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>
45 * ntfs_test_inode - compare two (possibly fake) inodes for equality
46 * @vi: vfs inode which to test
47 * @na: ntfs attribute which is being tested with
49 * Compare the ntfs attribute embedded in the ntfs specific part of the vfs
50 * inode @vi for equality with the ntfs attribute @na.
52 * If searching for the normal file/directory inode, set @na->type to AT_UNUSED.
53 * @na->name and @na->name_len are then ignored.
55 * Return 1 if the attributes match and 0 if not.
57 * NOTE: This function runs with the inode_hash_lock spin lock held so it is not
60 int ntfs_test_inode(struct inode *vi, ntfs_attr *na)
64 if (vi->i_ino != na->mft_no)
67 /* If !NInoAttr(ni), @vi is a normal file or directory inode. */
68 if (likely(!NInoAttr(ni))) {
69 /* If not looking for a normal inode this is a mismatch. */
70 if (unlikely(na->type != AT_UNUSED))
73 /* A fake inode describing an attribute. */
74 if (ni->type != na->type)
76 if (ni->name_len != na->name_len)
78 if (na->name_len && memcmp(ni->name, na->name,
79 na->name_len * sizeof(ntfschar)))
87 * ntfs_init_locked_inode - initialize an inode
88 * @vi: vfs inode to initialize
89 * @na: ntfs attribute which to initialize @vi to
91 * Initialize the vfs inode @vi with the values from the ntfs attribute @na in
92 * order to enable ntfs_test_inode() to do its work.
94 * If initializing the normal file/directory inode, set @na->type to AT_UNUSED.
95 * In that case, @na->name and @na->name_len should be set to NULL and 0,
96 * respectively. Although that is not strictly necessary as
97 * ntfs_read_locked_inode() will fill them in later.
99 * Return 0 on success and -errno on error.
101 * NOTE: This function runs with the inode->i_lock spin lock held so it is not
102 * allowed to sleep. (Hence the GFP_ATOMIC allocation.)
104 static int ntfs_init_locked_inode(struct inode *vi, ntfs_attr *na)
106 ntfs_inode *ni = NTFS_I(vi);
108 vi->i_ino = na->mft_no;
111 if (na->type == AT_INDEX_ALLOCATION)
112 NInoSetMstProtected(ni);
115 ni->name_len = na->name_len;
117 /* If initializing a normal inode, we are done. */
118 if (likely(na->type == AT_UNUSED)) {
120 BUG_ON(na->name_len);
124 /* It is a fake inode. */
128 * We have I30 global constant as an optimization as it is the name
129 * in >99.9% of named attributes! The other <0.1% incur a GFP_ATOMIC
130 * allocation but that is ok. And most attributes are unnamed anyway,
131 * thus the fraction of named attributes with name != I30 is actually
134 if (na->name_len && na->name != I30) {
138 i = na->name_len * sizeof(ntfschar);
139 ni->name = kmalloc(i + sizeof(ntfschar), GFP_ATOMIC);
142 memcpy(ni->name, na->name, i);
143 ni->name[na->name_len] = 0;
148 typedef int (*set_t)(struct inode *, void *);
149 static int ntfs_read_locked_inode(struct inode *vi);
150 static int ntfs_read_locked_attr_inode(struct inode *base_vi, struct inode *vi);
151 static int ntfs_read_locked_index_inode(struct inode *base_vi,
155 * ntfs_iget - obtain a struct inode corresponding to a specific normal inode
156 * @sb: super block of mounted volume
157 * @mft_no: mft record number / inode number to obtain
159 * Obtain the struct inode corresponding to a specific normal inode (i.e. a
160 * file or directory).
162 * If the inode is in the cache, it is just returned with an increased
163 * reference count. Otherwise, a new struct inode is allocated and initialized,
164 * and finally ntfs_read_locked_inode() is called to read in the inode and
165 * fill in the remainder of the inode structure.
167 * Return the struct inode on success. Check the return value with IS_ERR() and
168 * if true, the function failed and the error code is obtained from PTR_ERR().
170 struct inode *ntfs_iget(struct super_block *sb, unsigned long mft_no)
181 vi = iget5_locked(sb, mft_no, (test_t)ntfs_test_inode,
182 (set_t)ntfs_init_locked_inode, &na);
184 return ERR_PTR(-ENOMEM);
188 /* If this is a freshly allocated inode, need to read it now. */
189 if (vi->i_state & I_NEW) {
190 err = ntfs_read_locked_inode(vi);
191 unlock_new_inode(vi);
194 * There is no point in keeping bad inodes around if the failure was
195 * due to ENOMEM. We want to be able to retry again later.
197 if (unlikely(err == -ENOMEM)) {
205 * ntfs_attr_iget - obtain a struct inode corresponding to an attribute
206 * @base_vi: vfs base inode containing the attribute
207 * @type: attribute type
208 * @name: Unicode name of the attribute (NULL if unnamed)
209 * @name_len: length of @name in Unicode characters (0 if unnamed)
211 * Obtain the (fake) struct inode corresponding to the attribute specified by
212 * @type, @name, and @name_len, which is present in the base mft record
213 * specified by the vfs inode @base_vi.
215 * If the attribute inode is in the cache, it is just returned with an
216 * increased reference count. Otherwise, a new struct inode is allocated and
217 * initialized, and finally ntfs_read_locked_attr_inode() is called to read the
218 * attribute and fill in the inode structure.
220 * Note, for index allocation attributes, you need to use ntfs_index_iget()
221 * instead of ntfs_attr_iget() as working with indices is a lot more complex.
223 * Return the struct inode of the attribute inode on success. Check the return
224 * value with IS_ERR() and if true, the function failed and the error code is
225 * obtained from PTR_ERR().
227 struct inode *ntfs_attr_iget(struct inode *base_vi, ATTR_TYPE type,
228 ntfschar *name, u32 name_len)
234 /* Make sure no one calls ntfs_attr_iget() for indices. */
235 BUG_ON(type == AT_INDEX_ALLOCATION);
237 na.mft_no = base_vi->i_ino;
240 na.name_len = name_len;
242 vi = iget5_locked(base_vi->i_sb, na.mft_no, (test_t)ntfs_test_inode,
243 (set_t)ntfs_init_locked_inode, &na);
245 return ERR_PTR(-ENOMEM);
249 /* If this is a freshly allocated inode, need to read it now. */
250 if (vi->i_state & I_NEW) {
251 err = ntfs_read_locked_attr_inode(base_vi, vi);
252 unlock_new_inode(vi);
255 * There is no point in keeping bad attribute inodes around. This also
256 * simplifies things in that we never need to check for bad attribute
267 * ntfs_index_iget - obtain a struct inode corresponding to an index
268 * @base_vi: vfs base inode containing the index related attributes
269 * @name: Unicode name of the index
270 * @name_len: length of @name in Unicode characters
272 * Obtain the (fake) struct inode corresponding to the index specified by @name
273 * and @name_len, which is present in the base mft record specified by the vfs
276 * If the index inode is in the cache, it is just returned with an increased
277 * reference count. Otherwise, a new struct inode is allocated and
278 * initialized, and finally ntfs_read_locked_index_inode() is called to read
279 * the index related attributes and fill in the inode structure.
281 * Return the struct inode of the index inode on success. Check the return
282 * value with IS_ERR() and if true, the function failed and the error code is
283 * obtained from PTR_ERR().
285 struct inode *ntfs_index_iget(struct inode *base_vi, ntfschar *name,
292 na.mft_no = base_vi->i_ino;
293 na.type = AT_INDEX_ALLOCATION;
295 na.name_len = name_len;
297 vi = iget5_locked(base_vi->i_sb, na.mft_no, (test_t)ntfs_test_inode,
298 (set_t)ntfs_init_locked_inode, &na);
300 return ERR_PTR(-ENOMEM);
304 /* If this is a freshly allocated inode, need to read it now. */
305 if (vi->i_state & I_NEW) {
306 err = ntfs_read_locked_index_inode(base_vi, vi);
307 unlock_new_inode(vi);
310 * There is no point in keeping bad index inodes around. This also
311 * simplifies things in that we never need to check for bad index
321 struct inode *ntfs_alloc_big_inode(struct super_block *sb)
325 ntfs_debug("Entering.");
326 ni = kmem_cache_alloc(ntfs_big_inode_cache, GFP_NOFS);
327 if (likely(ni != NULL)) {
331 ntfs_error(sb, "Allocation of NTFS big inode structure failed.");
335 static void ntfs_i_callback(struct rcu_head *head)
337 struct inode *inode = container_of(head, struct inode, i_rcu);
338 kmem_cache_free(ntfs_big_inode_cache, NTFS_I(inode));
341 void ntfs_destroy_big_inode(struct inode *inode)
343 ntfs_inode *ni = NTFS_I(inode);
345 ntfs_debug("Entering.");
347 if (!atomic_dec_and_test(&ni->count))
349 call_rcu(&inode->i_rcu, ntfs_i_callback);
352 static inline ntfs_inode *ntfs_alloc_extent_inode(void)
356 ntfs_debug("Entering.");
357 ni = kmem_cache_alloc(ntfs_inode_cache, GFP_NOFS);
358 if (likely(ni != NULL)) {
362 ntfs_error(NULL, "Allocation of NTFS inode structure failed.");
366 static void ntfs_destroy_extent_inode(ntfs_inode *ni)
368 ntfs_debug("Entering.");
370 if (!atomic_dec_and_test(&ni->count))
372 kmem_cache_free(ntfs_inode_cache, ni);
376 * The attribute runlist lock has separate locking rules from the
377 * normal runlist lock, so split the two lock-classes:
379 static struct lock_class_key attr_list_rl_lock_class;
382 * __ntfs_init_inode - initialize ntfs specific part of an inode
383 * @sb: super block of mounted volume
384 * @ni: freshly allocated ntfs inode which to initialize
386 * Initialize an ntfs inode to defaults.
388 * NOTE: ni->mft_no, ni->state, ni->type, ni->name, and ni->name_len are left
389 * untouched. Make sure to initialize them elsewhere.
391 * Return zero on success and -ENOMEM on error.
393 void __ntfs_init_inode(struct super_block *sb, ntfs_inode *ni)
395 ntfs_debug("Entering.");
396 rwlock_init(&ni->size_lock);
397 ni->initialized_size = ni->allocated_size = 0;
399 atomic_set(&ni->count, 1);
400 ni->vol = NTFS_SB(sb);
401 ntfs_init_runlist(&ni->runlist);
402 mutex_init(&ni->mrec_lock);
405 ni->attr_list_size = 0;
406 ni->attr_list = NULL;
407 ntfs_init_runlist(&ni->attr_list_rl);
408 lockdep_set_class(&ni->attr_list_rl.lock,
409 &attr_list_rl_lock_class);
410 ni->itype.index.block_size = 0;
411 ni->itype.index.vcn_size = 0;
412 ni->itype.index.collation_rule = 0;
413 ni->itype.index.block_size_bits = 0;
414 ni->itype.index.vcn_size_bits = 0;
415 mutex_init(&ni->extent_lock);
417 ni->ext.base_ntfs_ino = NULL;
421 * Extent inodes get MFT-mapped in a nested way, while the base inode
422 * is still mapped. Teach this nesting to the lock validator by creating
423 * a separate class for nested inode's mrec_lock's:
425 static struct lock_class_key extent_inode_mrec_lock_key;
427 inline ntfs_inode *ntfs_new_extent_inode(struct super_block *sb,
428 unsigned long mft_no)
430 ntfs_inode *ni = ntfs_alloc_extent_inode();
432 ntfs_debug("Entering.");
433 if (likely(ni != NULL)) {
434 __ntfs_init_inode(sb, ni);
435 lockdep_set_class(&ni->mrec_lock, &extent_inode_mrec_lock_key);
437 ni->type = AT_UNUSED;
445 * ntfs_is_extended_system_file - check if a file is in the $Extend directory
446 * @ctx: initialized attribute search context
448 * Search all file name attributes in the inode described by the attribute
449 * search context @ctx and check if any of the names are in the $Extend system
453 * 1: file is in $Extend directory
454 * 0: file is not in $Extend directory
455 * -errno: failed to determine if the file is in the $Extend directory
457 static int ntfs_is_extended_system_file(ntfs_attr_search_ctx *ctx)
461 /* Restart search. */
462 ntfs_attr_reinit_search_ctx(ctx);
464 /* Get number of hard links. */
465 nr_links = le16_to_cpu(ctx->mrec->link_count);
467 /* Loop through all hard links. */
468 while (!(err = ntfs_attr_lookup(AT_FILE_NAME, NULL, 0, 0, 0, NULL, 0,
470 FILE_NAME_ATTR *file_name_attr;
471 ATTR_RECORD *attr = ctx->attr;
476 * Maximum sanity checking as we are called on an inode that
477 * we suspect might be corrupt.
479 p = (u8*)attr + le32_to_cpu(attr->length);
480 if (p < (u8*)ctx->mrec || (u8*)p > (u8*)ctx->mrec +
481 le32_to_cpu(ctx->mrec->bytes_in_use)) {
483 ntfs_error(ctx->ntfs_ino->vol->sb, "Corrupt file name "
484 "attribute. You should run chkdsk.");
487 if (attr->non_resident) {
488 ntfs_error(ctx->ntfs_ino->vol->sb, "Non-resident file "
489 "name. You should run chkdsk.");
493 ntfs_error(ctx->ntfs_ino->vol->sb, "File name with "
494 "invalid flags. You should run "
498 if (!(attr->data.resident.flags & RESIDENT_ATTR_IS_INDEXED)) {
499 ntfs_error(ctx->ntfs_ino->vol->sb, "Unindexed file "
500 "name. You should run chkdsk.");
503 file_name_attr = (FILE_NAME_ATTR*)((u8*)attr +
504 le16_to_cpu(attr->data.resident.value_offset));
505 p2 = (u8 *)file_name_attr + le32_to_cpu(attr->data.resident.value_length);
506 if (p2 < (u8*)attr || p2 > p)
507 goto err_corrupt_attr;
508 /* This attribute is ok, but is it in the $Extend directory? */
509 if (MREF_LE(file_name_attr->parent_directory) == FILE_Extend)
510 return 1; /* YES, it's an extended system file. */
512 if (unlikely(err != -ENOENT))
514 if (unlikely(nr_links)) {
515 ntfs_error(ctx->ntfs_ino->vol->sb, "Inode hard link count "
516 "doesn't match number of name attributes. You "
517 "should run chkdsk.");
520 return 0; /* NO, it is not an extended system file. */
524 * ntfs_read_locked_inode - read an inode from its device
527 * ntfs_read_locked_inode() is called from ntfs_iget() to read the inode
528 * described by @vi into memory from the device.
530 * The only fields in @vi that we need to/can look at when the function is
531 * called are i_sb, pointing to the mounted device's super block, and i_ino,
532 * the number of the inode to load.
534 * ntfs_read_locked_inode() maps, pins and locks the mft record number i_ino
535 * for reading and sets up the necessary @vi fields as well as initializing
538 * Q: What locks are held when the function is called?
539 * A: i_state has I_NEW set, hence the inode is locked, also
540 * i_count is set to 1, so it is not going to go away
541 * i_flags is set to 0 and we have no business touching it. Only an ioctl()
542 * is allowed to write to them. We should of course be honouring them but
543 * we need to do that using the IS_* macros defined in include/linux/fs.h.
544 * In any case ntfs_read_locked_inode() has nothing to do with i_flags.
546 * Return 0 on success and -errno on error. In the error case, the inode will
547 * have had make_bad_inode() executed on it.
549 static int ntfs_read_locked_inode(struct inode *vi)
551 ntfs_volume *vol = NTFS_SB(vi->i_sb);
556 STANDARD_INFORMATION *si;
557 ntfs_attr_search_ctx *ctx;
560 ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino);
562 /* Setup the generic vfs inode parts now. */
563 vi->i_uid = vol->uid;
564 vi->i_gid = vol->gid;
568 * Initialize the ntfs specific part of @vi special casing
569 * FILE_MFT which we need to do at mount time.
571 if (vi->i_ino != FILE_MFT)
572 ntfs_init_big_inode(vi);
575 m = map_mft_record(ni);
580 ctx = ntfs_attr_get_search_ctx(ni, m);
586 if (!(m->flags & MFT_RECORD_IN_USE)) {
587 ntfs_error(vi->i_sb, "Inode is not in use!");
590 if (m->base_mft_record) {
591 ntfs_error(vi->i_sb, "Inode is an extent inode!");
595 /* Transfer information from mft record into vfs and ntfs inodes. */
596 vi->i_generation = ni->seq_no = le16_to_cpu(m->sequence_number);
599 * FIXME: Keep in mind that link_count is two for files which have both
600 * a long file name and a short file name as separate entries, so if
601 * we are hiding short file names this will be too high. Either we need
602 * to account for the short file names by subtracting them or we need
603 * to make sure we delete files even though i_nlink is not zero which
604 * might be tricky due to vfs interactions. Need to think about this
605 * some more when implementing the unlink command.
607 set_nlink(vi, le16_to_cpu(m->link_count));
609 * FIXME: Reparse points can have the directory bit set even though
610 * they would be S_IFLNK. Need to deal with this further below when we
611 * implement reparse points / symbolic links but it will do for now.
612 * Also if not a directory, it could be something else, rather than
613 * a regular file. But again, will do for now.
615 /* Everyone gets all permissions. */
616 vi->i_mode |= S_IRWXUGO;
617 /* If read-only, no one gets write permissions. */
619 vi->i_mode &= ~S_IWUGO;
620 if (m->flags & MFT_RECORD_IS_DIRECTORY) {
621 vi->i_mode |= S_IFDIR;
623 * Apply the directory permissions mask set in the mount
626 vi->i_mode &= ~vol->dmask;
627 /* Things break without this kludge! */
631 vi->i_mode |= S_IFREG;
632 /* Apply the file permissions mask set in the mount options. */
633 vi->i_mode &= ~vol->fmask;
636 * Find the standard information attribute in the mft record. At this
637 * stage we haven't setup the attribute list stuff yet, so this could
638 * in fact fail if the standard information is in an extent record, but
639 * I don't think this actually ever happens.
641 err = ntfs_attr_lookup(AT_STANDARD_INFORMATION, NULL, 0, 0, 0, NULL, 0,
644 if (err == -ENOENT) {
646 * TODO: We should be performing a hot fix here (if the
647 * recover mount option is set) by creating a new
650 ntfs_error(vi->i_sb, "$STANDARD_INFORMATION attribute "
656 /* Get the standard information attribute value. */
657 if ((u8 *)a + le16_to_cpu(a->data.resident.value_offset)
658 + le32_to_cpu(a->data.resident.value_length) >
659 (u8 *)ctx->mrec + vol->mft_record_size) {
660 ntfs_error(vi->i_sb, "Corrupt standard information attribute in inode.");
663 si = (STANDARD_INFORMATION*)((u8*)a +
664 le16_to_cpu(a->data.resident.value_offset));
666 /* Transfer information from the standard information into vi. */
668 * Note: The i_?times do not quite map perfectly onto the NTFS times,
669 * but they are close enough, and in the end it doesn't really matter
673 * mtime is the last change of the data within the file. Not changed
674 * when only metadata is changed, e.g. a rename doesn't affect mtime.
676 vi->i_mtime = ntfs2utc(si->last_data_change_time);
678 * ctime is the last change of the metadata of the file. This obviously
679 * always changes, when mtime is changed. ctime can be changed on its
680 * own, mtime is then not changed, e.g. when a file is renamed.
682 vi->i_ctime = ntfs2utc(si->last_mft_change_time);
684 * Last access to the data within the file. Not changed during a rename
685 * for example but changed whenever the file is written to.
687 vi->i_atime = ntfs2utc(si->last_access_time);
689 /* Find the attribute list attribute if present. */
690 ntfs_attr_reinit_search_ctx(ctx);
691 err = ntfs_attr_lookup(AT_ATTRIBUTE_LIST, NULL, 0, 0, 0, NULL, 0, ctx);
693 if (unlikely(err != -ENOENT)) {
694 ntfs_error(vi->i_sb, "Failed to lookup attribute list "
698 } else /* if (!err) */ {
699 if (vi->i_ino == FILE_MFT)
700 goto skip_attr_list_load;
701 ntfs_debug("Attribute list found in inode 0x%lx.", vi->i_ino);
704 if (a->flags & ATTR_COMPRESSION_MASK) {
705 ntfs_error(vi->i_sb, "Attribute list attribute is "
709 if (a->flags & ATTR_IS_ENCRYPTED ||
710 a->flags & ATTR_IS_SPARSE) {
711 if (a->non_resident) {
712 ntfs_error(vi->i_sb, "Non-resident attribute "
713 "list attribute is encrypted/"
717 ntfs_warning(vi->i_sb, "Resident attribute list "
718 "attribute in inode 0x%lx is marked "
719 "encrypted/sparse which is not true. "
720 "However, Windows allows this and "
721 "chkdsk does not detect or correct it "
722 "so we will just ignore the invalid "
723 "flags and pretend they are not set.",
726 /* Now allocate memory for the attribute list. */
727 ni->attr_list_size = (u32)ntfs_attr_size(a);
728 ni->attr_list = ntfs_malloc_nofs(ni->attr_list_size);
729 if (!ni->attr_list) {
730 ntfs_error(vi->i_sb, "Not enough memory to allocate "
731 "buffer for attribute list.");
735 if (a->non_resident) {
736 NInoSetAttrListNonResident(ni);
737 if (a->data.non_resident.lowest_vcn) {
738 ntfs_error(vi->i_sb, "Attribute list has non "
743 * Setup the runlist. No need for locking as we have
744 * exclusive access to the inode at this time.
746 ni->attr_list_rl.rl = ntfs_mapping_pairs_decompress(vol,
748 if (IS_ERR(ni->attr_list_rl.rl)) {
749 err = PTR_ERR(ni->attr_list_rl.rl);
750 ni->attr_list_rl.rl = NULL;
751 ntfs_error(vi->i_sb, "Mapping pairs "
752 "decompression failed.");
755 /* Now load the attribute list. */
756 if ((err = load_attribute_list(vol, &ni->attr_list_rl,
757 ni->attr_list, ni->attr_list_size,
758 sle64_to_cpu(a->data.non_resident.
759 initialized_size)))) {
760 ntfs_error(vi->i_sb, "Failed to load "
761 "attribute list attribute.");
764 } else /* if (!a->non_resident) */ {
765 if ((u8*)a + le16_to_cpu(a->data.resident.value_offset)
767 a->data.resident.value_length) >
768 (u8*)ctx->mrec + vol->mft_record_size) {
769 ntfs_error(vi->i_sb, "Corrupt attribute list "
773 /* Now copy the attribute list. */
774 memcpy(ni->attr_list, (u8*)a + le16_to_cpu(
775 a->data.resident.value_offset),
777 a->data.resident.value_length));
782 * If an attribute list is present we now have the attribute list value
783 * in ntfs_ino->attr_list and it is ntfs_ino->attr_list_size bytes.
785 if (S_ISDIR(vi->i_mode)) {
789 u8 *ir_end, *index_end;
791 /* It is a directory, find index root attribute. */
792 ntfs_attr_reinit_search_ctx(ctx);
793 err = ntfs_attr_lookup(AT_INDEX_ROOT, I30, 4, CASE_SENSITIVE,
796 if (err == -ENOENT) {
797 // FIXME: File is corrupt! Hot-fix with empty
798 // index root attribute if recovery option is
800 ntfs_error(vi->i_sb, "$INDEX_ROOT attribute "
806 /* Set up the state. */
807 if (unlikely(a->non_resident)) {
808 ntfs_error(vol->sb, "$INDEX_ROOT attribute is not "
812 /* Ensure the attribute name is placed before the value. */
813 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
814 le16_to_cpu(a->data.resident.value_offset)))) {
815 ntfs_error(vol->sb, "$INDEX_ROOT attribute name is "
816 "placed after the attribute value.");
820 * Compressed/encrypted index root just means that the newly
821 * created files in that directory should be created compressed/
822 * encrypted. However index root cannot be both compressed and
825 if (a->flags & ATTR_COMPRESSION_MASK)
826 NInoSetCompressed(ni);
827 if (a->flags & ATTR_IS_ENCRYPTED) {
828 if (a->flags & ATTR_COMPRESSION_MASK) {
829 ntfs_error(vi->i_sb, "Found encrypted and "
830 "compressed attribute.");
833 NInoSetEncrypted(ni);
835 if (a->flags & ATTR_IS_SPARSE)
837 ir = (INDEX_ROOT*)((u8*)a +
838 le16_to_cpu(a->data.resident.value_offset));
839 ir_end = (u8*)ir + le32_to_cpu(a->data.resident.value_length);
840 if (ir_end > (u8*)ctx->mrec + vol->mft_record_size) {
841 ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is "
845 index_end = (u8*)&ir->index +
846 le32_to_cpu(ir->index.index_length);
847 if (index_end > ir_end) {
848 ntfs_error(vi->i_sb, "Directory index is corrupt.");
851 if (ir->type != AT_FILE_NAME) {
852 ntfs_error(vi->i_sb, "Indexed attribute is not "
856 if (ir->collation_rule != COLLATION_FILE_NAME) {
857 ntfs_error(vi->i_sb, "Index collation rule is not "
858 "COLLATION_FILE_NAME.");
861 ni->itype.index.collation_rule = ir->collation_rule;
862 ni->itype.index.block_size = le32_to_cpu(ir->index_block_size);
863 if (ni->itype.index.block_size &
864 (ni->itype.index.block_size - 1)) {
865 ntfs_error(vi->i_sb, "Index block size (%u) is not a "
867 ni->itype.index.block_size);
870 if (ni->itype.index.block_size > PAGE_SIZE) {
871 ntfs_error(vi->i_sb, "Index block size (%u) > "
872 "PAGE_SIZE (%ld) is not "
874 ni->itype.index.block_size,
879 if (ni->itype.index.block_size < NTFS_BLOCK_SIZE) {
880 ntfs_error(vi->i_sb, "Index block size (%u) < "
881 "NTFS_BLOCK_SIZE (%i) is not "
883 ni->itype.index.block_size,
888 ni->itype.index.block_size_bits =
889 ffs(ni->itype.index.block_size) - 1;
890 /* Determine the size of a vcn in the directory index. */
891 if (vol->cluster_size <= ni->itype.index.block_size) {
892 ni->itype.index.vcn_size = vol->cluster_size;
893 ni->itype.index.vcn_size_bits = vol->cluster_size_bits;
895 ni->itype.index.vcn_size = vol->sector_size;
896 ni->itype.index.vcn_size_bits = vol->sector_size_bits;
899 /* Setup the index allocation attribute, even if not present. */
900 NInoSetMstProtected(ni);
901 ni->type = AT_INDEX_ALLOCATION;
905 if (!(ir->index.flags & LARGE_INDEX)) {
906 /* No index allocation. */
907 vi->i_size = ni->initialized_size =
908 ni->allocated_size = 0;
909 /* We are done with the mft record, so we release it. */
910 ntfs_attr_put_search_ctx(ctx);
911 unmap_mft_record(ni);
914 goto skip_large_dir_stuff;
915 } /* LARGE_INDEX: Index allocation present. Setup state. */
916 NInoSetIndexAllocPresent(ni);
917 /* Find index allocation attribute. */
918 ntfs_attr_reinit_search_ctx(ctx);
919 err = ntfs_attr_lookup(AT_INDEX_ALLOCATION, I30, 4,
920 CASE_SENSITIVE, 0, NULL, 0, ctx);
923 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION "
924 "attribute is not present but "
925 "$INDEX_ROOT indicated it is.");
927 ntfs_error(vi->i_sb, "Failed to lookup "
933 if (!a->non_resident) {
934 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
939 * Ensure the attribute name is placed before the mapping pairs
942 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
944 a->data.non_resident.mapping_pairs_offset)))) {
945 ntfs_error(vol->sb, "$INDEX_ALLOCATION attribute name "
946 "is placed after the mapping pairs "
950 if (a->flags & ATTR_IS_ENCRYPTED) {
951 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
955 if (a->flags & ATTR_IS_SPARSE) {
956 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
960 if (a->flags & ATTR_COMPRESSION_MASK) {
961 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
965 if (a->data.non_resident.lowest_vcn) {
966 ntfs_error(vi->i_sb, "First extent of "
967 "$INDEX_ALLOCATION attribute has non "
971 vi->i_size = sle64_to_cpu(a->data.non_resident.data_size);
972 ni->initialized_size = sle64_to_cpu(
973 a->data.non_resident.initialized_size);
974 ni->allocated_size = sle64_to_cpu(
975 a->data.non_resident.allocated_size);
977 * We are done with the mft record, so we release it. Otherwise
978 * we would deadlock in ntfs_attr_iget().
980 ntfs_attr_put_search_ctx(ctx);
981 unmap_mft_record(ni);
984 /* Get the index bitmap attribute inode. */
985 bvi = ntfs_attr_iget(vi, AT_BITMAP, I30, 4);
987 ntfs_error(vi->i_sb, "Failed to get bitmap attribute.");
992 if (NInoCompressed(bni) || NInoEncrypted(bni) ||
994 ntfs_error(vi->i_sb, "$BITMAP attribute is compressed "
995 "and/or encrypted and/or sparse.");
996 goto iput_unm_err_out;
998 /* Consistency check bitmap size vs. index allocation size. */
999 bvi_size = i_size_read(bvi);
1000 if ((bvi_size << 3) < (vi->i_size >>
1001 ni->itype.index.block_size_bits)) {
1002 ntfs_error(vi->i_sb, "Index bitmap too small (0x%llx) "
1003 "for index allocation (0x%llx).",
1004 bvi_size << 3, vi->i_size);
1005 goto iput_unm_err_out;
1007 /* No longer need the bitmap attribute inode. */
1009 skip_large_dir_stuff:
1010 /* Setup the operations for this inode. */
1011 vi->i_op = &ntfs_dir_inode_ops;
1012 vi->i_fop = &ntfs_dir_ops;
1013 vi->i_mapping->a_ops = &ntfs_mst_aops;
1016 ntfs_attr_reinit_search_ctx(ctx);
1018 /* Setup the data attribute, even if not present. */
1023 /* Find first extent of the unnamed data attribute. */
1024 err = ntfs_attr_lookup(AT_DATA, NULL, 0, 0, 0, NULL, 0, ctx);
1025 if (unlikely(err)) {
1026 vi->i_size = ni->initialized_size =
1027 ni->allocated_size = 0;
1028 if (err != -ENOENT) {
1029 ntfs_error(vi->i_sb, "Failed to lookup $DATA "
1034 * FILE_Secure does not have an unnamed $DATA
1035 * attribute, so we special case it here.
1037 if (vi->i_ino == FILE_Secure)
1038 goto no_data_attr_special_case;
1040 * Most if not all the system files in the $Extend
1041 * system directory do not have unnamed data
1042 * attributes so we need to check if the parent
1043 * directory of the file is FILE_Extend and if it is
1044 * ignore this error. To do this we need to get the
1045 * name of this inode from the mft record as the name
1046 * contains the back reference to the parent directory.
1048 if (ntfs_is_extended_system_file(ctx) > 0)
1049 goto no_data_attr_special_case;
1050 // FIXME: File is corrupt! Hot-fix with empty data
1051 // attribute if recovery option is set.
1052 ntfs_error(vi->i_sb, "$DATA attribute is missing.");
1056 /* Setup the state. */
1057 if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_SPARSE)) {
1058 if (a->flags & ATTR_COMPRESSION_MASK) {
1059 NInoSetCompressed(ni);
1060 if (vol->cluster_size > 4096) {
1061 ntfs_error(vi->i_sb, "Found "
1062 "compressed data but "
1065 "cluster size (%i) > "
1070 if ((a->flags & ATTR_COMPRESSION_MASK)
1071 != ATTR_IS_COMPRESSED) {
1072 ntfs_error(vi->i_sb, "Found unknown "
1073 "compression method "
1074 "or corrupt file.");
1078 if (a->flags & ATTR_IS_SPARSE)
1081 if (a->flags & ATTR_IS_ENCRYPTED) {
1082 if (NInoCompressed(ni)) {
1083 ntfs_error(vi->i_sb, "Found encrypted and "
1084 "compressed data.");
1087 NInoSetEncrypted(ni);
1089 if (a->non_resident) {
1090 NInoSetNonResident(ni);
1091 if (NInoCompressed(ni) || NInoSparse(ni)) {
1092 if (NInoCompressed(ni) && a->data.non_resident.
1093 compression_unit != 4) {
1094 ntfs_error(vi->i_sb, "Found "
1096 "compression unit (%u "
1098 "Cannot handle this.",
1099 a->data.non_resident.
1104 if (a->data.non_resident.compression_unit) {
1105 ni->itype.compressed.block_size = 1U <<
1106 (a->data.non_resident.
1108 vol->cluster_size_bits);
1109 ni->itype.compressed.block_size_bits =
1113 ni->itype.compressed.block_clusters =
1118 ni->itype.compressed.block_size = 0;
1119 ni->itype.compressed.block_size_bits =
1121 ni->itype.compressed.block_clusters =
1124 ni->itype.compressed.size = sle64_to_cpu(
1125 a->data.non_resident.
1128 if (a->data.non_resident.lowest_vcn) {
1129 ntfs_error(vi->i_sb, "First extent of $DATA "
1130 "attribute has non zero "
1134 vi->i_size = sle64_to_cpu(
1135 a->data.non_resident.data_size);
1136 ni->initialized_size = sle64_to_cpu(
1137 a->data.non_resident.initialized_size);
1138 ni->allocated_size = sle64_to_cpu(
1139 a->data.non_resident.allocated_size);
1140 } else { /* Resident attribute. */
1141 vi->i_size = ni->initialized_size = le32_to_cpu(
1142 a->data.resident.value_length);
1143 ni->allocated_size = le32_to_cpu(a->length) -
1145 a->data.resident.value_offset);
1146 if (vi->i_size > ni->allocated_size) {
1147 ntfs_error(vi->i_sb, "Resident data attribute "
1148 "is corrupt (size exceeds "
1153 no_data_attr_special_case:
1154 /* We are done with the mft record, so we release it. */
1155 ntfs_attr_put_search_ctx(ctx);
1156 unmap_mft_record(ni);
1159 /* Setup the operations for this inode. */
1160 vi->i_op = &ntfs_file_inode_ops;
1161 vi->i_fop = &ntfs_file_ops;
1162 vi->i_mapping->a_ops = &ntfs_normal_aops;
1163 if (NInoMstProtected(ni))
1164 vi->i_mapping->a_ops = &ntfs_mst_aops;
1165 else if (NInoCompressed(ni))
1166 vi->i_mapping->a_ops = &ntfs_compressed_aops;
1169 * The number of 512-byte blocks used on disk (for stat). This is in so
1170 * far inaccurate as it doesn't account for any named streams or other
1171 * special non-resident attributes, but that is how Windows works, too,
1172 * so we are at least consistent with Windows, if not entirely
1173 * consistent with the Linux Way. Doing it the Linux Way would cause a
1174 * significant slowdown as it would involve iterating over all
1175 * attributes in the mft record and adding the allocated/compressed
1176 * sizes of all non-resident attributes present to give us the Linux
1177 * correct size that should go into i_blocks (after division by 512).
1179 if (S_ISREG(vi->i_mode) && (NInoCompressed(ni) || NInoSparse(ni)))
1180 vi->i_blocks = ni->itype.compressed.size >> 9;
1182 vi->i_blocks = ni->allocated_size >> 9;
1183 ntfs_debug("Done.");
1191 ntfs_attr_put_search_ctx(ctx);
1193 unmap_mft_record(ni);
1195 ntfs_error(vol->sb, "Failed with error code %i. Marking corrupt "
1196 "inode 0x%lx as bad. Run chkdsk.", err, vi->i_ino);
1198 if (err != -EOPNOTSUPP && err != -ENOMEM)
1204 * ntfs_read_locked_attr_inode - read an attribute inode from its base inode
1205 * @base_vi: base inode
1206 * @vi: attribute inode to read
1208 * ntfs_read_locked_attr_inode() is called from ntfs_attr_iget() to read the
1209 * attribute inode described by @vi into memory from the base mft record
1210 * described by @base_ni.
1212 * ntfs_read_locked_attr_inode() maps, pins and locks the base inode for
1213 * reading and looks up the attribute described by @vi before setting up the
1214 * necessary fields in @vi as well as initializing the ntfs inode.
1216 * Q: What locks are held when the function is called?
1217 * A: i_state has I_NEW set, hence the inode is locked, also
1218 * i_count is set to 1, so it is not going to go away
1220 * Return 0 on success and -errno on error. In the error case, the inode will
1221 * have had make_bad_inode() executed on it.
1223 * Note this cannot be called for AT_INDEX_ALLOCATION.
1225 static int ntfs_read_locked_attr_inode(struct inode *base_vi, struct inode *vi)
1227 ntfs_volume *vol = NTFS_SB(vi->i_sb);
1228 ntfs_inode *ni, *base_ni;
1231 ntfs_attr_search_ctx *ctx;
1234 ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino);
1236 ntfs_init_big_inode(vi);
1239 base_ni = NTFS_I(base_vi);
1241 /* Just mirror the values from the base inode. */
1242 vi->i_uid = base_vi->i_uid;
1243 vi->i_gid = base_vi->i_gid;
1244 set_nlink(vi, base_vi->i_nlink);
1245 vi->i_mtime = base_vi->i_mtime;
1246 vi->i_ctime = base_vi->i_ctime;
1247 vi->i_atime = base_vi->i_atime;
1248 vi->i_generation = ni->seq_no = base_ni->seq_no;
1250 /* Set inode type to zero but preserve permissions. */
1251 vi->i_mode = base_vi->i_mode & ~S_IFMT;
1253 m = map_mft_record(base_ni);
1258 ctx = ntfs_attr_get_search_ctx(base_ni, m);
1263 /* Find the attribute. */
1264 err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
1265 CASE_SENSITIVE, 0, NULL, 0, ctx);
1269 if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_SPARSE)) {
1270 if (a->flags & ATTR_COMPRESSION_MASK) {
1271 NInoSetCompressed(ni);
1272 if ((ni->type != AT_DATA) || (ni->type == AT_DATA &&
1274 ntfs_error(vi->i_sb, "Found compressed "
1275 "non-data or named data "
1276 "attribute. Please report "
1277 "you saw this message to "
1278 "linux-ntfs-dev@lists."
1282 if (vol->cluster_size > 4096) {
1283 ntfs_error(vi->i_sb, "Found compressed "
1284 "attribute but compression is "
1285 "disabled due to cluster size "
1290 if ((a->flags & ATTR_COMPRESSION_MASK) !=
1291 ATTR_IS_COMPRESSED) {
1292 ntfs_error(vi->i_sb, "Found unknown "
1293 "compression method.");
1298 * The compressed/sparse flag set in an index root just means
1299 * to compress all files.
1301 if (NInoMstProtected(ni) && ni->type != AT_INDEX_ROOT) {
1302 ntfs_error(vi->i_sb, "Found mst protected attribute "
1303 "but the attribute is %s. Please "
1304 "report you saw this message to "
1305 "linux-ntfs-dev@lists.sourceforge.net",
1306 NInoCompressed(ni) ? "compressed" :
1310 if (a->flags & ATTR_IS_SPARSE)
1313 if (a->flags & ATTR_IS_ENCRYPTED) {
1314 if (NInoCompressed(ni)) {
1315 ntfs_error(vi->i_sb, "Found encrypted and compressed "
1320 * The encryption flag set in an index root just means to
1321 * encrypt all files.
1323 if (NInoMstProtected(ni) && ni->type != AT_INDEX_ROOT) {
1324 ntfs_error(vi->i_sb, "Found mst protected attribute "
1325 "but the attribute is encrypted. "
1326 "Please report you saw this message "
1327 "to linux-ntfs-dev@lists.sourceforge."
1331 if (ni->type != AT_DATA) {
1332 ntfs_error(vi->i_sb, "Found encrypted non-data "
1336 NInoSetEncrypted(ni);
1338 if (!a->non_resident) {
1339 /* Ensure the attribute name is placed before the value. */
1340 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1341 le16_to_cpu(a->data.resident.value_offset)))) {
1342 ntfs_error(vol->sb, "Attribute name is placed after "
1343 "the attribute value.");
1346 if (NInoMstProtected(ni)) {
1347 ntfs_error(vi->i_sb, "Found mst protected attribute "
1348 "but the attribute is resident. "
1349 "Please report you saw this message to "
1350 "linux-ntfs-dev@lists.sourceforge.net");
1353 vi->i_size = ni->initialized_size = le32_to_cpu(
1354 a->data.resident.value_length);
1355 ni->allocated_size = le32_to_cpu(a->length) -
1356 le16_to_cpu(a->data.resident.value_offset);
1357 if (vi->i_size > ni->allocated_size) {
1358 ntfs_error(vi->i_sb, "Resident attribute is corrupt "
1359 "(size exceeds allocation).");
1363 NInoSetNonResident(ni);
1365 * Ensure the attribute name is placed before the mapping pairs
1368 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1370 a->data.non_resident.mapping_pairs_offset)))) {
1371 ntfs_error(vol->sb, "Attribute name is placed after "
1372 "the mapping pairs array.");
1375 if (NInoCompressed(ni) || NInoSparse(ni)) {
1376 if (NInoCompressed(ni) && a->data.non_resident.
1377 compression_unit != 4) {
1378 ntfs_error(vi->i_sb, "Found non-standard "
1379 "compression unit (%u instead "
1380 "of 4). Cannot handle this.",
1381 a->data.non_resident.
1386 if (a->data.non_resident.compression_unit) {
1387 ni->itype.compressed.block_size = 1U <<
1388 (a->data.non_resident.
1390 vol->cluster_size_bits);
1391 ni->itype.compressed.block_size_bits =
1392 ffs(ni->itype.compressed.
1394 ni->itype.compressed.block_clusters = 1U <<
1395 a->data.non_resident.
1398 ni->itype.compressed.block_size = 0;
1399 ni->itype.compressed.block_size_bits = 0;
1400 ni->itype.compressed.block_clusters = 0;
1402 ni->itype.compressed.size = sle64_to_cpu(
1403 a->data.non_resident.compressed_size);
1405 if (a->data.non_resident.lowest_vcn) {
1406 ntfs_error(vi->i_sb, "First extent of attribute has "
1407 "non-zero lowest_vcn.");
1410 vi->i_size = sle64_to_cpu(a->data.non_resident.data_size);
1411 ni->initialized_size = sle64_to_cpu(
1412 a->data.non_resident.initialized_size);
1413 ni->allocated_size = sle64_to_cpu(
1414 a->data.non_resident.allocated_size);
1416 vi->i_mapping->a_ops = &ntfs_normal_aops;
1417 if (NInoMstProtected(ni))
1418 vi->i_mapping->a_ops = &ntfs_mst_aops;
1419 else if (NInoCompressed(ni))
1420 vi->i_mapping->a_ops = &ntfs_compressed_aops;
1421 if ((NInoCompressed(ni) || NInoSparse(ni)) && ni->type != AT_INDEX_ROOT)
1422 vi->i_blocks = ni->itype.compressed.size >> 9;
1424 vi->i_blocks = ni->allocated_size >> 9;
1426 * Make sure the base inode does not go away and attach it to the
1430 ni->ext.base_ntfs_ino = base_ni;
1431 ni->nr_extents = -1;
1433 ntfs_attr_put_search_ctx(ctx);
1434 unmap_mft_record(base_ni);
1436 ntfs_debug("Done.");
1443 ntfs_attr_put_search_ctx(ctx);
1444 unmap_mft_record(base_ni);
1446 ntfs_error(vol->sb, "Failed with error code %i while reading attribute "
1447 "inode (mft_no 0x%lx, type 0x%x, name_len %i). "
1448 "Marking corrupt inode and base inode 0x%lx as bad. "
1449 "Run chkdsk.", err, vi->i_ino, ni->type, ni->name_len,
1458 * ntfs_read_locked_index_inode - read an index inode from its base inode
1459 * @base_vi: base inode
1460 * @vi: index inode to read
1462 * ntfs_read_locked_index_inode() is called from ntfs_index_iget() to read the
1463 * index inode described by @vi into memory from the base mft record described
1466 * ntfs_read_locked_index_inode() maps, pins and locks the base inode for
1467 * reading and looks up the attributes relating to the index described by @vi
1468 * before setting up the necessary fields in @vi as well as initializing the
1471 * Note, index inodes are essentially attribute inodes (NInoAttr() is true)
1472 * with the attribute type set to AT_INDEX_ALLOCATION. Apart from that, they
1473 * are setup like directory inodes since directories are a special case of
1474 * indices ao they need to be treated in much the same way. Most importantly,
1475 * for small indices the index allocation attribute might not actually exist.
1476 * However, the index root attribute always exists but this does not need to
1477 * have an inode associated with it and this is why we define a new inode type
1478 * index. Also, like for directories, we need to have an attribute inode for
1479 * the bitmap attribute corresponding to the index allocation attribute and we
1480 * can store this in the appropriate field of the inode, just like we do for
1481 * normal directory inodes.
1483 * Q: What locks are held when the function is called?
1484 * A: i_state has I_NEW set, hence the inode is locked, also
1485 * i_count is set to 1, so it is not going to go away
1487 * Return 0 on success and -errno on error. In the error case, the inode will
1488 * have had make_bad_inode() executed on it.
1490 static int ntfs_read_locked_index_inode(struct inode *base_vi, struct inode *vi)
1493 ntfs_volume *vol = NTFS_SB(vi->i_sb);
1494 ntfs_inode *ni, *base_ni, *bni;
1498 ntfs_attr_search_ctx *ctx;
1500 u8 *ir_end, *index_end;
1503 ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino);
1504 ntfs_init_big_inode(vi);
1506 base_ni = NTFS_I(base_vi);
1507 /* Just mirror the values from the base inode. */
1508 vi->i_uid = base_vi->i_uid;
1509 vi->i_gid = base_vi->i_gid;
1510 set_nlink(vi, base_vi->i_nlink);
1511 vi->i_mtime = base_vi->i_mtime;
1512 vi->i_ctime = base_vi->i_ctime;
1513 vi->i_atime = base_vi->i_atime;
1514 vi->i_generation = ni->seq_no = base_ni->seq_no;
1515 /* Set inode type to zero but preserve permissions. */
1516 vi->i_mode = base_vi->i_mode & ~S_IFMT;
1517 /* Map the mft record for the base inode. */
1518 m = map_mft_record(base_ni);
1523 ctx = ntfs_attr_get_search_ctx(base_ni, m);
1528 /* Find the index root attribute. */
1529 err = ntfs_attr_lookup(AT_INDEX_ROOT, ni->name, ni->name_len,
1530 CASE_SENSITIVE, 0, NULL, 0, ctx);
1531 if (unlikely(err)) {
1533 ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is "
1538 /* Set up the state. */
1539 if (unlikely(a->non_resident)) {
1540 ntfs_error(vol->sb, "$INDEX_ROOT attribute is not resident.");
1543 /* Ensure the attribute name is placed before the value. */
1544 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1545 le16_to_cpu(a->data.resident.value_offset)))) {
1546 ntfs_error(vol->sb, "$INDEX_ROOT attribute name is placed "
1547 "after the attribute value.");
1551 * Compressed/encrypted/sparse index root is not allowed, except for
1552 * directories of course but those are not dealt with here.
1554 if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_ENCRYPTED |
1556 ntfs_error(vi->i_sb, "Found compressed/encrypted/sparse index "
1560 ir = (INDEX_ROOT*)((u8*)a + le16_to_cpu(a->data.resident.value_offset));
1561 ir_end = (u8*)ir + le32_to_cpu(a->data.resident.value_length);
1562 if (ir_end > (u8*)ctx->mrec + vol->mft_record_size) {
1563 ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is corrupt.");
1566 index_end = (u8*)&ir->index + le32_to_cpu(ir->index.index_length);
1567 if (index_end > ir_end) {
1568 ntfs_error(vi->i_sb, "Index is corrupt.");
1572 ntfs_error(vi->i_sb, "Index type is not 0 (type is 0x%x).",
1573 le32_to_cpu(ir->type));
1576 ni->itype.index.collation_rule = ir->collation_rule;
1577 ntfs_debug("Index collation rule is 0x%x.",
1578 le32_to_cpu(ir->collation_rule));
1579 ni->itype.index.block_size = le32_to_cpu(ir->index_block_size);
1580 if (!is_power_of_2(ni->itype.index.block_size)) {
1581 ntfs_error(vi->i_sb, "Index block size (%u) is not a power of "
1582 "two.", ni->itype.index.block_size);
1585 if (ni->itype.index.block_size > PAGE_SIZE) {
1586 ntfs_error(vi->i_sb, "Index block size (%u) > PAGE_SIZE "
1587 "(%ld) is not supported. Sorry.",
1588 ni->itype.index.block_size, PAGE_SIZE);
1592 if (ni->itype.index.block_size < NTFS_BLOCK_SIZE) {
1593 ntfs_error(vi->i_sb, "Index block size (%u) < NTFS_BLOCK_SIZE "
1594 "(%i) is not supported. Sorry.",
1595 ni->itype.index.block_size, NTFS_BLOCK_SIZE);
1599 ni->itype.index.block_size_bits = ffs(ni->itype.index.block_size) - 1;
1600 /* Determine the size of a vcn in the index. */
1601 if (vol->cluster_size <= ni->itype.index.block_size) {
1602 ni->itype.index.vcn_size = vol->cluster_size;
1603 ni->itype.index.vcn_size_bits = vol->cluster_size_bits;
1605 ni->itype.index.vcn_size = vol->sector_size;
1606 ni->itype.index.vcn_size_bits = vol->sector_size_bits;
1608 /* Check for presence of index allocation attribute. */
1609 if (!(ir->index.flags & LARGE_INDEX)) {
1610 /* No index allocation. */
1611 vi->i_size = ni->initialized_size = ni->allocated_size = 0;
1612 /* We are done with the mft record, so we release it. */
1613 ntfs_attr_put_search_ctx(ctx);
1614 unmap_mft_record(base_ni);
1617 goto skip_large_index_stuff;
1618 } /* LARGE_INDEX: Index allocation present. Setup state. */
1619 NInoSetIndexAllocPresent(ni);
1620 /* Find index allocation attribute. */
1621 ntfs_attr_reinit_search_ctx(ctx);
1622 err = ntfs_attr_lookup(AT_INDEX_ALLOCATION, ni->name, ni->name_len,
1623 CASE_SENSITIVE, 0, NULL, 0, ctx);
1624 if (unlikely(err)) {
1626 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1627 "not present but $INDEX_ROOT "
1628 "indicated it is.");
1630 ntfs_error(vi->i_sb, "Failed to lookup "
1631 "$INDEX_ALLOCATION attribute.");
1635 if (!a->non_resident) {
1636 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1641 * Ensure the attribute name is placed before the mapping pairs array.
1643 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1645 a->data.non_resident.mapping_pairs_offset)))) {
1646 ntfs_error(vol->sb, "$INDEX_ALLOCATION attribute name is "
1647 "placed after the mapping pairs array.");
1650 if (a->flags & ATTR_IS_ENCRYPTED) {
1651 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1655 if (a->flags & ATTR_IS_SPARSE) {
1656 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is sparse.");
1659 if (a->flags & ATTR_COMPRESSION_MASK) {
1660 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1664 if (a->data.non_resident.lowest_vcn) {
1665 ntfs_error(vi->i_sb, "First extent of $INDEX_ALLOCATION "
1666 "attribute has non zero lowest_vcn.");
1669 vi->i_size = sle64_to_cpu(a->data.non_resident.data_size);
1670 ni->initialized_size = sle64_to_cpu(
1671 a->data.non_resident.initialized_size);
1672 ni->allocated_size = sle64_to_cpu(a->data.non_resident.allocated_size);
1674 * We are done with the mft record, so we release it. Otherwise
1675 * we would deadlock in ntfs_attr_iget().
1677 ntfs_attr_put_search_ctx(ctx);
1678 unmap_mft_record(base_ni);
1681 /* Get the index bitmap attribute inode. */
1682 bvi = ntfs_attr_iget(base_vi, AT_BITMAP, ni->name, ni->name_len);
1684 ntfs_error(vi->i_sb, "Failed to get bitmap attribute.");
1689 if (NInoCompressed(bni) || NInoEncrypted(bni) ||
1691 ntfs_error(vi->i_sb, "$BITMAP attribute is compressed and/or "
1692 "encrypted and/or sparse.");
1693 goto iput_unm_err_out;
1695 /* Consistency check bitmap size vs. index allocation size. */
1696 bvi_size = i_size_read(bvi);
1697 if ((bvi_size << 3) < (vi->i_size >> ni->itype.index.block_size_bits)) {
1698 ntfs_error(vi->i_sb, "Index bitmap too small (0x%llx) for "
1699 "index allocation (0x%llx).", bvi_size << 3,
1701 goto iput_unm_err_out;
1704 skip_large_index_stuff:
1705 /* Setup the operations for this index inode. */
1706 vi->i_mapping->a_ops = &ntfs_mst_aops;
1707 vi->i_blocks = ni->allocated_size >> 9;
1709 * Make sure the base inode doesn't go away and attach it to the
1713 ni->ext.base_ntfs_ino = base_ni;
1714 ni->nr_extents = -1;
1716 ntfs_debug("Done.");
1724 ntfs_attr_put_search_ctx(ctx);
1726 unmap_mft_record(base_ni);
1728 ntfs_error(vi->i_sb, "Failed with error code %i while reading index "
1729 "inode (mft_no 0x%lx, name_len %i.", err, vi->i_ino,
1732 if (err != -EOPNOTSUPP && err != -ENOMEM)
1738 * The MFT inode has special locking, so teach the lock validator
1739 * about this by splitting off the locking rules of the MFT from
1740 * the locking rules of other inodes. The MFT inode can never be
1741 * accessed from the VFS side (or even internally), only by the
1742 * map_mft functions.
1744 static struct lock_class_key mft_ni_runlist_lock_key, mft_ni_mrec_lock_key;
1747 * ntfs_read_inode_mount - special read_inode for mount time use only
1748 * @vi: inode to read
1750 * Read inode FILE_MFT at mount time, only called with super_block lock
1751 * held from within the read_super() code path.
1753 * This function exists because when it is called the page cache for $MFT/$DATA
1754 * is not initialized and hence we cannot get at the contents of mft records
1755 * by calling map_mft_record*().
1757 * Further it needs to cope with the circular references problem, i.e. cannot
1758 * load any attributes other than $ATTRIBUTE_LIST until $DATA is loaded, because
1759 * we do not know where the other extent mft records are yet and again, because
1760 * we cannot call map_mft_record*() yet. Obviously this applies only when an
1761 * attribute list is actually present in $MFT inode.
1763 * We solve these problems by starting with the $DATA attribute before anything
1764 * else and iterating using ntfs_attr_lookup($DATA) over all extents. As each
1765 * extent is found, we ntfs_mapping_pairs_decompress() including the implied
1766 * ntfs_runlists_merge(). Each step of the iteration necessarily provides
1767 * sufficient information for the next step to complete.
1769 * This should work but there are two possible pit falls (see inline comments
1770 * below), but only time will tell if they are real pits or just smoke...
1772 int ntfs_read_inode_mount(struct inode *vi)
1774 VCN next_vcn, last_vcn, highest_vcn;
1776 struct super_block *sb = vi->i_sb;
1777 ntfs_volume *vol = NTFS_SB(sb);
1778 struct buffer_head *bh;
1780 MFT_RECORD *m = NULL;
1782 ntfs_attr_search_ctx *ctx;
1783 unsigned int i, nr_blocks;
1786 ntfs_debug("Entering.");
1788 /* Initialize the ntfs specific part of @vi. */
1789 ntfs_init_big_inode(vi);
1793 /* Setup the data attribute. It is special as it is mst protected. */
1794 NInoSetNonResident(ni);
1795 NInoSetMstProtected(ni);
1796 NInoSetSparseDisabled(ni);
1801 * This sets up our little cheat allowing us to reuse the async read io
1802 * completion handler for directories.
1804 ni->itype.index.block_size = vol->mft_record_size;
1805 ni->itype.index.block_size_bits = vol->mft_record_size_bits;
1807 /* Very important! Needed to be able to call map_mft_record*(). */
1810 /* Allocate enough memory to read the first mft record. */
1811 if (vol->mft_record_size > 64 * 1024) {
1812 ntfs_error(sb, "Unsupported mft record size %i (max 64kiB).",
1813 vol->mft_record_size);
1816 i = vol->mft_record_size;
1817 if (i < sb->s_blocksize)
1818 i = sb->s_blocksize;
1819 m = (MFT_RECORD*)ntfs_malloc_nofs(i);
1821 ntfs_error(sb, "Failed to allocate buffer for $MFT record 0.");
1825 /* Determine the first block of the $MFT/$DATA attribute. */
1826 block = vol->mft_lcn << vol->cluster_size_bits >>
1827 sb->s_blocksize_bits;
1828 nr_blocks = vol->mft_record_size >> sb->s_blocksize_bits;
1832 /* Load $MFT/$DATA's first mft record. */
1833 for (i = 0; i < nr_blocks; i++) {
1834 bh = sb_bread(sb, block++);
1836 ntfs_error(sb, "Device read failed.");
1839 memcpy((char*)m + (i << sb->s_blocksize_bits), bh->b_data,
1844 if (le32_to_cpu(m->bytes_allocated) != vol->mft_record_size) {
1845 ntfs_error(sb, "Incorrect mft record size %u in superblock, should be %u.",
1846 le32_to_cpu(m->bytes_allocated), vol->mft_record_size);
1850 /* Apply the mst fixups. */
1851 if (post_read_mst_fixup((NTFS_RECORD*)m, vol->mft_record_size)) {
1852 /* FIXME: Try to use the $MFTMirr now. */
1853 ntfs_error(sb, "MST fixup failed. $MFT is corrupt.");
1857 /* Need this to sanity check attribute list references to $MFT. */
1858 vi->i_generation = ni->seq_no = le16_to_cpu(m->sequence_number);
1860 /* Provides readpage() for map_mft_record(). */
1861 vi->i_mapping->a_ops = &ntfs_mst_aops;
1863 ctx = ntfs_attr_get_search_ctx(ni, m);
1869 /* Find the attribute list attribute if present. */
1870 err = ntfs_attr_lookup(AT_ATTRIBUTE_LIST, NULL, 0, 0, 0, NULL, 0, ctx);
1872 if (unlikely(err != -ENOENT)) {
1873 ntfs_error(sb, "Failed to lookup attribute list "
1874 "attribute. You should run chkdsk.");
1877 } else /* if (!err) */ {
1878 ATTR_LIST_ENTRY *al_entry, *next_al_entry;
1880 static const char *es = " Not allowed. $MFT is corrupt. "
1881 "You should run chkdsk.";
1883 ntfs_debug("Attribute list attribute found in $MFT.");
1884 NInoSetAttrList(ni);
1886 if (a->flags & ATTR_COMPRESSION_MASK) {
1887 ntfs_error(sb, "Attribute list attribute is "
1888 "compressed.%s", es);
1891 if (a->flags & ATTR_IS_ENCRYPTED ||
1892 a->flags & ATTR_IS_SPARSE) {
1893 if (a->non_resident) {
1894 ntfs_error(sb, "Non-resident attribute list "
1895 "attribute is encrypted/"
1899 ntfs_warning(sb, "Resident attribute list attribute "
1900 "in $MFT system file is marked "
1901 "encrypted/sparse which is not true. "
1902 "However, Windows allows this and "
1903 "chkdsk does not detect or correct it "
1904 "so we will just ignore the invalid "
1905 "flags and pretend they are not set.");
1907 /* Now allocate memory for the attribute list. */
1908 ni->attr_list_size = (u32)ntfs_attr_size(a);
1909 ni->attr_list = ntfs_malloc_nofs(ni->attr_list_size);
1910 if (!ni->attr_list) {
1911 ntfs_error(sb, "Not enough memory to allocate buffer "
1912 "for attribute list.");
1915 if (a->non_resident) {
1916 NInoSetAttrListNonResident(ni);
1917 if (a->data.non_resident.lowest_vcn) {
1918 ntfs_error(sb, "Attribute list has non zero "
1919 "lowest_vcn. $MFT is corrupt. "
1920 "You should run chkdsk.");
1923 /* Setup the runlist. */
1924 ni->attr_list_rl.rl = ntfs_mapping_pairs_decompress(vol,
1926 if (IS_ERR(ni->attr_list_rl.rl)) {
1927 err = PTR_ERR(ni->attr_list_rl.rl);
1928 ni->attr_list_rl.rl = NULL;
1929 ntfs_error(sb, "Mapping pairs decompression "
1930 "failed with error code %i.",
1934 /* Now load the attribute list. */
1935 if ((err = load_attribute_list(vol, &ni->attr_list_rl,
1936 ni->attr_list, ni->attr_list_size,
1937 sle64_to_cpu(a->data.
1938 non_resident.initialized_size)))) {
1939 ntfs_error(sb, "Failed to load attribute list "
1940 "attribute with error code %i.",
1944 } else /* if (!ctx.attr->non_resident) */ {
1945 if ((u8*)a + le16_to_cpu(
1946 a->data.resident.value_offset) +
1948 a->data.resident.value_length) >
1949 (u8*)ctx->mrec + vol->mft_record_size) {
1950 ntfs_error(sb, "Corrupt attribute list "
1954 /* Now copy the attribute list. */
1955 memcpy(ni->attr_list, (u8*)a + le16_to_cpu(
1956 a->data.resident.value_offset),
1958 a->data.resident.value_length));
1960 /* The attribute list is now setup in memory. */
1962 * FIXME: I don't know if this case is actually possible.
1963 * According to logic it is not possible but I have seen too
1964 * many weird things in MS software to rely on logic... Thus we
1965 * perform a manual search and make sure the first $MFT/$DATA
1966 * extent is in the base inode. If it is not we abort with an
1967 * error and if we ever see a report of this error we will need
1968 * to do some magic in order to have the necessary mft record
1969 * loaded and in the right place in the page cache. But
1970 * hopefully logic will prevail and this never happens...
1972 al_entry = (ATTR_LIST_ENTRY*)ni->attr_list;
1973 al_end = (u8*)al_entry + ni->attr_list_size;
1974 for (;; al_entry = next_al_entry) {
1975 /* Out of bounds check. */
1976 if ((u8*)al_entry < ni->attr_list ||
1977 (u8*)al_entry > al_end)
1978 goto em_put_err_out;
1979 /* Catch the end of the attribute list. */
1980 if ((u8*)al_entry == al_end)
1981 goto em_put_err_out;
1982 if (!al_entry->length)
1983 goto em_put_err_out;
1984 if ((u8*)al_entry + 6 > al_end || (u8*)al_entry +
1985 le16_to_cpu(al_entry->length) > al_end)
1986 goto em_put_err_out;
1987 next_al_entry = (ATTR_LIST_ENTRY*)((u8*)al_entry +
1988 le16_to_cpu(al_entry->length));
1989 if (le32_to_cpu(al_entry->type) > le32_to_cpu(AT_DATA))
1990 goto em_put_err_out;
1991 if (AT_DATA != al_entry->type)
1993 /* We want an unnamed attribute. */
1994 if (al_entry->name_length)
1995 goto em_put_err_out;
1996 /* Want the first entry, i.e. lowest_vcn == 0. */
1997 if (al_entry->lowest_vcn)
1998 goto em_put_err_out;
1999 /* First entry has to be in the base mft record. */
2000 if (MREF_LE(al_entry->mft_reference) != vi->i_ino) {
2001 /* MFT references do not match, logic fails. */
2002 ntfs_error(sb, "BUG: The first $DATA extent "
2003 "of $MFT is not in the base "
2004 "mft record. Please report "
2005 "you saw this message to "
2006 "linux-ntfs-dev@lists."
2010 /* Sequence numbers must match. */
2011 if (MSEQNO_LE(al_entry->mft_reference) !=
2013 goto em_put_err_out;
2014 /* Got it. All is ok. We can stop now. */
2020 ntfs_attr_reinit_search_ctx(ctx);
2022 /* Now load all attribute extents. */
2024 next_vcn = last_vcn = highest_vcn = 0;
2025 while (!(err = ntfs_attr_lookup(AT_DATA, NULL, 0, 0, next_vcn, NULL, 0,
2027 runlist_element *nrl;
2029 /* Cache the current attribute. */
2031 /* $MFT must be non-resident. */
2032 if (!a->non_resident) {
2033 ntfs_error(sb, "$MFT must be non-resident but a "
2034 "resident extent was found. $MFT is "
2035 "corrupt. Run chkdsk.");
2038 /* $MFT must be uncompressed and unencrypted. */
2039 if (a->flags & ATTR_COMPRESSION_MASK ||
2040 a->flags & ATTR_IS_ENCRYPTED ||
2041 a->flags & ATTR_IS_SPARSE) {
2042 ntfs_error(sb, "$MFT must be uncompressed, "
2043 "non-sparse, and unencrypted but a "
2044 "compressed/sparse/encrypted extent "
2045 "was found. $MFT is corrupt. Run "
2050 * Decompress the mapping pairs array of this extent and merge
2051 * the result into the existing runlist. No need for locking
2052 * as we have exclusive access to the inode at this time and we
2053 * are a mount in progress task, too.
2055 nrl = ntfs_mapping_pairs_decompress(vol, a, ni->runlist.rl);
2057 ntfs_error(sb, "ntfs_mapping_pairs_decompress() "
2058 "failed with error code %ld. $MFT is "
2059 "corrupt.", PTR_ERR(nrl));
2062 ni->runlist.rl = nrl;
2064 /* Are we in the first extent? */
2066 if (a->data.non_resident.lowest_vcn) {
2067 ntfs_error(sb, "First extent of $DATA "
2068 "attribute has non zero "
2069 "lowest_vcn. $MFT is corrupt. "
2070 "You should run chkdsk.");
2073 /* Get the last vcn in the $DATA attribute. */
2074 last_vcn = sle64_to_cpu(
2075 a->data.non_resident.allocated_size)
2076 >> vol->cluster_size_bits;
2077 /* Fill in the inode size. */
2078 vi->i_size = sle64_to_cpu(
2079 a->data.non_resident.data_size);
2080 ni->initialized_size = sle64_to_cpu(
2081 a->data.non_resident.initialized_size);
2082 ni->allocated_size = sle64_to_cpu(
2083 a->data.non_resident.allocated_size);
2085 * Verify the number of mft records does not exceed
2088 if ((vi->i_size >> vol->mft_record_size_bits) >=
2090 ntfs_error(sb, "$MFT is too big! Aborting.");
2094 * We have got the first extent of the runlist for
2095 * $MFT which means it is now relatively safe to call
2096 * the normal ntfs_read_inode() function.
2097 * Complete reading the inode, this will actually
2098 * re-read the mft record for $MFT, this time entering
2099 * it into the page cache with which we complete the
2100 * kick start of the volume. It should be safe to do
2101 * this now as the first extent of $MFT/$DATA is
2102 * already known and we would hope that we don't need
2103 * further extents in order to find the other
2104 * attributes belonging to $MFT. Only time will tell if
2105 * this is really the case. If not we will have to play
2106 * magic at this point, possibly duplicating a lot of
2107 * ntfs_read_inode() at this point. We will need to
2108 * ensure we do enough of its work to be able to call
2109 * ntfs_read_inode() on extents of $MFT/$DATA. But lets
2110 * hope this never happens...
2112 ntfs_read_locked_inode(vi);
2113 if (is_bad_inode(vi)) {
2114 ntfs_error(sb, "ntfs_read_inode() of $MFT "
2115 "failed. BUG or corrupt $MFT. "
2116 "Run chkdsk and if no errors "
2117 "are found, please report you "
2118 "saw this message to "
2119 "linux-ntfs-dev@lists."
2121 ntfs_attr_put_search_ctx(ctx);
2122 /* Revert to the safe super operations. */
2127 * Re-initialize some specifics about $MFT's inode as
2128 * ntfs_read_inode() will have set up the default ones.
2130 /* Set uid and gid to root. */
2131 vi->i_uid = GLOBAL_ROOT_UID;
2132 vi->i_gid = GLOBAL_ROOT_GID;
2133 /* Regular file. No access for anyone. */
2134 vi->i_mode = S_IFREG;
2135 /* No VFS initiated operations allowed for $MFT. */
2136 vi->i_op = &ntfs_empty_inode_ops;
2137 vi->i_fop = &ntfs_empty_file_ops;
2140 /* Get the lowest vcn for the next extent. */
2141 highest_vcn = sle64_to_cpu(a->data.non_resident.highest_vcn);
2142 next_vcn = highest_vcn + 1;
2144 /* Only one extent or error, which we catch below. */
2148 /* Avoid endless loops due to corruption. */
2149 if (next_vcn < sle64_to_cpu(
2150 a->data.non_resident.lowest_vcn)) {
2151 ntfs_error(sb, "$MFT has corrupt attribute list "
2152 "attribute. Run chkdsk.");
2156 if (err != -ENOENT) {
2157 ntfs_error(sb, "Failed to lookup $MFT/$DATA attribute extent. "
2158 "$MFT is corrupt. Run chkdsk.");
2162 ntfs_error(sb, "$MFT/$DATA attribute not found. $MFT is "
2163 "corrupt. Run chkdsk.");
2166 if (highest_vcn && highest_vcn != last_vcn - 1) {
2167 ntfs_error(sb, "Failed to load the complete runlist for "
2168 "$MFT/$DATA. Driver bug or corrupt $MFT. "
2170 ntfs_debug("highest_vcn = 0x%llx, last_vcn - 1 = 0x%llx",
2171 (unsigned long long)highest_vcn,
2172 (unsigned long long)last_vcn - 1);
2175 ntfs_attr_put_search_ctx(ctx);
2176 ntfs_debug("Done.");
2180 * Split the locking rules of the MFT inode from the
2181 * locking rules of other inodes:
2183 lockdep_set_class(&ni->runlist.lock, &mft_ni_runlist_lock_key);
2184 lockdep_set_class(&ni->mrec_lock, &mft_ni_mrec_lock_key);
2189 ntfs_error(sb, "Couldn't find first extent of $DATA attribute in "
2190 "attribute list. $MFT is corrupt. Run chkdsk.");
2192 ntfs_attr_put_search_ctx(ctx);
2194 ntfs_error(sb, "Failed. Marking inode as bad.");
2200 static void __ntfs_clear_inode(ntfs_inode *ni)
2202 /* Free all alocated memory. */
2203 down_write(&ni->runlist.lock);
2204 if (ni->runlist.rl) {
2205 ntfs_free(ni->runlist.rl);
2206 ni->runlist.rl = NULL;
2208 up_write(&ni->runlist.lock);
2210 if (ni->attr_list) {
2211 ntfs_free(ni->attr_list);
2212 ni->attr_list = NULL;
2215 down_write(&ni->attr_list_rl.lock);
2216 if (ni->attr_list_rl.rl) {
2217 ntfs_free(ni->attr_list_rl.rl);
2218 ni->attr_list_rl.rl = NULL;
2220 up_write(&ni->attr_list_rl.lock);
2222 if (ni->name_len && ni->name != I30) {
2229 void ntfs_clear_extent_inode(ntfs_inode *ni)
2231 ntfs_debug("Entering for inode 0x%lx.", ni->mft_no);
2233 BUG_ON(NInoAttr(ni));
2234 BUG_ON(ni->nr_extents != -1);
2237 if (NInoDirty(ni)) {
2238 if (!is_bad_inode(VFS_I(ni->ext.base_ntfs_ino)))
2239 ntfs_error(ni->vol->sb, "Clearing dirty extent inode! "
2240 "Losing data! This is a BUG!!!");
2241 // FIXME: Do something!!!
2243 #endif /* NTFS_RW */
2245 __ntfs_clear_inode(ni);
2248 ntfs_destroy_extent_inode(ni);
2252 * ntfs_evict_big_inode - clean up the ntfs specific part of an inode
2253 * @vi: vfs inode pending annihilation
2255 * When the VFS is going to remove an inode from memory, ntfs_clear_big_inode()
2256 * is called, which deallocates all memory belonging to the NTFS specific part
2257 * of the inode and returns.
2259 * If the MFT record is dirty, we commit it before doing anything else.
2261 void ntfs_evict_big_inode(struct inode *vi)
2263 ntfs_inode *ni = NTFS_I(vi);
2265 truncate_inode_pages_final(&vi->i_data);
2269 if (NInoDirty(ni)) {
2270 bool was_bad = (is_bad_inode(vi));
2272 /* Committing the inode also commits all extent inodes. */
2273 ntfs_commit_inode(vi);
2275 if (!was_bad && (is_bad_inode(vi) || NInoDirty(ni))) {
2276 ntfs_error(vi->i_sb, "Failed to commit dirty inode "
2277 "0x%lx. Losing data!", vi->i_ino);
2278 // FIXME: Do something!!!
2281 #endif /* NTFS_RW */
2283 /* No need to lock at this stage as no one else has a reference. */
2284 if (ni->nr_extents > 0) {
2287 for (i = 0; i < ni->nr_extents; i++)
2288 ntfs_clear_extent_inode(ni->ext.extent_ntfs_inos[i]);
2289 kfree(ni->ext.extent_ntfs_inos);
2292 __ntfs_clear_inode(ni);
2295 /* Release the base inode if we are holding it. */
2296 if (ni->nr_extents == -1) {
2297 iput(VFS_I(ni->ext.base_ntfs_ino));
2299 ni->ext.base_ntfs_ino = NULL;
2306 * ntfs_show_options - show mount options in /proc/mounts
2307 * @sf: seq_file in which to write our mount options
2308 * @root: root of the mounted tree whose mount options to display
2310 * Called by the VFS once for each mounted ntfs volume when someone reads
2311 * /proc/mounts in order to display the NTFS specific mount options of each
2312 * mount. The mount options of fs specified by @root are written to the seq file
2313 * @sf and success is returned.
2315 int ntfs_show_options(struct seq_file *sf, struct dentry *root)
2317 ntfs_volume *vol = NTFS_SB(root->d_sb);
2320 seq_printf(sf, ",uid=%i", from_kuid_munged(&init_user_ns, vol->uid));
2321 seq_printf(sf, ",gid=%i", from_kgid_munged(&init_user_ns, vol->gid));
2322 if (vol->fmask == vol->dmask)
2323 seq_printf(sf, ",umask=0%o", vol->fmask);
2325 seq_printf(sf, ",fmask=0%o", vol->fmask);
2326 seq_printf(sf, ",dmask=0%o", vol->dmask);
2328 seq_printf(sf, ",nls=%s", vol->nls_map->charset);
2329 if (NVolCaseSensitive(vol))
2330 seq_printf(sf, ",case_sensitive");
2331 if (NVolShowSystemFiles(vol))
2332 seq_printf(sf, ",show_sys_files");
2333 if (!NVolSparseEnabled(vol))
2334 seq_printf(sf, ",disable_sparse");
2335 for (i = 0; on_errors_arr[i].val; i++) {
2336 if (on_errors_arr[i].val & vol->on_errors)
2337 seq_printf(sf, ",errors=%s", on_errors_arr[i].str);
2339 seq_printf(sf, ",mft_zone_multiplier=%i", vol->mft_zone_multiplier);
2345 static const char *es = " Leaving inconsistent metadata. Unmount and run "
2349 * ntfs_truncate - called when the i_size of an ntfs inode is changed
2350 * @vi: inode for which the i_size was changed
2352 * We only support i_size changes for normal files at present, i.e. not
2353 * compressed and not encrypted. This is enforced in ntfs_setattr(), see
2356 * The kernel guarantees that @vi is a regular file (S_ISREG() is true) and
2357 * that the change is allowed.
2359 * This implies for us that @vi is a file inode rather than a directory, index,
2360 * or attribute inode as well as that @vi is a base inode.
2362 * Returns 0 on success or -errno on error.
2364 * Called with ->i_mutex held.
2366 int ntfs_truncate(struct inode *vi)
2368 s64 new_size, old_size, nr_freed, new_alloc_size, old_alloc_size;
2370 unsigned long flags;
2371 ntfs_inode *base_ni, *ni = NTFS_I(vi);
2372 ntfs_volume *vol = ni->vol;
2373 ntfs_attr_search_ctx *ctx;
2376 const char *te = " Leaving file length out of sync with i_size.";
2377 int err, mp_size, size_change, alloc_change;
2380 ntfs_debug("Entering for inode 0x%lx.", vi->i_ino);
2381 BUG_ON(NInoAttr(ni));
2382 BUG_ON(S_ISDIR(vi->i_mode));
2383 BUG_ON(NInoMstProtected(ni));
2384 BUG_ON(ni->nr_extents < 0);
2387 * Lock the runlist for writing and map the mft record to ensure it is
2388 * safe to mess with the attribute runlist and sizes.
2390 down_write(&ni->runlist.lock);
2394 base_ni = ni->ext.base_ntfs_ino;
2395 m = map_mft_record(base_ni);
2398 ntfs_error(vi->i_sb, "Failed to map mft record for inode 0x%lx "
2399 "(error code %d).%s", vi->i_ino, err, te);
2404 ctx = ntfs_attr_get_search_ctx(base_ni, m);
2405 if (unlikely(!ctx)) {
2406 ntfs_error(vi->i_sb, "Failed to allocate a search context for "
2407 "inode 0x%lx (not enough memory).%s",
2412 err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
2413 CASE_SENSITIVE, 0, NULL, 0, ctx);
2414 if (unlikely(err)) {
2415 if (err == -ENOENT) {
2416 ntfs_error(vi->i_sb, "Open attribute is missing from "
2417 "mft record. Inode 0x%lx is corrupt. "
2418 "Run chkdsk.%s", vi->i_ino, te);
2421 ntfs_error(vi->i_sb, "Failed to lookup attribute in "
2422 "inode 0x%lx (error code %d).%s",
2423 vi->i_ino, err, te);
2429 * The i_size of the vfs inode is the new size for the attribute value.
2431 new_size = i_size_read(vi);
2432 /* The current size of the attribute value is the old size. */
2433 old_size = ntfs_attr_size(a);
2434 /* Calculate the new allocated size. */
2435 if (NInoNonResident(ni))
2436 new_alloc_size = (new_size + vol->cluster_size - 1) &
2437 ~(s64)vol->cluster_size_mask;
2439 new_alloc_size = (new_size + 7) & ~7;
2440 /* The current allocated size is the old allocated size. */
2441 read_lock_irqsave(&ni->size_lock, flags);
2442 old_alloc_size = ni->allocated_size;
2443 read_unlock_irqrestore(&ni->size_lock, flags);
2445 * The change in the file size. This will be 0 if no change, >0 if the
2446 * size is growing, and <0 if the size is shrinking.
2449 if (new_size - old_size >= 0) {
2451 if (new_size == old_size)
2454 /* As above for the allocated size. */
2456 if (new_alloc_size - old_alloc_size >= 0) {
2458 if (new_alloc_size == old_alloc_size)
2462 * If neither the size nor the allocation are being changed there is
2465 if (!size_change && !alloc_change)
2467 /* If the size is changing, check if new size is allowed in $AttrDef. */
2469 err = ntfs_attr_size_bounds_check(vol, ni->type, new_size);
2470 if (unlikely(err)) {
2471 if (err == -ERANGE) {
2472 ntfs_error(vol->sb, "Truncate would cause the "
2473 "inode 0x%lx to %simum size "
2474 "for its attribute type "
2475 "(0x%x). Aborting truncate.",
2477 new_size > old_size ? "exceed "
2478 "the max" : "go under the min",
2479 le32_to_cpu(ni->type));
2482 ntfs_error(vol->sb, "Inode 0x%lx has unknown "
2483 "attribute type 0x%x. "
2484 "Aborting truncate.",
2486 le32_to_cpu(ni->type));
2489 /* Reset the vfs inode size to the old size. */
2490 i_size_write(vi, old_size);
2494 if (NInoCompressed(ni) || NInoEncrypted(ni)) {
2495 ntfs_warning(vi->i_sb, "Changes in inode size are not "
2496 "supported yet for %s files, ignoring.",
2497 NInoCompressed(ni) ? "compressed" :
2502 if (a->non_resident)
2503 goto do_non_resident_truncate;
2504 BUG_ON(NInoNonResident(ni));
2505 /* Resize the attribute record to best fit the new attribute size. */
2506 if (new_size < vol->mft_record_size &&
2507 !ntfs_resident_attr_value_resize(m, a, new_size)) {
2508 /* The resize succeeded! */
2509 flush_dcache_mft_record_page(ctx->ntfs_ino);
2510 mark_mft_record_dirty(ctx->ntfs_ino);
2511 write_lock_irqsave(&ni->size_lock, flags);
2512 /* Update the sizes in the ntfs inode and all is done. */
2513 ni->allocated_size = le32_to_cpu(a->length) -
2514 le16_to_cpu(a->data.resident.value_offset);
2516 * Note ntfs_resident_attr_value_resize() has already done any
2517 * necessary data clearing in the attribute record. When the
2518 * file is being shrunk vmtruncate() will already have cleared
2519 * the top part of the last partial page, i.e. since this is
2520 * the resident case this is the page with index 0. However,
2521 * when the file is being expanded, the page cache page data
2522 * between the old data_size, i.e. old_size, and the new_size
2523 * has not been zeroed. Fortunately, we do not need to zero it
2524 * either since on one hand it will either already be zero due
2525 * to both readpage and writepage clearing partial page data
2526 * beyond i_size in which case there is nothing to do or in the
2527 * case of the file being mmap()ped at the same time, POSIX
2528 * specifies that the behaviour is unspecified thus we do not
2529 * have to do anything. This means that in our implementation
2530 * in the rare case that the file is mmap()ped and a write
2531 * occurred into the mmap()ped region just beyond the file size
2532 * and writepage has not yet been called to write out the page
2533 * (which would clear the area beyond the file size) and we now
2534 * extend the file size to incorporate this dirty region
2535 * outside the file size, a write of the page would result in
2536 * this data being written to disk instead of being cleared.
2537 * Given both POSIX and the Linux mmap(2) man page specify that
2538 * this corner case is undefined, we choose to leave it like
2539 * that as this is much simpler for us as we cannot lock the
2540 * relevant page now since we are holding too many ntfs locks
2541 * which would result in a lock reversal deadlock.
2543 ni->initialized_size = new_size;
2544 write_unlock_irqrestore(&ni->size_lock, flags);
2547 /* If the above resize failed, this must be an attribute extension. */
2548 BUG_ON(size_change < 0);
2550 * We have to drop all the locks so we can call
2551 * ntfs_attr_make_non_resident(). This could be optimised by try-
2552 * locking the first page cache page and only if that fails dropping
2553 * the locks, locking the page, and redoing all the locking and
2554 * lookups. While this would be a huge optimisation, it is not worth
2555 * it as this is definitely a slow code path as it only ever can happen
2556 * once for any given file.
2558 ntfs_attr_put_search_ctx(ctx);
2559 unmap_mft_record(base_ni);
2560 up_write(&ni->runlist.lock);
2562 * Not enough space in the mft record, try to make the attribute
2563 * non-resident and if successful restart the truncation process.
2565 err = ntfs_attr_make_non_resident(ni, old_size);
2567 goto retry_truncate;
2569 * Could not make non-resident. If this is due to this not being
2570 * permitted for this attribute type or there not being enough space,
2571 * try to make other attributes non-resident. Otherwise fail.
2573 if (unlikely(err != -EPERM && err != -ENOSPC)) {
2574 ntfs_error(vol->sb, "Cannot truncate inode 0x%lx, attribute "
2575 "type 0x%x, because the conversion from "
2576 "resident to non-resident attribute failed "
2577 "with error code %i.", vi->i_ino,
2578 (unsigned)le32_to_cpu(ni->type), err);
2583 /* TODO: Not implemented from here, abort. */
2585 ntfs_error(vol->sb, "Not enough space in the mft record/on "
2586 "disk for the non-resident attribute value. "
2587 "This case is not implemented yet.");
2588 else /* if (err == -EPERM) */
2589 ntfs_error(vol->sb, "This attribute type may not be "
2590 "non-resident. This case is not implemented "
2595 // TODO: Attempt to make other attributes non-resident.
2597 goto do_resident_extend;
2599 * Both the attribute list attribute and the standard information
2600 * attribute must remain in the base inode. Thus, if this is one of
2601 * these attributes, we have to try to move other attributes out into
2602 * extent mft records instead.
2604 if (ni->type == AT_ATTRIBUTE_LIST ||
2605 ni->type == AT_STANDARD_INFORMATION) {
2606 // TODO: Attempt to move other attributes into extent mft
2610 goto do_resident_extend;
2613 // TODO: Attempt to move this attribute to an extent mft record, but
2614 // only if it is not already the only attribute in an mft record in
2615 // which case there would be nothing to gain.
2618 goto do_resident_extend;
2619 /* There is nothing we can do to make enough space. )-: */
2622 do_non_resident_truncate:
2623 BUG_ON(!NInoNonResident(ni));
2624 if (alloc_change < 0) {
2625 highest_vcn = sle64_to_cpu(a->data.non_resident.highest_vcn);
2626 if (highest_vcn > 0 &&
2627 old_alloc_size >> vol->cluster_size_bits >
2630 * This attribute has multiple extents. Not yet
2633 ntfs_error(vol->sb, "Cannot truncate inode 0x%lx, "
2634 "attribute type 0x%x, because the "
2635 "attribute is highly fragmented (it "
2636 "consists of multiple extents) and "
2637 "this case is not implemented yet.",
2639 (unsigned)le32_to_cpu(ni->type));
2645 * If the size is shrinking, need to reduce the initialized_size and
2646 * the data_size before reducing the allocation.
2648 if (size_change < 0) {
2650 * Make the valid size smaller (i_size is already up-to-date).
2652 write_lock_irqsave(&ni->size_lock, flags);
2653 if (new_size < ni->initialized_size) {
2654 ni->initialized_size = new_size;
2655 a->data.non_resident.initialized_size =
2656 cpu_to_sle64(new_size);
2658 a->data.non_resident.data_size = cpu_to_sle64(new_size);
2659 write_unlock_irqrestore(&ni->size_lock, flags);
2660 flush_dcache_mft_record_page(ctx->ntfs_ino);
2661 mark_mft_record_dirty(ctx->ntfs_ino);
2662 /* If the allocated size is not changing, we are done. */
2666 * If the size is shrinking it makes no sense for the
2667 * allocation to be growing.
2669 BUG_ON(alloc_change > 0);
2670 } else /* if (size_change >= 0) */ {
2672 * The file size is growing or staying the same but the
2673 * allocation can be shrinking, growing or staying the same.
2675 if (alloc_change > 0) {
2677 * We need to extend the allocation and possibly update
2678 * the data size. If we are updating the data size,
2679 * since we are not touching the initialized_size we do
2680 * not need to worry about the actual data on disk.
2681 * And as far as the page cache is concerned, there
2682 * will be no pages beyond the old data size and any
2683 * partial region in the last page between the old and
2684 * new data size (or the end of the page if the new
2685 * data size is outside the page) does not need to be
2686 * modified as explained above for the resident
2687 * attribute truncate case. To do this, we simply drop
2688 * the locks we hold and leave all the work to our
2689 * friendly helper ntfs_attr_extend_allocation().
2691 ntfs_attr_put_search_ctx(ctx);
2692 unmap_mft_record(base_ni);
2693 up_write(&ni->runlist.lock);
2694 err = ntfs_attr_extend_allocation(ni, new_size,
2695 size_change > 0 ? new_size : -1, -1);
2697 * ntfs_attr_extend_allocation() will have done error
2705 /* alloc_change < 0 */
2706 /* Free the clusters. */
2707 nr_freed = ntfs_cluster_free(ni, new_alloc_size >>
2708 vol->cluster_size_bits, -1, ctx);
2711 if (unlikely(nr_freed < 0)) {
2712 ntfs_error(vol->sb, "Failed to release cluster(s) (error code "
2713 "%lli). Unmount and run chkdsk to recover "
2714 "the lost cluster(s).", (long long)nr_freed);
2718 /* Truncate the runlist. */
2719 err = ntfs_rl_truncate_nolock(vol, &ni->runlist,
2720 new_alloc_size >> vol->cluster_size_bits);
2722 * If the runlist truncation failed and/or the search context is no
2723 * longer valid, we cannot resize the attribute record or build the
2724 * mapping pairs array thus we mark the inode bad so that no access to
2725 * the freed clusters can happen.
2727 if (unlikely(err || IS_ERR(m))) {
2728 ntfs_error(vol->sb, "Failed to %s (error code %li).%s",
2730 "restore attribute search context" :
2731 "truncate attribute runlist",
2732 IS_ERR(m) ? PTR_ERR(m) : err, es);
2736 /* Get the size for the shrunk mapping pairs array for the runlist. */
2737 mp_size = ntfs_get_size_for_mapping_pairs(vol, ni->runlist.rl, 0, -1);
2738 if (unlikely(mp_size <= 0)) {
2739 ntfs_error(vol->sb, "Cannot shrink allocation of inode 0x%lx, "
2740 "attribute type 0x%x, because determining the "
2741 "size for the mapping pairs failed with error "
2742 "code %i.%s", vi->i_ino,
2743 (unsigned)le32_to_cpu(ni->type), mp_size, es);
2748 * Shrink the attribute record for the new mapping pairs array. Note,
2749 * this cannot fail since we are making the attribute smaller thus by
2750 * definition there is enough space to do so.
2752 attr_len = le32_to_cpu(a->length);
2753 err = ntfs_attr_record_resize(m, a, mp_size +
2754 le16_to_cpu(a->data.non_resident.mapping_pairs_offset));
2757 * Generate the mapping pairs array directly into the attribute record.
2759 err = ntfs_mapping_pairs_build(vol, (u8*)a +
2760 le16_to_cpu(a->data.non_resident.mapping_pairs_offset),
2761 mp_size, ni->runlist.rl, 0, -1, NULL);
2762 if (unlikely(err)) {
2763 ntfs_error(vol->sb, "Cannot shrink allocation of inode 0x%lx, "
2764 "attribute type 0x%x, because building the "
2765 "mapping pairs failed with error code %i.%s",
2766 vi->i_ino, (unsigned)le32_to_cpu(ni->type),
2771 /* Update the allocated/compressed size as well as the highest vcn. */
2772 a->data.non_resident.highest_vcn = cpu_to_sle64((new_alloc_size >>
2773 vol->cluster_size_bits) - 1);
2774 write_lock_irqsave(&ni->size_lock, flags);
2775 ni->allocated_size = new_alloc_size;
2776 a->data.non_resident.allocated_size = cpu_to_sle64(new_alloc_size);
2777 if (NInoSparse(ni) || NInoCompressed(ni)) {
2779 ni->itype.compressed.size -= nr_freed <<
2780 vol->cluster_size_bits;
2781 BUG_ON(ni->itype.compressed.size < 0);
2782 a->data.non_resident.compressed_size = cpu_to_sle64(
2783 ni->itype.compressed.size);
2784 vi->i_blocks = ni->itype.compressed.size >> 9;
2787 vi->i_blocks = new_alloc_size >> 9;
2788 write_unlock_irqrestore(&ni->size_lock, flags);
2790 * We have shrunk the allocation. If this is a shrinking truncate we
2791 * have already dealt with the initialized_size and the data_size above
2792 * and we are done. If the truncate is only changing the allocation
2793 * and not the data_size, we are also done. If this is an extending
2794 * truncate, need to extend the data_size now which is ensured by the
2795 * fact that @size_change is positive.
2799 * If the size is growing, need to update it now. If it is shrinking,
2800 * we have already updated it above (before the allocation change).
2802 if (size_change > 0)
2803 a->data.non_resident.data_size = cpu_to_sle64(new_size);
2804 /* Ensure the modified mft record is written out. */
2805 flush_dcache_mft_record_page(ctx->ntfs_ino);
2806 mark_mft_record_dirty(ctx->ntfs_ino);
2808 ntfs_attr_put_search_ctx(ctx);
2809 unmap_mft_record(base_ni);
2810 up_write(&ni->runlist.lock);
2812 /* Update the mtime and ctime on the base inode. */
2813 /* normally ->truncate shouldn't update ctime or mtime,
2814 * but ntfs did before so it got a copy & paste version
2815 * of file_update_time. one day someone should fix this
2818 if (!IS_NOCMTIME(VFS_I(base_ni)) && !IS_RDONLY(VFS_I(base_ni))) {
2819 struct timespec64 now = current_time(VFS_I(base_ni));
2822 if (!timespec64_equal(&VFS_I(base_ni)->i_mtime, &now) ||
2823 !timespec64_equal(&VFS_I(base_ni)->i_ctime, &now))
2825 VFS_I(base_ni)->i_mtime = now;
2826 VFS_I(base_ni)->i_ctime = now;
2829 mark_inode_dirty_sync(VFS_I(base_ni));
2833 NInoClearTruncateFailed(ni);
2834 ntfs_debug("Done.");
2840 if (err != -ENOMEM && err != -EOPNOTSUPP)
2842 if (err != -EOPNOTSUPP)
2843 NInoSetTruncateFailed(ni);
2844 else if (old_size >= 0)
2845 i_size_write(vi, old_size);
2848 ntfs_attr_put_search_ctx(ctx);
2850 unmap_mft_record(base_ni);
2851 up_write(&ni->runlist.lock);
2853 ntfs_debug("Failed. Returning error code %i.", err);
2856 if (err != -ENOMEM && err != -EOPNOTSUPP)
2858 if (err != -EOPNOTSUPP)
2859 NInoSetTruncateFailed(ni);
2861 i_size_write(vi, old_size);
2866 * ntfs_truncate_vfs - wrapper for ntfs_truncate() that has no return value
2867 * @vi: inode for which the i_size was changed
2869 * Wrapper for ntfs_truncate() that has no return value.
2871 * See ntfs_truncate() description above for details.
2874 void ntfs_truncate_vfs(struct inode *vi) {
2880 * ntfs_setattr - called from notify_change() when an attribute is being changed
2881 * @dentry: dentry whose attributes to change
2882 * @attr: structure describing the attributes and the changes
2884 * We have to trap VFS attempts to truncate the file described by @dentry as
2885 * soon as possible, because we do not implement changes in i_size yet. So we
2886 * abort all i_size changes here.
2888 * We also abort all changes of user, group, and mode as we do not implement
2889 * the NTFS ACLs yet.
2891 * Called with ->i_mutex held.
2893 int ntfs_setattr(struct dentry *dentry, struct iattr *attr)
2895 struct inode *vi = d_inode(dentry);
2897 unsigned int ia_valid = attr->ia_valid;
2899 err = setattr_prepare(dentry, attr);
2902 /* We do not support NTFS ACLs yet. */
2903 if (ia_valid & (ATTR_UID | ATTR_GID | ATTR_MODE)) {
2904 ntfs_warning(vi->i_sb, "Changes in user/group/mode are not "
2905 "supported yet, ignoring.");
2909 if (ia_valid & ATTR_SIZE) {
2910 if (attr->ia_size != i_size_read(vi)) {
2911 ntfs_inode *ni = NTFS_I(vi);
2913 * FIXME: For now we do not support resizing of
2914 * compressed or encrypted files yet.
2916 if (NInoCompressed(ni) || NInoEncrypted(ni)) {
2917 ntfs_warning(vi->i_sb, "Changes in inode size "
2918 "are not supported yet for "
2919 "%s files, ignoring.",
2920 NInoCompressed(ni) ?
2921 "compressed" : "encrypted");
2924 truncate_setsize(vi, attr->ia_size);
2925 ntfs_truncate_vfs(vi);
2927 if (err || ia_valid == ATTR_SIZE)
2931 * We skipped the truncate but must still update
2934 ia_valid |= ATTR_MTIME | ATTR_CTIME;
2937 if (ia_valid & ATTR_ATIME)
2938 vi->i_atime = timespec64_trunc(attr->ia_atime,
2939 vi->i_sb->s_time_gran);
2940 if (ia_valid & ATTR_MTIME)
2941 vi->i_mtime = timespec64_trunc(attr->ia_mtime,
2942 vi->i_sb->s_time_gran);
2943 if (ia_valid & ATTR_CTIME)
2944 vi->i_ctime = timespec64_trunc(attr->ia_ctime,
2945 vi->i_sb->s_time_gran);
2946 mark_inode_dirty(vi);
2952 * ntfs_write_inode - write out a dirty inode
2953 * @vi: inode to write out
2954 * @sync: if true, write out synchronously
2956 * Write out a dirty inode to disk including any extent inodes if present.
2958 * If @sync is true, commit the inode to disk and wait for io completion. This
2959 * is done using write_mft_record().
2961 * If @sync is false, just schedule the write to happen but do not wait for i/o
2962 * completion. In 2.6 kernels, scheduling usually happens just by virtue of
2963 * marking the page (and in this case mft record) dirty but we do not implement
2964 * this yet as write_mft_record() largely ignores the @sync parameter and
2965 * always performs synchronous writes.
2967 * Return 0 on success and -errno on error.
2969 int __ntfs_write_inode(struct inode *vi, int sync)
2972 ntfs_inode *ni = NTFS_I(vi);
2973 ntfs_attr_search_ctx *ctx;
2975 STANDARD_INFORMATION *si;
2977 bool modified = false;
2979 ntfs_debug("Entering for %sinode 0x%lx.", NInoAttr(ni) ? "attr " : "",
2982 * Dirty attribute inodes are written via their real inodes so just
2983 * clean them here. Access time updates are taken care off when the
2984 * real inode is written.
2988 ntfs_debug("Done.");
2991 /* Map, pin, and lock the mft record belonging to the inode. */
2992 m = map_mft_record(ni);
2997 /* Update the access times in the standard information attribute. */
2998 ctx = ntfs_attr_get_search_ctx(ni, m);
2999 if (unlikely(!ctx)) {
3003 err = ntfs_attr_lookup(AT_STANDARD_INFORMATION, NULL, 0,
3004 CASE_SENSITIVE, 0, NULL, 0, ctx);
3005 if (unlikely(err)) {
3006 ntfs_attr_put_search_ctx(ctx);
3009 si = (STANDARD_INFORMATION*)((u8*)ctx->attr +
3010 le16_to_cpu(ctx->attr->data.resident.value_offset));
3011 /* Update the access times if they have changed. */
3012 nt = utc2ntfs(vi->i_mtime);
3013 if (si->last_data_change_time != nt) {
3014 ntfs_debug("Updating mtime for inode 0x%lx: old = 0x%llx, "
3015 "new = 0x%llx", vi->i_ino, (long long)
3016 sle64_to_cpu(si->last_data_change_time),
3017 (long long)sle64_to_cpu(nt));
3018 si->last_data_change_time = nt;
3021 nt = utc2ntfs(vi->i_ctime);
3022 if (si->last_mft_change_time != nt) {
3023 ntfs_debug("Updating ctime for inode 0x%lx: old = 0x%llx, "
3024 "new = 0x%llx", vi->i_ino, (long long)
3025 sle64_to_cpu(si->last_mft_change_time),
3026 (long long)sle64_to_cpu(nt));
3027 si->last_mft_change_time = nt;
3030 nt = utc2ntfs(vi->i_atime);
3031 if (si->last_access_time != nt) {
3032 ntfs_debug("Updating atime for inode 0x%lx: old = 0x%llx, "
3033 "new = 0x%llx", vi->i_ino,
3034 (long long)sle64_to_cpu(si->last_access_time),
3035 (long long)sle64_to_cpu(nt));
3036 si->last_access_time = nt;
3040 * If we just modified the standard information attribute we need to
3041 * mark the mft record it is in dirty. We do this manually so that
3042 * mark_inode_dirty() is not called which would redirty the inode and
3043 * hence result in an infinite loop of trying to write the inode.
3044 * There is no need to mark the base inode nor the base mft record
3045 * dirty, since we are going to write this mft record below in any case
3046 * and the base mft record may actually not have been modified so it
3047 * might not need to be written out.
3048 * NOTE: It is not a problem when the inode for $MFT itself is being
3049 * written out as mark_ntfs_record_dirty() will only set I_DIRTY_PAGES
3050 * on the $MFT inode and hence ntfs_write_inode() will not be
3051 * re-invoked because of it which in turn is ok since the dirtied mft
3052 * record will be cleaned and written out to disk below, i.e. before
3053 * this function returns.
3056 flush_dcache_mft_record_page(ctx->ntfs_ino);
3057 if (!NInoTestSetDirty(ctx->ntfs_ino))
3058 mark_ntfs_record_dirty(ctx->ntfs_ino->page,
3059 ctx->ntfs_ino->page_ofs);
3061 ntfs_attr_put_search_ctx(ctx);
3062 /* Now the access times are updated, write the base mft record. */
3064 err = write_mft_record(ni, m, sync);
3065 /* Write all attached extent mft records. */
3066 mutex_lock(&ni->extent_lock);
3067 if (ni->nr_extents > 0) {
3068 ntfs_inode **extent_nis = ni->ext.extent_ntfs_inos;
3071 ntfs_debug("Writing %i extent inodes.", ni->nr_extents);
3072 for (i = 0; i < ni->nr_extents; i++) {
3073 ntfs_inode *tni = extent_nis[i];
3075 if (NInoDirty(tni)) {
3076 MFT_RECORD *tm = map_mft_record(tni);
3080 if (!err || err == -ENOMEM)
3084 ret = write_mft_record(tni, tm, sync);
3085 unmap_mft_record(tni);
3086 if (unlikely(ret)) {
3087 if (!err || err == -ENOMEM)
3093 mutex_unlock(&ni->extent_lock);
3094 unmap_mft_record(ni);
3097 ntfs_debug("Done.");
3100 unmap_mft_record(ni);
3102 if (err == -ENOMEM) {
3103 ntfs_warning(vi->i_sb, "Not enough memory to write inode. "
3104 "Marking the inode dirty again, so the VFS "
3106 mark_inode_dirty(vi);
3108 ntfs_error(vi->i_sb, "Failed (error %i): Run chkdsk.", -err);
3109 NVolSetErrors(ni->vol);
3114 #endif /* NTFS_RW */