_inode(vi
);
297 * There is no point in keeping bad index inodes around. This also
298 * simplifies things in that we never need to check for bad index
308 struct inode
*ntfs_alloc_big_inode(struct super_block
*sb
)
312 ntfs_debug("Entering.");
313 ni
= kmem_cache_alloc(ntfs_big_inode_cache
, GFP_NOFS
);
314 if (likely(ni
!= NULL
)) {
318 ntfs_error(sb
, "Allocation of NTFS big inode structure failed.");
322 void ntfs_free_big_inode(struct inode
*inode
)
324 kmem_cache_free(ntfs_big_inode_cache
, NTFS_I(inode
));
327 static inline ntfs_inode
*ntfs_alloc_extent_inode(void)
331 ntfs_debug("Entering.");
332 ni
= kmem_cache_alloc(ntfs_inode_cache
, GFP_NOFS
);
333 if (likely(ni
!= NULL
)) {
337 ntfs_error(NULL
, "Allocation of NTFS inode structure failed.");
341 static void ntfs_destroy_extent_inode(ntfs_inode
*ni
)
343 ntfs_debug("Entering.");
345 if (!atomic_dec_and_test(&ni
->count
))
347 kmem_cache_free(ntfs_inode_cache
, ni
);
351 * The attribute runlist lock has separate locking rules from the
352 * normal runlist lock, so split the two lock-classes:
354 static struct lock_class_key attr_list_rl_lock_class
;
357 * __ntfs_init_inode - initialize ntfs specific part of an inode
358 * @sb: super block of mounted volume
359 * @ni: freshly allocated ntfs inode which to initialize
361 * Initialize an ntfs inode to defaults.
363 * NOTE: ni->mft_no, ni->state, ni->type, ni->name, and ni->name_len are left
364 * untouched. Make sure to initialize them elsewhere.
366 * Return zero on success and -ENOMEM on error.
368 void __ntfs_init_inode(struct super_block
*sb
, ntfs_inode
*ni
)
370 ntfs_debug("Entering.");
371 rwlock_init(&ni
->size_lock
);
372 ni
->initialized_size
= ni
->allocated_size
= 0;
374 atomic_set(&ni
->count
, 1);
375 ni
->vol
= NTFS_SB(sb
);
376 ntfs_init_runlist(&ni
->runlist
);
377 mutex_init(&ni
->mrec_lock
);
380 ni
->attr_list_size
= 0;
381 ni
->attr_list
= NULL
;
382 ntfs_init_runlist(&ni
->attr_list_rl
);
383 lockdep_set_class(&ni
->attr_list_rl
.lock
,
384 &attr_list_rl_lock_class
);
385 ni
->itype
.index
.block_size
= 0;
386 ni
->itype
.index
.vcn_size
= 0;
387 ni
->itype
.index
.collation_rule
= 0;
388 ni
->itype
.index
.block_size_bits
= 0;
389 ni
->itype
.index
.vcn_size_bits
= 0;
390 mutex_init(&ni
->extent_lock
);
392 ni
->ext
.base_ntfs_ino
= NULL
;
396 * Extent inodes get MFT-mapped in a nested way, while the base inode
397 * is still mapped. Teach this nesting to the lock validator by creating
398 * a separate class for nested inode's mrec_lock's:
400 static struct lock_class_key extent_inode_mrec_lock_key
;
402 inline ntfs_inode
*ntfs_new_extent_inode(struct super_block
*sb
,
403 unsigned long mft_no
)
405 ntfs_inode
*ni
= ntfs_alloc_extent_inode();
407 ntfs_debug("Entering.");
408 if (likely(ni
!= NULL
)) {
409 __ntfs_init_inode(sb
, ni
);
410 lockdep_set_class(&ni
->mrec_lock
, &extent_inode_mrec_lock_key
);
412 ni
->type
= AT_UNUSED
;
420 * ntfs_is_extended_system_file - check if a file is in the $Extend directory
421 * @ctx: initialized attribute search context
423 * Search all file name attributes in the inode described by the attribute
424 * search context @ctx and check if any of the names are in the $Extend system
428 * 1: file is in $Extend directory
429 * 0: file is not in $Extend directory
430 * -errno: failed to determine if the file is in the $Extend directory
432 static int ntfs_is_extended_system_file(ntfs_attr_search_ctx
*ctx
)
436 /* Restart search. */
437 ntfs_attr_reinit_search_ctx(ctx
);
439 /* Get number of hard links. */
440 nr_links
= le16_to_cpu(ctx
->mrec
->link_count
);
442 /* Loop through all hard links. */
443 while (!(err
= ntfs_attr_lookup(AT_FILE_NAME
, NULL
, 0, 0, 0, NULL
, 0,
445 FILE_NAME_ATTR
*file_name_attr
;
446 ATTR_RECORD
*attr
= ctx
->attr
;
451 * Maximum sanity checking as we are called on an inode that
452 * we suspect might be corrupt.
454 p
= (u8
*)attr
+ le32_to_cpu(attr
->length
);
455 if (p
< (u8
*)ctx
->mrec
|| (u8
*)p
> (u8
*)ctx
->mrec
+
456 le32_to_cpu(ctx
->mrec
->bytes_in_use
)) {
458 ntfs_error(ctx
->ntfs_ino
->vol
->sb
, "Corrupt file name "
459 "attribute. You should run chkdsk.");
462 if (attr
->non_resident
) {
463 ntfs_error(ctx
->ntfs_ino
->vol
->sb
, "Non-resident file "
464 "name. You should run chkdsk.");
468 ntfs_error(ctx
->ntfs_ino
->vol
->sb
, "File name with "
469 "invalid flags. You should run "
473 if (!(attr
->data
.resident
.flags
& RESIDENT_ATTR_IS_INDEXED
)) {
474 ntfs_error(ctx
->ntfs_ino
->vol
->sb
, "Unindexed file "
475 "name. You should run chkdsk.");
478 file_name_attr
= (FILE_NAME_ATTR
*)((u8
*)attr
+
479 le16_to_cpu(attr
->data
.resident
.value_offset
));
480 p2
= (u8
*)file_name_attr
+ le32_to_cpu(attr
->data
.resident
.value_length
);
481 if (p2
< (u8
*)attr
|| p2
> p
)
482 goto err_corrupt_attr
;
483 /* This attribute is ok, but is it in the $Extend directory? */
484 if (MREF_LE(file_name_attr
->parent_directory
) == FILE_Extend
)
485 return 1; /* YES, it's an extended system file. */
487 if (unlikely(err
!= -ENOENT
))
489 if (unlikely(nr_links
)) {
490 ntfs_error(ctx
->ntfs_ino
->vol
->sb
, "Inode hard link count "
491 "doesn't match number of name attributes. You "
492 "should run chkdsk.");
495 return 0; /* NO, it is not an extended system file. */
499 * ntfs_read_locked_inode - read an inode from its device
502 * ntfs_read_locked_inode() is called from ntfs_iget() to read the inode
503 * described by @vi into memory from the device.
505 * The only fields in @vi that we need to/can look at when the function is
506 * called are i_sb, pointing to the mounted device's super block, and i_ino,
507 * the number of the inode to load.
509 * ntfs_read_locked_inode() maps, pins and locks the mft record number i_ino
510 * for reading and sets up the necessary @vi fields as well as initializing
513 * Q: What locks are held when the function is called?
514 * A: i_state has I_NEW set, hence the inode is locked, also
515 * i_count is set to 1, so it is not going to go away
516 * i_flags is set to 0 and we have no business touching it. Only an ioctl()
517 * is allowed to write to them. We should of course be honouring them but
518 * we need to do that using the IS_* macros defined in include/linux/fs.h.
519 * In any case ntfs_read_locked_inode() has nothing to do with i_flags.
521 * Return 0 on success and -errno on error. In the error case, the inode will
522 * have had make_bad_inode() executed on it.
524 static int ntfs_read_locked_inode(struct inode
*vi
)
526 ntfs_volume
*vol
= NTFS_SB(vi
->i_sb
);
531 STANDARD_INFORMATION
*si
;
532 ntfs_attr_search_ctx
*ctx
;
535 ntfs_debug("Entering for i_ino 0x%lx.", vi
->i_ino
);
537 /* Setup the generic vfs inode parts now. */
538 vi
->i_uid
= vol
->uid
;
539 vi
->i_gid
= vol
->gid
;
543 * Initialize the ntfs specific part of @vi special casing
544 * FILE_MFT which we need to do at mount time.
546 if (vi
->i_ino
!= FILE_MFT
)
547 ntfs_init_big_inode(vi
);
550 m
= map_mft_record(ni
);
555 ctx
= ntfs_attr_get_search_ctx(ni
, m
);
561 if (!(m
->flags
& MFT_RECORD_IN_USE
)) {
562 ntfs_error(vi
->i_sb
, "Inode is not in use!");
565 if (m
->base_mft_record
) {
566 ntfs_error(vi
->i_sb
, "Inode is an extent inode!");
570 /* Transfer information from mft record into vfs and ntfs inodes. */
571 vi
->i_generation
= ni
->seq_no
= le16_to_cpu(m
->sequence_number
);
574 * FIXME: Keep in mind that link_count is two for files which have both
575 * a long file name and a short file name as separate entries, so if
576 * we are hiding short file names this will be too high. Either we need
577 * to account for the short file names by subtracting them or we need
578 * to make sure we delete files even though i_nlink is not zero which
579 * might be tricky due to vfs interactions. Need to think about this
580 * some more when implementing the unlink command.
582 set_nlink(vi
, le16_to_cpu(m
->link_count
));
584 * FIXME: Reparse points can have the directory bit set even though
585 * they would be S_IFLNK. Need to deal with this further below when we
586 * implement reparse points / symbolic links but it will do for now.
587 * Also if not a directory, it could be something else, rather than
588 * a regular file. But again, will do for now.
590 /* Everyone gets all permissions. */
591 vi
->i_mode
|= S_IRWXUGO
;
592 /* If read-only, no one gets write permissions. */
594 vi
->i_mode
&= ~S_IWUGO
;
595 if (m
->flags
& MFT_RECORD_IS_DIRECTORY
) {
596 vi
->i_mode
|= S_IFDIR
;
598 * Apply the directory permissions mask set in the mount
601 vi
->i_mode
&= ~vol
->dmask
;
602 /* Things break without this kludge! */
606 vi
->i_mode
|= S_IFREG
;
607 /* Apply the file permissions mask set in the mount options. */
608 vi
->i_mode
&= ~vol
->fmask
;
611 * Find the standard information attribute in the mft record. At this
612 * stage we haven't setup the attribute list stuff yet, so this could
613 * in fact fail if the standard information is in an extent record, but
614 * I don't think this actually ever happens.
616 err
= ntfs_attr_lookup(AT_STANDARD_INFORMATION
, NULL
, 0, 0, 0, NULL
, 0,
619 if (err
== -ENOENT
) {
621 * TODO: We should be performing a hot fix here (if the
622 * recover mount option is set) by creating a new
625 ntfs_error(vi
->i_sb
, "$STANDARD_INFORMATION attribute "
631 /* Get the standard information attribute value. */
632 if ((u8
*)a
+ le16_to_cpu(a
->data
.resident
.value_offset
)
633 + le32_to_cpu(a
->data
.resident
.value_length
) >
634 (u8
*)ctx
->mrec
+ vol
->mft_record_size
) {
635 ntfs_error(vi
->i_sb
, "Corrupt standard information attribute in inode.");
638 si
= (STANDARD_INFORMATION
*)((u8
*)a
+
639 le16_to_cpu(a
->data
.resident
.value_offset
));
641 /* Transfer information from the standard information into vi. */
643 * Note: The i_?times do not quite map perfectly onto the NTFS times,
644 * but they are close enough, and in the end it doesn't really matter
648 * mtime is the last change of the data within the file. Not changed
649 * when only metadata is changed, e.g. a rename doesn't affect mtime.
651 vi
->i_mtime
= ntfs2utc(si
->last_data_change_time
);
653 * ctime is the last change of the metadata of the file. This obviously
654 * always changes, when mtime is changed. ctime can be changed on its
655 * own, mtime is then not changed, e.g. when a file is renamed.
657 vi
->i_ctime
= ntfs2utc(si
->last_mft_change_time
);
659 * Last access to the data within the file. Not changed during a rename
660 * for example but changed whenever the file is written to.
662 vi
->i_atime
= ntfs2utc(si
->last_access_time
);
664 /* Find the attribute list attribute if present. */
665 ntfs_attr_reinit_search_ctx(ctx
);
666 err
= ntfs_attr_lookup(AT_ATTRIBUTE_LIST
, NULL
, 0, 0, 0, NULL
, 0, ctx
);
668 if (unlikely(err
!= -ENOENT
)) {
669 ntfs_error(vi
->i_sb
, "Failed to lookup attribute list "
673 } else /* if (!err) */ {
674 if (vi
->i_ino
== FILE_MFT
)
675 goto skip_attr_list_load
;
676 ntfs_debug("Attribute list found in inode 0x%lx.", vi
->i_ino
);
679 if (a
->flags
& ATTR_COMPRESSION_MASK
) {
680 ntfs_error(vi
->i_sb
, "Attribute list attribute is "
684 if (a
->flags
& ATTR_IS_ENCRYPTED
||
685 a
->flags
& ATTR_IS_SPARSE
) {
686 if (a
->non_resident
) {
687 ntfs_error(vi
->i_sb
, "Non-resident attribute "
688 "list attribute is encrypted/"
692 ntfs_warning(vi
->i_sb
, "Resident attribute list "
693 "attribute in inode 0x%lx is marked "
694 "encrypted/sparse which is not true. "
695 "However, Windows allows this and "
696 "chkdsk does not detect or correct it "
697 "so we will just ignore the invalid "
698 "flags and pretend they are not set.",
701 /* Now allocate memory for the attribute list. */
702 ni
->attr_list_size
= (u32
)ntfs_attr_size(a
);
703 ni
->attr_list
= ntfs_malloc_nofs(ni
->attr_list_size
);
704 if (!ni
->attr_list
) {
705 ntfs_error(vi
->i_sb
, "Not enough memory to allocate "
706 "buffer for attribute list.");
710 if (a
->non_resident
) {
711 NInoSetAttrListNonResident(ni
);
712 if (a
->data
.non_resident
.lowest_vcn
) {
713 ntfs_error(vi
->i_sb
, "Attribute list has non "
718 * Setup the runlist. No need for locking as we have
719 * exclusive access to the inode at this time.
721 ni
->attr_list_rl
.rl
= ntfs_mapping_pairs_decompress(vol
,
723 if (IS_ERR(ni
->attr_list_rl
.rl
)) {
724 err
= PTR_ERR(ni
->attr_list_rl
.rl
);
725 ni
->attr_list_rl
.rl
= NULL
;
726 ntfs_error(vi
->i_sb
, "Mapping pairs "
727 "decompression failed.");
730 /* Now load the attribute list. */
731 if ((err
= load_attribute_list(vol
, &ni
->attr_list_rl
,
732 ni
->attr_list
, ni
->attr_list_size
,
733 sle64_to_cpu(a
->data
.non_resident
.
734 initialized_size
)))) {
735 ntfs_error(vi
->i_sb
, "Failed to load "
736 "attribute list attribute.");
739 } else /* if (!a->non_resident) */ {
740 if ((u8
*)a
+ le16_to_cpu(a
->data
.resident
.value_offset
)
742 a
->data
.resident
.value_length
) >
743 (u8
*)ctx
->mrec
+ vol
->mft_record_size
) {
744 ntfs_error(vi
->i_sb
, "Corrupt attribute list "
748 /* Now copy the attribute list. */
749 memcpy(ni
->attr_list
, (u8
*)a
+ le16_to_cpu(
750 a
->data
.resident
.value_offset
),
752 a
->data
.resident
.value_length
));
757 * If an attribute list is present we now have the attribute list value
758 * in ntfs_ino->attr_list and it is ntfs_ino->attr_list_size bytes.
760 if (S_ISDIR(vi
->i_mode
)) {
764 u8
*ir_end
, *index_end
;
766 /* It is a directory, find index root attribute. */
767 ntfs_attr_reinit_search_ctx(ctx
);
768 err
= ntfs_attr_lookup(AT_INDEX_ROOT
, I30
, 4, CASE_SENSITIVE
,
771 if (err
== -ENOENT
) {
772 // FIXME: File is corrupt! Hot-fix with empty
773 // index root attribute if recovery option is
775 ntfs_error(vi
->i_sb
, "$INDEX_ROOT attribute "
781 /* Set up the state. */
782 if (unlikely(a
->non_resident
)) {
783 ntfs_error(vol
->sb
, "$INDEX_ROOT attribute is not "
787 /* Ensure the attribute name is placed before the value. */
788 if (unlikely(a
->name_length
&& (le16_to_cpu(a
->name_offset
) >=
789 le16_to_cpu(a
->data
.resident
.value_offset
)))) {
790 ntfs_error(vol
->sb
, "$INDEX_ROOT attribute name is "
791 "placed after the attribute value.");
795 * Compressed/encrypted index root just means that the newly
796 * created files in that directory should be created compressed/
797 * encrypted. However index root cannot be both compressed and
800 if (a
->flags
& ATTR_COMPRESSION_MASK
)
801 NInoSetCompressed(ni
);
802 if (a
->flags
& ATTR_IS_ENCRYPTED
) {
803 if (a
->flags
& ATTR_COMPRESSION_MASK
) {
804 ntfs_error(vi
->i_sb
, "Found encrypted and "
805 "compressed attribute.");
808 NInoSetEncrypted(ni
);
810 if (a
->flags
& ATTR_IS_SPARSE
)
812 ir
= (INDEX_ROOT
*)((u8
*)a
+
813 le16_to_cpu(a
->data
.resident
.value_offset
));
814 ir_end
= (u8
*)ir
+ le32_to_cpu(a
->data
.resident
.value_length
);
815 if (ir_end
> (u8
*)ctx
->mrec
+ vol
->mft_record_size
) {
816 ntfs_error(vi
->i_sb
, "$INDEX_ROOT attribute is "
820 index_end
= (u8
*)&ir
->index
+
821 le32_to_cpu(ir
->index
.index_length
);
822 if (index_end
> ir_end
) {
823 ntfs_error(vi
->i_sb
, "Directory index is corrupt.");
826 if (ir
->type
!= AT_FILE_NAME
) {
827 ntfs_error(vi
->i_sb
, "Indexed attribute is not "
831 if (ir
->collation_rule
!= COLLATION_FILE_NAME
) {
832 ntfs_error(vi
->i_sb
, "Index collation rule is not "
833 "COLLATION_FILE_NAME.");
836 ni
->itype
.index
.collation_rule
= ir
->collation_rule
;
837 ni
->itype
.index
.block_size
= le32_to_cpu(ir
->index_block_size
);
838 if (ni
->itype
.index
.block_size
&
839 (ni
->itype
.index
.block_size
- 1)) {
840 ntfs_error(vi
->i_sb
, "Index block size (%u) is not a "
842 ni
->itype
.index
.block_size
);
845 if (ni
->itype
.index
.block_size
> PAGE_SIZE
) {
846 ntfs_error(vi
->i_sb
, "Index block size (%u) > "
847 "PAGE_SIZE (%ld) is not "
849 ni
->itype
.index
.block_size
,
854 if (ni
->itype
.index
.block_size
< NTFS_BLOCK_SIZE
) {
855 ntfs_error(vi
->i_sb
, "Index block size (%u) < "
856 "NTFS_BLOCK_SIZE (%i) is not "
858 ni
->itype
.index
.block_size
,
863 ni
->itype
.index
.block_size_bits
=
864 ffs(ni
->itype
.index
.block_size
) - 1;
865 /* Determine the size of a vcn in the directory index. */
866 if (vol
->cluster_size
<= ni
->itype
.index
.block_size
) {
867 ni
->itype
.index
.vcn_size
= vol
->cluster_size
;
868 ni
->itype
.index
.vcn_size_bits
= vol
->cluster_size_bits
;
870 ni
->itype
.index
.vcn_size
= vol
->sector_size
;
871 ni
->itype
.index
.vcn_size_bits
= vol
->sector_size_bits
;
874 /* Setup the index allocation attribute, even if not present. */
875 NInoSetMstProtected(ni
);
876 ni
->type
= AT_INDEX_ALLOCATION
;
880 if (!(ir
->index
.flags
& LARGE_INDEX
)) {
881 /* No index allocation. */
882 vi
->i_size
= ni
->initialized_size
=
883 ni
->allocated_size
= 0;
884 /* We are done with the mft record, so we release it. */
885 ntfs_attr_put_search_ctx(ctx
);
886 unmap_mft_record(ni
);
889 goto skip_large_dir_stuff
;
890 } /* LARGE_INDEX: Index allocation present. Setup state. */
891 NInoSetIndexAllocPresent(ni
);
892 /* Find index allocation attribute. */
893 ntfs_attr_reinit_search_ctx(ctx
);
894 err
= ntfs_attr_lookup(AT_INDEX_ALLOCATION
, I30
, 4,
895 CASE_SENSITIVE
, 0, NULL
, 0, ctx
);
898 ntfs_error(vi
->i_sb
, "$INDEX_ALLOCATION "
899 "attribute is not present but "
900 "$INDEX_ROOT indicated it is.");
902 ntfs_error(vi
->i_sb
, "Failed to lookup "
908 if (!a
->non_resident
) {
909 ntfs_error(vi
->i_sb
, "$INDEX_ALLOCATION attribute "
914 * Ensure the attribute name is placed before the mapping pairs
917 if (unlikely(a
->name_length
&& (le16_to_cpu(a
->name_offset
) >=
919 a
->data
.non_resident
.mapping_pairs_offset
)))) {
920 ntfs_error(vol
->sb
, "$INDEX_ALLOCATION attribute name "
921 "is placed after the mapping pairs "
925 if (a
->flags
& ATTR_IS_ENCRYPTED
) {
926 ntfs_error(vi
->i_sb
, "$INDEX_ALLOCATION attribute "
930 if (a
->flags
& ATTR_IS_SPARSE
) {
931 ntfs_error(vi
->i_sb
, "$INDEX_ALLOCATION attribute "
935 if (a
->flags
& ATTR_COMPRESSION_MASK
) {
936 ntfs_error(vi
->i_sb
, "$INDEX_ALLOCATION attribute "
940 if (a
->data
.non_resident
.lowest_vcn
) {
941 ntfs_error(vi
->i_sb
, "First extent of "
942 "$INDEX_ALLOCATION attribute has non "
946 vi
->i_size
= sle64_to_cpu(a
->data
.non_resident
.data_size
);
947 ni
->initialized_size
= sle64_to_cpu(
948 a
->data
.non_resident
.initialized_size
);
949 ni
->allocated_size
= sle64_to_cpu(
950 a
->data
.non_resident
.allocated_size
);
952 * We are done with the mft record, so we release it. Otherwise
953 * we would deadlock in ntfs_attr_iget().
955 ntfs_attr_put_search_ctx(ctx
);
956 unmap_mft_record(ni
);
959 /* Get the index bitmap attribute inode. */
960 bvi
= ntfs_attr_iget(vi
, AT_BITMAP
, I30
, 4);
962 ntfs_error(vi
->i_sb
, "Failed to get bitmap attribute.");
967 if (NInoCompressed(bni
) || NInoEncrypted(bni
) ||
969 ntfs_error(vi
->i_sb
, "$BITMAP attribute is compressed "
970 "and/or encrypted and/or sparse.");
971 goto iput_unm_err_out
;
973 /* Consistency check bitmap size vs. index allocation size. */
974 bvi_size
= i_size_read(bvi
);
975 if ((bvi_size
<< 3) < (vi
->i_size
>>
976 ni
->itype
.index
.block_size_bits
)) {
977 ntfs_error(vi
->i_sb
, "Index bitmap too small (0x%llx) "
978 "for index allocation (0x%llx).",
979 bvi_size
<< 3, vi
->i_size
);
980 goto iput_unm_err_out
;
982 /* No longer need the bitmap attribute inode. */
984 skip_large_dir_stuff
:
985 /* Setup the operations for this inode. */
986 vi
->i_op
= &ntfs_dir_inode_ops
;
987 vi
->i_fop
= &ntfs_dir_ops
;
988 vi
->i_mapping
->a_ops
= &ntfs_mst_aops
;
991 ntfs_attr_reinit_search_ctx(ctx
);
993 /* Setup the data attribute, even if not present. */
998 /* Find first extent of the unnamed data attribute. */
999 err
= ntfs_attr_lookup(AT_DATA
, NULL
, 0, 0, 0, NULL
, 0, ctx
);
1000 if (unlikely(err
)) {
1001 vi
->i_size
= ni
->initialized_size
=
1002 ni
->allocated_size
= 0;
1003 if (err
!= -ENOENT
) {
1004 ntfs_error(vi
->i_sb
, "Failed to lookup $DATA "
1009 * FILE_Secure does not have an unnamed $DATA
1010 * attribute, so we special case it here.
1012 if (vi
->i_ino
== FILE_Secure
)
1013 goto no_data_attr_special_case
;
1015 * Most if not all the system files in the $Extend
1016 * system directory do not have unnamed data
1017 * attributes so we need to check if the parent
1018 * directory of the file is FILE_Extend and if it is
1019 * ignore this error. To do this we need to get the
1020 * name of this inode from the mft record as the name
1021 * contains the back reference to the parent directory.
1023 if (ntfs_is_extended_system_file(ctx
) > 0)
1024 goto no_data_attr_special_case
;
1025 // FIXME: File is corrupt! Hot-fix with empty data
1026 // attribute if recovery option is set.
1027 ntfs_error(vi
->i_sb
, "$DATA attribute is missing.");
1031 /* Setup the state. */
1032 if (a
->flags
& (ATTR_COMPRESSION_MASK
| ATTR_IS_SPARSE
)) {
1033 if (a
->flags
& ATTR_COMPRESSION_MASK
) {
1034 NInoSetCompressed(ni
);
1035 if (vol
->cluster_size
> 4096) {
1036 ntfs_error(vi
->i_sb
, "Found "
1037 "compressed data but "
1040 "cluster size (%i) > "
1045 if ((a
->flags
& ATTR_COMPRESSION_MASK
)
1046 != ATTR_IS_COMPRESSED
) {
1047 ntfs_error(vi
->i_sb
, "Found unknown "
1048 "compression method "
1049 "or corrupt file.");
1053 if (a
->flags
& ATTR_IS_SPARSE
)
1056 if (a
->flags
& ATTR_IS_ENCRYPTED
) {
1057 if (NInoCompressed(ni
)) {
1058 ntfs_error(vi
->i_sb
, "Found encrypted and "
1059 "compressed data.");
1062 NInoSetEncrypted(ni
);
1064 if (a
->non_resident
) {
1065 NInoSetNonResident(ni
);
1066 if (NInoCompressed(ni
) || NInoSparse(ni
)) {
1067 if (NInoCompressed(ni
) && a
->data
.non_resident
.
1068 compression_unit
!= 4) {
1069 ntfs_error(vi
->i_sb
, "Found "
1071 "compression unit (%u "
1073 "Cannot handle this.",
1074 a
->data
.non_resident
.
1079 if (a
->data
.non_resident
.compression_unit
) {
1080 ni
->itype
.compressed
.block_size
= 1U <<
1081 (a
->data
.non_resident
.
1083 vol
->cluster_size_bits
);
1084 ni
->itype
.compressed
.block_size_bits
=
1088 ni
->itype
.compressed
.block_clusters
=
1093 ni
->itype
.compressed
.block_size
= 0;
1094 ni
->itype
.compressed
.block_size_bits
=
1096 ni
->itype
.compressed
.block_clusters
=
1099 ni
->itype
.compressed
.size
= sle64_to_cpu(
1100 a
->data
.non_resident
.
1103 if (a
->data
.non_resident
.lowest_vcn
) {
1104 ntfs_error(vi
->i_sb
, "First extent of $DATA "
1105 "attribute has non zero "
1109 vi
->i_size
= sle64_to_cpu(
1110 a
->data
.non_resident
.data_size
);
1111 ni
->initialized_size
= sle64_to_cpu(
1112 a
->data
.non_resident
.initialized_size
);
1113 ni
->allocated_size
= sle64_to_cpu(
1114 a
->data
.non_resident
.allocated_size
);
1115 } else { /* Resident attribute. */
1116 vi
->i_size
= ni
->initialized_size
= le32_to_cpu(
1117 a
->data
.resident
.value_length
);
1118 ni
->allocated_size
= le32_to_cpu(a
->length
) -
1120 a
->data
.resident
.value_offset
);
1121 if (vi
->i_size
> ni
->allocated_size
) {
1122 ntfs_error(vi
->i_sb
, "Resident data attribute "
1123 "is corrupt (size exceeds "
1128 no_data_attr_special_case
:
1129 /* We are done with the mft record, so we release it. */
1130 ntfs_attr_put_search_ctx(ctx
);
1131 unmap_mft_record(ni
);
1134 /* Setup the operations for this inode. */
1135 vi
->i_op
= &ntfs_file_inode_ops
;
1136 vi
->i_fop
= &ntfs_file_ops
;
1137 vi
->i_mapping
->a_ops
= &ntfs_normal_aops
;
1138 if (NInoMstProtected(ni
))
1139 vi
->i_mapping
->a_ops
= &ntfs_mst_aops
;
1140 else if (NInoCompressed(ni
))
1141 vi
->i_mapping
->a_ops
= &ntfs_compressed_aops
;
1144 * The number of 512-byte blocks used on disk (for stat). This is in so
1145 * far inaccurate as it doesn't account for any named streams or other
1146 * special non-resident attributes, but that is how Windows works, too,
1147 * so we are at least consistent with Windows, if not entirely
1148 * consistent with the Linux Way. Doing it the Linux Way would cause a
1149 * significant slowdown as it would involve iterating over all
1150 * attributes in the mft record and adding the allocated/compressed
1151 * sizes of all non-resident attributes present to give us the Linux
1152 * correct size that should go into i_blocks (after division by 512).
1154 if (S_ISREG(vi
->i_mode
) && (NInoCompressed(ni
) || NInoSparse(ni
)))
1155 vi
->i_blocks
= ni
->itype
.compressed
.size
>> 9;
1157 vi
->i_blocks
= ni
->allocated_size
>> 9;
1158 ntfs_debug("Done.");
1166 ntfs_attr_put_search_ctx(ctx
);
1168 unmap_mft_record(ni
);
1170 ntfs_error(vol
->sb
, "Failed with error code %i. Marking corrupt "
1171 "inode 0x%lx as bad. Run chkdsk.", err
, vi
->i_ino
);
1173 if (err
!= -EOPNOTSUPP
&& err
!= -ENOMEM
)
1179 * ntfs_read_locked_attr_inode - read an attribute inode from its base inode
1180 * @base_vi: base inode
1181 * @vi: attribute inode to read
1183 * ntfs_read_locked_attr_inode() is called from ntfs_attr_iget() to read the
1184 * attribute inode described by @vi into memory from the base mft record
1185 * described by @base_ni.
1187 * ntfs_read_locked_attr_inode() maps, pins and locks the base inode for
1188 * reading and looks up the attribute described by @vi before setting up the
1189 * necessary fields in @vi as well as initializing the ntfs inode.
1191 * Q: What locks are held when the function is called?
1192 * A: i_state has I_NEW set, hence the inode is locked, also
1193 * i_count is set to 1, so it is not going to go away
1195 * Return 0 on success and -errno on error. In the error case, the inode will
1196 * have had make_bad_inode() executed on it.
1198 * Note this cannot be called for AT_INDEX_ALLOCATION.
1200 static int ntfs_read_locked_attr_inode(struct inode
*base_vi
, struct inode
*vi
)
1202 ntfs_volume
*vol
= NTFS_SB(vi
->i_sb
);
1203 ntfs_inode
*ni
, *base_ni
;
1206 ntfs_attr_search_ctx
*ctx
;
1209 ntfs_debug("Entering for i_ino 0x%lx.", vi
->i_ino
);
1211 ntfs_init_big_inode(vi
);
1214 base_ni
= NTFS_I(base_vi
);
1216 /* Just mirror the values from the base inode. */
1217 vi
->i_uid
= base_vi
->i_uid
;
1218 vi
->i_gid
= base_vi
->i_gid
;
1219 set_nlink(vi
, base_vi
->i_nlink
);
1220 vi
->i_mtime
= base_vi
->i_mtime
;
1221 vi
->i_ctime
= base_vi
->i_ctime
;
1222 vi
->i_atime
= base_vi
->i_atime
;
1223 vi
->i_generation
= ni
->seq_no
= base_ni
->seq_no
;
1225 /* Set inode type to zero but preserve permissions. */
1226 vi
->i_mode
= base_vi
->i_mode
& ~S_IFMT
;
1228 m
= map_mft_record(base_ni
);
1233 ctx
= ntfs_attr_get_search_ctx(base_ni
, m
);
1238 /* Find the attribute. */
1239 err
= ntfs_attr_lookup(ni
->type
, ni
->name
, ni
->name_len
,
1240 CASE_SENSITIVE
, 0, NULL
, 0, ctx
);
1244 if (a
->flags
& (ATTR_COMPRESSION_MASK
| ATTR_IS_SPARSE
)) {
1245 if (a
->flags
& ATTR_COMPRESSION_MASK
) {
1246 NInoSetCompressed(ni
);
1247 if ((ni
->type
!= AT_DATA
) || (ni
->type
== AT_DATA
&&
1249 ntfs_error(vi
->i_sb
, "Found compressed "
1250 "non-data or named data "
1251 "attribute. Please report "
1252 "you saw this message to "
1253 "linux-ntfs-dev@lists."
1257 if (vol
->cluster_size
> 4096) {
1258 ntfs_error(vi
->i_sb
, "Found compressed "
1259 "attribute but compression is "
1260 "disabled due to cluster size "
1265 if ((a
->flags
& ATTR_COMPRESSION_MASK
) !=
1266 ATTR_IS_COMPRESSED
) {
1267 ntfs_error(vi
->i_sb
, "Found unknown "
1268 "compression method.");
1273 * The compressed/sparse flag set in an index root just means
1274 * to compress all files.
1276 if (NInoMstProtected(ni
) && ni
->type
!= AT_INDEX_ROOT
) {
1277 ntfs_error(vi
->i_sb
, "Found mst protected attribute "
1278 "but the attribute is %s. Please "
1279 "report you saw this message to "
1280 "linux-ntfs-dev@lists.sourceforge.net",
1281 NInoCompressed(ni
) ? "compressed" :
1285 if (a
->flags
& ATTR_IS_SPARSE
)
1288 if (a
->flags
& ATTR_IS_ENCRYPTED
) {
1289 if (NInoCompressed(ni
)) {
1290 ntfs_error(vi
->i_sb
, "Found encrypted and compressed "
1295 * The encryption flag set in an index root just means to
1296 * encrypt all files.
1298 if (NInoMstProtected(ni
) && ni
->type
!= AT_INDEX_ROOT
) {
1299 ntfs_error(vi
->i_sb
, "Found mst protected attribute "
1300 "but the attribute is encrypted. "
1301 "Please report you saw this message "
1302 "to linux-ntfs-dev@lists.sourceforge."
1306 if (ni
->type
!= AT_DATA
) {
1307 ntfs_error(vi
->i_sb
, "Found encrypted non-data "
1311 NInoSetEncrypted(ni
);
1313 if (!a
->non_resident
) {
1314 /* Ensure the attribute name is placed before the value. */
1315 if (unlikely(a
->name_length
&& (le16_to_cpu(a
->name_offset
) >=
1316 le16_to_cpu(a
->data
.resident
.value_offset
)))) {
1317 ntfs_error(vol
->sb
, "Attribute name is placed after "
1318 "the attribute value.");
1321 if (NInoMstProtected(ni
)) {
1322 ntfs_error(vi
->i_sb
, "Found mst protected attribute "
1323 "but the attribute is resident. "
1324 "Please report you saw this message to "
1325 "linux-ntfs-dev@lists.sourceforge.net");
1328 vi
->i_size
= ni
->initialized_size
= le32_to_cpu(
1329 a
->data
.resident
.value_length
);
1330 ni
->allocated_size
= le32_to_cpu(a
->length
) -
1331 le16_to_cpu(a
->data
.resident
.value_offset
);
1332 if (vi
->i_size
> ni
->allocated_size
) {
1333 ntfs_error(vi
->i_sb
, "Resident attribute is corrupt "
1334 "(size exceeds allocation).");
1338 NInoSetNonResident(ni
);
1340 * Ensure the attribute name is placed before the mapping pairs
1343 if (unlikely(a
->name_length
&& (le16_to_cpu(a
->name_offset
) >=
1345 a
->data
.non_resident
.mapping_pairs_offset
)))) {
1346 ntfs_error(vol
->sb
, "Attribute name is placed after "
1347 "the mapping pairs array.");
1350 if (NInoCompressed(ni
) || NInoSparse(ni
)) {
1351 if (NInoCompressed(ni
) && a
->data
.non_resident
.
1352 compression_unit
!= 4) {
1353 ntfs_error(vi
->i_sb
, "Found non-standard "
1354 "compression unit (%u instead "
1355 "of 4). Cannot handle this.",
1356 a
->data
.non_resident
.
1361 if (a
->data
.non_resident
.compression_unit
) {
1362 ni
->itype
.compressed
.block_size
= 1U <<
1363 (a
->data
.non_resident
.
1365 vol
->cluster_size_bits
);
1366 ni
->itype
.compressed
.block_size_bits
=
1367 ffs(ni
->itype
.compressed
.
1369 ni
->itype
.compressed
.block_clusters
= 1U <<
1370 a
->data
.non_resident
.
1373 ni
->itype
.compressed
.block_size
= 0;
1374 ni
->itype
.compressed
.block_size_bits
= 0;
1375 ni
->itype
.compressed
.block_clusters
= 0;
1377 ni
->itype
.compressed
.size
= sle64_to_cpu(
1378 a
->data
.non_resident
.compressed_size
);
1380 if (a
->data
.non_resident
.lowest_vcn
) {
1381 ntfs_error(vi
->i_sb
, "First extent of attribute has "
1382 "non-zero lowest_vcn.");
1385 vi
->i_size
= sle64_to_cpu(a
->data
.non_resident
.data_size
);
1386 ni
->initialized_size
= sle64_to_cpu(
1387 a
->data
.non_resident
.initialized_size
);
1388 ni
->allocated_size
= sle64_to_cpu(
1389 a
->data
.non_resident
.allocated_size
);
1391 vi
->i_mapping
->a_ops
= &ntfs_normal_aops
;
1392 if (NInoMstProtected(ni
))
1393 vi
->i_mapping
->a_ops
= &ntfs_mst_aops
;
1394 else if (NInoCompressed(ni
))
1395 vi
->i_mapping
->a_ops
= &ntfs_compressed_aops
;
1396 if ((NInoCompressed(ni
) || NInoSparse(ni
)) && ni
->type
!= AT_INDEX_ROOT
)
1397 vi
->i_blocks
= ni
->itype
.compressed
.size
>> 9;
1399 vi
->i_blocks
= ni
->allocated_size
>> 9;
1401 * Make sure the base inode does not go away and attach it to the
1405 ni
->ext
.base_ntfs_ino
= base_ni
;
1406 ni
->nr_extents
= -1;
1408 ntfs_attr_put_search_ctx(ctx
);
1409 unmap_mft_record(base_ni
);
1411 ntfs_debug("Done.");
1418 ntfs_attr_put_search_ctx(ctx
);
1419 unmap_mft_record(base_ni
);
1421 ntfs_error(vol
->sb
, "Failed with error code %i while reading attribute "
1422 "inode (mft_no 0x%lx, type 0x%x, name_len %i). "
1423 "Marking corrupt inode and base inode 0x%lx as bad. "
1424 "Run chkdsk.", err
, vi
->i_ino
, ni
->type
, ni
->name_len
,
1433 * ntfs_read_locked_index_inode - read an index inode from its base inode
1434 * @base_vi: base inode
1435 * @vi: index inode to read
1437 * ntfs_read_locked_index_inode() is called from ntfs_index_iget() to read the
1438 * index inode described by @vi into memory from the base mft record described
1441 * ntfs_read_locked_index_inode() maps, pins and locks the base inode for
1442 * reading and looks up the attributes relating to the index described by @vi
1443 * before setting up the necessary fields in @vi as well as initializing the
1446 * Note, index inodes are essentially attribute inodes (NInoAttr() is true)
1447 * with the attribute type set to AT_INDEX_ALLOCATION. Apart from that, they
1448 * are setup like directory inodes since directories are a special case of
1449 * indices ao they need to be treated in much the same way. Most importantly,
1450 * for small indices the index allocation attribute might not actually exist.
1451 * However, the index root attribute always exists but this does not need to
1452 * have an inode associated with it and this is why we define a new inode type
1453 * index. Also, like for directories, we need to have an attribute inode for
1454 * the bitmap attribute corresponding to the index allocation attribute and we
1455 * can store this in the appropriate field of the inode, just like we do for
1456 * normal directory inodes.
1458 * Q: What locks are held when the function is called?
1459 * A: i_state has I_NEW set, hence the inode is locked, also
1460 * i_count is set to 1, so it is not going to go away
1462 * Return 0 on success and -errno on error. In the error case, the inode will
1463 * have had make_bad_inode() executed on it.
1465 static int ntfs_read_locked_index_inode(struct inode
*base_vi
, struct inode
*vi
)
1468 ntfs_volume
*vol
= NTFS_SB(vi
->i_sb
);
1469 ntfs_inode
*ni
, *base_ni
, *bni
;
1473 ntfs_attr_search_ctx
*ctx
;
1475 u8
*ir_end
, *index_end
;
1478 ntfs_debug("Entering for i_ino 0x%lx.", vi
->i_ino
);
1479 ntfs_init_big_inode(vi
);
1481 base_ni
= NTFS_I(base_vi
);
1482 /* Just mirror the values from the base inode. */
1483 vi
->i_uid
= base_vi
->i_uid
;
1484 vi
->i_gid
= base_vi
->i_gid
;
1485 set_nlink(vi
, base_vi
->i_nlink
);
1486 vi
->i_mtime
= base_vi
->i_mtime
;
1487 vi
->i_ctime
= base_vi
->i_ctime
;
1488 vi
->i_atime
= base_vi
->i_atime
;
1489 vi
->i_generation
= ni
->seq_no
= base_ni
->seq_no
;
1490 /* Set inode type to zero but preserve permissions. */
1491 vi
->i_mode
= base_vi
->i_mode
& ~S_IFMT
;
1492 /* Map the mft record for the base inode. */
1493 m
= map_mft_record(base_ni
);
1498 ctx
= ntfs_attr_get_search_ctx(base_ni
, m
);
1503 /* Find the index root attribute. */
1504 err
= ntfs_attr_lookup(AT_INDEX_ROOT
, ni
->name
, ni
->name_len
,
1505 CASE_SENSITIVE
, 0, NULL
, 0, ctx
);
1506 if (unlikely(err
)) {
1508 ntfs_error(vi
->i_sb
, "$INDEX_ROOT attribute is "
1513 /* Set up the state. */
1514 if (unlikely(a
->non_resident
)) {
1515 ntfs_error(vol
->sb
, "$INDEX_ROOT attribute is not resident.");
1518 /* Ensure the attribute name is placed before the value. */
1519 if (unlikely(a
->name_length
&& (le16_to_cpu(a
->name_offset
) >=
1520 le16_to_cpu(a
->data
.resident
.value_offset
)))) {
1521 ntfs_error(vol
->sb
, "$INDEX_ROOT attribute name is placed "
1522 "after the attribute value.");
1526 * Compressed/encrypted/sparse index root is not allowed, except for
1527 * directories of course but those are not dealt with here.
1529 if (a
->flags
& (ATTR_COMPRESSION_MASK
| ATTR_IS_ENCRYPTED
|
1531 ntfs_error(vi
->i_sb
, "Found compressed/encrypted/sparse index "
1535 ir
= (INDEX_ROOT
*)((u8
*)a
+ le16_to_cpu(a
->data
.resident
.value_offset
));
1536 ir_end
= (u8
*)ir
+ le32_to_cpu(a
->data
.resident
.value_length
);
1537 if (ir_end
> (u8
*)ctx
->mrec
+ vol
->mft_record_size
) {
1538 ntfs_error(vi
->i_sb
, "$INDEX_ROOT attribute is corrupt.");
1541 index_end
= (u8
*)&ir
->index
+ le32_to_cpu(ir
->index
.index_length
);
1542 if (index_end
> ir_end
) {
1543 ntfs_error(vi
->i_sb
, "Index is corrupt.");
1547 ntfs_error(vi
->i_sb
, "Index type is not 0 (type is 0x%x).",
1548 le32_to_cpu(ir
->type
));
1551 ni
->itype
.index
.collation_rule
= ir
->collation_rule
;
1552 ntfs_debug("Index collation rule is 0x%x.",
1553 le32_to_cpu(ir
->collation_rule
));
1554 ni
->itype
.index
.block_size
= le32_to_cpu(ir
->index_block_size
);
1555 if (!is_power_of_2(ni
->itype
.index
.block_size
)) {
1556 ntfs_error(vi
->i_sb
, "Index block size (%u) is not a power of "
1557 "two.", ni
->itype
.index
.block_size
);
1560 if (ni
->itype
.index
.block_size
> PAGE_SIZE
) {
1561 ntfs_error(vi
->i_sb
, "Index block size (%u) > PAGE_SIZE "
1562 "(%ld) is not supported. Sorry.",
1563 ni
->itype
.index
.block_size
, PAGE_SIZE
);
1567 if (ni
->itype
.index
.block_size
< NTFS_BLOCK_SIZE
) {
1568 ntfs_error(vi
->i_sb
, "Index block size (%u) < NTFS_BLOCK_SIZE "
1569 "(%i) is not supported. Sorry.",
1570 ni
->itype
.index
.block_size
, NTFS_BLOCK_SIZE
);
1574 ni
->itype
.index
.block_size_bits
= ffs(ni
->itype
.index
.block_size
) - 1;
1575 /* Determine the size of a vcn in the index. */
1576 if (vol
->cluster_size
<= ni
->itype
.index
.block_size
) {
1577 ni
->itype
.index
.vcn_size
= vol
->cluster_size
;
1578 ni
->itype
.index
.vcn_size_bits
= vol
->cluster_size_bits
;
1580 ni
->itype
.index
.vcn_size
= vol
->sector_size
;
1581 ni
->itype
.index
.vcn_size_bits
= vol
->sector_size_bits
;
1583 /* Check for presence of index allocation attribute. */
1584 if (!(ir
->index
.flags
& LARGE_INDEX
)) {
1585 /* No index allocation. */
1586 vi
->i_size
= ni
->initialized_size
= ni
->allocated_size
= 0;
1587 /* We are done with the mft record, so we release it. */
1588 ntfs_attr_put_search_ctx(ctx
);
1589 unmap_mft_record(base_ni
);
1592 goto skip_large_index_stuff
;
1593 } /* LARGE_INDEX: Index allocation present. Setup state. */
1594 NInoSetIndexAllocPresent(ni
);
1595 /* Find index allocation attribute. */
1596 ntfs_attr_reinit_search_ctx(ctx
);
1597 err
= ntfs_attr_lookup(AT_INDEX_ALLOCATION
, ni
->name
, ni
->name_len
,
1598 CASE_SENSITIVE
, 0, NULL
, 0, ctx
);
1599 if (unlikely(err
)) {
1601 ntfs_error(vi
->i_sb
, "$INDEX_ALLOCATION attribute is "
1602 "not present but $INDEX_ROOT "
1603 "indicated it is.");
1605 ntfs_error(vi
->i_sb
, "Failed to lookup "
1606 "$INDEX_ALLOCATION attribute.");
1610 if (!a
->non_resident
) {
1611 ntfs_error(vi
->i_sb
, "$INDEX_ALLOCATION attribute is "
1616 * Ensure the attribute name is placed before the mapping pairs array.
1618 if (unlikely(a
->name_length
&& (le16_to_cpu(a
->name_offset
) >=
1620 a
->data
.non_resident
.mapping_pairs_offset
)))) {
1621 ntfs_error(vol
->sb
, "$INDEX_ALLOCATION attribute name is "
1622 "placed after the mapping pairs array.");
1625 if (a
->flags
& ATTR_IS_ENCRYPTED
) {
1626 ntfs_error(vi
->i_sb
, "$INDEX_ALLOCATION attribute is "
1630 if (a
->flags
& ATTR_IS_SPARSE
) {
1631 ntfs_error(vi
->i_sb
, "$INDEX_ALLOCATION attribute is sparse.");
1634 if (a
->flags
& ATTR_COMPRESSION_MASK
) {
1635 ntfs_error(vi
->i_sb
, "$INDEX_ALLOCATION attribute is "
1639 if (a
->data
.non_resident
.lowest_vcn
) {
1640 ntfs_error(vi
->i_sb
, "First extent of $INDEX_ALLOCATION "
1641 "attribute has non zero lowest_vcn.");
1644 vi
->i_size
= sle64_to_cpu(a
->data
.non_resident
.data_size
);
1645 ni
->initialized_size
= sle64_to_cpu(
1646 a
->data
.non_resident
.initialized_size
);
1647 ni
->allocated_size
= sle64_to_cpu(a
->data
.non_resident
.allocated_size
);
1649 * We are done with the mft record, so we release it. Otherwise
1650 * we would deadlock in ntfs_attr_iget().
1652 ntfs_attr_put_search_ctx(ctx
);
1653 unmap_mft_record(base_ni
);
1656 /* Get the index bitmap attribute inode. */
1657 bvi
= ntfs_attr_iget(base_vi
, AT_BITMAP
, ni
->name
, ni
->name_len
);
1659 ntfs_error(vi
->i_sb
, "Failed to get bitmap attribute.");
1664 if (NInoCompressed(bni
) || NInoEncrypted(bni
) ||
1666 ntfs_error(vi
->i_sb
, "$BITMAP attribute is compressed and/or "
1667 "encrypted and/or sparse.");
1668 goto iput_unm_err_out
;
1670 /* Consistency check bitmap size vs. index allocation size. */
1671 bvi_size
= i_size_read(bvi
);
1672 if ((bvi_size
<< 3) < (vi
->i_size
>> ni
->itype
.index
.block_size_bits
)) {
1673 ntfs_error(vi
->i_sb
, "Index bitmap too small (0x%llx) for "
1674 "index allocation (0x%llx).", bvi_size
<< 3,
1676 goto iput_unm_err_out
;
1679 skip_large_index_stuff
:
1680 /* Setup the operations for this index inode. */
1681 vi
->i_mapping
->a_ops
= &ntfs_mst_aops
;
1682 vi
->i_blocks
= ni
->allocated_size
>> 9;
1684 * Make sure the base inode doesn't go away and attach it to the
1688 ni
->ext
.base_ntfs_ino
= base_ni
;
1689 ni
->nr_extents
= -1;
1691 ntfs_debug("Done.");
1699 ntfs_attr_put_search_ctx(ctx
);
1701 unmap_mft_record(base_ni
);
1703 ntfs_error(vi
->i_sb
, "Failed with error code %i while reading index "
1704 "inode (mft_no 0x%lx, name_len %i.", err
, vi
->i_ino
,
1707 if (err
!= -EOPNOTSUPP
&& err
!= -ENOMEM
)
1713 * The MFT inode has special locking, so teach the lock validator
1714 * about this by splitting off the locking rules of the MFT from
1715 * the locking rules of other inodes. The MFT inode can never be
1716 * accessed from the VFS side (or even internally), only by the
1717 * map_mft functions.
1719 static struct lock_class_key mft_ni_runlist_lock_key
, mft_ni_mrec_lock_key
;
1722 * ntfs_read_inode_mount - special read_inode for mount time use only
1723 * @vi: inode to read
1725 * Read inode FILE_MFT at mount time, only called with super_block lock
1726 * held from within the read_super() code path.
1728 * This function exists because when it is called the page cache for $MFT/$DATA
1729 * is not initialized and hence we cannot get at the contents of mft records
1730 * by calling map_mft_record*().
1732 * Further it needs to cope with the circular references problem, i.e. cannot
1733 * load any attributes other than $ATTRIBUTE_LIST until $DATA is loaded, because
1734 * we do not know where the other extent mft records are yet and again, because
1735 * we cannot call map_mft_record*() yet. Obviously this applies only when an
1736 * attribute list is actually present in $MFT inode.
1738 * We solve these problems by starting with the $DATA attribute before anything
1739 * else and iterating using ntfs_attr_lookup($DATA) over all extents. As each
1740 * extent is found, we ntfs_mapping_pairs_decompress() including the implied
1741 * ntfs_runlists_merge(). Each step of the iteration necessarily provides
1742 * sufficient information for the next step to complete.
1744 * This should work but there are two possible pit falls (see inline comments
1745 * below), but only time will tell if they are real pits or just smoke...
1747 int ntfs_read_inode_mount(struct inode
*vi
)
1749 VCN next_vcn
, last_vcn
, highest_vcn
;
1751 struct super_block
*sb
= vi
->i_sb
;
1752 ntfs_volume
*vol
= NTFS_SB(sb
);
1753 struct buffer_head
*bh
;
1755 MFT_RECORD
*m
= NULL
;
1757 ntfs_attr_search_ctx
*ctx
;
1758 unsigned int i
, nr_blocks
;
1761 ntfs_debug("Entering.");
1763 /* Initialize the ntfs specific part of @vi. */
1764 ntfs_init_big_inode(vi
);
1768 /* Setup the data attribute. It is special as it is mst protected. */
1769 NInoSetNonResident(ni
);
1770 NInoSetMstProtected(ni
);
1771 NInoSetSparseDisabled(ni
);
1776 * This sets up our little cheat allowing us to reuse the async read io
1777 * completion handler for directories.
1779 ni
->itype
.index
.block_size
= vol
->mft_record_size
;
1780 ni
->itype
.index
.block_size_bits
= vol
->mft_record_size_bits
;
1782 /* Very important! Needed to be able to call map_mft_record*(). */
1785 /* Allocate enough memory to read the first mft record. */
1786 if (vol
->mft_record_size
> 64 * 1024) {
1787 ntfs_error(sb
, "Unsupported mft record size %i (max 64kiB).",
1788 vol
->mft_record_size
);
1791 i
= vol
->mft_record_size
;
1792 if (i
< sb
->s_blocksize
)
1793 i
= sb
->s_blocksize
;
1794 m
= (MFT_RECORD
*)ntfs_malloc_nofs(i
);
1796 ntfs_error(sb
, "Failed to allocate buffer for $MFT record 0.");
1800 /* Determine the first block of the $MFT/$DATA attribute. */
1801 block
= vol
->mft_lcn
<< vol
->cluster_size_bits
>>
1802 sb
->s_blocksize_bits
;
1803 nr_blocks
= vol
->mft_record_size
>> sb
->s_blocksize_bits
;
1807 /* Load $MFT/$DATA's first mft record. */
1808 for (i
= 0; i
< nr_blocks
; i
++) {
1809 bh
= sb_bread(sb
, block
++);
1811 ntfs_error(sb
, "Device read failed.");
1814 memcpy((char*)m
+ (i
<< sb
->s_blocksize_bits
), bh
->b_data
,
1819 if (le32_to_cpu(m
->bytes_allocated
) != vol
->mft_record_size
) {
1820 ntfs_error(sb
, "Incorrect mft record size %u in superblock, should be %u.",
1821 le32_to_cpu(m
->bytes_allocated
), vol
->mft_record_size
);
1825 /* Apply the mst fixups. */
1826 if (post_read_mst_fixup((NTFS_RECORD
*)m
, vol
->mft_record_size
)) {
1827 /* FIXME: Try to use the $MFTMirr now. */
1828 ntfs_error(sb
, "MST fixup failed. $MFT is corrupt.");
1832 /* Need this to sanity check attribute list references to $MFT. */
1833 vi
->i_generation
= ni
->seq_no
= le16_to_cpu(m
->sequence_number
);
1835 /* Provides readpage() for map_mft_record(). */
1836 vi
->i_mapping
->a_ops
= &ntfs_mst_aops
;
1838 ctx
= ntfs_attr_get_search_ctx(ni
, m
);
1844 /* Find the attribute list attribute if present. */
1845 err
= ntfs_attr_lookup(AT_ATTRIBUTE_LIST
, NULL
, 0, 0, 0, NULL
, 0, ctx
);
1847 if (unlikely(err
!= -ENOENT
)) {
1848 ntfs_error(sb
, "Failed to lookup attribute list "
1849 "attribute. You should run chkdsk.");
1852 } else /* if (!err) */ {
1853 ATTR_LIST_ENTRY
*al_entry
, *next_al_entry
;
1855 static const char *es
= " Not allowed. $MFT is corrupt. "
1856 "You should run chkdsk.";
1858 ntfs_debug("Attribute list attribute found in $MFT.");
1859 NInoSetAttrList(ni
);
1861 if (a
->flags
& ATTR_COMPRESSION_MASK
) {
1862 ntfs_error(sb
, "Attribute list attribute is "
1863 "compressed.%s", es
);
1866 if (a
->flags
& ATTR_IS_ENCRYPTED
||
1867 a
->flags
& ATTR_IS_SPARSE
) {
1868 if (a
->non_resident
) {
1869 ntfs_error(sb
, "Non-resident attribute list "
1870 "attribute is encrypted/"
1874 ntfs_warning(sb
, "Resident attribute list attribute "
1875 "in $MFT system file is marked "
1876 "encrypted/sparse which is not true. "
1877 "However, Windows allows this and "
1878 "chkdsk does not detect or correct it "
1879 "so we will just ignore the invalid "
1880 "flags and pretend they are not set.");
1882 /* Now allocate memory for the attribute list. */
1883 ni
->attr_list_size
= (u32
)ntfs_attr_size(a
);
1884 ni
->attr_list
= ntfs_malloc_nofs(ni
->attr_list_size
);
1885 if (!ni
->attr_list
) {
1886 ntfs_error(sb
, "Not enough memory to allocate buffer "
1887 "for attribute list.");
1890 if (a
->non_resident
) {
1891 NInoSetAttrListNonResident(ni
);
1892 if (a
->data
.non_resident
.lowest_vcn
) {
1893 ntfs_error(sb
, "Attribute list has non zero "
1894 "lowest_vcn. $MFT is corrupt. "
1895 "You should run chkdsk.");
1898 /* Setup the runlist. */
1899 ni
->attr_list_rl
.rl
= ntfs_mapping_pairs_decompress(vol
,
1901 if (IS_ERR(ni
->attr_list_rl
.rl
)) {
1902 err
= PTR_ERR(ni
->attr_list_rl
.rl
);
1903 ni
->attr_list_rl
.rl
= NULL
;
1904 ntfs_error(sb
, "Mapping pairs decompression "
1905 "failed with error code %i.",
1909 /* Now load the attribute list. */
1910 if ((err
= load_attribute_list(vol
, &ni
->attr_list_rl
,
1911 ni
->attr_list
, ni
->attr_list_size
,
1912 sle64_to_cpu(a
->data
.
1913 non_resident
.initialized_size
)))) {
1914 ntfs_error(sb
, "Failed to load attribute list "
1915 "attribute with error code %i.",
1919 } else /* if (!ctx.attr->non_resident) */ {
1920 if ((u8
*)a
+ le16_to_cpu(
1921 a
->data
.resident
.value_offset
) +
1923 a
->data
.resident
.value_length
) >
1924 (u8
*)ctx
->mrec
+ vol
->mft_record_size
) {
1925 ntfs_error(sb
, "Corrupt attribute list "
1929 /* Now copy the attribute list. */
1930 memcpy(ni
->attr_list
, (u8
*)a
+ le16_to_cpu(
1931 a
->data
.resident
.value_offset
),
1933 a
->data
.resident
.value_length
));
1935 /* The attribute list is now setup in memory. */
1937 * FIXME: I don't know if this case is actually possible.
1938 * According to logic it is not possible but I have seen too
1939 * many weird things in MS software to rely on logic... Thus we
1940 * perform a manual search and make sure the first $MFT/$DATA
1941 * extent is in the base inode. If it is not we abort with an
1942 * error and if we ever see a report of this error we will need
1943 * to do some magic in order to have the necessary mft record
1944 * loaded and in the right place in the page cache. But
1945 * hopefully logic will prevail and this never happens...
1947 al_entry
= (ATTR_LIST_ENTRY
*)ni
->attr_list
;
1948 al_end
= (u8
*)al_entry
+ ni
->attr_list_size
;
1949 for (;; al_entry
= next_al_entry
) {
1950 /* Out of bounds check. */
1951 if ((u8
*)al_entry
< ni
->attr_list
||
1952 (u8
*)al_entry
> al_end
)
1953 goto em_put_err_out
;
1954 /* Catch the end of the attribute list. */
1955 if ((u8
*)al_entry
== al_end
)
1956 goto em_put_err_out
;
1957 if (!al_entry
->length
)
1958 goto em_put_err_out
;
1959 if ((u8
*)al_entry
+ 6 > al_end
|| (u8
*)al_entry
+
1960 le16_to_cpu(al_entry
->length
) > al_end
)
1961 goto em_put_err_out
;
1962 next_al_entry
= (ATTR_LIST_ENTRY
*)((u8
*)al_entry
+
1963 le16_to_cpu(al_entry
->length
));
1964 if (le32_to_cpu(al_entry
->type
) > le32_to_cpu(AT_DATA
))
1965 goto em_put_err_out
;
1966 if (AT_DATA
!= al_entry
->type
)
1968 /* We want an unnamed attribute. */
1969 if (al_entry
->name_length
)
1970 goto em_put_err_out
;
1971 /* Want the first entry, i.e. lowest_vcn == 0. */
1972 if (al_entry
->lowest_vcn
)
1973 goto em_put_err_out
;
1974 /* First entry has to be in the base mft record. */
1975 if (MREF_LE(al_entry
->mft_reference
) != vi
->i_ino
) {
1976 /* MFT references do not match, logic fails. */
1977 ntfs_error(sb
, "BUG: The first $DATA extent "
1978 "of $MFT is not in the base "
1979 "mft record. Please report "
1980 "you saw this message to "
1981 "linux-ntfs-dev@lists."
1985 /* Sequence numbers must match. */
1986 if (MSEQNO_LE(al_entry
->mft_reference
) !=
1988 goto em_put_err_out
;
1989 /* Got it. All is ok. We can stop now. */
1995 ntfs_attr_reinit_search_ctx(ctx
);
1997 /* Now load all attribute extents. */
1999 next_vcn
= last_vcn
= highest_vcn
= 0;
2000 while (!(err
= ntfs_attr_lookup(AT_DATA
, NULL
, 0, 0, next_vcn
, NULL
, 0,
2002 runlist_element
*nrl
;
2004 /* Cache the current attribute. */
2006 /* $MFT must be non-resident. */
2007 if (!a
->non_resident
) {
2008 ntfs_error(sb
, "$MFT must be non-resident but a "
2009 "resident extent was found. $MFT is "
2010 "corrupt. Run chkdsk.");
2013 /* $MFT must be uncompressed and unencrypted. */
2014 if (a
->flags
& ATTR_COMPRESSION_MASK
||
2015 a
->flags
& ATTR_IS_ENCRYPTED
||
2016 a
->flags
& ATTR_IS_SPARSE
) {
2017 ntfs_error(sb
, "$MFT must be uncompressed, "
2018 "non-sparse, and unencrypted but a "
2019 "compressed/sparse/encrypted extent "
2020 "was found. $MFT is corrupt. Run "
2025 * Decompress the mapping pairs array of this extent and merge
2026 * the result into the existing runlist. No need for locking
2027 * as we have exclusive access to the inode at this time and we
2028 * are a mount in progress task, too.
2030 nrl
= ntfs_mapping_pairs_decompress(vol
, a
, ni
->runlist
.rl
);
2032 ntfs_error(sb
, "ntfs_mapping_pairs_decompress() "
2033 "failed with error code %ld. $MFT is "
2034 "corrupt.", PTR_ERR(nrl
));
2037 ni
->runlist
.rl
= nrl
;
2039 /* Are we in the first extent? */
2041 if (a
->data
.non_resident
.lowest_vcn
) {
2042 ntfs_error(sb
, "First extent of $DATA "
2043 "attribute has non zero "
2044 "lowest_vcn. $MFT is corrupt. "
2045 "You should run chkdsk.");
2048 /* Get the last vcn in the $DATA attribute. */
2049 last_vcn
= sle64_to_cpu(
2050 a
->data
.non_resident
.allocated_size
)
2051 >> vol
->cluster_size_bits
;
2052 /* Fill in the inode size. */
2053 vi
->i_size
= sle64_to_cpu(
2054 a
->data
.non_resident
.data_size
);
2055 ni
->initialized_size
= sle64_to_cpu(
2056 a
->data
.non_resident
.initialized_size
);
2057 ni
->allocated_size
= sle64_to_cpu(
2058 a
->data
.non_resident
.allocated_size
);
2060 * Verify the number of mft records does not exceed
2063 if ((vi
->i_size
>> vol
->mft_record_size_bits
) >=
2065 ntfs_error(sb
, "$MFT is too big! Aborting.");
2069 * We have got the first extent of the runlist for
2070 * $MFT which means it is now relatively safe to call
2071 * the normal ntfs_read_inode() function.
2072 * Complete reading the inode, this will actually
2073 * re-read the mft record for $MFT, this time entering
2074 * it into the page cache with which we complete the
2075 * kick start of the volume. It should be safe to do
2076 * this now as the first extent of $MFT/$DATA is
2077 * already known and we would hope that we don't need
2078 * further extents in order to find the other
2079 * attributes belonging to $MFT. Only time will tell if
2080 * this is really the case. If not we will have to play
2081 * magic at this point, possibly duplicating a lot of
2082 * ntfs_read_inode() at this point. We will need to
2083 * ensure we do enough of its work to be able to call
2084 * ntfs_read_inode() on extents of $MFT/$DATA. But lets
2085 * hope this never happens...
2087 ntfs_read_locked_inode(vi
);
2088 if (is_bad_inode(vi
)) {
2089 ntfs_error(sb
, "ntfs_read_inode() of $MFT "
2090 "failed. BUG or corrupt $MFT. "
2091 "Run chkdsk and if no errors "
2092 "are found, please report you "
2093 "saw this message to "
2094 "linux-ntfs-dev@lists."
2096 ntfs_attr_put_search_ctx(ctx
);
2097 /* Revert to the safe super operations. */
2102 * Re-initialize some specifics about $MFT's inode as
2103 * ntfs_read_inode() will have set up the default ones.
2105 /* Set uid and gid to root. */
2106 vi
->i_uid
= GLOBAL_ROOT_UID
;
2107 vi
->i_gid
= GLOBAL_ROOT_GID
;
2108 /* Regular file. No access for anyone. */
2109 vi
->i_mode
= S_IFREG
;
2110 /* No VFS initiated operations allowed for $MFT. */
2111 vi
->i_op
= &ntfs_empty_inode_ops
;
2112 vi
->i_fop
= &ntfs_empty_file_ops
;
2115 /* Get the lowest vcn for the next extent. */
2116 highest_vcn
= sle64_to_cpu(a
->data
.non_resident
.highest_vcn
);
2117 next_vcn
= highest_vcn
+ 1;
2119 /* Only one extent or error, which we catch below. */
2123 /* Avoid endless loops due to corruption. */
2124 if (next_vcn
< sle64_to_cpu(
2125 a
->data
.non_resident
.lowest_vcn
)) {
2126 ntfs_error(sb
, "$MFT has corrupt attribute list "
2127 "attribute. Run chkdsk.");
2131 if (err
!= -ENOENT
) {
2132 ntfs_error(sb
, "Failed to lookup $MFT/$DATA attribute extent. "
2133 "$MFT is corrupt. Run chkdsk.");
2137 ntfs_error(sb
, "$MFT/$DATA attribute not found. $MFT is "
2138 "corrupt. Run chkdsk.");
2141 if (highest_vcn
&& highest_vcn
!= last_vcn
- 1) {
2142 ntfs_error(sb
, "Failed to load the complete runlist for "
2143 "$MFT/$DATA. Driver bug or corrupt $MFT. "
2145 ntfs_debug("highest_vcn = 0x%llx, last_vcn - 1 = 0x%llx",
2146 (unsigned long long)highest_vcn
,
2147 (unsigned long long)last_vcn
- 1);
2150 ntfs_attr_put_search_ctx(ctx
);
2151 ntfs_debug("Done.");
2155 * Split the locking rules of the MFT inode from the
2156 * locking rules of other inodes:
2158 lockdep_set_class(&ni
->runlist
.lock
, &mft_ni_runlist_lock_key
);
2159 lockdep_set_class(&ni
->mrec_lock
, &mft_ni_mrec_lock_key
);
2164 ntfs_error(sb
, "Couldn't find first extent of $DATA attribute in "
2165 "attribute list. $MFT is corrupt. Run chkdsk.");
2167 ntfs_attr_put_search_ctx(ctx
);
2169 ntfs_error(sb
, "Failed. Marking inode as bad.");
2175 static void __ntfs_clear_inode(ntfs_inode
*ni
)
2177 /* Free all alocated memory. */
2178 down_write(&ni
->runlist
.lock
);
2179 if (ni
->runlist
.rl
) {
2180 ntfs_free(ni
->runlist
.rl
);
2181 ni
->runlist
.rl
= NULL
;
2183 up_write(&ni
->runlist
.lock
);
2185 if (ni
->attr_list
) {
2186 ntfs_free(ni
->attr_list
);
2187 ni
->attr_list
= NULL
;
2190 down_write(&ni
->attr_list_rl
.lock
);
2191 if (ni
->attr_list_rl
.rl
) {
2192 ntfs_free(ni
->attr_list_rl
.rl
);
2193 ni
->attr_list_rl
.rl
= NULL
;
2195 up_write(&ni
->attr_list_rl
.lock
);
2197 if (ni
->name_len
&& ni
->name
!= I30
) {
2204 void ntfs_clear_extent_inode(ntfs_inode
*ni
)
2206 ntfs_debug("Entering for inode 0x%lx.", ni
->mft_no
);
2208 BUG_ON(NInoAttr(ni
));
2209 BUG_ON(ni
->nr_extents
!= -1);
2212 if (NInoDirty(ni
)) {
2213 if (!is_bad_inode(VFS_I(ni
->ext
.base_ntfs_ino
)))
2214 ntfs_error(ni
->vol
->sb
, "Clearing dirty extent inode! "
2215 "Losing data! This is a BUG!!!");
2216 // FIXME: Do something!!!
2218 #endif /* NTFS_RW */
2220 __ntfs_clear_inode(ni
);
2223 ntfs_destroy_extent_inode(ni
);
2227 * ntfs_evict_big_inode - clean up the ntfs specific part of an inode
2228 * @vi: vfs inode pending annihilation
2230 * When the VFS is going to remove an inode from memory, ntfs_clear_big_inode()
2231 * is called, which deallocates all memory belonging to the NTFS specific part
2232 * of the inode and returns.
2234 * If the MFT record is dirty, we commit it before doing anything else.
2236 void ntfs_evict_big_inode(struct inode
*vi
)
2238 ntfs_inode
*ni
= NTFS_I(vi
);
2240 truncate_inode_pages_final(&vi
->i_data
);
2244 if (NInoDirty(ni
)) {
2245 bool was_bad
= (is_bad_inode(vi
));
2247 /* Committing the inode also commits all extent inodes. */
2248 ntfs_commit_inode(vi
);
2250 if (!was_bad
&& (is_bad_inode(vi
) || NInoDirty(ni
))) {
2251 ntfs_error(vi
->i_sb
, "Failed to commit dirty inode "
2252 "0x%lx. Losing data!", vi
->i_ino
);
2253 // FIXME: Do something!!!
2256 #endif /* NTFS_RW */
2258 /* No need to lock at this stage as no one else has a reference. */
2259 if (ni
->nr_extents
> 0) {
2262 for (i
= 0; i
< ni
->nr_extents
; i
++)
2263 ntfs_clear_extent_inode(ni
->ext
.extent_ntfs_inos
[i
]);
2264 kfree(ni
->ext
.extent_ntfs_inos
);
2267 __ntfs_clear_inode(ni
);
2270 /* Release the base inode if we are holding it. */
2271 if (ni
->nr_extents
== -1) {
2272 iput(VFS_I(ni
->ext
.base_ntfs_ino
));
2274 ni
->ext
.base_ntfs_ino
= NULL
;
2278 if (!atomic_dec_and_test(&ni
->count
))
2284 * ntfs_show_options - show mount options in /proc/mounts
2285 * @sf: seq_file in which to write our mount options
2286 * @root: root of the mounted tree whose mount options to display
2288 * Called by the VFS once for each mounted ntfs volume when someone reads
2289 * /proc/mounts in order to display the NTFS specific mount options of each
2290 * mount. The mount options of fs specified by @root are written to the seq file
2291 * @sf and success is returned.
2293 int ntfs_show_options(struct seq_file
*sf
, struct dentry
*root
)
2295 ntfs_volume
*vol
= NTFS_SB(root
->d_sb
);
2298 seq_printf(sf
, ",uid=%i", from_kuid_munged(&init_user_ns
, vol
->uid
));
2299 seq_printf(sf
, ",gid=%i", from_kgid_munged(&init_user_ns
, vol
->gid
));
2300 if (vol
->fmask
== vol
->dmask
)
2301 seq_printf(sf
, ",umask=0%o", vol
->fmask
);
2303 seq_printf(sf
, ",fmask=0%o", vol
->fmask
);
2304 seq_printf(sf
, ",dmask=0%o", vol
->dmask
);
2306 seq_printf(sf
, ",nls=%s", vol
->nls_map
->charset
);
2307 if (NVolCaseSensitive(vol
))
2308 seq_printf(sf
, ",case_sensitive");
2309 if (NVolShowSystemFiles(vol
))
2310 seq_printf(sf
, ",show_sys_files");
2311 if (!NVolSparseEnabled(vol
))
2312 seq_printf(sf
, ",disable_sparse");
2313 for (i
= 0; on_errors_arr
[i
].val
; i
++) {
2314 if (on_errors_arr
[i
].val
& vol
->on_errors
)
2315 seq_printf(sf
, ",errors=%s", on_errors_arr
[i
].str
);
2317 seq_printf(sf
, ",mft_zone_multiplier=%i", vol
->mft_zone_multiplier
);
2323 static const char *es
= " Leaving inconsistent metadata. Unmount and run "
2327 * ntfs_truncate - called when the i_size of an ntfs inode is changed
2328 * @vi: inode for which the i_size was changed
2330 * We only support i_size changes for normal files at present, i.e. not
2331 * compressed and not encrypted. This is enforced in ntfs_setattr(), see
2334 * The kernel guarantees that @vi is a regular file (S_ISREG() is true) and
2335 * that the change is allowed.
2337 * This implies for us that @vi is a file inode rather than a directory, index,
2338 * or attribute inode as well as that @vi is a base inode.
2340 * Returns 0 on success or -errno on error.
2342 * Called with ->i_mutex held.
2344 int ntfs_truncate(struct inode
*vi
)
2346 s64 new_size
, old_size
, nr_freed
, new_alloc_size
, old_alloc_size
;
2348 unsigned long flags
;
2349 ntfs_inode
*base_ni
, *ni
= NTFS_I(vi
);
2350 ntfs_volume
*vol
= ni
->vol
;
2351 ntfs_attr_search_ctx
*ctx
;
2354 const char *te
= " Leaving file length out of sync with i_size.";
2355 int err
, mp_size
, size_change
, alloc_change
;
2357 ntfs_debug("Entering for inode 0x%lx.", vi
->i_ino
);
2358 BUG_ON(NInoAttr(ni
));
2359 BUG_ON(S_ISDIR(vi
->i_mode
));
2360 BUG_ON(NInoMstProtected(ni
));
2361 BUG_ON(ni
->nr_extents
< 0);
2364 * Lock the runlist for writing and map the mft record to ensure it is
2365 * safe to mess with the attribute runlist and sizes.
2367 down_write(&ni
->runlist
.lock
);
2371 base_ni
= ni
->ext
.base_ntfs_ino
;
2372 m
= map_mft_record(base_ni
);
2375 ntfs_error(vi
->i_sb
, "Failed to map mft record for inode 0x%lx "
2376 "(error code %d).%s", vi
->i_ino
, err
, te
);
2381 ctx
= ntfs_attr_get_search_ctx(base_ni
, m
);
2382 if (unlikely(!ctx
)) {
2383 ntfs_error(vi
->i_sb
, "Failed to allocate a search context for "
2384 "inode 0x%lx (not enough memory).%s",
2389 err
= ntfs_attr_lookup(ni
->type
, ni
->name
, ni
->name_len
,
2390 CASE_SENSITIVE
, 0, NULL
, 0, ctx
);
2391 if (unlikely(err
)) {
2392 if (err
== -ENOENT
) {
2393 ntfs_error(vi
->i_sb
, "Open attribute is missing from "
2394 "mft record. Inode 0x%lx is corrupt. "
2395 "Run chkdsk.%s", vi
->i_ino
, te
);
2398 ntfs_error(vi
->i_sb
, "Failed to lookup attribute in "
2399 "inode 0x%lx (error code %d).%s",
2400 vi
->i_ino
, err
, te
);
2406 * The i_size of the vfs inode is the new size for the attribute value.
2408 new_size
= i_size_read(vi
);
2409 /* The current size of the attribute value is the old size. */
2410 old_size
= ntfs_attr_size(a
);
2411 /* Calculate the new allocated size. */
2412 if (NInoNonResident(ni
))
2413 new_alloc_size
= (new_size
+ vol
->cluster_size
- 1) &
2414 ~(s64
)vol
->cluster_size_mask
;
2416 new_alloc_size
= (new_size
+ 7) & ~7;
2417 /* The current allocated size is the old allocated size. */
2418 read_lock_irqsave(&ni
->size_lock
, flags
);
2419 old_alloc_size
= ni
->allocated_size
;
2420 read_unlock_irqrestore(&ni
->size_lock
, flags
);
2422 * The change in the file size. This will be 0 if no change, >0 if the
2423 * size is growing, and <0 if the size is shrinking.
2426 if (new_size
- old_size
>= 0) {
2428 if (new_size
== old_size
)
2431 /* As above for the allocated size. */
2433 if (new_alloc_size
- old_alloc_size
>= 0) {
2435 if (new_alloc_size
== old_alloc_size
)
2439 * If neither the size nor the allocation are being changed there is
2442 if (!size_change
&& !alloc_change
)
2444 /* If the size is changing, check if new size is allowed in $AttrDef. */
2446 err
= ntfs_attr_size_bounds_check(vol
, ni
->type
, new_size
);
2447 if (unlikely(err
)) {
2448 if (err
== -ERANGE
) {
2449 ntfs_error(vol
->sb
, "Truncate would cause the "
2450 "inode 0x%lx to %simum size "
2451 "for its attribute type "
2452 "(0x%x). Aborting truncate.",
2454 new_size
> old_size
? "exceed "
2455 "the max" : "go under the min",
2456 le32_to_cpu(ni
->type
));
2459 ntfs_error(vol
->sb
, "Inode 0x%lx has unknown "
2460 "attribute type 0x%x. "
2461 "Aborting truncate.",
2463 le32_to_cpu(ni
->type
));
2466 /* Reset the vfs inode size to the old size. */
2467 i_size_write(vi
, old_size
);
2471 if (NInoCompressed(ni
) || NInoEncrypted(ni
)) {
2472 ntfs_warning(vi
->i_sb
, "Changes in inode size are not "
2473 "supported yet for %s files, ignoring.",
2474 NInoCompressed(ni
) ? "compressed" :
2479 if (a
->non_resident
)
2480 goto do_non_resident_truncate
;
2481 BUG_ON(NInoNonResident(ni
));
2482 /* Resize the attribute record to best fit the new attribute size. */
2483 if (new_size
< vol
->mft_record_size
&&
2484 !ntfs_resident_attr_value_resize(m
, a
, new_size
)) {
2485 /* The resize succeeded! */
2486 flush_dcache_mft_record_page(ctx
->ntfs_ino
);
2487 mark_mft_record_dirty(ctx
->ntfs_ino
);
2488 write_lock_irqsave(&ni
->size_lock
, flags
);
2489 /* Update the sizes in the ntfs inode and all is done. */
2490 ni
->allocated_size
= le32_to_cpu(a
->length
) -
2491 le16_to_cpu(a
->data
.resident
.value_offset
);
2493 * Note ntfs_resident_attr_value_resize() has already done any
2494 * necessary data clearing in the attribute record. When the
2495 * file is being shrunk vmtruncate() will already have cleared
2496 * the top part of the last partial page, i.e. since this is
2497 * the resident case this is the page with index 0. However,
2498 * when the file is being expanded, the page cache page data
2499 * between the old data_size, i.e. old_size, and the new_size
2500 * has not been zeroed. Fortunately, we do not need to zero it
2501 * either since on one hand it will either already be zero due
2502 * to both readpage and writepage clearing partial page data
2503 * beyond i_size in which case there is nothing to do or in the
2504 * case of the file being mmap()ped at the same time, POSIX
2505 * specifies that the behaviour is unspecified thus we do not
2506 * have to do anything. This means that in our implementation
2507 * in the rare case that the file is mmap()ped and a write
2508 * occurred into the mmap()ped region just beyond the file size
2509 * and writepage has not yet been called to write out the page
2510 * (which would clear the area beyond the file size) and we now
2511 * extend the file size to incorporate this dirty region
2512 * outside the file size, a write of the page would result in
2513 * this data being written to disk instead of being cleared.
2514 * Given both POSIX and the Linux mmap(2) man page specify that
2515 * this corner case is undefined, we choose to leave it like
2516 * that as this is much simpler for us as we cannot lock the
2517 * relevant page now since we are holding too many ntfs locks
2518 * which would result in a lock reversal deadlock.
2520 ni
->initialized_size
= new_size
;
2521 write_unlock_irqrestore(&ni
->size_lock
, flags
);
2524 /* If the above resize failed, this must be an attribute extension. */
2525 BUG_ON(size_change
< 0);
2527 * We have to drop all the locks so we can call
2528 * ntfs_attr_make_non_resident(). This could be optimised by try-
2529 * locking the first page cache page and only if that fails dropping
2530 * the locks, locking the page, and redoing all the locking and
2531 * lookups. While this would be a huge optimisation, it is not worth
2532 * it as this is definitely a slow code path as it only ever can happen
2533 * once for any given file.
2535 ntfs_attr_put_search_ctx(ctx
);
2536 unmap_mft_record(base_ni
);
2537 up_write(&ni
->runlist
.lock
);
2539 * Not enough space in the mft record, try to make the attribute
2540 * non-resident and if successful restart the truncation process.
2542 err
= ntfs_attr_make_non_resident(ni
, old_size
);
2544 goto retry_truncate
;
2546 * Could not make non-resident. If this is due to this not being
2547 * permitted for this attribute type or there not being enough space,
2548 * try to make other attributes non-resident. Otherwise fail.
2550 if (unlikely(err
!= -EPERM
&& err
!= -ENOSPC
)) {
2551 ntfs_error(vol
->sb
, "Cannot truncate inode 0x%lx, attribute "
2552 "type 0x%x, because the conversion from "
2553 "resident to non-resident attribute failed "
2554 "with error code %i.", vi
->i_ino
,
2555 (unsigned)le32_to_cpu(ni
->type
), err
);
2560 /* TODO: Not implemented from here, abort. */
2562 ntfs_error(vol
->sb
, "Not enough space in the mft record/on "
2563 "disk for the non-resident attribute value. "
2564 "This case is not implemented yet.");
2565 else /* if (err == -EPERM) */
2566 ntfs_error(vol
->sb
, "This attribute type may not be "
2567 "non-resident. This case is not implemented "
2572 // TODO: Attempt to make other attributes non-resident.
2574 goto do_resident_extend
;
2576 * Both the attribute list attribute and the standard information
2577 * attribute must remain in the base inode. Thus, if this is one of
2578 * these attributes, we have to try to move other attributes out into
2579 * extent mft records instead.
2581 if (ni
->type
== AT_ATTRIBUTE_LIST
||
2582 ni
->type
== AT_STANDARD_INFORMATION
) {
2583 // TODO: Attempt to move other attributes into extent mft
2587 goto do_resident_extend
;
2590 // TODO: Attempt to move this attribute to an extent mft record, but
2591 // only if it is not already the only attribute in an mft record in
2592 // which case there would be nothing to gain.
2595 goto do_resident_extend
;
2596 /* There is nothing we can do to make enough space. )-: */
2599 do_non_resident_truncate
:
2600 BUG_ON(!NInoNonResident(ni
));
2601 if (alloc_change
< 0) {
2602 highest_vcn
= sle64_to_cpu(a
->data
.non_resident
.highest_vcn
);
2603 if (highest_vcn
> 0 &&
2604 old_alloc_size
>> vol
->cluster_size_bits
>
2607 * This attribute has multiple extents. Not yet
2610 ntfs_error(vol
->sb
, "Cannot truncate inode 0x%lx, "
2611 "attribute type 0x%x, because the "
2612 "attribute is highly fragmented (it "
2613 "consists of multiple extents) and "
2614 "this case is not implemented yet.",
2616 (unsigned)le32_to_cpu(ni
->type
));
2622 * If the size is shrinking, need to reduce the initialized_size and
2623 * the data_size before reducing the allocation.
2625 if (size_change
< 0) {
2627 * Make the valid size smaller (i_size is already up-to-date).
2629 write_lock_irqsave(&ni
->size_lock
, flags
);
2630 if (new_size
< ni
->initialized_size
) {
2631 ni
->initialized_size
= new_size
;
2632 a
->data
.non_resident
.initialized_size
=
2633 cpu_to_sle64(new_size
);
2635 a
->data
.non_resident
.data_size
= cpu_to_sle64(new_size
);
2636 write_unlock_irqrestore(&ni
->size_lock
, flags
);
2637 flush_dcache_mft_record_page(ctx
->ntfs_ino
);
2638 mark_mft_record_dirty(ctx
->ntfs_ino
);
2639 /* If the allocated size is not changing, we are done. */
2643 * If the size is shrinking it makes no sense for the
2644 * allocation to be growing.
2646 BUG_ON(alloc_change
> 0);
2647 } else /* if (size_change >= 0) */ {
2649 * The file size is growing or staying the same but the
2650 * allocation can be shrinking, growing or staying the same.
2652 if (alloc_change
> 0) {
2654 * We need to extend the allocation and possibly update
2655 * the data size. If we are updating the data size,
2656 * since we are not touching the initialized_size we do
2657 * not need to worry about the actual data on disk.
2658 * And as far as the page cache is concerned, there
2659 * will be no pages beyond the old data size and any
2660 * partial region in the last page between the old and
2661 * new data size (or the end of the page if the new
2662 * data size is outside the page) does not need to be
2663 * modified as explained above for the resident
2664 * attribute truncate case. To do this, we simply drop
2665 * the locks we hold and leave all the work to our
2666 * friendly helper ntfs_attr_extend_allocation().
2668 ntfs_attr_put_search_ctx(ctx
);
2669 unmap_mft_record(base_ni
);
2670 up_write(&ni
->runlist
.lock
);
2671 err
= ntfs_attr_extend_allocation(ni
, new_size
,
2672 size_change
> 0 ? new_size
: -1, -1);
2674 * ntfs_attr_extend_allocation() will have done error
2682 /* alloc_change < 0 */
2683 /* Free the clusters. */
2684 nr_freed
= ntfs_cluster_free(ni
, new_alloc_size
>>
2685 vol
->cluster_size_bits
, -1, ctx
);
2688 if (unlikely(nr_freed
< 0)) {
2689 ntfs_error(vol
->sb
, "Failed to release cluster(s) (error code "
2690 "%lli). Unmount and run chkdsk to recover "
2691 "the lost cluster(s).", (long long)nr_freed
);
2695 /* Truncate the runlist. */
2696 err
= ntfs_rl_truncate_nolock(vol
, &ni
->runlist
,
2697 new_alloc_size
>> vol
->cluster_size_bits
);
2699 * If the runlist truncation failed and/or the search context is no
2700 * longer valid, we cannot resize the attribute record or build the
2701 * mapping pairs array thus we mark the inode bad so that no access to
2702 * the freed clusters can happen.
2704 if (unlikely(err
|| IS_ERR(m
))) {
2705 ntfs_error(vol
->sb
, "Failed to %s (error code %li).%s",
2707 "restore attribute search context" :
2708 "truncate attribute runlist",
2709 IS_ERR(m
) ? PTR_ERR(m
) : err
, es
);
2713 /* Get the size for the shrunk mapping pairs array for the runlist. */
2714 mp_size
= ntfs_get_size_for_mapping_pairs(vol
, ni
->runlist
.rl
, 0, -1);
2715 if (unlikely(mp_size
<= 0)) {
2716 ntfs_error(vol
->sb
, "Cannot shrink allocation of inode 0x%lx, "
2717 "attribute type 0x%x, because determining the "
2718 "size for the mapping pairs failed with error "
2719 "code %i.%s", vi
->i_ino
,
2720 (unsigned)le32_to_cpu(ni
->type
), mp_size
, es
);
2725 * Shrink the attribute record for the new mapping pairs array. Note,
2726 * this cannot fail since we are making the attribute smaller thus by
2727 * definition there is enough space to do so.
2729 err
= ntfs_attr_record_resize(m
, a
, mp_size
+
2730 le16_to_cpu(a
->data
.non_resident
.mapping_pairs_offset
));
2733 * Generate the mapping pairs array directly into the attribute record.
2735 err
= ntfs_mapping_pairs_build(vol
, (u8
*)a
+
2736 le16_to_cpu(a
->data
.non_resident
.mapping_pairs_offset
),
2737 mp_size
, ni
->runlist
.rl
, 0, -1, NULL
);
2738 if (unlikely(err
)) {
2739 ntfs_error(vol
->sb
, "Cannot shrink allocation of inode 0x%lx, "
2740 "attribute type 0x%x, because building the "
2741 "mapping pairs failed with error code %i.%s",
2742 vi
->i_ino
, (unsigned)le32_to_cpu(ni
->type
),
2747 /* Update the allocated/compressed size as well as the highest vcn. */
2748 a
->data
.non_resident
.highest_vcn
= cpu_to_sle64((new_alloc_size
>>
2749 vol
->cluster_size_bits
) - 1);
2750 write_lock_irqsave(&ni
->size_lock
, flags
);
2751 ni
->allocated_size
= new_alloc_size
;
2752 a
->data
.non_resident
.allocated_size
= cpu_to_sle64(new_alloc_size
);
2753 if (NInoSparse(ni
) || NInoCompressed(ni
)) {
2755 ni
->itype
.compressed
.size
-= nr_freed
<<
2756 vol
->cluster_size_bits
;
2757 BUG_ON(ni
->itype
.compressed
.size
< 0);
2758 a
->data
.non_resident
.compressed_size
= cpu_to_sle64(
2759 ni
->itype
.compressed
.size
);
2760 vi
->i_blocks
= ni
->itype
.compressed
.size
>> 9;
2763 vi
->i_blocks
= new_alloc_size
>> 9;
2764 write_unlock_irqrestore(&ni
->size_lock
, flags
);
2766 * We have shrunk the allocation. If this is a shrinking truncate we
2767 * have already dealt with the initialized_size and the data_size above
2768 * and we are done. If the truncate is only changing the allocation
2769 * and not the data_size, we are also done. If this is an extending
2770 * truncate, need to extend the data_size now which is ensured by the
2771 * fact that @size_change is positive.
2775 * If the size is growing, need to update it now. If it is shrinking,
2776 * we have already updated it above (before the allocation change).
2778 if (size_change
> 0)
2779 a
->data
.non_resident
.data_size
= cpu_to_sle64(new_size
);
2780 /* Ensure the modified mft record is written out. */
2781 flush_dcache_mft_record_page(ctx
->ntfs_ino
);
2782 mark_mft_record_dirty(ctx
->ntfs_ino
);
2784 ntfs_attr_put_search_ctx(ctx
);
2785 unmap_mft_record(base_ni
);
2786 up_write(&ni
->runlist
.lock
);
2788 /* Update the mtime and ctime on the base inode. */
2789 /* normally ->truncate shouldn't update ctime or mtime,
2790 * but ntfs did before so it got a copy & paste version
2791 * of file_update_time. one day someone should fix this
2794 if (!IS_NOCMTIME(VFS_I(base_ni
)) && !IS_RDONLY(VFS_I(base_ni
))) {
2795 struct timespec64 now
= current_time(VFS_I(base_ni
));
2798 if (!timespec64_equal(&VFS_I(base_ni
)->i_mtime
, &now
) ||
2799 !timespec64_equal(&VFS_I(base_ni
)->i_ctime
, &now
))
2801 VFS_I(base_ni
)->i_mtime
= now
;
2802 VFS_I(base_ni
)->i_ctime
= now
;
2805 mark_inode_dirty_sync(VFS_I(base_ni
));
2809 NInoClearTruncateFailed(ni
);
2810 ntfs_debug("Done.");
2816 if (err
!= -ENOMEM
&& err
!= -EOPNOTSUPP
)
2818 if (err
!= -EOPNOTSUPP
)
2819 NInoSetTruncateFailed(ni
);
2820 else if (old_size
>= 0)
2821 i_size_write(vi
, old_size
);
2824 ntfs_attr_put_search_ctx(ctx
);
2826 unmap_mft_record(base_ni
);
2827 up_write(&ni
->runlist
.lock
);
2829 ntfs_debug("Failed. Returning error code %i.", err
);
2832 if (err
!= -ENOMEM
&& err
!= -EOPNOTSUPP
)
2834 if (err
!= -EOPNOTSUPP
)
2835 NInoSetTruncateFailed(ni
);
2837 i_size_write(vi
, old_size
);
2842 * ntfs_truncate_vfs - wrapper for ntfs_truncate() that has no return value
2843 * @vi: inode for which the i_size was changed
2845 * Wrapper for ntfs_truncate() that has no return value.
2847 * See ntfs_truncate() description above for details.
2850 void ntfs_truncate_vfs(struct inode
*vi
) {
2856 * ntfs_setattr - called from notify_change() when an attribute is being changed
2857 * @mnt_userns: user namespace of the mount the inode was found from
2858 * @dentry: dentry whose attributes to change
2859 * @attr: structure describing the attributes and the changes
2861 * We have to trap VFS attempts to truncate the file described by @dentry as
2862 * soon as possible, because we do not implement changes in i_size yet. So we
2863 * abort all i_size changes here.
2865 * We also abort all changes of user, group, and mode as we do not implement
2866 * the NTFS ACLs yet.
2868 * Called with ->i_mutex held.
2870 int ntfs_setattr(struct user_namespace
*mnt_userns
, struct dentry
*dentry
,
2873 struct inode
*vi
= d_inode(dentry
);
2875 unsigned int ia_valid
= attr
->ia_valid
;
2877 err
= setattr_prepare(&init_user_ns
, dentry
, attr
);
2880 /* We do not support NTFS ACLs yet. */
2881 if (ia_valid
& (ATTR_UID
| ATTR_GID
| ATTR_MODE
)) {
2882 ntfs_warning(vi
->i_sb
, "Changes in user/group/mode are not "
2883 "supported yet, ignoring.");
2887 if (ia_valid
& ATTR_SIZE
) {
2888 if (attr
->ia_size
!= i_size_read(vi
)) {
2889 ntfs_inode
*ni
= NTFS_I(vi
);
2891 * FIXME: For now we do not support resizing of
2892 * compressed or encrypted files yet.
2894 if (NInoCompressed(ni
) || NInoEncrypted(ni
)) {
2895 ntfs_warning(vi
->i_sb
, "Changes in inode size "
2896 "are not supported yet for "
2897 "%s files, ignoring.",
2898 NInoCompressed(ni
) ?
2899 "compressed" : "encrypted");
2902 truncate_setsize(vi
, attr
->ia_size
);
2903 ntfs_truncate_vfs(vi
);
2905 if (err
|| ia_valid
== ATTR_SIZE
)
2909 * We skipped the truncate but must still update
2912 ia_valid
|= ATTR_MTIME
| ATTR_CTIME
;
2915 if (ia_valid
& ATTR_ATIME
)
2916 vi
->i_atime
= attr
->ia_atime
;
2917 if (ia_valid
& ATTR_MTIME
)
2918 vi
->i_mtime
= attr
->ia_mtime
;
2919 if (ia_valid
& ATTR_CTIME
)
2920 vi
->i_ctime
= attr
->ia_ctime
;
2921 mark_inode_dirty(vi
);
2927 * ntfs_write_inode - write out a dirty inode
2928 * @vi: inode to write out
2929 * @sync: if true, write out synchronously
2931 * Write out a dirty inode to disk including any extent inodes if present.
2933 * If @sync is true, commit the inode to disk and wait for io completion. This
2934 * is done using write_mft_record().
2936 * If @sync is false, just schedule the write to happen but do not wait for i/o
2937 * completion. In 2.6 kernels, scheduling usually happens just by virtue of
2938 * marking the page (and in this case mft record) dirty but we do not implement
2939 * this yet as write_mft_record() largely ignores the @sync parameter and
2940 * always performs synchronous writes.
2942 * Return 0 on success and -errno on error.
2944 int __ntfs_write_inode(struct inode
*vi
, int sync
)
2947 ntfs_inode
*ni
= NTFS_I(vi
);
2948 ntfs_attr_search_ctx
*ctx
;
2950 STANDARD_INFORMATION
*si
;
2952 bool modified
= false;
2954 ntfs_debug("Entering for %sinode 0x%lx.", NInoAttr(ni
) ? "attr " : "",
2957 * Dirty attribute inodes are written via their real inodes so just
2958 * clean them here. Access time updates are taken care off when the
2959 * real inode is written.
2963 ntfs_debug("Done.");
2966 /* Map, pin, and lock the mft record belonging to the inode. */
2967 m
= map_mft_record(ni
);
2972 /* Update the access times in the standard information attribute. */
2973 ctx
= ntfs_attr_get_search_ctx(ni
, m
);
2974 if (unlikely(!ctx
)) {
2978 err
= ntfs_attr_lookup(AT_STANDARD_INFORMATION
, NULL
, 0,
2979 CASE_SENSITIVE
, 0, NULL
, 0, ctx
);
2980 if (unlikely(err
)) {
2981 ntfs_attr_put_search_ctx(ctx
);
2984 si
= (STANDARD_INFORMATION
*)((u8
*)ctx
->attr
+
2985 le16_to_cpu(ctx
->attr
->data
.resident
.value_offset
));
2986 /* Update the access times if they have changed. */
2987 nt
= utc2ntfs(vi
->i_mtime
);
2988 if (si
->last_data_change_time
!= nt
) {
2989 ntfs_debug("Updating mtime for inode 0x%lx: old = 0x%llx, "
2990 "new = 0x%llx", vi
->i_ino
, (long long)
2991 sle64_to_cpu(si
->last_data_change_time
),
2992 (long long)sle64_to_cpu(nt
));
2993 si
->last_data_change_time
= nt
;
2996 nt
= utc2ntfs(vi
->i_ctime
);
2997 if (si
->last_mft_change_time
!= nt
) {
2998 ntfs_debug("Updating ctime for inode 0x%lx: old = 0x%llx, "
2999 "new = 0x%llx", vi
->i_ino
, (long long)
3000 sle64_to_cpu(si
->last_mft_change_time
),
3001 (long long)sle64_to_cpu(nt
));
3002 si
->last_mft_change_time
= nt
;
3005 nt
= utc2ntfs(vi
->i_atime
);
3006 if (si
->last_access_time
!= nt
) {
3007 ntfs_debug("Updating atime for inode 0x%lx: old = 0x%llx, "
3008 "new = 0x%llx", vi
->i_ino
,
3009 (long long)sle64_to_cpu(si
->last_access_time
),
3010 (long long)sle64_to_cpu(nt
));
3011 si
->last_access_time
= nt
;
3015 * If we just modified the standard information attribute we need to
3016 * mark the mft record it is in dirty. We do this manually so that
3017 * mark_inode_dirty() is not called which would redirty the inode and
3018 * hence result in an infinite loop of trying to write the inode.
3019 * There is no need to mark the base inode nor the base mft record
3020 * dirty, since we are going to write this mft record below in any case
3021 * and the base mft record may actually not have been modified so it
3022 * might not need to be written out.
3023 * NOTE: It is not a problem when the inode for $MFT itself is being
3024 * written out as mark_ntfs_record_dirty() will only set I_DIRTY_PAGES
3025 * on the $MFT inode and hence ntfs_write_inode() will not be
3026 * re-invoked because of it which in turn is ok since the dirtied mft
3027 * record will be cleaned and written out to disk below, i.e. before
3028 * this function returns.
3031 flush_dcache_mft_record_page(ctx
->ntfs_ino
);
3032 if (!NInoTestSetDirty(ctx
->ntfs_ino
))
3033 mark_ntfs_record_dirty(ctx
->ntfs_ino
->page
,
3034 ctx
->ntfs_ino
->page_ofs
);
3036 ntfs_attr_put_search_ctx(ctx
);
3037 /* Now the access times are updated, write the base mft record. */
3039 err
= write_mft_record(ni
, m
, sync
);
3040 /* Write all attached extent mft records. */
3041 mutex_lock(&ni
->extent_lock
);
3042 if (ni
->nr_extents
> 0) {
3043 ntfs_inode
**extent_nis
= ni
->ext
.extent_ntfs_inos
;
3046 ntfs_debug("Writing %i extent inodes.", ni
->nr_extents
);
3047 for (i
= 0; i
< ni
->nr_extents
; i
++) {
3048 ntfs_inode
*tni
= extent_nis
[i
];
3050 if (NInoDirty(tni
)) {
3051 MFT_RECORD
*tm
= map_mft_record(tni
);
3055 if (!err
|| err
== -ENOMEM
)
3059 ret
= write_mft_record(tni
, tm
, sync
);
3060 unmap_mft_record(tni
);
3061 if (unlikely(ret
)) {
3062 if (!err
|| err
== -ENOMEM
)
3068 mutex_unlock(&ni
->extent_lock
);
3069 unmap_mft_record(ni
);
3072 ntfs_debug("Done.");
3075 unmap_mft_record(ni
);
3077 if (err
== -ENOMEM
) {
3078 ntfs_warning(vi
->i_sb
, "Not enough memory to write inode. "
3079 "Marking the inode dirty again, so the VFS "
3081 mark_inode_dirty(vi
);
3083 ntfs_error(vi
->i_sb
, "Failed (error %i): Run chkdsk.", -err
);
3084 NVolSetErrors(ni
->vol
);
3089 #endif /* NTFS_RW */