2 * inode.c - NTFS kernel inode handling. Part of the Linux-NTFS project.
4 * Copyright (c) 2001-2006 Anton Altaparmakov
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/pagemap.h>
23 #include <linux/buffer_head.h>
24 #include <linux/smp_lock.h>
25 #include <linux/quotaops.h>
26 #include <linux/mount.h>
27 #include <linux/mutex.h>
42 * ntfs_test_inode - compare two (possibly fake) inodes for equality
43 * @vi: vfs inode which to test
44 * @na: ntfs attribute which is being tested with
46 * Compare the ntfs attribute embedded in the ntfs specific part of the vfs
47 * inode @vi for equality with the ntfs attribute @na.
49 * If searching for the normal file/directory inode, set @na->type to AT_UNUSED.
50 * @na->name and @na->name_len are then ignored.
52 * Return 1 if the attributes match and 0 if not.
54 * NOTE: This function runs with the inode_lock spin lock held so it is not
57 int ntfs_test_inode(struct inode *vi, ntfs_attr *na)
61 if (vi->i_ino != na->mft_no)
64 /* If !NInoAttr(ni), @vi is a normal file or directory inode. */
65 if (likely(!NInoAttr(ni))) {
66 /* If not looking for a normal inode this is a mismatch. */
67 if (unlikely(na->type != AT_UNUSED))
70 /* A fake inode describing an attribute. */
71 if (ni->type != na->type)
73 if (ni->name_len != na->name_len)
75 if (na->name_len && memcmp(ni->name, na->name,
76 na->name_len * sizeof(ntfschar)))
84 * ntfs_init_locked_inode - initialize an inode
85 * @vi: vfs inode to initialize
86 * @na: ntfs attribute which to initialize @vi to
88 * Initialize the vfs inode @vi with the values from the ntfs attribute @na in
89 * order to enable ntfs_test_inode() to do its work.
91 * If initializing the normal file/directory inode, set @na->type to AT_UNUSED.
92 * In that case, @na->name and @na->name_len should be set to NULL and 0,
93 * respectively. Although that is not strictly necessary as
94 * ntfs_read_inode_locked() will fill them in later.
96 * Return 0 on success and -errno on error.
98 * NOTE: This function runs with the inode_lock spin lock held so it is not
99 * allowed to sleep. (Hence the GFP_ATOMIC allocation.)
101 static int ntfs_init_locked_inode(struct inode *vi, ntfs_attr *na)
103 ntfs_inode *ni = NTFS_I(vi);
105 vi->i_ino = na->mft_no;
108 if (na->type == AT_INDEX_ALLOCATION)
109 NInoSetMstProtected(ni);
112 ni->name_len = na->name_len;
114 /* If initializing a normal inode, we are done. */
115 if (likely(na->type == AT_UNUSED)) {
117 BUG_ON(na->name_len);
121 /* It is a fake inode. */
125 * We have I30 global constant as an optimization as it is the name
126 * in >99.9% of named attributes! The other <0.1% incur a GFP_ATOMIC
127 * allocation but that is ok. And most attributes are unnamed anyway,
128 * thus the fraction of named attributes with name != I30 is actually
131 if (na->name_len && na->name != I30) {
135 i = na->name_len * sizeof(ntfschar);
136 ni->name = (ntfschar*)kmalloc(i + sizeof(ntfschar), GFP_ATOMIC);
139 memcpy(ni->name, na->name, i);
145 typedef int (*set_t)(struct inode *, void *);
146 static int ntfs_read_locked_inode(struct inode *vi);
147 static int ntfs_read_locked_attr_inode(struct inode *base_vi, struct inode *vi);
148 static int ntfs_read_locked_index_inode(struct inode *base_vi,
152 * ntfs_iget - obtain a struct inode corresponding to a specific normal inode
153 * @sb: super block of mounted volume
154 * @mft_no: mft record number / inode number to obtain
156 * Obtain the struct inode corresponding to a specific normal inode (i.e. a
157 * file or directory).
159 * If the inode is in the cache, it is just returned with an increased
160 * reference count. Otherwise, a new struct inode is allocated and initialized,
161 * and finally ntfs_read_locked_inode() is called to read in the inode and
162 * fill in the remainder of the inode structure.
164 * Return the struct inode on success. Check the return value with IS_ERR() and
165 * if true, the function failed and the error code is obtained from PTR_ERR().
167 struct inode *ntfs_iget(struct super_block *sb, unsigned long mft_no)
178 vi = iget5_locked(sb, mft_no, (test_t)ntfs_test_inode,
179 (set_t)ntfs_init_locked_inode, &na);
181 return ERR_PTR(-ENOMEM);
185 /* If this is a freshly allocated inode, need to read it now. */
186 if (vi->i_state & I_NEW) {
187 err = ntfs_read_locked_inode(vi);
188 unlock_new_inode(vi);
191 * There is no point in keeping bad inodes around if the failure was
192 * due to ENOMEM. We want to be able to retry again later.
194 if (unlikely(err == -ENOMEM)) {
202 * ntfs_attr_iget - obtain a struct inode corresponding to an attribute
203 * @base_vi: vfs base inode containing the attribute
204 * @type: attribute type
205 * @name: Unicode name of the attribute (NULL if unnamed)
206 * @name_len: length of @name in Unicode characters (0 if unnamed)
208 * Obtain the (fake) struct inode corresponding to the attribute specified by
209 * @type, @name, and @name_len, which is present in the base mft record
210 * specified by the vfs inode @base_vi.
212 * If the attribute inode is in the cache, it is just returned with an
213 * increased reference count. Otherwise, a new struct inode is allocated and
214 * initialized, and finally ntfs_read_locked_attr_inode() is called to read the
215 * attribute and fill in the inode structure.
217 * Note, for index allocation attributes, you need to use ntfs_index_iget()
218 * instead of ntfs_attr_iget() as working with indices is a lot more complex.
220 * Return the struct inode of the attribute inode on success. Check the return
221 * value with IS_ERR() and if true, the function failed and the error code is
222 * obtained from PTR_ERR().
224 struct inode *ntfs_attr_iget(struct inode *base_vi, ATTR_TYPE type,
225 ntfschar *name, u32 name_len)
231 /* Make sure no one calls ntfs_attr_iget() for indices. */
232 BUG_ON(type == AT_INDEX_ALLOCATION);
234 na.mft_no = base_vi->i_ino;
237 na.name_len = name_len;
239 vi = iget5_locked(base_vi->i_sb, na.mft_no, (test_t)ntfs_test_inode,
240 (set_t)ntfs_init_locked_inode, &na);
242 return ERR_PTR(-ENOMEM);
246 /* If this is a freshly allocated inode, need to read it now. */
247 if (vi->i_state & I_NEW) {
248 err = ntfs_read_locked_attr_inode(base_vi, vi);
249 unlock_new_inode(vi);
252 * There is no point in keeping bad attribute inodes around. This also
253 * simplifies things in that we never need to check for bad attribute
264 * ntfs_index_iget - obtain a struct inode corresponding to an index
265 * @base_vi: vfs base inode containing the index related attributes
266 * @name: Unicode name of the index
267 * @name_len: length of @name in Unicode characters
269 * Obtain the (fake) struct inode corresponding to the index specified by @name
270 * and @name_len, which is present in the base mft record specified by the vfs
273 * If the index inode is in the cache, it is just returned with an increased
274 * reference count. Otherwise, a new struct inode is allocated and
275 * initialized, and finally ntfs_read_locked_index_inode() is called to read
276 * the index related attributes and fill in the inode structure.
278 * Return the struct inode of the index inode on success. Check the return
279 * value with IS_ERR() and if true, the function failed and the error code is
280 * obtained from PTR_ERR().
282 struct inode *ntfs_index_iget(struct inode *base_vi, ntfschar *name,
289 na.mft_no = base_vi->i_ino;
290 na.type = AT_INDEX_ALLOCATION;
292 na.name_len = name_len;
294 vi = iget5_locked(base_vi->i_sb, na.mft_no, (test_t)ntfs_test_inode,
295 (set_t)ntfs_init_locked_inode, &na);
297 return ERR_PTR(-ENOMEM);
301 /* If this is a freshly allocated inode, need to read it now. */
302 if (vi->i_state & I_NEW) {
303 err = ntfs_read_locked_index_inode(base_vi, vi);
304 unlock_new_inode(vi);
307 * There is no point in keeping bad index inodes around. This also
308 * simplifies things in that we never need to check for bad index
318 struct inode *ntfs_alloc_big_inode(struct super_block *sb)
322 ntfs_debug("Entering.");
323 ni = kmem_cache_alloc(ntfs_big_inode_cache, SLAB_NOFS);
324 if (likely(ni != NULL)) {
328 ntfs_error(sb, "Allocation of NTFS big inode structure failed.");
332 void ntfs_destroy_big_inode(struct inode *inode)
334 ntfs_inode *ni = NTFS_I(inode);
336 ntfs_debug("Entering.");
338 if (!atomic_dec_and_test(&ni->count))
340 kmem_cache_free(ntfs_big_inode_cache, NTFS_I(inode));
343 static inline ntfs_inode *ntfs_alloc_extent_inode(void)
347 ntfs_debug("Entering.");
348 ni = kmem_cache_alloc(ntfs_inode_cache, SLAB_NOFS);
349 if (likely(ni != NULL)) {
353 ntfs_error(NULL, "Allocation of NTFS inode structure failed.");
357 static void ntfs_destroy_extent_inode(ntfs_inode *ni)
359 ntfs_debug("Entering.");
361 if (!atomic_dec_and_test(&ni->count))
363 kmem_cache_free(ntfs_inode_cache, ni);
367 * __ntfs_init_inode - initialize ntfs specific part of an inode
368 * @sb: super block of mounted volume
369 * @ni: freshly allocated ntfs inode which to initialize
371 * Initialize an ntfs inode to defaults.
373 * NOTE: ni->mft_no, ni->state, ni->type, ni->name, and ni->name_len are left
374 * untouched. Make sure to initialize them elsewhere.
376 * Return zero on success and -ENOMEM on error.
378 void __ntfs_init_inode(struct super_block *sb, ntfs_inode *ni)
380 ntfs_debug("Entering.");
381 rwlock_init(&ni->size_lock);
382 ni->initialized_size = ni->allocated_size = 0;
384 atomic_set(&ni->count, 1);
385 ni->vol = NTFS_SB(sb);
386 ntfs_init_runlist(&ni->runlist);
387 init_MUTEX(&ni->mrec_lock);
390 ni->attr_list_size = 0;
391 ni->attr_list = NULL;
392 ntfs_init_runlist(&ni->attr_list_rl);
393 ni->itype.index.bmp_ino = NULL;
394 ni->itype.index.block_size = 0;
395 ni->itype.index.vcn_size = 0;
396 ni->itype.index.collation_rule = 0;
397 ni->itype.index.block_size_bits = 0;
398 ni->itype.index.vcn_size_bits = 0;
399 init_MUTEX(&ni->extent_lock);
401 ni->ext.base_ntfs_ino = NULL;
404 inline ntfs_inode *ntfs_new_extent_inode(struct super_block *sb,
405 unsigned long mft_no)
407 ntfs_inode *ni = ntfs_alloc_extent_inode();
409 ntfs_debug("Entering.");
410 if (likely(ni != NULL)) {
411 __ntfs_init_inode(sb, ni);
413 ni->type = AT_UNUSED;
421 * ntfs_is_extended_system_file - check if a file is in the $Extend directory
422 * @ctx: initialized attribute search context
424 * Search all file name attributes in the inode described by the attribute
425 * search context @ctx and check if any of the names are in the $Extend system
429 * 1: file is in $Extend directory
430 * 0: file is not in $Extend directory
431 * -errno: failed to determine if the file is in the $Extend directory
433 static int ntfs_is_extended_system_file(ntfs_attr_search_ctx *ctx)
437 /* Restart search. */
438 ntfs_attr_reinit_search_ctx(ctx);
440 /* Get number of hard links. */
441 nr_links = le16_to_cpu(ctx->mrec->link_count);
443 /* Loop through all hard links. */
444 while (!(err = ntfs_attr_lookup(AT_FILE_NAME, NULL, 0, 0, 0, NULL, 0,
446 FILE_NAME_ATTR *file_name_attr;
447 ATTR_RECORD *attr = ctx->attr;
452 * Maximum sanity checking as we are called on an inode that
453 * we suspect might be corrupt.
455 p = (u8*)attr + le32_to_cpu(attr->length);
456 if (p < (u8*)ctx->mrec || (u8*)p > (u8*)ctx->mrec +
457 le32_to_cpu(ctx->mrec->bytes_in_use)) {
459 ntfs_error(ctx->ntfs_ino->vol->sb, "Corrupt file name "
460 "attribute. You should run chkdsk.");
463 if (attr->non_resident) {
464 ntfs_error(ctx->ntfs_ino->vol->sb, "Non-resident file "
465 "name. You should run chkdsk.");
469 ntfs_error(ctx->ntfs_ino->vol->sb, "File name with "
470 "invalid flags. You should run "
474 if (!(attr->data.resident.flags & RESIDENT_ATTR_IS_INDEXED)) {
475 ntfs_error(ctx->ntfs_ino->vol->sb, "Unindexed file "
476 "name. You should run chkdsk.");
479 file_name_attr = (FILE_NAME_ATTR*)((u8*)attr +
480 le16_to_cpu(attr->data.resident.value_offset));
481 p2 = (u8*)attr + le32_to_cpu(attr->data.resident.value_length);
482 if (p2 < (u8*)attr || p2 > p)
483 goto err_corrupt_attr;
484 /* This attribute is ok, but is it in the $Extend directory? */
485 if (MREF_LE(file_name_attr->parent_directory) == FILE_Extend)
486 return 1; /* YES, it's an extended system file. */
488 if (unlikely(err != -ENOENT))
490 if (unlikely(nr_links)) {
491 ntfs_error(ctx->ntfs_ino->vol->sb, "Inode hard link count "
492 "doesn't match number of name attributes. You "
493 "should run chkdsk.");
496 return 0; /* NO, it is not an extended system file. */
500 * ntfs_read_locked_inode - read an inode from its device
503 * ntfs_read_locked_inode() is called from ntfs_iget() to read the inode
504 * described by @vi into memory from the device.
506 * The only fields in @vi that we need to/can look at when the function is
507 * called are i_sb, pointing to the mounted device's super block, and i_ino,
508 * the number of the inode to load.
510 * ntfs_read_locked_inode() maps, pins and locks the mft record number i_ino
511 * for reading and sets up the necessary @vi fields as well as initializing
514 * Q: What locks are held when the function is called?
515 * A: i_state has I_LOCK set, hence the inode is locked, also
516 * i_count is set to 1, so it is not going to go away
517 * i_flags is set to 0 and we have no business touching it. Only an ioctl()
518 * is allowed to write to them. We should of course be honouring them but
519 * we need to do that using the IS_* macros defined in include/linux/fs.h.
520 * In any case ntfs_read_locked_inode() has nothing to do with i_flags.
522 * Return 0 on success and -errno on error. In the error case, the inode will
523 * have had make_bad_inode() executed on it.
525 static int ntfs_read_locked_inode(struct inode *vi)
527 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. */
539 /* This is the optimal IO size (for stat), not the fs block size. */
540 vi->i_blksize = PAGE_CACHE_SIZE;
542 * This is for checking whether an inode has changed w.r.t. a file so
543 * that the file can be updated if necessary (compare with f_version).
547 vi->i_uid = vol->uid;
548 vi->i_gid = vol->gid;
552 * Initialize the ntfs specific part of @vi special casing
553 * FILE_MFT which we need to do at mount time.
555 if (vi->i_ino != FILE_MFT)
556 ntfs_init_big_inode(vi);
559 m = map_mft_record(ni);
564 ctx = ntfs_attr_get_search_ctx(ni, m);
570 if (!(m->flags & MFT_RECORD_IN_USE)) {
571 ntfs_error(vi->i_sb, "Inode is not in use!");
574 if (m->base_mft_record) {
575 ntfs_error(vi->i_sb, "Inode is an extent inode!");
579 /* Transfer information from mft record into vfs and ntfs inodes. */
580 vi->i_generation = ni->seq_no = le16_to_cpu(m->sequence_number);
583 * FIXME: Keep in mind that link_count is two for files which have both
584 * a long file name and a short file name as separate entries, so if
585 * we are hiding short file names this will be too high. Either we need
586 * to account for the short file names by subtracting them or we need
587 * to make sure we delete files even though i_nlink is not zero which
588 * might be tricky due to vfs interactions. Need to think about this
589 * some more when implementing the unlink command.
591 vi->i_nlink = le16_to_cpu(m->link_count);
593 * FIXME: Reparse points can have the directory bit set even though
594 * they would be S_IFLNK. Need to deal with this further below when we
595 * implement reparse points / symbolic links but it will do for now.
596 * Also if not a directory, it could be something else, rather than
597 * a regular file. But again, will do for now.
599 /* Everyone gets all permissions. */
600 vi->i_mode |= S_IRWXUGO;
601 /* If read-only, noone gets write permissions. */
603 vi->i_mode &= ~S_IWUGO;
604 if (m->flags & MFT_RECORD_IS_DIRECTORY) {
605 vi->i_mode |= S_IFDIR;
607 * Apply the directory permissions mask set in the mount
610 vi->i_mode &= ~vol->dmask;
611 /* Things break without this kludge! */
615 vi->i_mode |= S_IFREG;
616 /* Apply the file permissions mask set in the mount options. */
617 vi->i_mode &= ~vol->fmask;
620 * Find the standard information attribute in the mft record. At this
621 * stage we haven't setup the attribute list stuff yet, so this could
622 * in fact fail if the standard information is in an extent record, but
623 * I don't think this actually ever happens.
625 err = ntfs_attr_lookup(AT_STANDARD_INFORMATION, NULL, 0, 0, 0, NULL, 0,
628 if (err == -ENOENT) {
630 * TODO: We should be performing a hot fix here (if the
631 * recover mount option is set) by creating a new
634 ntfs_error(vi->i_sb, "$STANDARD_INFORMATION attribute "
640 /* Get the standard information attribute value. */
641 si = (STANDARD_INFORMATION*)((u8*)a +
642 le16_to_cpu(a->data.resident.value_offset));
644 /* Transfer information from the standard information into vi. */
646 * Note: The i_?times do not quite map perfectly onto the NTFS times,
647 * but they are close enough, and in the end it doesn't really matter
651 * mtime is the last change of the data within the file. Not changed
652 * when only metadata is changed, e.g. a rename doesn't affect mtime.
654 vi->i_mtime = ntfs2utc(si->last_data_change_time);
656 * ctime is the last change of the metadata of the file. This obviously
657 * always changes, when mtime is changed. ctime can be changed on its
658 * own, mtime is then not changed, e.g. when a file is renamed.
660 vi->i_ctime = ntfs2utc(si->last_mft_change_time);
662 * Last access to the data within the file. Not changed during a rename
663 * for example but changed whenever the file is written to.
665 vi->i_atime = ntfs2utc(si->last_access_time);
667 /* Find the attribute list attribute if present. */
668 ntfs_attr_reinit_search_ctx(ctx);
669 err = ntfs_attr_lookup(AT_ATTRIBUTE_LIST, NULL, 0, 0, 0, NULL, 0, ctx);
671 if (unlikely(err != -ENOENT)) {
672 ntfs_error(vi->i_sb, "Failed to lookup attribute list "
676 } else /* if (!err) */ {
677 if (vi->i_ino == FILE_MFT)
678 goto skip_attr_list_load;
679 ntfs_debug("Attribute list found in inode 0x%lx.", vi->i_ino);
682 if (a->flags & ATTR_COMPRESSION_MASK) {
683 ntfs_error(vi->i_sb, "Attribute list attribute is "
687 if (a->flags & ATTR_IS_ENCRYPTED ||
688 a->flags & ATTR_IS_SPARSE) {
689 if (a->non_resident) {
690 ntfs_error(vi->i_sb, "Non-resident attribute "
691 "list attribute is encrypted/"
695 ntfs_warning(vi->i_sb, "Resident attribute list "
696 "attribute in inode 0x%lx is marked "
697 "encrypted/sparse which is not true. "
698 "However, Windows allows this and "
699 "chkdsk does not detect or correct it "
700 "so we will just ignore the invalid "
701 "flags and pretend they are not set.",
704 /* Now allocate memory for the attribute list. */
705 ni->attr_list_size = (u32)ntfs_attr_size(a);
706 ni->attr_list = ntfs_malloc_nofs(ni->attr_list_size);
707 if (!ni->attr_list) {
708 ntfs_error(vi->i_sb, "Not enough memory to allocate "
709 "buffer for attribute list.");
713 if (a->non_resident) {
714 NInoSetAttrListNonResident(ni);
715 if (a->data.non_resident.lowest_vcn) {
716 ntfs_error(vi->i_sb, "Attribute list has non "
721 * Setup the runlist. No need for locking as we have
722 * exclusive access to the inode at this time.
724 ni->attr_list_rl.rl = ntfs_mapping_pairs_decompress(vol,
726 if (IS_ERR(ni->attr_list_rl.rl)) {
727 err = PTR_ERR(ni->attr_list_rl.rl);
728 ni->attr_list_rl.rl = NULL;
729 ntfs_error(vi->i_sb, "Mapping pairs "
730 "decompression failed.");
733 /* Now load the attribute list. */
734 if ((err = load_attribute_list(vol, &ni->attr_list_rl,
735 ni->attr_list, ni->attr_list_size,
736 sle64_to_cpu(a->data.non_resident.
737 initialized_size)))) {
738 ntfs_error(vi->i_sb, "Failed to load "
739 "attribute list attribute.");
742 } else /* if (!a->non_resident) */ {
743 if ((u8*)a + le16_to_cpu(a->data.resident.value_offset)
745 a->data.resident.value_length) >
746 (u8*)ctx->mrec + vol->mft_record_size) {
747 ntfs_error(vi->i_sb, "Corrupt attribute list "
751 /* Now copy the attribute list. */
752 memcpy(ni->attr_list, (u8*)a + le16_to_cpu(
753 a->data.resident.value_offset),
755 a->data.resident.value_length));
760 * If an attribute list is present we now have the attribute list value
761 * in ntfs_ino->attr_list and it is ntfs_ino->attr_list_size bytes.
763 if (S_ISDIR(vi->i_mode)) {
768 u8 *ir_end, *index_end;
770 /* It is a directory, find index root attribute. */
771 ntfs_attr_reinit_search_ctx(ctx);
772 err = ntfs_attr_lookup(AT_INDEX_ROOT, I30, 4, CASE_SENSITIVE,
775 if (err == -ENOENT) {
776 // FIXME: File is corrupt! Hot-fix with empty
777 // index root attribute if recovery option is
779 ntfs_error(vi->i_sb, "$INDEX_ROOT attribute "
785 /* Set up the state. */
786 if (unlikely(a->non_resident)) {
787 ntfs_error(vol->sb, "$INDEX_ROOT attribute is not "
791 /* Ensure the attribute name is placed before the value. */
792 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
793 le16_to_cpu(a->data.resident.value_offset)))) {
794 ntfs_error(vol->sb, "$INDEX_ROOT attribute name is "
795 "placed after the attribute value.");
799 * Compressed/encrypted index root just means that the newly
800 * created files in that directory should be created compressed/
801 * encrypted. However index root cannot be both compressed and
804 if (a->flags & ATTR_COMPRESSION_MASK)
805 NInoSetCompressed(ni);
806 if (a->flags & ATTR_IS_ENCRYPTED) {
807 if (a->flags & ATTR_COMPRESSION_MASK) {
808 ntfs_error(vi->i_sb, "Found encrypted and "
809 "compressed attribute.");
812 NInoSetEncrypted(ni);
814 if (a->flags & ATTR_IS_SPARSE)
816 ir = (INDEX_ROOT*)((u8*)a +
817 le16_to_cpu(a->data.resident.value_offset));
818 ir_end = (u8*)ir + le32_to_cpu(a->data.resident.value_length);
819 if (ir_end > (u8*)ctx->mrec + vol->mft_record_size) {
820 ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is "
824 index_end = (u8*)&ir->index +
825 le32_to_cpu(ir->index.index_length);
826 if (index_end > ir_end) {
827 ntfs_error(vi->i_sb, "Directory index is corrupt.");
830 if (ir->type != AT_FILE_NAME) {
831 ntfs_error(vi->i_sb, "Indexed attribute is not "
835 if (ir->collation_rule != COLLATION_FILE_NAME) {
836 ntfs_error(vi->i_sb, "Index collation rule is not "
837 "COLLATION_FILE_NAME.");
840 ni->itype.index.collation_rule = ir->collation_rule;
841 ni->itype.index.block_size = le32_to_cpu(ir->index_block_size);
842 if (ni->itype.index.block_size &
843 (ni->itype.index.block_size - 1)) {
844 ntfs_error(vi->i_sb, "Index block size (%u) is not a "
846 ni->itype.index.block_size);
849 if (ni->itype.index.block_size > PAGE_CACHE_SIZE) {
850 ntfs_error(vi->i_sb, "Index block size (%u) > "
851 "PAGE_CACHE_SIZE (%ld) is not "
853 ni->itype.index.block_size,
858 if (ni->itype.index.block_size < NTFS_BLOCK_SIZE) {
859 ntfs_error(vi->i_sb, "Index block size (%u) < "
860 "NTFS_BLOCK_SIZE (%i) is not "
862 ni->itype.index.block_size,
867 ni->itype.index.block_size_bits =
868 ffs(ni->itype.index.block_size) - 1;
869 /* Determine the size of a vcn in the directory index. */
870 if (vol->cluster_size <= ni->itype.index.block_size) {
871 ni->itype.index.vcn_size = vol->cluster_size;
872 ni->itype.index.vcn_size_bits = vol->cluster_size_bits;
874 ni->itype.index.vcn_size = vol->sector_size;
875 ni->itype.index.vcn_size_bits = vol->sector_size_bits;
878 /* Setup the index allocation attribute, even if not present. */
879 NInoSetMstProtected(ni);
880 ni->type = AT_INDEX_ALLOCATION;
884 if (!(ir->index.flags & LARGE_INDEX)) {
885 /* No index allocation. */
886 vi->i_size = ni->initialized_size =
887 ni->allocated_size = 0;
888 /* We are done with the mft record, so we release it. */
889 ntfs_attr_put_search_ctx(ctx);
890 unmap_mft_record(ni);
893 goto skip_large_dir_stuff;
894 } /* LARGE_INDEX: Index allocation present. Setup state. */
895 NInoSetIndexAllocPresent(ni);
896 /* Find index allocation attribute. */
897 ntfs_attr_reinit_search_ctx(ctx);
898 err = ntfs_attr_lookup(AT_INDEX_ALLOCATION, I30, 4,
899 CASE_SENSITIVE, 0, NULL, 0, ctx);
902 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION "
903 "attribute is not present but "
904 "$INDEX_ROOT indicated it is.");
906 ntfs_error(vi->i_sb, "Failed to lookup "
912 if (!a->non_resident) {
913 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
918 * Ensure the attribute name is placed before the mapping pairs
921 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
923 a->data.non_resident.mapping_pairs_offset)))) {
924 ntfs_error(vol->sb, "$INDEX_ALLOCATION attribute name "
925 "is placed after the mapping pairs "
929 if (a->flags & ATTR_IS_ENCRYPTED) {
930 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
934 if (a->flags & ATTR_IS_SPARSE) {
935 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
939 if (a->flags & ATTR_COMPRESSION_MASK) {
940 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
944 if (a->data.non_resident.lowest_vcn) {
945 ntfs_error(vi->i_sb, "First extent of "
946 "$INDEX_ALLOCATION attribute has non "
950 vi->i_size = sle64_to_cpu(a->data.non_resident.data_size);
951 ni->initialized_size = sle64_to_cpu(
952 a->data.non_resident.initialized_size);
953 ni->allocated_size = sle64_to_cpu(
954 a->data.non_resident.allocated_size);
956 * We are done with the mft record, so we release it. Otherwise
957 * we would deadlock in ntfs_attr_iget().
959 ntfs_attr_put_search_ctx(ctx);
960 unmap_mft_record(ni);
963 /* Get the index bitmap attribute inode. */
964 bvi = ntfs_attr_iget(vi, AT_BITMAP, I30, 4);
966 ntfs_error(vi->i_sb, "Failed to get bitmap attribute.");
970 ni->itype.index.bmp_ino = bvi;
972 if (NInoCompressed(bni) || NInoEncrypted(bni) ||
974 ntfs_error(vi->i_sb, "$BITMAP attribute is compressed "
975 "and/or encrypted and/or sparse.");
978 /* Consistency check bitmap size vs. index allocation size. */
979 bvi_size = i_size_read(bvi);
980 if ((bvi_size << 3) < (vi->i_size >>
981 ni->itype.index.block_size_bits)) {
982 ntfs_error(vi->i_sb, "Index bitmap too small (0x%llx) "
983 "for index allocation (0x%llx).",
984 bvi_size << 3, vi->i_size);
987 skip_large_dir_stuff:
988 /* Setup the operations for this inode. */
989 vi->i_op = &ntfs_dir_inode_ops;
990 vi->i_fop = &ntfs_dir_ops;
993 ntfs_attr_reinit_search_ctx(ctx);
995 /* Setup the data attribute, even if not present. */
1000 /* Find first extent of the unnamed data attribute. */
1001 err = ntfs_attr_lookup(AT_DATA, NULL, 0, 0, 0, NULL, 0, ctx);
1002 if (unlikely(err)) {
1003 vi->i_size = ni->initialized_size =
1004 ni->allocated_size = 0;
1005 if (err != -ENOENT) {
1006 ntfs_error(vi->i_sb, "Failed to lookup $DATA "
1011 * FILE_Secure does not have an unnamed $DATA
1012 * attribute, so we special case it here.
1014 if (vi->i_ino == FILE_Secure)
1015 goto no_data_attr_special_case;
1017 * Most if not all the system files in the $Extend
1018 * system directory do not have unnamed data
1019 * attributes so we need to check if the parent
1020 * directory of the file is FILE_Extend and if it is
1021 * ignore this error. To do this we need to get the
1022 * name of this inode from the mft record as the name
1023 * contains the back reference to the parent directory.
1025 if (ntfs_is_extended_system_file(ctx) > 0)
1026 goto no_data_attr_special_case;
1027 // FIXME: File is corrupt! Hot-fix with empty data
1028 // attribute if recovery option is set.
1029 ntfs_error(vi->i_sb, "$DATA attribute is missing.");
1033 /* Setup the state. */
1034 if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_SPARSE)) {
1035 if (a->flags & ATTR_COMPRESSION_MASK) {
1036 NInoSetCompressed(ni);
1037 if (vol->cluster_size > 4096) {
1038 ntfs_error(vi->i_sb, "Found "
1039 "compressed data but "
1042 "cluster size (%i) > "
1047 if ((a->flags & ATTR_COMPRESSION_MASK)
1048 != ATTR_IS_COMPRESSED) {
1049 ntfs_error(vi->i_sb, "Found unknown "
1050 "compression method "
1051 "or corrupt file.");
1055 if (a->flags & ATTR_IS_SPARSE)
1058 if (a->flags & ATTR_IS_ENCRYPTED) {
1059 if (NInoCompressed(ni)) {
1060 ntfs_error(vi->i_sb, "Found encrypted and "
1061 "compressed data.");
1064 NInoSetEncrypted(ni);
1066 if (a->non_resident) {
1067 NInoSetNonResident(ni);
1068 if (NInoCompressed(ni) || NInoSparse(ni)) {
1069 if (NInoCompressed(ni) && a->data.non_resident.
1070 compression_unit != 4) {
1071 ntfs_error(vi->i_sb, "Found "
1073 "compression unit (%u "
1075 "Cannot handle this.",
1076 a->data.non_resident.
1081 if (a->data.non_resident.compression_unit) {
1082 ni->itype.compressed.block_size = 1U <<
1083 (a->data.non_resident.
1085 vol->cluster_size_bits);
1086 ni->itype.compressed.block_size_bits =
1090 ni->itype.compressed.block_clusters =
1095 ni->itype.compressed.block_size = 0;
1096 ni->itype.compressed.block_size_bits =
1098 ni->itype.compressed.block_clusters =
1101 ni->itype.compressed.size = sle64_to_cpu(
1102 a->data.non_resident.
1105 if (a->data.non_resident.lowest_vcn) {
1106 ntfs_error(vi->i_sb, "First extent of $DATA "
1107 "attribute has non zero "
1111 vi->i_size = sle64_to_cpu(
1112 a->data.non_resident.data_size);
1113 ni->initialized_size = sle64_to_cpu(
1114 a->data.non_resident.initialized_size);
1115 ni->allocated_size = sle64_to_cpu(
1116 a->data.non_resident.allocated_size);
1117 } else { /* Resident attribute. */
1118 vi->i_size = ni->initialized_size = le32_to_cpu(
1119 a->data.resident.value_length);
1120 ni->allocated_size = le32_to_cpu(a->length) -
1122 a->data.resident.value_offset);
1123 if (vi->i_size > ni->allocated_size) {
1124 ntfs_error(vi->i_sb, "Resident data attribute "
1125 "is corrupt (size exceeds "
1130 no_data_attr_special_case:
1131 /* We are done with the mft record, so we release it. */
1132 ntfs_attr_put_search_ctx(ctx);
1133 unmap_mft_record(ni);
1136 /* Setup the operations for this inode. */
1137 vi->i_op = &ntfs_file_inode_ops;
1138 vi->i_fop = &ntfs_file_ops;
1140 if (NInoMstProtected(ni))
1141 vi->i_mapping->a_ops = &ntfs_mst_aops;
1143 vi->i_mapping->a_ops = &ntfs_aops;
1145 * The number of 512-byte blocks used on disk (for stat). This is in so
1146 * far inaccurate as it doesn't account for any named streams or other
1147 * special non-resident attributes, but that is how Windows works, too,
1148 * so we are at least consistent with Windows, if not entirely
1149 * consistent with the Linux Way. Doing it the Linux Way would cause a
1150 * significant slowdown as it would involve iterating over all
1151 * attributes in the mft record and adding the allocated/compressed
1152 * sizes of all non-resident attributes present to give us the Linux
1153 * correct size that should go into i_blocks (after division by 512).
1155 if (S_ISREG(vi->i_mode) && (NInoCompressed(ni) || NInoSparse(ni)))
1156 vi->i_blocks = ni->itype.compressed.size >> 9;
1158 vi->i_blocks = ni->allocated_size >> 9;
1159 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_LOCK 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_blksize = base_vi->i_blksize;
1218 vi->i_version = base_vi->i_version;
1219 vi->i_uid = base_vi->i_uid;
1220 vi->i_gid = base_vi->i_gid;
1221 vi->i_nlink = base_vi->i_nlink;
1222 vi->i_mtime = base_vi->i_mtime;
1223 vi->i_ctime = base_vi->i_ctime;
1224 vi->i_atime = base_vi->i_atime;
1225 vi->i_generation = ni->seq_no = base_ni->seq_no;
1227 /* Set inode type to zero but preserve permissions. */
1228 vi->i_mode = base_vi->i_mode & ~S_IFMT;
1230 m = map_mft_record(base_ni);
1235 ctx = ntfs_attr_get_search_ctx(base_ni, m);
1240 /* Find the attribute. */
1241 err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
1242 CASE_SENSITIVE, 0, NULL, 0, ctx);
1246 if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_SPARSE)) {
1247 if (a->flags & ATTR_COMPRESSION_MASK) {
1248 NInoSetCompressed(ni);
1249 if ((ni->type != AT_DATA) || (ni->type == AT_DATA &&
1251 ntfs_error(vi->i_sb, "Found compressed "
1252 "non-data or named data "
1253 "attribute. Please report "
1254 "you saw this message to "
1255 "linux-ntfs-dev@lists."
1259 if (vol->cluster_size > 4096) {
1260 ntfs_error(vi->i_sb, "Found compressed "
1261 "attribute but compression is "
1262 "disabled due to cluster size "
1267 if ((a->flags & ATTR_COMPRESSION_MASK) !=
1268 ATTR_IS_COMPRESSED) {
1269 ntfs_error(vi->i_sb, "Found unknown "
1270 "compression method.");
1275 * The compressed/sparse flag set in an index root just means
1276 * to compress all files.
1278 if (NInoMstProtected(ni) && ni->type != AT_INDEX_ROOT) {
1279 ntfs_error(vi->i_sb, "Found mst protected attribute "
1280 "but the attribute is %s. Please "
1281 "report you saw this message to "
1282 "linux-ntfs-dev@lists.sourceforge.net",
1283 NInoCompressed(ni) ? "compressed" :
1287 if (a->flags & ATTR_IS_SPARSE)
1290 if (a->flags & ATTR_IS_ENCRYPTED) {
1291 if (NInoCompressed(ni)) {
1292 ntfs_error(vi->i_sb, "Found encrypted and compressed "
1297 * The encryption flag set in an index root just means to
1298 * encrypt all files.
1300 if (NInoMstProtected(ni) && ni->type != AT_INDEX_ROOT) {
1301 ntfs_error(vi->i_sb, "Found mst protected attribute "
1302 "but the attribute is encrypted. "
1303 "Please report you saw this message "
1304 "to linux-ntfs-dev@lists.sourceforge."
1308 if (ni->type != AT_DATA) {
1309 ntfs_error(vi->i_sb, "Found encrypted non-data "
1313 NInoSetEncrypted(ni);
1315 if (!a->non_resident) {
1316 /* Ensure the attribute name is placed before the value. */
1317 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1318 le16_to_cpu(a->data.resident.value_offset)))) {
1319 ntfs_error(vol->sb, "Attribute name is placed after "
1320 "the attribute value.");
1323 if (NInoMstProtected(ni)) {
1324 ntfs_error(vi->i_sb, "Found mst protected attribute "
1325 "but the attribute is resident. "
1326 "Please report you saw this message to "
1327 "linux-ntfs-dev@lists.sourceforge.net");
1330 vi->i_size = ni->initialized_size = le32_to_cpu(
1331 a->data.resident.value_length);
1332 ni->allocated_size = le32_to_cpu(a->length) -
1333 le16_to_cpu(a->data.resident.value_offset);
1334 if (vi->i_size > ni->allocated_size) {
1335 ntfs_error(vi->i_sb, "Resident attribute is corrupt "
1336 "(size exceeds allocation).");
1340 NInoSetNonResident(ni);
1342 * Ensure the attribute name is placed before the mapping pairs
1345 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1347 a->data.non_resident.mapping_pairs_offset)))) {
1348 ntfs_error(vol->sb, "Attribute name is placed after "
1349 "the mapping pairs array.");
1352 if (NInoCompressed(ni) || NInoSparse(ni)) {
1353 if (NInoCompressed(ni) && a->data.non_resident.
1354 compression_unit != 4) {
1355 ntfs_error(vi->i_sb, "Found non-standard "
1356 "compression unit (%u instead "
1357 "of 4). Cannot handle this.",
1358 a->data.non_resident.
1363 if (a->data.non_resident.compression_unit) {
1364 ni->itype.compressed.block_size = 1U <<
1365 (a->data.non_resident.
1367 vol->cluster_size_bits);
1368 ni->itype.compressed.block_size_bits =
1369 ffs(ni->itype.compressed.
1371 ni->itype.compressed.block_clusters = 1U <<
1372 a->data.non_resident.
1375 ni->itype.compressed.block_size = 0;
1376 ni->itype.compressed.block_size_bits = 0;
1377 ni->itype.compressed.block_clusters = 0;
1379 ni->itype.compressed.size = sle64_to_cpu(
1380 a->data.non_resident.compressed_size);
1382 if (a->data.non_resident.lowest_vcn) {
1383 ntfs_error(vi->i_sb, "First extent of attribute has "
1384 "non-zero lowest_vcn.");
1387 vi->i_size = sle64_to_cpu(a->data.non_resident.data_size);
1388 ni->initialized_size = sle64_to_cpu(
1389 a->data.non_resident.initialized_size);
1390 ni->allocated_size = sle64_to_cpu(
1391 a->data.non_resident.allocated_size);
1393 /* Setup the operations for this attribute inode. */
1396 if (NInoMstProtected(ni))
1397 vi->i_mapping->a_ops = &ntfs_mst_aops;
1399 vi->i_mapping->a_ops = &ntfs_aops;
1400 if ((NInoCompressed(ni) || NInoSparse(ni)) && ni->type != AT_INDEX_ROOT)
1401 vi->i_blocks = ni->itype.compressed.size >> 9;
1403 vi->i_blocks = ni->allocated_size >> 9;
1405 * Make sure the base inode does not go away and attach it to the
1409 ni->ext.base_ntfs_ino = base_ni;
1410 ni->nr_extents = -1;
1412 ntfs_attr_put_search_ctx(ctx);
1413 unmap_mft_record(base_ni);
1415 ntfs_debug("Done.");
1422 ntfs_attr_put_search_ctx(ctx);
1423 unmap_mft_record(base_ni);
1425 ntfs_error(vol->sb, "Failed with error code %i while reading attribute "
1426 "inode (mft_no 0x%lx, type 0x%x, name_len %i). "
1427 "Marking corrupt inode and base inode 0x%lx as bad. "
1428 "Run chkdsk.", err, vi->i_ino, ni->type, ni->name_len,
1431 make_bad_inode(base_vi);
1438 * ntfs_read_locked_index_inode - read an index inode from its base inode
1439 * @base_vi: base inode
1440 * @vi: index inode to read
1442 * ntfs_read_locked_index_inode() is called from ntfs_index_iget() to read the
1443 * index inode described by @vi into memory from the base mft record described
1446 * ntfs_read_locked_index_inode() maps, pins and locks the base inode for
1447 * reading and looks up the attributes relating to the index described by @vi
1448 * before setting up the necessary fields in @vi as well as initializing the
1451 * Note, index inodes are essentially attribute inodes (NInoAttr() is true)
1452 * with the attribute type set to AT_INDEX_ALLOCATION. Apart from that, they
1453 * are setup like directory inodes since directories are a special case of
1454 * indices ao they need to be treated in much the same way. Most importantly,
1455 * for small indices the index allocation attribute might not actually exist.
1456 * However, the index root attribute always exists but this does not need to
1457 * have an inode associated with it and this is why we define a new inode type
1458 * index. Also, like for directories, we need to have an attribute inode for
1459 * the bitmap attribute corresponding to the index allocation attribute and we
1460 * can store this in the appropriate field of the inode, just like we do for
1461 * normal directory inodes.
1463 * Q: What locks are held when the function is called?
1464 * A: i_state has I_LOCK set, hence the inode is locked, also
1465 * i_count is set to 1, so it is not going to go away
1467 * Return 0 on success and -errno on error. In the error case, the inode will
1468 * have had make_bad_inode() executed on it.
1470 static int ntfs_read_locked_index_inode(struct inode *base_vi, struct inode *vi)
1473 ntfs_volume *vol = NTFS_SB(vi->i_sb);
1474 ntfs_inode *ni, *base_ni, *bni;
1478 ntfs_attr_search_ctx *ctx;
1480 u8 *ir_end, *index_end;
1483 ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino);
1484 ntfs_init_big_inode(vi);
1486 base_ni = NTFS_I(base_vi);
1487 /* Just mirror the values from the base inode. */
1488 vi->i_blksize = base_vi->i_blksize;
1489 vi->i_version = base_vi->i_version;
1490 vi->i_uid = base_vi->i_uid;
1491 vi->i_gid = base_vi->i_gid;
1492 vi->i_nlink = base_vi->i_nlink;
1493 vi->i_mtime = base_vi->i_mtime;
1494 vi->i_ctime = base_vi->i_ctime;
1495 vi->i_atime = base_vi->i_atime;
1496 vi->i_generation = ni->seq_no = base_ni->seq_no;
1497 /* Set inode type to zero but preserve permissions. */
1498 vi->i_mode = base_vi->i_mode & ~S_IFMT;
1499 /* Map the mft record for the base inode. */
1500 m = map_mft_record(base_ni);
1505 ctx = ntfs_attr_get_search_ctx(base_ni, m);
1510 /* Find the index root attribute. */
1511 err = ntfs_attr_lookup(AT_INDEX_ROOT, ni->name, ni->name_len,
1512 CASE_SENSITIVE, 0, NULL, 0, ctx);
1513 if (unlikely(err)) {
1515 ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is "
1520 /* Set up the state. */
1521 if (unlikely(a->non_resident)) {
1522 ntfs_error(vol->sb, "$INDEX_ROOT attribute is not resident.");
1525 /* Ensure the attribute name is placed before the value. */
1526 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1527 le16_to_cpu(a->data.resident.value_offset)))) {
1528 ntfs_error(vol->sb, "$INDEX_ROOT attribute name is placed "
1529 "after the attribute value.");
1533 * Compressed/encrypted/sparse index root is not allowed, except for
1534 * directories of course but those are not dealt with here.
1536 if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_ENCRYPTED |
1538 ntfs_error(vi->i_sb, "Found compressed/encrypted/sparse index "
1542 ir = (INDEX_ROOT*)((u8*)a + le16_to_cpu(a->data.resident.value_offset));
1543 ir_end = (u8*)ir + le32_to_cpu(a->data.resident.value_length);
1544 if (ir_end > (u8*)ctx->mrec + vol->mft_record_size) {
1545 ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is corrupt.");
1548 index_end = (u8*)&ir->index + le32_to_cpu(ir->index.index_length);
1549 if (index_end > ir_end) {
1550 ntfs_error(vi->i_sb, "Index is corrupt.");
1554 ntfs_error(vi->i_sb, "Index type is not 0 (type is 0x%x).",
1555 le32_to_cpu(ir->type));
1558 ni->itype.index.collation_rule = ir->collation_rule;
1559 ntfs_debug("Index collation rule is 0x%x.",
1560 le32_to_cpu(ir->collation_rule));
1561 ni->itype.index.block_size = le32_to_cpu(ir->index_block_size);
1562 if (ni->itype.index.block_size & (ni->itype.index.block_size - 1)) {
1563 ntfs_error(vi->i_sb, "Index block size (%u) is not a power of "
1564 "two.", ni->itype.index.block_size);
1567 if (ni->itype.index.block_size > PAGE_CACHE_SIZE) {
1568 ntfs_error(vi->i_sb, "Index block size (%u) > PAGE_CACHE_SIZE "
1569 "(%ld) is not supported. Sorry.",
1570 ni->itype.index.block_size, PAGE_CACHE_SIZE);
1574 if (ni->itype.index.block_size < NTFS_BLOCK_SIZE) {
1575 ntfs_error(vi->i_sb, "Index block size (%u) < NTFS_BLOCK_SIZE "
1576 "(%i) is not supported. Sorry.",
1577 ni->itype.index.block_size, NTFS_BLOCK_SIZE);
1581 ni->itype.index.block_size_bits = ffs(ni->itype.index.block_size) - 1;
1582 /* Determine the size of a vcn in the index. */
1583 if (vol->cluster_size <= ni->itype.index.block_size) {
1584 ni->itype.index.vcn_size = vol->cluster_size;
1585 ni->itype.index.vcn_size_bits = vol->cluster_size_bits;
1587 ni->itype.index.vcn_size = vol->sector_size;
1588 ni->itype.index.vcn_size_bits = vol->sector_size_bits;
1590 /* Check for presence of index allocation attribute. */
1591 if (!(ir->index.flags & LARGE_INDEX)) {
1592 /* No index allocation. */
1593 vi->i_size = ni->initialized_size = ni->allocated_size = 0;
1594 /* We are done with the mft record, so we release it. */
1595 ntfs_attr_put_search_ctx(ctx);
1596 unmap_mft_record(base_ni);
1599 goto skip_large_index_stuff;
1600 } /* LARGE_INDEX: Index allocation present. Setup state. */
1601 NInoSetIndexAllocPresent(ni);
1602 /* Find index allocation attribute. */
1603 ntfs_attr_reinit_search_ctx(ctx);
1604 err = ntfs_attr_lookup(AT_INDEX_ALLOCATION, ni->name, ni->name_len,
1605 CASE_SENSITIVE, 0, NULL, 0, ctx);
1606 if (unlikely(err)) {
1608 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1609 "not present but $INDEX_ROOT "
1610 "indicated it is.");
1612 ntfs_error(vi->i_sb, "Failed to lookup "
1613 "$INDEX_ALLOCATION attribute.");
1616 if (!a->non_resident) {
1617 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1622 * Ensure the attribute name is placed before the mapping pairs array.
1624 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1626 a->data.non_resident.mapping_pairs_offset)))) {
1627 ntfs_error(vol->sb, "$INDEX_ALLOCATION attribute name is "
1628 "placed after the mapping pairs array.");
1631 if (a->flags & ATTR_IS_ENCRYPTED) {
1632 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1636 if (a->flags & ATTR_IS_SPARSE) {
1637 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is sparse.");
1640 if (a->flags & ATTR_COMPRESSION_MASK) {
1641 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1645 if (a->data.non_resident.lowest_vcn) {
1646 ntfs_error(vi->i_sb, "First extent of $INDEX_ALLOCATION "
1647 "attribute has non zero lowest_vcn.");
1650 vi->i_size = sle64_to_cpu(a->data.non_resident.data_size);
1651 ni->initialized_size = sle64_to_cpu(
1652 a->data.non_resident.initialized_size);
1653 ni->allocated_size = sle64_to_cpu(a->data.non_resident.allocated_size);
1655 * We are done with the mft record, so we release it. Otherwise
1656 * we would deadlock in ntfs_attr_iget().
1658 ntfs_attr_put_search_ctx(ctx);
1659 unmap_mft_record(base_ni);
1662 /* Get the index bitmap attribute inode. */
1663 bvi = ntfs_attr_iget(base_vi, AT_BITMAP, ni->name, ni->name_len);
1665 ntfs_error(vi->i_sb, "Failed to get bitmap attribute.");
1670 if (NInoCompressed(bni) || NInoEncrypted(bni) ||
1672 ntfs_error(vi->i_sb, "$BITMAP attribute is compressed and/or "
1673 "encrypted and/or sparse.");
1674 goto iput_unm_err_out;
1676 /* Consistency check bitmap size vs. index allocation size. */
1677 bvi_size = i_size_read(bvi);
1678 if ((bvi_size << 3) < (vi->i_size >> ni->itype.index.block_size_bits)) {
1679 ntfs_error(vi->i_sb, "Index bitmap too small (0x%llx) for "
1680 "index allocation (0x%llx).", bvi_size << 3,
1682 goto iput_unm_err_out;
1684 ni->itype.index.bmp_ino = bvi;
1685 skip_large_index_stuff:
1686 /* Setup the operations for this index inode. */
1689 vi->i_mapping->a_ops = &ntfs_mst_aops;
1690 vi->i_blocks = ni->allocated_size >> 9;
1692 * Make sure the base inode doesn't go away and attach it to the
1696 ni->ext.base_ntfs_ino = base_ni;
1697 ni->nr_extents = -1;
1699 ntfs_debug("Done.");
1708 ntfs_attr_put_search_ctx(ctx);
1710 unmap_mft_record(base_ni);
1712 ntfs_error(vi->i_sb, "Failed with error code %i while reading index "
1713 "inode (mft_no 0x%lx, name_len %i.", err, vi->i_ino,
1716 if (err != -EOPNOTSUPP && err != -ENOMEM)
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 /* Apply the mst fixups. */
1820 if (post_read_mst_fixup((NTFS_RECORD*)m, vol->mft_record_size)) {
1821 /* FIXME: Try to use the $MFTMirr now. */
1822 ntfs_error(sb, "MST fixup failed. $MFT is corrupt.");
1826 /* Need this to sanity check attribute list references to $MFT. */
1827 vi->i_generation = ni->seq_no = le16_to_cpu(m->sequence_number);
1829 /* Provides readpage() and sync_page() for map_mft_record(). */
1830 vi->i_mapping->a_ops = &ntfs_mst_aops;
1832 ctx = ntfs_attr_get_search_ctx(ni, m);
1838 /* Find the attribute list attribute if present. */
1839 err = ntfs_attr_lookup(AT_ATTRIBUTE_LIST, NULL, 0, 0, 0, NULL, 0, ctx);
1841 if (unlikely(err != -ENOENT)) {
1842 ntfs_error(sb, "Failed to lookup attribute list "
1843 "attribute. You should run chkdsk.");
1846 } else /* if (!err) */ {
1847 ATTR_LIST_ENTRY *al_entry, *next_al_entry;
1849 static const char *es = " Not allowed. $MFT is corrupt. "
1850 "You should run chkdsk.";
1852 ntfs_debug("Attribute list attribute found in $MFT.");
1853 NInoSetAttrList(ni);
1855 if (a->flags & ATTR_COMPRESSION_MASK) {
1856 ntfs_error(sb, "Attribute list attribute is "
1857 "compressed.%s", es);
1860 if (a->flags & ATTR_IS_ENCRYPTED ||
1861 a->flags & ATTR_IS_SPARSE) {
1862 if (a->non_resident) {
1863 ntfs_error(sb, "Non-resident attribute list "
1864 "attribute is encrypted/"
1868 ntfs_warning(sb, "Resident attribute list attribute "
1869 "in $MFT system file is marked "
1870 "encrypted/sparse which is not true. "
1871 "However, Windows allows this and "
1872 "chkdsk does not detect or correct it "
1873 "so we will just ignore the invalid "
1874 "flags and pretend they are not set.");
1876 /* Now allocate memory for the attribute list. */
1877 ni->attr_list_size = (u32)ntfs_attr_size(a);
1878 ni->attr_list = ntfs_malloc_nofs(ni->attr_list_size);
1879 if (!ni->attr_list) {
1880 ntfs_error(sb, "Not enough memory to allocate buffer "
1881 "for attribute list.");
1884 if (a->non_resident) {
1885 NInoSetAttrListNonResident(ni);
1886 if (a->data.non_resident.lowest_vcn) {
1887 ntfs_error(sb, "Attribute list has non zero "
1888 "lowest_vcn. $MFT is corrupt. "
1889 "You should run chkdsk.");
1892 /* Setup the runlist. */
1893 ni->attr_list_rl.rl = ntfs_mapping_pairs_decompress(vol,
1895 if (IS_ERR(ni->attr_list_rl.rl)) {
1896 err = PTR_ERR(ni->attr_list_rl.rl);
1897 ni->attr_list_rl.rl = NULL;
1898 ntfs_error(sb, "Mapping pairs decompression "
1899 "failed with error code %i.",
1903 /* Now load the attribute list. */
1904 if ((err = load_attribute_list(vol, &ni->attr_list_rl,
1905 ni->attr_list, ni->attr_list_size,
1906 sle64_to_cpu(a->data.
1907 non_resident.initialized_size)))) {
1908 ntfs_error(sb, "Failed to load attribute list "
1909 "attribute with error code %i.",
1913 } else /* if (!ctx.attr->non_resident) */ {
1914 if ((u8*)a + le16_to_cpu(
1915 a->data.resident.value_offset) +
1917 a->data.resident.value_length) >
1918 (u8*)ctx->mrec + vol->mft_record_size) {
1919 ntfs_error(sb, "Corrupt attribute list "
1923 /* Now copy the attribute list. */
1924 memcpy(ni->attr_list, (u8*)a + le16_to_cpu(
1925 a->data.resident.value_offset),
1927 a->data.resident.value_length));
1929 /* The attribute list is now setup in memory. */
1931 * FIXME: I don't know if this case is actually possible.
1932 * According to logic it is not possible but I have seen too
1933 * many weird things in MS software to rely on logic... Thus we
1934 * perform a manual search and make sure the first $MFT/$DATA
1935 * extent is in the base inode. If it is not we abort with an
1936 * error and if we ever see a report of this error we will need
1937 * to do some magic in order to have the necessary mft record
1938 * loaded and in the right place in the page cache. But
1939 * hopefully logic will prevail and this never happens...
1941 al_entry = (ATTR_LIST_ENTRY*)ni->attr_list;
1942 al_end = (u8*)al_entry + ni->attr_list_size;
1943 for (;; al_entry = next_al_entry) {
1944 /* Out of bounds check. */
1945 if ((u8*)al_entry < ni->attr_list ||
1946 (u8*)al_entry > al_end)
1947 goto em_put_err_out;
1948 /* Catch the end of the attribute list. */
1949 if ((u8*)al_entry == al_end)
1950 goto em_put_err_out;
1951 if (!al_entry->length)
1952 goto em_put_err_out;
1953 if ((u8*)al_entry + 6 > al_end || (u8*)al_entry +
1954 le16_to_cpu(al_entry->length) > al_end)
1955 goto em_put_err_out;
1956 next_al_entry = (ATTR_LIST_ENTRY*)((u8*)al_entry +
1957 le16_to_cpu(al_entry->length));
1958 if (le32_to_cpu(al_entry->type) >
1959 const_le32_to_cpu(AT_DATA))
1960 goto em_put_err_out;
1961 if (AT_DATA != al_entry->type)
1963 /* We want an unnamed attribute. */
1964 if (al_entry->name_length)
1965 goto em_put_err_out;
1966 /* Want the first entry, i.e. lowest_vcn == 0. */
1967 if (al_entry->lowest_vcn)
1968 goto em_put_err_out;
1969 /* First entry has to be in the base mft record. */
1970 if (MREF_LE(al_entry->mft_reference) != vi->i_ino) {
1971 /* MFT references do not match, logic fails. */
1972 ntfs_error(sb, "BUG: The first $DATA extent "
1973 "of $MFT is not in the base "
1974 "mft record. Please report "
1975 "you saw this message to "
1976 "linux-ntfs-dev@lists."
1980 /* Sequence numbers must match. */
1981 if (MSEQNO_LE(al_entry->mft_reference) !=
1983 goto em_put_err_out;
1984 /* Got it. All is ok. We can stop now. */
1990 ntfs_attr_reinit_search_ctx(ctx);
1992 /* Now load all attribute extents. */
1994 next_vcn = last_vcn = highest_vcn = 0;
1995 while (!(err = ntfs_attr_lookup(AT_DATA, NULL, 0, 0, next_vcn, NULL, 0,
1997 runlist_element *nrl;
1999 /* Cache the current attribute. */
2001 /* $MFT must be non-resident. */
2002 if (!a->non_resident) {
2003 ntfs_error(sb, "$MFT must be non-resident but a "
2004 "resident extent was found. $MFT is "
2005 "corrupt. Run chkdsk.");
2008 /* $MFT must be uncompressed and unencrypted. */
2009 if (a->flags & ATTR_COMPRESSION_MASK ||
2010 a->flags & ATTR_IS_ENCRYPTED ||
2011 a->flags & ATTR_IS_SPARSE) {
2012 ntfs_error(sb, "$MFT must be uncompressed, "
2013 "non-sparse, and unencrypted but a "
2014 "compressed/sparse/encrypted extent "
2015 "was found. $MFT is corrupt. Run "
2020 * Decompress the mapping pairs array of this extent and merge
2021 * the result into the existing runlist. No need for locking
2022 * as we have exclusive access to the inode at this time and we
2023 * are a mount in progress task, too.
2025 nrl = ntfs_mapping_pairs_decompress(vol, a, ni->runlist.rl);
2027 ntfs_error(sb, "ntfs_mapping_pairs_decompress() "
2028 "failed with error code %ld. $MFT is "
2029 "corrupt.", PTR_ERR(nrl));
2032 ni->runlist.rl = nrl;
2034 /* Are we in the first extent? */
2036 if (a->data.non_resident.lowest_vcn) {
2037 ntfs_error(sb, "First extent of $DATA "
2038 "attribute has non zero "
2039 "lowest_vcn. $MFT is corrupt. "
2040 "You should run chkdsk.");
2043 /* Get the last vcn in the $DATA attribute. */
2044 last_vcn = sle64_to_cpu(
2045 a->data.non_resident.allocated_size)
2046 >> vol->cluster_size_bits;
2047 /* Fill in the inode size. */
2048 vi->i_size = sle64_to_cpu(
2049 a->data.non_resident.data_size);
2050 ni->initialized_size = sle64_to_cpu(
2051 a->data.non_resident.initialized_size);
2052 ni->allocated_size = sle64_to_cpu(
2053 a->data.non_resident.allocated_size);
2055 * Verify the number of mft records does not exceed
2058 if ((vi->i_size >> vol->mft_record_size_bits) >=
2060 ntfs_error(sb, "$MFT is too big! Aborting.");
2064 * We have got the first extent of the runlist for
2065 * $MFT which means it is now relatively safe to call
2066 * the normal ntfs_read_inode() function.
2067 * Complete reading the inode, this will actually
2068 * re-read the mft record for $MFT, this time entering
2069 * it into the page cache with which we complete the
2070 * kick start of the volume. It should be safe to do
2071 * this now as the first extent of $MFT/$DATA is
2072 * already known and we would hope that we don't need
2073 * further extents in order to find the other
2074 * attributes belonging to $MFT. Only time will tell if
2075 * this is really the case. If not we will have to play
2076 * magic at this point, possibly duplicating a lot of
2077 * ntfs_read_inode() at this point. We will need to
2078 * ensure we do enough of its work to be able to call
2079 * ntfs_read_inode() on extents of $MFT/$DATA. But lets
2080 * hope this never happens...
2082 ntfs_read_locked_inode(vi);
2083 if (is_bad_inode(vi)) {
2084 ntfs_error(sb, "ntfs_read_inode() of $MFT "
2085 "failed. BUG or corrupt $MFT. "
2086 "Run chkdsk and if no errors "
2087 "are found, please report you "
2088 "saw this message to "
2089 "linux-ntfs-dev@lists."
2091 ntfs_attr_put_search_ctx(ctx);
2092 /* Revert to the safe super operations. */
2097 * Re-initialize some specifics about $MFT's inode as
2098 * ntfs_read_inode() will have set up the default ones.
2100 /* Set uid and gid to root. */
2101 vi->i_uid = vi->i_gid = 0;
2102 /* Regular file. No access for anyone. */
2103 vi->i_mode = S_IFREG;
2104 /* No VFS initiated operations allowed for $MFT. */
2105 vi->i_op = &ntfs_empty_inode_ops;
2106 vi->i_fop = &ntfs_empty_file_ops;
2109 /* Get the lowest vcn for the next extent. */
2110 highest_vcn = sle64_to_cpu(a->data.non_resident.highest_vcn);
2111 next_vcn = highest_vcn + 1;
2113 /* Only one extent or error, which we catch below. */
2117 /* Avoid endless loops due to corruption. */
2118 if (next_vcn < sle64_to_cpu(
2119 a->data.non_resident.lowest_vcn)) {
2120 ntfs_error(sb, "$MFT has corrupt attribute list "
2121 "attribute. Run chkdsk.");
2125 if (err != -ENOENT) {
2126 ntfs_error(sb, "Failed to lookup $MFT/$DATA attribute extent. "
2127 "$MFT is corrupt. Run chkdsk.");
2131 ntfs_error(sb, "$MFT/$DATA attribute not found. $MFT is "
2132 "corrupt. Run chkdsk.");
2135 if (highest_vcn && highest_vcn != last_vcn - 1) {
2136 ntfs_error(sb, "Failed to load the complete runlist for "
2137 "$MFT/$DATA. Driver bug or corrupt $MFT. "
2139 ntfs_debug("highest_vcn = 0x%llx, last_vcn - 1 = 0x%llx",
2140 (unsigned long long)highest_vcn,
2141 (unsigned long long)last_vcn - 1);
2144 ntfs_attr_put_search_ctx(ctx);
2145 ntfs_debug("Done.");
2150 ntfs_error(sb, "Couldn't find first extent of $DATA attribute in "
2151 "attribute list. $MFT is corrupt. Run chkdsk.");
2153 ntfs_attr_put_search_ctx(ctx);
2155 ntfs_error(sb, "Failed. Marking inode as bad.");
2162 * ntfs_put_inode - handler for when the inode reference count is decremented
2165 * The VFS calls ntfs_put_inode() every time the inode reference count (i_count)
2166 * is about to be decremented (but before the decrement itself.
2168 * If the inode @vi is a directory with two references, one of which is being
2169 * dropped, we need to put the attribute inode for the directory index bitmap,
2170 * if it is present, otherwise the directory inode would remain pinned for
2173 void ntfs_put_inode(struct inode *vi)
2175 if (S_ISDIR(vi->i_mode) && atomic_read(&vi->i_count) == 2) {
2176 ntfs_inode *ni = NTFS_I(vi);
2177 if (NInoIndexAllocPresent(ni)) {
2178 struct inode *bvi = NULL;
2179 mutex_lock(&vi->i_mutex);
2180 if (atomic_read(&vi->i_count) == 2) {
2181 bvi = ni->itype.index.bmp_ino;
2183 ni->itype.index.bmp_ino = NULL;
2185 mutex_unlock(&vi->i_mutex);
2192 static void __ntfs_clear_inode(ntfs_inode *ni)
2194 /* Free all alocated memory. */
2195 down_write(&ni->runlist.lock);
2196 if (ni->runlist.rl) {
2197 ntfs_free(ni->runlist.rl);
2198 ni->runlist.rl = NULL;
2200 up_write(&ni->runlist.lock);
2202 if (ni->attr_list) {
2203 ntfs_free(ni->attr_list);
2204 ni->attr_list = NULL;
2207 down_write(&ni->attr_list_rl.lock);
2208 if (ni->attr_list_rl.rl) {
2209 ntfs_free(ni->attr_list_rl.rl);
2210 ni->attr_list_rl.rl = NULL;
2212 up_write(&ni->attr_list_rl.lock);
2214 if (ni->name_len && ni->name != I30) {
2221 void ntfs_clear_extent_inode(ntfs_inode *ni)
2223 ntfs_debug("Entering for inode 0x%lx.", ni->mft_no);
2225 BUG_ON(NInoAttr(ni));
2226 BUG_ON(ni->nr_extents != -1);
2229 if (NInoDirty(ni)) {
2230 if (!is_bad_inode(VFS_I(ni->ext.base_ntfs_ino)))
2231 ntfs_error(ni->vol->sb, "Clearing dirty extent inode! "
2232 "Losing data! This is a BUG!!!");
2233 // FIXME: Do something!!!
2235 #endif /* NTFS_RW */
2237 __ntfs_clear_inode(ni);
2240 ntfs_destroy_extent_inode(ni);
2244 * ntfs_clear_big_inode - clean up the ntfs specific part of an inode
2245 * @vi: vfs inode pending annihilation
2247 * When the VFS is going to remove an inode from memory, ntfs_clear_big_inode()
2248 * is called, which deallocates all memory belonging to the NTFS specific part
2249 * of the inode and returns.
2251 * If the MFT record is dirty, we commit it before doing anything else.
2253 void ntfs_clear_big_inode(struct inode *vi)
2255 ntfs_inode *ni = NTFS_I(vi);
2258 * If the inode @vi is an index inode we need to put the attribute
2259 * inode for the index bitmap, if it is present, otherwise the index
2260 * inode would disappear and the attribute inode for the index bitmap
2261 * would no longer be referenced from anywhere and thus it would remain
2264 if (NInoAttr(ni) && (ni->type == AT_INDEX_ALLOCATION) &&
2265 NInoIndexAllocPresent(ni) && ni->itype.index.bmp_ino) {
2266 iput(ni->itype.index.bmp_ino);
2267 ni->itype.index.bmp_ino = NULL;
2270 if (NInoDirty(ni)) {
2271 BOOL was_bad = (is_bad_inode(vi));
2273 /* Committing the inode also commits all extent inodes. */
2274 ntfs_commit_inode(vi);
2276 if (!was_bad && (is_bad_inode(vi) || NInoDirty(ni))) {
2277 ntfs_error(vi->i_sb, "Failed to commit dirty inode "
2278 "0x%lx. Losing data!", vi->i_ino);
2279 // FIXME: Do something!!!
2282 #endif /* NTFS_RW */
2284 /* No need to lock at this stage as no one else has a reference. */
2285 if (ni->nr_extents > 0) {
2288 for (i = 0; i < ni->nr_extents; i++)
2289 ntfs_clear_extent_inode(ni->ext.extent_ntfs_inos[i]);
2290 kfree(ni->ext.extent_ntfs_inos);
2293 __ntfs_clear_inode(ni);
2296 /* Release the base inode if we are holding it. */
2297 if (ni->nr_extents == -1) {
2298 iput(VFS_I(ni->ext.base_ntfs_ino));
2300 ni->ext.base_ntfs_ino = NULL;
2307 * ntfs_show_options - show mount options in /proc/mounts
2308 * @sf: seq_file in which to write our mount options
2309 * @mnt: vfs mount whose mount options to display
2311 * Called by the VFS once for each mounted ntfs volume when someone reads
2312 * /proc/mounts in order to display the NTFS specific mount options of each
2313 * mount. The mount options of the vfs mount @mnt are written to the seq file
2314 * @sf and success is returned.
2316 int ntfs_show_options(struct seq_file *sf, struct vfsmount *mnt)
2318 ntfs_volume *vol = NTFS_SB(mnt->mnt_sb);
2321 seq_printf(sf, ",uid=%i", vol->uid);
2322 seq_printf(sf, ",gid=%i", vol->gid);
2323 if (vol->fmask == vol->dmask)
2324 seq_printf(sf, ",umask=0%o", vol->fmask);
2326 seq_printf(sf, ",fmask=0%o", vol->fmask);
2327 seq_printf(sf, ",dmask=0%o", vol->dmask);
2329 seq_printf(sf, ",nls=%s", vol->nls_map->charset);
2330 if (NVolCaseSensitive(vol))
2331 seq_printf(sf, ",case_sensitive");
2332 if (NVolShowSystemFiles(vol))
2333 seq_printf(sf, ",show_sys_files");
2334 if (!NVolSparseEnabled(vol))
2335 seq_printf(sf, ",disable_sparse");
2336 for (i = 0; on_errors_arr[i].val; i++) {
2337 if (on_errors_arr[i].val & vol->on_errors)
2338 seq_printf(sf, ",errors=%s", on_errors_arr[i].str);
2340 seq_printf(sf, ",mft_zone_multiplier=%i", vol->mft_zone_multiplier);
2346 static const char *es = " Leaving inconsistent metadata. Unmount and run "
2350 * ntfs_truncate - called when the i_size of an ntfs inode is changed
2351 * @vi: inode for which the i_size was changed
2353 * We only support i_size changes for normal files at present, i.e. not
2354 * compressed and not encrypted. This is enforced in ntfs_setattr(), see
2357 * The kernel guarantees that @vi is a regular file (S_ISREG() is true) and
2358 * that the change is allowed.
2360 * This implies for us that @vi is a file inode rather than a directory, index,
2361 * or attribute inode as well as that @vi is a base inode.
2363 * Returns 0 on success or -errno on error.
2365 * Called with ->i_mutex held. In all but one case ->i_alloc_sem is held for
2366 * writing. The only case in the kernel where ->i_alloc_sem is not held is
2367 * mm/filemap.c::generic_file_buffered_write() where vmtruncate() is called
2368 * with the current i_size as the offset. The analogous place in NTFS is in
2369 * fs/ntfs/file.c::ntfs_file_buffered_write() where we call vmtruncate() again
2370 * without holding ->i_alloc_sem.
2372 int ntfs_truncate(struct inode *vi)
2374 s64 new_size, old_size, nr_freed, new_alloc_size, old_alloc_size;
2376 unsigned long flags;
2377 ntfs_inode *base_ni, *ni = NTFS_I(vi);
2378 ntfs_volume *vol = ni->vol;
2379 ntfs_attr_search_ctx *ctx;
2382 const char *te = " Leaving file length out of sync with i_size.";
2383 int err, mp_size, size_change, alloc_change;
2386 ntfs_debug("Entering for inode 0x%lx.", vi->i_ino);
2387 BUG_ON(NInoAttr(ni));
2388 BUG_ON(S_ISDIR(vi->i_mode));
2389 BUG_ON(NInoMstProtected(ni));
2390 BUG_ON(ni->nr_extents < 0);
2393 * Lock the runlist for writing and map the mft record to ensure it is
2394 * safe to mess with the attribute runlist and sizes.
2396 down_write(&ni->runlist.lock);
2400 base_ni = ni->ext.base_ntfs_ino;
2401 m = map_mft_record(base_ni);
2404 ntfs_error(vi->i_sb, "Failed to map mft record for inode 0x%lx "
2405 "(error code %d).%s", vi->i_ino, err, te);
2410 ctx = ntfs_attr_get_search_ctx(base_ni, m);
2411 if (unlikely(!ctx)) {
2412 ntfs_error(vi->i_sb, "Failed to allocate a search context for "
2413 "inode 0x%lx (not enough memory).%s",
2418 err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
2419 CASE_SENSITIVE, 0, NULL, 0, ctx);
2420 if (unlikely(err)) {
2421 if (err == -ENOENT) {
2422 ntfs_error(vi->i_sb, "Open attribute is missing from "
2423 "mft record. Inode 0x%lx is corrupt. "
2424 "Run chkdsk.%s", vi->i_ino, te);
2427 ntfs_error(vi->i_sb, "Failed to lookup attribute in "
2428 "inode 0x%lx (error code %d).%s",
2429 vi->i_ino, err, te);
2435 * The i_size of the vfs inode is the new size for the attribute value.
2437 new_size = i_size_read(vi);
2438 /* The current size of the attribute value is the old size. */
2439 old_size = ntfs_attr_size(a);
2440 /* Calculate the new allocated size. */
2441 if (NInoNonResident(ni))
2442 new_alloc_size = (new_size + vol->cluster_size - 1) &
2443 ~(s64)vol->cluster_size_mask;
2445 new_alloc_size = (new_size + 7) & ~7;
2446 /* The current allocated size is the old allocated size. */
2447 read_lock_irqsave(&ni->size_lock, flags);
2448 old_alloc_size = ni->allocated_size;
2449 read_unlock_irqrestore(&ni->size_lock, flags);
2451 * The change in the file size. This will be 0 if no change, >0 if the
2452 * size is growing, and <0 if the size is shrinking.
2455 if (new_size - old_size >= 0) {
2457 if (new_size == old_size)
2460 /* As above for the allocated size. */
2462 if (new_alloc_size - old_alloc_size >= 0) {
2464 if (new_alloc_size == old_alloc_size)
2468 * If neither the size nor the allocation are being changed there is
2471 if (!size_change && !alloc_change)
2473 /* If the size is changing, check if new size is allowed in $AttrDef. */
2475 err = ntfs_attr_size_bounds_check(vol, ni->type, new_size);
2476 if (unlikely(err)) {
2477 if (err == -ERANGE) {
2478 ntfs_error(vol->sb, "Truncate would cause the "
2479 "inode 0x%lx to %simum size "
2480 "for its attribute type "
2481 "(0x%x). Aborting truncate.",
2483 new_size > old_size ? "exceed "
2484 "the max" : "go under the min",
2485 le32_to_cpu(ni->type));
2488 ntfs_error(vol->sb, "Inode 0x%lx has unknown "
2489 "attribute type 0x%x. "
2490 "Aborting truncate.",
2492 le32_to_cpu(ni->type));
2495 /* Reset the vfs inode size to the old size. */
2496 i_size_write(vi, old_size);
2500 if (NInoCompressed(ni) || NInoEncrypted(ni)) {
2501 ntfs_warning(vi->i_sb, "Changes in inode size are not "
2502 "supported yet for %s files, ignoring.",
2503 NInoCompressed(ni) ? "compressed" :
2508 if (a->non_resident)
2509 goto do_non_resident_truncate;
2510 BUG_ON(NInoNonResident(ni));
2511 /* Resize the attribute record to best fit the new attribute size. */
2512 if (new_size < vol->mft_record_size &&
2513 !ntfs_resident_attr_value_resize(m, a, new_size)) {
2514 unsigned long flags;
2516 /* The resize succeeded! */
2517 flush_dcache_mft_record_page(ctx->ntfs_ino);
2518 mark_mft_record_dirty(ctx->ntfs_ino);
2519 write_lock_irqsave(&ni->size_lock, flags);
2520 /* Update the sizes in the ntfs inode and all is done. */
2521 ni->allocated_size = le32_to_cpu(a->length) -
2522 le16_to_cpu(a->data.resident.value_offset);
2524 * Note ntfs_resident_attr_value_resize() has already done any
2525 * necessary data clearing in the attribute record. When the
2526 * file is being shrunk vmtruncate() will already have cleared
2527 * the top part of the last partial page, i.e. since this is
2528 * the resident case this is the page with index 0. However,
2529 * when the file is being expanded, the page cache page data
2530 * between the old data_size, i.e. old_size, and the new_size
2531 * has not been zeroed. Fortunately, we do not need to zero it
2532 * either since on one hand it will either already be zero due
2533 * to both readpage and writepage clearing partial page data
2534 * beyond i_size in which case there is nothing to do or in the
2535 * case of the file being mmap()ped at the same time, POSIX
2536 * specifies that the behaviour is unspecified thus we do not
2537 * have to do anything. This means that in our implementation
2538 * in the rare case that the file is mmap()ped and a write
2539 * occured into the mmap()ped region just beyond the file size
2540 * and writepage has not yet been called to write out the page
2541 * (which would clear the area beyond the file size) and we now
2542 * extend the file size to incorporate this dirty region
2543 * outside the file size, a write of the page would result in
2544 * this data being written to disk instead of being cleared.
2545 * Given both POSIX and the Linux mmap(2) man page specify that
2546 * this corner case is undefined, we choose to leave it like
2547 * that as this is much simpler for us as we cannot lock the
2548 * relevant page now since we are holding too many ntfs locks
2549 * which would result in a lock reversal deadlock.
2551 ni->initialized_size = new_size;
2552 write_unlock_irqrestore(&ni->size_lock, flags);
2555 /* If the above resize failed, this must be an attribute extension. */
2556 BUG_ON(size_change < 0);
2558 * We have to drop all the locks so we can call
2559 * ntfs_attr_make_non_resident(). This could be optimised by try-
2560 * locking the first page cache page and only if that fails dropping
2561 * the locks, locking the page, and redoing all the locking and
2562 * lookups. While this would be a huge optimisation, it is not worth
2563 * it as this is definitely a slow code path as it only ever can happen
2564 * once for any given file.
2566 ntfs_attr_put_search_ctx(ctx);
2567 unmap_mft_record(base_ni);
2568 up_write(&ni->runlist.lock);
2570 * Not enough space in the mft record, try to make the attribute
2571 * non-resident and if successful restart the truncation process.
2573 err = ntfs_attr_make_non_resident(ni, old_size);
2575 goto retry_truncate;
2577 * Could not make non-resident. If this is due to this not being
2578 * permitted for this attribute type or there not being enough space,
2579 * try to make other attributes non-resident. Otherwise fail.
2581 if (unlikely(err != -EPERM && err != -ENOSPC)) {
2582 ntfs_error(vol->sb, "Cannot truncate inode 0x%lx, attribute "
2583 "type 0x%x, because the conversion from "
2584 "resident to non-resident attribute failed "
2585 "with error code %i.", vi->i_ino,
2586 (unsigned)le32_to_cpu(ni->type), err);
2591 /* TODO: Not implemented from here, abort. */
2593 ntfs_error(vol->sb, "Not enough space in the mft record/on "
2594 "disk for the non-resident attribute value. "
2595 "This case is not implemented yet.");
2596 else /* if (err == -EPERM) */
2597 ntfs_error(vol->sb, "This attribute type may not be "
2598 "non-resident. This case is not implemented "
2603 // TODO: Attempt to make other attributes non-resident.
2605 goto do_resident_extend;
2607 * Both the attribute list attribute and the standard information
2608 * attribute must remain in the base inode. Thus, if this is one of
2609 * these attributes, we have to try to move other attributes out into
2610 * extent mft records instead.
2612 if (ni->type == AT_ATTRIBUTE_LIST ||
2613 ni->type == AT_STANDARD_INFORMATION) {
2614 // TODO: Attempt to move other attributes into extent mft
2618 goto do_resident_extend;
2621 // TODO: Attempt to move this attribute to an extent mft record, but
2622 // only if it is not already the only attribute in an mft record in
2623 // which case there would be nothing to gain.
2626 goto do_resident_extend;
2627 /* There is nothing we can do to make enough space. )-: */
2630 do_non_resident_truncate:
2631 BUG_ON(!NInoNonResident(ni));
2632 if (alloc_change < 0) {
2633 highest_vcn = sle64_to_cpu(a->data.non_resident.highest_vcn);
2634 if (highest_vcn > 0 &&
2635 old_alloc_size >> vol->cluster_size_bits >
2638 * This attribute has multiple extents. Not yet
2641 ntfs_error(vol->sb, "Cannot truncate inode 0x%lx, "
2642 "attribute type 0x%x, because the "
2643 "attribute is highly fragmented (it "
2644 "consists of multiple extents) and "
2645 "this case is not implemented yet.",
2647 (unsigned)le32_to_cpu(ni->type));
2653 * If the size is shrinking, need to reduce the initialized_size and
2654 * the data_size before reducing the allocation.
2656 if (size_change < 0) {
2658 * Make the valid size smaller (i_size is already up-to-date).
2660 write_lock_irqsave(&ni->size_lock, flags);
2661 if (new_size < ni->initialized_size) {
2662 ni->initialized_size = new_size;
2663 a->data.non_resident.initialized_size =
2664 cpu_to_sle64(new_size);
2666 a->data.non_resident.data_size = cpu_to_sle64(new_size);
2667 write_unlock_irqrestore(&ni->size_lock, flags);
2668 flush_dcache_mft_record_page(ctx->ntfs_ino);
2669 mark_mft_record_dirty(ctx->ntfs_ino);
2670 /* If the allocated size is not changing, we are done. */
2674 * If the size is shrinking it makes no sense for the
2675 * allocation to be growing.
2677 BUG_ON(alloc_change > 0);
2678 } else /* if (size_change >= 0) */ {
2680 * The file size is growing or staying the same but the
2681 * allocation can be shrinking, growing or staying the same.
2683 if (alloc_change > 0) {
2685 * We need to extend the allocation and possibly update
2686 * the data size. If we are updating the data size,
2687 * since we are not touching the initialized_size we do
2688 * not need to worry about the actual data on disk.
2689 * And as far as the page cache is concerned, there
2690 * will be no pages beyond the old data size and any
2691 * partial region in the last page between the old and
2692 * new data size (or the end of the page if the new
2693 * data size is outside the page) does not need to be
2694 * modified as explained above for the resident
2695 * attribute truncate case. To do this, we simply drop
2696 * the locks we hold and leave all the work to our
2697 * friendly helper ntfs_attr_extend_allocation().
2699 ntfs_attr_put_search_ctx(ctx);
2700 unmap_mft_record(base_ni);
2701 up_write(&ni->runlist.lock);
2702 err = ntfs_attr_extend_allocation(ni, new_size,
2703 size_change > 0 ? new_size : -1, -1);
2705 * ntfs_attr_extend_allocation() will have done error
2713 /* alloc_change < 0 */
2714 /* Free the clusters. */
2715 nr_freed = ntfs_cluster_free(ni, new_alloc_size >>
2716 vol->cluster_size_bits, -1, ctx);
2719 if (unlikely(nr_freed < 0)) {
2720 ntfs_error(vol->sb, "Failed to release cluster(s) (error code "
2721 "%lli). Unmount and run chkdsk to recover "
2722 "the lost cluster(s).", (long long)nr_freed);
2726 /* Truncate the runlist. */
2727 err = ntfs_rl_truncate_nolock(vol, &ni->runlist,
2728 new_alloc_size >> vol->cluster_size_bits);
2730 * If the runlist truncation failed and/or the search context is no
2731 * longer valid, we cannot resize the attribute record or build the
2732 * mapping pairs array thus we mark the inode bad so that no access to
2733 * the freed clusters can happen.
2735 if (unlikely(err || IS_ERR(m))) {
2736 ntfs_error(vol->sb, "Failed to %s (error code %li).%s",
2738 "restore attribute search context" :
2739 "truncate attribute runlist",
2740 IS_ERR(m) ? PTR_ERR(m) : err, es);
2744 /* Get the size for the shrunk mapping pairs array for the runlist. */
2745 mp_size = ntfs_get_size_for_mapping_pairs(vol, ni->runlist.rl, 0, -1);
2746 if (unlikely(mp_size <= 0)) {
2747 ntfs_error(vol->sb, "Cannot shrink allocation of inode 0x%lx, "
2748 "attribute type 0x%x, because determining the "
2749 "size for the mapping pairs failed with error "
2750 "code %i.%s", vi->i_ino,
2751 (unsigned)le32_to_cpu(ni->type), mp_size, es);
2756 * Shrink the attribute record for the new mapping pairs array. Note,
2757 * this cannot fail since we are making the attribute smaller thus by
2758 * definition there is enough space to do so.
2760 attr_len = le32_to_cpu(a->length);
2761 err = ntfs_attr_record_resize(m, a, mp_size +
2762 le16_to_cpu(a->data.non_resident.mapping_pairs_offset));
2765 * Generate the mapping pairs array directly into the attribute record.
2767 err = ntfs_mapping_pairs_build(vol, (u8*)a +
2768 le16_to_cpu(a->data.non_resident.mapping_pairs_offset),
2769 mp_size, ni->runlist.rl, 0, -1, NULL);
2770 if (unlikely(err)) {
2771 ntfs_error(vol->sb, "Cannot shrink allocation of inode 0x%lx, "
2772 "attribute type 0x%x, because building the "
2773 "mapping pairs failed with error code %i.%s",
2774 vi->i_ino, (unsigned)le32_to_cpu(ni->type),
2779 /* Update the allocated/compressed size as well as the highest vcn. */
2780 a->data.non_resident.highest_vcn = cpu_to_sle64((new_alloc_size >>
2781 vol->cluster_size_bits) - 1);
2782 write_lock_irqsave(&ni->size_lock, flags);
2783 ni->allocated_size = new_alloc_size;
2784 a->data.non_resident.allocated_size = cpu_to_sle64(new_alloc_size);
2785 if (NInoSparse(ni) || NInoCompressed(ni)) {
2787 ni->itype.compressed.size -= nr_freed <<
2788 vol->cluster_size_bits;
2789 BUG_ON(ni->itype.compressed.size < 0);
2790 a->data.non_resident.compressed_size = cpu_to_sle64(
2791 ni->itype.compressed.size);
2792 vi->i_blocks = ni->itype.compressed.size >> 9;
2795 vi->i_blocks = new_alloc_size >> 9;
2796 write_unlock_irqrestore(&ni->size_lock, flags);
2798 * We have shrunk the allocation. If this is a shrinking truncate we
2799 * have already dealt with the initialized_size and the data_size above
2800 * and we are done. If the truncate is only changing the allocation
2801 * and not the data_size, we are also done. If this is an extending
2802 * truncate, need to extend the data_size now which is ensured by the
2803 * fact that @size_change is positive.
2807 * If the size is growing, need to update it now. If it is shrinking,
2808 * we have already updated it above (before the allocation change).
2810 if (size_change > 0)
2811 a->data.non_resident.data_size = cpu_to_sle64(new_size);
2812 /* Ensure the modified mft record is written out. */
2813 flush_dcache_mft_record_page(ctx->ntfs_ino);
2814 mark_mft_record_dirty(ctx->ntfs_ino);
2816 ntfs_attr_put_search_ctx(ctx);
2817 unmap_mft_record(base_ni);
2818 up_write(&ni->runlist.lock);
2820 /* Update the mtime and ctime on the base inode. */
2821 /* normally ->truncate shouldn't update ctime or mtime,
2822 * but ntfs did before so it got a copy & paste version
2823 * of file_update_time. one day someone should fix this
2826 if (!IS_NOCMTIME(VFS_I(base_ni)) && !IS_RDONLY(VFS_I(base_ni))) {
2827 struct timespec now = current_fs_time(VFS_I(base_ni)->i_sb);
2830 if (!timespec_equal(&VFS_I(base_ni)->i_mtime, &now) ||
2831 !timespec_equal(&VFS_I(base_ni)->i_ctime, &now))
2833 VFS_I(base_ni)->i_mtime = now;
2834 VFS_I(base_ni)->i_ctime = now;
2837 mark_inode_dirty_sync(VFS_I(base_ni));
2841 NInoClearTruncateFailed(ni);
2842 ntfs_debug("Done.");
2848 if (err != -ENOMEM && err != -EOPNOTSUPP) {
2850 make_bad_inode(VFS_I(base_ni));
2853 if (err != -EOPNOTSUPP)
2854 NInoSetTruncateFailed(ni);
2855 else if (old_size >= 0)
2856 i_size_write(vi, old_size);
2859 ntfs_attr_put_search_ctx(ctx);
2861 unmap_mft_record(base_ni);
2862 up_write(&ni->runlist.lock);
2864 ntfs_debug("Failed. Returning error code %i.", err);
2867 if (err != -ENOMEM && err != -EOPNOTSUPP) {
2869 make_bad_inode(VFS_I(base_ni));
2872 if (err != -EOPNOTSUPP)
2873 NInoSetTruncateFailed(ni);
2875 i_size_write(vi, old_size);
2880 * ntfs_truncate_vfs - wrapper for ntfs_truncate() that has no return value
2881 * @vi: inode for which the i_size was changed
2883 * Wrapper for ntfs_truncate() that has no return value.
2885 * See ntfs_truncate() description above for details.
2887 void ntfs_truncate_vfs(struct inode *vi) {
2892 * ntfs_setattr - called from notify_change() when an attribute is being changed
2893 * @dentry: dentry whose attributes to change
2894 * @attr: structure describing the attributes and the changes
2896 * We have to trap VFS attempts to truncate the file described by @dentry as
2897 * soon as possible, because we do not implement changes in i_size yet. So we
2898 * abort all i_size changes here.
2900 * We also abort all changes of user, group, and mode as we do not implement
2901 * the NTFS ACLs yet.
2903 * Called with ->i_mutex held. For the ATTR_SIZE (i.e. ->truncate) case, also
2904 * called with ->i_alloc_sem held for writing.
2906 * Basically this is a copy of generic notify_change() and inode_setattr()
2907 * functionality, except we intercept and abort changes in i_size.
2909 int ntfs_setattr(struct dentry *dentry, struct iattr *attr)
2911 struct inode *vi = dentry->d_inode;
2913 unsigned int ia_valid = attr->ia_valid;
2915 err = inode_change_ok(vi, attr);
2918 /* We do not support NTFS ACLs yet. */
2919 if (ia_valid & (ATTR_UID | ATTR_GID | ATTR_MODE)) {
2920 ntfs_warning(vi->i_sb, "Changes in user/group/mode are not "
2921 "supported yet, ignoring.");
2925 if (ia_valid & ATTR_SIZE) {
2926 if (attr->ia_size != i_size_read(vi)) {
2927 ntfs_inode *ni = NTFS_I(vi);
2929 * FIXME: For now we do not support resizing of
2930 * compressed or encrypted files yet.
2932 if (NInoCompressed(ni) || NInoEncrypted(ni)) {
2933 ntfs_warning(vi->i_sb, "Changes in inode size "
2934 "are not supported yet for "
2935 "%s files, ignoring.",
2936 NInoCompressed(ni) ?
2937 "compressed" : "encrypted");
2940 err = vmtruncate(vi, attr->ia_size);
2941 if (err || ia_valid == ATTR_SIZE)
2945 * We skipped the truncate but must still update
2948 ia_valid |= ATTR_MTIME | ATTR_CTIME;
2951 if (ia_valid & ATTR_ATIME)
2952 vi->i_atime = timespec_trunc(attr->ia_atime,
2953 vi->i_sb->s_time_gran);
2954 if (ia_valid & ATTR_MTIME)
2955 vi->i_mtime = timespec_trunc(attr->ia_mtime,
2956 vi->i_sb->s_time_gran);
2957 if (ia_valid & ATTR_CTIME)
2958 vi->i_ctime = timespec_trunc(attr->ia_ctime,
2959 vi->i_sb->s_time_gran);
2960 mark_inode_dirty(vi);
2966 * ntfs_write_inode - write out a dirty inode
2967 * @vi: inode to write out
2968 * @sync: if true, write out synchronously
2970 * Write out a dirty inode to disk including any extent inodes if present.
2972 * If @sync is true, commit the inode to disk and wait for io completion. This
2973 * is done using write_mft_record().
2975 * If @sync is false, just schedule the write to happen but do not wait for i/o
2976 * completion. In 2.6 kernels, scheduling usually happens just by virtue of
2977 * marking the page (and in this case mft record) dirty but we do not implement
2978 * this yet as write_mft_record() largely ignores the @sync parameter and
2979 * always performs synchronous writes.
2981 * Return 0 on success and -errno on error.
2983 int ntfs_write_inode(struct inode *vi, int sync)
2986 ntfs_inode *ni = NTFS_I(vi);
2987 ntfs_attr_search_ctx *ctx;
2989 STANDARD_INFORMATION *si;
2991 BOOL modified = FALSE;
2993 ntfs_debug("Entering for %sinode 0x%lx.", NInoAttr(ni) ? "attr " : "",
2996 * Dirty attribute inodes are written via their real inodes so just
2997 * clean them here. Access time updates are taken care off when the
2998 * real inode is written.
3002 ntfs_debug("Done.");
3005 /* Map, pin, and lock the mft record belonging to the inode. */
3006 m = map_mft_record(ni);
3011 /* Update the access times in the standard information attribute. */
3012 ctx = ntfs_attr_get_search_ctx(ni, m);
3013 if (unlikely(!ctx)) {
3017 err = ntfs_attr_lookup(AT_STANDARD_INFORMATION, NULL, 0,
3018 CASE_SENSITIVE, 0, NULL, 0, ctx);
3019 if (unlikely(err)) {
3020 ntfs_attr_put_search_ctx(ctx);
3023 si = (STANDARD_INFORMATION*)((u8*)ctx->attr +
3024 le16_to_cpu(ctx->attr->data.resident.value_offset));
3025 /* Update the access times if they have changed. */
3026 nt = utc2ntfs(vi->i_mtime);
3027 if (si->last_data_change_time != nt) {
3028 ntfs_debug("Updating mtime for inode 0x%lx: old = 0x%llx, "
3029 "new = 0x%llx", vi->i_ino, (long long)
3030 sle64_to_cpu(si->last_data_change_time),
3031 (long long)sle64_to_cpu(nt));
3032 si->last_data_change_time = nt;
3035 nt = utc2ntfs(vi->i_ctime);
3036 if (si->last_mft_change_time != nt) {
3037 ntfs_debug("Updating ctime for inode 0x%lx: old = 0x%llx, "
3038 "new = 0x%llx", vi->i_ino, (long long)
3039 sle64_to_cpu(si->last_mft_change_time),
3040 (long long)sle64_to_cpu(nt));
3041 si->last_mft_change_time = nt;
3044 nt = utc2ntfs(vi->i_atime);
3045 if (si->last_access_time != nt) {
3046 ntfs_debug("Updating atime for inode 0x%lx: old = 0x%llx, "
3047 "new = 0x%llx", vi->i_ino,
3048 (long long)sle64_to_cpu(si->last_access_time),
3049 (long long)sle64_to_cpu(nt));
3050 si->last_access_time = nt;
3054 * If we just modified the standard information attribute we need to
3055 * mark the mft record it is in dirty. We do this manually so that
3056 * mark_inode_dirty() is not called which would redirty the inode and
3057 * hence result in an infinite loop of trying to write the inode.
3058 * There is no need to mark the base inode nor the base mft record
3059 * dirty, since we are going to write this mft record below in any case
3060 * and the base mft record may actually not have been modified so it
3061 * might not need to be written out.
3062 * NOTE: It is not a problem when the inode for $MFT itself is being
3063 * written out as mark_ntfs_record_dirty() will only set I_DIRTY_PAGES
3064 * on the $MFT inode and hence ntfs_write_inode() will not be
3065 * re-invoked because of it which in turn is ok since the dirtied mft
3066 * record will be cleaned and written out to disk below, i.e. before
3067 * this function returns.
3069 if (modified && !NInoTestSetDirty(ctx->ntfs_ino))
3070 mark_ntfs_record_dirty(ctx->ntfs_ino->page,
3071 ctx->ntfs_ino->page_ofs);
3072 ntfs_attr_put_search_ctx(ctx);
3073 /* Now the access times are updated, write the base mft record. */
3075 err = write_mft_record(ni, m, sync);
3076 /* Write all attached extent mft records. */
3077 down(&ni->extent_lock);
3078 if (ni->nr_extents > 0) {
3079 ntfs_inode **extent_nis = ni->ext.extent_ntfs_inos;
3082 ntfs_debug("Writing %i extent inodes.", ni->nr_extents);
3083 for (i = 0; i < ni->nr_extents; i++) {
3084 ntfs_inode *tni = extent_nis[i];
3086 if (NInoDirty(tni)) {
3087 MFT_RECORD *tm = map_mft_record(tni);
3091 if (!err || err == -ENOMEM)
3095 ret = write_mft_record(tni, tm, sync);
3096 unmap_mft_record(tni);
3097 if (unlikely(ret)) {
3098 if (!err || err == -ENOMEM)
3104 up(&ni->extent_lock);
3105 unmap_mft_record(ni);
3108 ntfs_debug("Done.");
3111 unmap_mft_record(ni);
3113 if (err == -ENOMEM) {
3114 ntfs_warning(vi->i_sb, "Not enough memory to write inode. "
3115 "Marking the inode dirty again, so the VFS "
3117 mark_inode_dirty(vi);
3119 ntfs_error(vi->i_sb, "Failed (error code %i): Marking inode "
3120 "as bad. You should run chkdsk.", -err);
3122 NVolSetErrors(ni->vol);
3127 #endif /* NTFS_RW */