2 * inode.c - NTFS kernel inode handling. Part of the Linux-NTFS project.
4 * Copyright (c) 2001-2005 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>
39 * ntfs_test_inode - compare two (possibly fake) inodes for equality
40 * @vi: vfs inode which to test
41 * @na: ntfs attribute which is being tested with
43 * Compare the ntfs attribute embedded in the ntfs specific part of the vfs
44 * inode @vi for equality with the ntfs attribute @na.
46 * If searching for the normal file/directory inode, set @na->type to AT_UNUSED.
47 * @na->name and @na->name_len are then ignored.
49 * Return 1 if the attributes match and 0 if not.
51 * NOTE: This function runs with the inode_lock spin lock held so it is not
54 int ntfs_test_inode(struct inode *vi, ntfs_attr *na)
58 if (vi->i_ino != na->mft_no)
61 /* If !NInoAttr(ni), @vi is a normal file or directory inode. */
62 if (likely(!NInoAttr(ni))) {
63 /* If not looking for a normal inode this is a mismatch. */
64 if (unlikely(na->type != AT_UNUSED))
67 /* A fake inode describing an attribute. */
68 if (ni->type != na->type)
70 if (ni->name_len != na->name_len)
72 if (na->name_len && memcmp(ni->name, na->name,
73 na->name_len * sizeof(ntfschar)))
81 * ntfs_init_locked_inode - initialize an inode
82 * @vi: vfs inode to initialize
83 * @na: ntfs attribute which to initialize @vi to
85 * Initialize the vfs inode @vi with the values from the ntfs attribute @na in
86 * order to enable ntfs_test_inode() to do its work.
88 * If initializing the normal file/directory inode, set @na->type to AT_UNUSED.
89 * In that case, @na->name and @na->name_len should be set to NULL and 0,
90 * respectively. Although that is not strictly necessary as
91 * ntfs_read_inode_locked() will fill them in later.
93 * Return 0 on success and -errno on error.
95 * NOTE: This function runs with the inode_lock spin lock held so it is not
96 * allowed to sleep. (Hence the GFP_ATOMIC allocation.)
98 static int ntfs_init_locked_inode(struct inode *vi, ntfs_attr *na)
100 ntfs_inode *ni = NTFS_I(vi);
102 vi->i_ino = na->mft_no;
105 if (na->type == AT_INDEX_ALLOCATION)
106 NInoSetMstProtected(ni);
109 ni->name_len = na->name_len;
111 /* If initializing a normal inode, we are done. */
112 if (likely(na->type == AT_UNUSED)) {
114 BUG_ON(na->name_len);
118 /* It is a fake inode. */
122 * We have I30 global constant as an optimization as it is the name
123 * in >99.9% of named attributes! The other <0.1% incur a GFP_ATOMIC
124 * allocation but that is ok. And most attributes are unnamed anyway,
125 * thus the fraction of named attributes with name != I30 is actually
128 if (na->name_len && na->name != I30) {
132 i = na->name_len * sizeof(ntfschar);
133 ni->name = (ntfschar*)kmalloc(i + sizeof(ntfschar), GFP_ATOMIC);
136 memcpy(ni->name, na->name, i);
142 typedef int (*set_t)(struct inode *, void *);
143 static int ntfs_read_locked_inode(struct inode *vi);
144 static int ntfs_read_locked_attr_inode(struct inode *base_vi, struct inode *vi);
145 static int ntfs_read_locked_index_inode(struct inode *base_vi,
149 * ntfs_iget - obtain a struct inode corresponding to a specific normal inode
150 * @sb: super block of mounted volume
151 * @mft_no: mft record number / inode number to obtain
153 * Obtain the struct inode corresponding to a specific normal inode (i.e. a
154 * file or directory).
156 * If the inode is in the cache, it is just returned with an increased
157 * reference count. Otherwise, a new struct inode is allocated and initialized,
158 * and finally ntfs_read_locked_inode() is called to read in the inode and
159 * fill in the remainder of the inode structure.
161 * Return the struct inode on success. Check the return value with IS_ERR() and
162 * if true, the function failed and the error code is obtained from PTR_ERR().
164 struct inode *ntfs_iget(struct super_block *sb, unsigned long mft_no)
175 vi = iget5_locked(sb, mft_no, (test_t)ntfs_test_inode,
176 (set_t)ntfs_init_locked_inode, &na);
178 return ERR_PTR(-ENOMEM);
182 /* If this is a freshly allocated inode, need to read it now. */
183 if (vi->i_state & I_NEW) {
184 err = ntfs_read_locked_inode(vi);
185 unlock_new_inode(vi);
188 * There is no point in keeping bad inodes around if the failure was
189 * due to ENOMEM. We want to be able to retry again later.
191 if (unlikely(err == -ENOMEM)) {
199 * ntfs_attr_iget - obtain a struct inode corresponding to an attribute
200 * @base_vi: vfs base inode containing the attribute
201 * @type: attribute type
202 * @name: Unicode name of the attribute (NULL if unnamed)
203 * @name_len: length of @name in Unicode characters (0 if unnamed)
205 * Obtain the (fake) struct inode corresponding to the attribute specified by
206 * @type, @name, and @name_len, which is present in the base mft record
207 * specified by the vfs inode @base_vi.
209 * If the attribute inode is in the cache, it is just returned with an
210 * increased reference count. Otherwise, a new struct inode is allocated and
211 * initialized, and finally ntfs_read_locked_attr_inode() is called to read the
212 * attribute and fill in the inode structure.
214 * Note, for index allocation attributes, you need to use ntfs_index_iget()
215 * instead of ntfs_attr_iget() as working with indices is a lot more complex.
217 * Return the struct inode of the attribute inode on success. Check the return
218 * value with IS_ERR() and if true, the function failed and the error code is
219 * obtained from PTR_ERR().
221 struct inode *ntfs_attr_iget(struct inode *base_vi, ATTR_TYPE type,
222 ntfschar *name, u32 name_len)
228 /* Make sure no one calls ntfs_attr_iget() for indices. */
229 BUG_ON(type == AT_INDEX_ALLOCATION);
231 na.mft_no = base_vi->i_ino;
234 na.name_len = name_len;
236 vi = iget5_locked(base_vi->i_sb, na.mft_no, (test_t)ntfs_test_inode,
237 (set_t)ntfs_init_locked_inode, &na);
239 return ERR_PTR(-ENOMEM);
243 /* If this is a freshly allocated inode, need to read it now. */
244 if (vi->i_state & I_NEW) {
245 err = ntfs_read_locked_attr_inode(base_vi, vi);
246 unlock_new_inode(vi);
249 * There is no point in keeping bad attribute inodes around. This also
250 * simplifies things in that we never need to check for bad attribute
261 * ntfs_index_iget - obtain a struct inode corresponding to an index
262 * @base_vi: vfs base inode containing the index related attributes
263 * @name: Unicode name of the index
264 * @name_len: length of @name in Unicode characters
266 * Obtain the (fake) struct inode corresponding to the index specified by @name
267 * and @name_len, which is present in the base mft record specified by the vfs
270 * If the index inode is in the cache, it is just returned with an increased
271 * reference count. Otherwise, a new struct inode is allocated and
272 * initialized, and finally ntfs_read_locked_index_inode() is called to read
273 * the index related attributes and fill in the inode structure.
275 * Return the struct inode of the index inode on success. Check the return
276 * value with IS_ERR() and if true, the function failed and the error code is
277 * obtained from PTR_ERR().
279 struct inode *ntfs_index_iget(struct inode *base_vi, ntfschar *name,
286 na.mft_no = base_vi->i_ino;
287 na.type = AT_INDEX_ALLOCATION;
289 na.name_len = name_len;
291 vi = iget5_locked(base_vi->i_sb, na.mft_no, (test_t)ntfs_test_inode,
292 (set_t)ntfs_init_locked_inode, &na);
294 return ERR_PTR(-ENOMEM);
298 /* If this is a freshly allocated inode, need to read it now. */
299 if (vi->i_state & I_NEW) {
300 err = ntfs_read_locked_index_inode(base_vi, vi);
301 unlock_new_inode(vi);
304 * There is no point in keeping bad index inodes around. This also
305 * simplifies things in that we never need to check for bad index
315 struct inode *ntfs_alloc_big_inode(struct super_block *sb)
319 ntfs_debug("Entering.");
320 ni = (ntfs_inode *)kmem_cache_alloc(ntfs_big_inode_cache,
322 if (likely(ni != NULL)) {
326 ntfs_error(sb, "Allocation of NTFS big inode structure failed.");
330 void ntfs_destroy_big_inode(struct inode *inode)
332 ntfs_inode *ni = NTFS_I(inode);
334 ntfs_debug("Entering.");
336 if (!atomic_dec_and_test(&ni->count))
338 kmem_cache_free(ntfs_big_inode_cache, NTFS_I(inode));
341 static inline ntfs_inode *ntfs_alloc_extent_inode(void)
345 ntfs_debug("Entering.");
346 ni = (ntfs_inode *)kmem_cache_alloc(ntfs_inode_cache, SLAB_NOFS);
347 if (likely(ni != NULL)) {
351 ntfs_error(NULL, "Allocation of NTFS inode structure failed.");
355 static void ntfs_destroy_extent_inode(ntfs_inode *ni)
357 ntfs_debug("Entering.");
359 if (!atomic_dec_and_test(&ni->count))
361 kmem_cache_free(ntfs_inode_cache, ni);
365 * __ntfs_init_inode - initialize ntfs specific part of an inode
366 * @sb: super block of mounted volume
367 * @ni: freshly allocated ntfs inode which to initialize
369 * Initialize an ntfs inode to defaults.
371 * NOTE: ni->mft_no, ni->state, ni->type, ni->name, and ni->name_len are left
372 * untouched. Make sure to initialize them elsewhere.
374 * Return zero on success and -ENOMEM on error.
376 void __ntfs_init_inode(struct super_block *sb, ntfs_inode *ni)
378 ntfs_debug("Entering.");
379 rwlock_init(&ni->size_lock);
380 ni->initialized_size = ni->allocated_size = 0;
382 atomic_set(&ni->count, 1);
383 ni->vol = NTFS_SB(sb);
384 ntfs_init_runlist(&ni->runlist);
385 init_MUTEX(&ni->mrec_lock);
388 ni->attr_list_size = 0;
389 ni->attr_list = NULL;
390 ntfs_init_runlist(&ni->attr_list_rl);
391 ni->itype.index.bmp_ino = NULL;
392 ni->itype.index.block_size = 0;
393 ni->itype.index.vcn_size = 0;
394 ni->itype.index.collation_rule = 0;
395 ni->itype.index.block_size_bits = 0;
396 ni->itype.index.vcn_size_bits = 0;
397 init_MUTEX(&ni->extent_lock);
399 ni->ext.base_ntfs_ino = NULL;
402 inline ntfs_inode *ntfs_new_extent_inode(struct super_block *sb,
403 unsigned long mft_no)
405 ntfs_inode *ni = ntfs_alloc_extent_inode();
407 ntfs_debug("Entering.");
408 if (likely(ni != NULL)) {
409 __ntfs_init_inode(sb, ni);
411 ni->type = AT_UNUSED;
419 * ntfs_is_extended_system_file - check if a file is in the $Extend directory
420 * @ctx: initialized attribute search context
422 * Search all file name attributes in the inode described by the attribute
423 * search context @ctx and check if any of the names are in the $Extend system
427 * 1: file is in $Extend directory
428 * 0: file is not in $Extend directory
429 * -errno: failed to determine if the file is in the $Extend directory
431 static int ntfs_is_extended_system_file(ntfs_attr_search_ctx *ctx)
435 /* Restart search. */
436 ntfs_attr_reinit_search_ctx(ctx);
438 /* Get number of hard links. */
439 nr_links = le16_to_cpu(ctx->mrec->link_count);
441 /* Loop through all hard links. */
442 while (!(err = ntfs_attr_lookup(AT_FILE_NAME, NULL, 0, 0, 0, NULL, 0,
444 FILE_NAME_ATTR *file_name_attr;
445 ATTR_RECORD *attr = ctx->attr;
450 * Maximum sanity checking as we are called on an inode that
451 * we suspect might be corrupt.
453 p = (u8*)attr + le32_to_cpu(attr->length);
454 if (p < (u8*)ctx->mrec || (u8*)p > (u8*)ctx->mrec +
455 le32_to_cpu(ctx->mrec->bytes_in_use)) {
457 ntfs_error(ctx->ntfs_ino->vol->sb, "Corrupt file name "
458 "attribute. You should run chkdsk.");
461 if (attr->non_resident) {
462 ntfs_error(ctx->ntfs_ino->vol->sb, "Non-resident file "
463 "name. You should run chkdsk.");
467 ntfs_error(ctx->ntfs_ino->vol->sb, "File name with "
468 "invalid flags. You should run "
472 if (!(attr->data.resident.flags & RESIDENT_ATTR_IS_INDEXED)) {
473 ntfs_error(ctx->ntfs_ino->vol->sb, "Unindexed file "
474 "name. You should run chkdsk.");
477 file_name_attr = (FILE_NAME_ATTR*)((u8*)attr +
478 le16_to_cpu(attr->data.resident.value_offset));
479 p2 = (u8*)attr + le32_to_cpu(attr->data.resident.value_length);
480 if (p2 < (u8*)attr || p2 > p)
481 goto err_corrupt_attr;
482 /* This attribute is ok, but is it in the $Extend directory? */
483 if (MREF_LE(file_name_attr->parent_directory) == FILE_Extend)
484 return 1; /* YES, it's an extended system file. */
486 if (unlikely(err != -ENOENT))
488 if (unlikely(nr_links)) {
489 ntfs_error(ctx->ntfs_ino->vol->sb, "Inode hard link count "
490 "doesn't match number of name attributes. You "
491 "should run chkdsk.");
494 return 0; /* NO, it is not an extended system file. */
498 * ntfs_read_locked_inode - read an inode from its device
501 * ntfs_read_locked_inode() is called from ntfs_iget() to read the inode
502 * described by @vi into memory from the device.
504 * The only fields in @vi that we need to/can look at when the function is
505 * called are i_sb, pointing to the mounted device's super block, and i_ino,
506 * the number of the inode to load.
508 * ntfs_read_locked_inode() maps, pins and locks the mft record number i_ino
509 * for reading and sets up the necessary @vi fields as well as initializing
512 * Q: What locks are held when the function is called?
513 * A: i_state has I_LOCK set, hence the inode is locked, also
514 * i_count is set to 1, so it is not going to go away
515 * i_flags is set to 0 and we have no business touching it. Only an ioctl()
516 * is allowed to write to them. We should of course be honouring them but
517 * we need to do that using the IS_* macros defined in include/linux/fs.h.
518 * In any case ntfs_read_locked_inode() has nothing to do with i_flags.
520 * Return 0 on success and -errno on error. In the error case, the inode will
521 * have had make_bad_inode() executed on it.
523 static int ntfs_read_locked_inode(struct inode *vi)
525 ntfs_volume *vol = NTFS_SB(vi->i_sb);
529 STANDARD_INFORMATION *si;
530 ntfs_attr_search_ctx *ctx;
533 ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino);
535 /* Setup the generic vfs inode parts now. */
537 /* This is the optimal IO size (for stat), not the fs block size. */
538 vi->i_blksize = PAGE_CACHE_SIZE;
540 * This is for checking whether an inode has changed w.r.t. a file so
541 * that the file can be updated if necessary (compare with f_version).
545 vi->i_uid = vol->uid;
546 vi->i_gid = vol->gid;
550 * Initialize the ntfs specific part of @vi special casing
551 * FILE_MFT which we need to do at mount time.
553 if (vi->i_ino != FILE_MFT)
554 ntfs_init_big_inode(vi);
557 m = map_mft_record(ni);
562 ctx = ntfs_attr_get_search_ctx(ni, m);
568 if (!(m->flags & MFT_RECORD_IN_USE)) {
569 ntfs_error(vi->i_sb, "Inode is not in use!");
572 if (m->base_mft_record) {
573 ntfs_error(vi->i_sb, "Inode is an extent inode!");
577 /* Transfer information from mft record into vfs and ntfs inodes. */
578 vi->i_generation = ni->seq_no = le16_to_cpu(m->sequence_number);
581 * FIXME: Keep in mind that link_count is two for files which have both
582 * a long file name and a short file name as separate entries, so if
583 * we are hiding short file names this will be too high. Either we need
584 * to account for the short file names by subtracting them or we need
585 * to make sure we delete files even though i_nlink is not zero which
586 * might be tricky due to vfs interactions. Need to think about this
587 * some more when implementing the unlink command.
589 vi->i_nlink = le16_to_cpu(m->link_count);
591 * FIXME: Reparse points can have the directory bit set even though
592 * they would be S_IFLNK. Need to deal with this further below when we
593 * implement reparse points / symbolic links but it will do for now.
594 * Also if not a directory, it could be something else, rather than
595 * a regular file. But again, will do for now.
597 /* Everyone gets all permissions. */
598 vi->i_mode |= S_IRWXUGO;
599 /* If read-only, noone gets write permissions. */
601 vi->i_mode &= ~S_IWUGO;
602 if (m->flags & MFT_RECORD_IS_DIRECTORY) {
603 vi->i_mode |= S_IFDIR;
605 * Apply the directory permissions mask set in the mount
608 vi->i_mode &= ~vol->dmask;
609 /* Things break without this kludge! */
613 vi->i_mode |= S_IFREG;
614 /* Apply the file permissions mask set in the mount options. */
615 vi->i_mode &= ~vol->fmask;
618 * Find the standard information attribute in the mft record. At this
619 * stage we haven't setup the attribute list stuff yet, so this could
620 * in fact fail if the standard information is in an extent record, but
621 * I don't think this actually ever happens.
623 err = ntfs_attr_lookup(AT_STANDARD_INFORMATION, NULL, 0, 0, 0, NULL, 0,
626 if (err == -ENOENT) {
628 * TODO: We should be performing a hot fix here (if the
629 * recover mount option is set) by creating a new
632 ntfs_error(vi->i_sb, "$STANDARD_INFORMATION attribute "
638 /* Get the standard information attribute value. */
639 si = (STANDARD_INFORMATION*)((u8*)a +
640 le16_to_cpu(a->data.resident.value_offset));
642 /* Transfer information from the standard information into vi. */
644 * Note: The i_?times do not quite map perfectly onto the NTFS times,
645 * but they are close enough, and in the end it doesn't really matter
649 * mtime is the last change of the data within the file. Not changed
650 * when only metadata is changed, e.g. a rename doesn't affect mtime.
652 vi->i_mtime = ntfs2utc(si->last_data_change_time);
654 * ctime is the last change of the metadata of the file. This obviously
655 * always changes, when mtime is changed. ctime can be changed on its
656 * own, mtime is then not changed, e.g. when a file is renamed.
658 vi->i_ctime = ntfs2utc(si->last_mft_change_time);
660 * Last access to the data within the file. Not changed during a rename
661 * for example but changed whenever the file is written to.
663 vi->i_atime = ntfs2utc(si->last_access_time);
665 /* Find the attribute list attribute if present. */
666 ntfs_attr_reinit_search_ctx(ctx);
667 err = ntfs_attr_lookup(AT_ATTRIBUTE_LIST, NULL, 0, 0, 0, NULL, 0, ctx);
669 if (unlikely(err != -ENOENT)) {
670 ntfs_error(vi->i_sb, "Failed to lookup attribute list "
674 } else /* if (!err) */ {
675 if (vi->i_ino == FILE_MFT)
676 goto skip_attr_list_load;
677 ntfs_debug("Attribute list found in inode 0x%lx.", vi->i_ino);
680 if (a->flags & ATTR_IS_ENCRYPTED ||
681 a->flags & ATTR_COMPRESSION_MASK ||
682 a->flags & ATTR_IS_SPARSE) {
683 ntfs_error(vi->i_sb, "Attribute list attribute is "
684 "compressed/encrypted/sparse.");
687 /* Now allocate memory for the attribute list. */
688 ni->attr_list_size = (u32)ntfs_attr_size(a);
689 ni->attr_list = ntfs_malloc_nofs(ni->attr_list_size);
690 if (!ni->attr_list) {
691 ntfs_error(vi->i_sb, "Not enough memory to allocate "
692 "buffer for attribute list.");
696 if (a->non_resident) {
697 NInoSetAttrListNonResident(ni);
698 if (a->data.non_resident.lowest_vcn) {
699 ntfs_error(vi->i_sb, "Attribute list has non "
704 * Setup the runlist. No need for locking as we have
705 * exclusive access to the inode at this time.
707 ni->attr_list_rl.rl = ntfs_mapping_pairs_decompress(vol,
709 if (IS_ERR(ni->attr_list_rl.rl)) {
710 err = PTR_ERR(ni->attr_list_rl.rl);
711 ni->attr_list_rl.rl = NULL;
712 ntfs_error(vi->i_sb, "Mapping pairs "
713 "decompression failed.");
716 /* Now load the attribute list. */
717 if ((err = load_attribute_list(vol, &ni->attr_list_rl,
718 ni->attr_list, ni->attr_list_size,
719 sle64_to_cpu(a->data.non_resident.
720 initialized_size)))) {
721 ntfs_error(vi->i_sb, "Failed to load "
722 "attribute list attribute.");
725 } else /* if (!a->non_resident) */ {
726 if ((u8*)a + le16_to_cpu(a->data.resident.value_offset)
728 a->data.resident.value_length) >
729 (u8*)ctx->mrec + vol->mft_record_size) {
730 ntfs_error(vi->i_sb, "Corrupt attribute list "
734 /* Now copy the attribute list. */
735 memcpy(ni->attr_list, (u8*)a + le16_to_cpu(
736 a->data.resident.value_offset),
738 a->data.resident.value_length));
743 * If an attribute list is present we now have the attribute list value
744 * in ntfs_ino->attr_list and it is ntfs_ino->attr_list_size bytes.
746 if (S_ISDIR(vi->i_mode)) {
751 u8 *ir_end, *index_end;
753 /* It is a directory, find index root attribute. */
754 ntfs_attr_reinit_search_ctx(ctx);
755 err = ntfs_attr_lookup(AT_INDEX_ROOT, I30, 4, CASE_SENSITIVE,
758 if (err == -ENOENT) {
759 // FIXME: File is corrupt! Hot-fix with empty
760 // index root attribute if recovery option is
762 ntfs_error(vi->i_sb, "$INDEX_ROOT attribute "
768 /* Set up the state. */
769 if (unlikely(a->non_resident)) {
770 ntfs_error(vol->sb, "$INDEX_ROOT attribute is not "
774 /* Ensure the attribute name is placed before the value. */
775 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
776 le16_to_cpu(a->data.resident.value_offset)))) {
777 ntfs_error(vol->sb, "$INDEX_ROOT attribute name is "
778 "placed after the attribute value.");
782 * Compressed/encrypted index root just means that the newly
783 * created files in that directory should be created compressed/
784 * encrypted. However index root cannot be both compressed and
787 if (a->flags & ATTR_COMPRESSION_MASK)
788 NInoSetCompressed(ni);
789 if (a->flags & ATTR_IS_ENCRYPTED) {
790 if (a->flags & ATTR_COMPRESSION_MASK) {
791 ntfs_error(vi->i_sb, "Found encrypted and "
792 "compressed attribute.");
795 NInoSetEncrypted(ni);
797 if (a->flags & ATTR_IS_SPARSE)
799 ir = (INDEX_ROOT*)((u8*)a +
800 le16_to_cpu(a->data.resident.value_offset));
801 ir_end = (u8*)ir + le32_to_cpu(a->data.resident.value_length);
802 if (ir_end > (u8*)ctx->mrec + vol->mft_record_size) {
803 ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is "
807 index_end = (u8*)&ir->index +
808 le32_to_cpu(ir->index.index_length);
809 if (index_end > ir_end) {
810 ntfs_error(vi->i_sb, "Directory index is corrupt.");
813 if (ir->type != AT_FILE_NAME) {
814 ntfs_error(vi->i_sb, "Indexed attribute is not "
818 if (ir->collation_rule != COLLATION_FILE_NAME) {
819 ntfs_error(vi->i_sb, "Index collation rule is not "
820 "COLLATION_FILE_NAME.");
823 ni->itype.index.collation_rule = ir->collation_rule;
824 ni->itype.index.block_size = le32_to_cpu(ir->index_block_size);
825 if (ni->itype.index.block_size &
826 (ni->itype.index.block_size - 1)) {
827 ntfs_error(vi->i_sb, "Index block size (%u) is not a "
829 ni->itype.index.block_size);
832 if (ni->itype.index.block_size > PAGE_CACHE_SIZE) {
833 ntfs_error(vi->i_sb, "Index block size (%u) > "
834 "PAGE_CACHE_SIZE (%ld) is not "
836 ni->itype.index.block_size,
841 if (ni->itype.index.block_size < NTFS_BLOCK_SIZE) {
842 ntfs_error(vi->i_sb, "Index block size (%u) < "
843 "NTFS_BLOCK_SIZE (%i) is not "
845 ni->itype.index.block_size,
850 ni->itype.index.block_size_bits =
851 ffs(ni->itype.index.block_size) - 1;
852 /* Determine the size of a vcn in the directory index. */
853 if (vol->cluster_size <= ni->itype.index.block_size) {
854 ni->itype.index.vcn_size = vol->cluster_size;
855 ni->itype.index.vcn_size_bits = vol->cluster_size_bits;
857 ni->itype.index.vcn_size = vol->sector_size;
858 ni->itype.index.vcn_size_bits = vol->sector_size_bits;
861 /* Setup the index allocation attribute, even if not present. */
862 NInoSetMstProtected(ni);
863 ni->type = AT_INDEX_ALLOCATION;
867 if (!(ir->index.flags & LARGE_INDEX)) {
868 /* No index allocation. */
869 vi->i_size = ni->initialized_size =
870 ni->allocated_size = 0;
871 /* We are done with the mft record, so we release it. */
872 ntfs_attr_put_search_ctx(ctx);
873 unmap_mft_record(ni);
876 goto skip_large_dir_stuff;
877 } /* LARGE_INDEX: Index allocation present. Setup state. */
878 NInoSetIndexAllocPresent(ni);
879 /* Find index allocation attribute. */
880 ntfs_attr_reinit_search_ctx(ctx);
881 err = ntfs_attr_lookup(AT_INDEX_ALLOCATION, I30, 4,
882 CASE_SENSITIVE, 0, NULL, 0, ctx);
885 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION "
886 "attribute is not present but "
887 "$INDEX_ROOT indicated it is.");
889 ntfs_error(vi->i_sb, "Failed to lookup "
895 if (!a->non_resident) {
896 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
901 * Ensure the attribute name is placed before the mapping pairs
904 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
906 a->data.non_resident.mapping_pairs_offset)))) {
907 ntfs_error(vol->sb, "$INDEX_ALLOCATION attribute name "
908 "is placed after the mapping pairs "
912 if (a->flags & ATTR_IS_ENCRYPTED) {
913 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
917 if (a->flags & ATTR_IS_SPARSE) {
918 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
922 if (a->flags & ATTR_COMPRESSION_MASK) {
923 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
927 if (a->data.non_resident.lowest_vcn) {
928 ntfs_error(vi->i_sb, "First extent of "
929 "$INDEX_ALLOCATION attribute has non "
933 vi->i_size = sle64_to_cpu(a->data.non_resident.data_size);
934 ni->initialized_size = sle64_to_cpu(
935 a->data.non_resident.initialized_size);
936 ni->allocated_size = sle64_to_cpu(
937 a->data.non_resident.allocated_size);
939 * We are done with the mft record, so we release it. Otherwise
940 * we would deadlock in ntfs_attr_iget().
942 ntfs_attr_put_search_ctx(ctx);
943 unmap_mft_record(ni);
946 /* Get the index bitmap attribute inode. */
947 bvi = ntfs_attr_iget(vi, AT_BITMAP, I30, 4);
949 ntfs_error(vi->i_sb, "Failed to get bitmap attribute.");
953 ni->itype.index.bmp_ino = bvi;
955 if (NInoCompressed(bni) || NInoEncrypted(bni) ||
957 ntfs_error(vi->i_sb, "$BITMAP attribute is compressed "
958 "and/or encrypted and/or sparse.");
961 /* Consistency check bitmap size vs. index allocation size. */
962 bvi_size = i_size_read(bvi);
963 if ((bvi_size << 3) < (vi->i_size >>
964 ni->itype.index.block_size_bits)) {
965 ntfs_error(vi->i_sb, "Index bitmap too small (0x%llx) "
966 "for index allocation (0x%llx).",
967 bvi_size << 3, vi->i_size);
970 skip_large_dir_stuff:
971 /* Setup the operations for this inode. */
972 vi->i_op = &ntfs_dir_inode_ops;
973 vi->i_fop = &ntfs_dir_ops;
976 ntfs_attr_reinit_search_ctx(ctx);
978 /* Setup the data attribute, even if not present. */
983 /* Find first extent of the unnamed data attribute. */
984 err = ntfs_attr_lookup(AT_DATA, NULL, 0, 0, 0, NULL, 0, ctx);
986 vi->i_size = ni->initialized_size =
987 ni->allocated_size = 0;
988 if (err != -ENOENT) {
989 ntfs_error(vi->i_sb, "Failed to lookup $DATA "
994 * FILE_Secure does not have an unnamed $DATA
995 * attribute, so we special case it here.
997 if (vi->i_ino == FILE_Secure)
998 goto no_data_attr_special_case;
1000 * Most if not all the system files in the $Extend
1001 * system directory do not have unnamed data
1002 * attributes so we need to check if the parent
1003 * directory of the file is FILE_Extend and if it is
1004 * ignore this error. To do this we need to get the
1005 * name of this inode from the mft record as the name
1006 * contains the back reference to the parent directory.
1008 if (ntfs_is_extended_system_file(ctx) > 0)
1009 goto no_data_attr_special_case;
1010 // FIXME: File is corrupt! Hot-fix with empty data
1011 // attribute if recovery option is set.
1012 ntfs_error(vi->i_sb, "$DATA attribute is missing.");
1016 /* Setup the state. */
1017 if (a->non_resident) {
1018 NInoSetNonResident(ni);
1019 if (a->flags & ATTR_COMPRESSION_MASK) {
1020 NInoSetCompressed(ni);
1021 if (vol->cluster_size > 4096) {
1022 ntfs_error(vi->i_sb, "Found "
1023 "compressed data but "
1024 "compression is disabled due "
1025 "to cluster size (%i) > 4kiB.",
1029 if ((a->flags & ATTR_COMPRESSION_MASK)
1030 != ATTR_IS_COMPRESSED) {
1031 ntfs_error(vi->i_sb, "Found "
1032 "unknown compression method or "
1036 ni->itype.compressed.block_clusters = 1U <<
1037 a->data.non_resident.
1039 if (a->data.non_resident.compression_unit !=
1041 ntfs_error(vi->i_sb, "Found "
1042 "nonstandard compression unit "
1043 "(%u instead of 4). Cannot "
1045 a->data.non_resident.
1050 ni->itype.compressed.block_size = 1U << (
1051 a->data.non_resident.
1053 vol->cluster_size_bits);
1054 ni->itype.compressed.block_size_bits = ffs(
1055 ni->itype.compressed.block_size) - 1;
1057 if (a->flags & ATTR_IS_ENCRYPTED) {
1058 if (a->flags & ATTR_COMPRESSION_MASK) {
1059 ntfs_error(vi->i_sb, "Found encrypted "
1060 "and compressed data.");
1063 NInoSetEncrypted(ni);
1065 if (a->flags & ATTR_IS_SPARSE)
1067 if (a->data.non_resident.lowest_vcn) {
1068 ntfs_error(vi->i_sb, "First extent of $DATA "
1069 "attribute has non zero "
1073 /* Setup all the sizes. */
1074 vi->i_size = sle64_to_cpu(
1075 a->data.non_resident.data_size);
1076 ni->initialized_size = sle64_to_cpu(
1077 a->data.non_resident.initialized_size);
1078 ni->allocated_size = sle64_to_cpu(
1079 a->data.non_resident.allocated_size);
1080 if (NInoCompressed(ni))
1081 ni->itype.compressed.size = sle64_to_cpu(
1082 a->data.non_resident.
1084 } else { /* Resident attribute. */
1085 /* Setup all the sizes. */
1086 vi->i_size = ni->initialized_size = le32_to_cpu(
1087 a->data.resident.value_length);
1088 ni->allocated_size = le32_to_cpu(a->length) -
1090 a->data.resident.value_offset);
1091 if (vi->i_size > ni->allocated_size) {
1092 ntfs_error(vi->i_sb, "Resident data attribute "
1093 "is corrupt (size exceeds "
1098 no_data_attr_special_case:
1099 /* We are done with the mft record, so we release it. */
1100 ntfs_attr_put_search_ctx(ctx);
1101 unmap_mft_record(ni);
1104 /* Setup the operations for this inode. */
1105 vi->i_op = &ntfs_file_inode_ops;
1106 vi->i_fop = &ntfs_file_ops;
1108 if (NInoMstProtected(ni))
1109 vi->i_mapping->a_ops = &ntfs_mst_aops;
1111 vi->i_mapping->a_ops = &ntfs_aops;
1113 * The number of 512-byte blocks used on disk (for stat). This is in so
1114 * far inaccurate as it doesn't account for any named streams or other
1115 * special non-resident attributes, but that is how Windows works, too,
1116 * so we are at least consistent with Windows, if not entirely
1117 * consistent with the Linux Way. Doing it the Linux Way would cause a
1118 * significant slowdown as it would involve iterating over all
1119 * attributes in the mft record and adding the allocated/compressed
1120 * sizes of all non-resident attributes present to give us the Linux
1121 * correct size that should go into i_blocks (after division by 512).
1123 if (S_ISDIR(vi->i_mode) || !NInoCompressed(ni))
1124 vi->i_blocks = ni->allocated_size >> 9;
1126 vi->i_blocks = ni->itype.compressed.size >> 9;
1128 ntfs_debug("Done.");
1135 ntfs_attr_put_search_ctx(ctx);
1137 unmap_mft_record(ni);
1139 ntfs_error(vol->sb, "Failed with error code %i. Marking corrupt "
1140 "inode 0x%lx as bad. Run chkdsk.", err, vi->i_ino);
1142 if (err != -EOPNOTSUPP && err != -ENOMEM)
1148 * ntfs_read_locked_attr_inode - read an attribute inode from its base inode
1149 * @base_vi: base inode
1150 * @vi: attribute inode to read
1152 * ntfs_read_locked_attr_inode() is called from ntfs_attr_iget() to read the
1153 * attribute inode described by @vi into memory from the base mft record
1154 * described by @base_ni.
1156 * ntfs_read_locked_attr_inode() maps, pins and locks the base inode for
1157 * reading and looks up the attribute described by @vi before setting up the
1158 * necessary fields in @vi as well as initializing the ntfs inode.
1160 * Q: What locks are held when the function is called?
1161 * A: i_state has I_LOCK set, hence the inode is locked, also
1162 * i_count is set to 1, so it is not going to go away
1164 * Return 0 on success and -errno on error. In the error case, the inode will
1165 * have had make_bad_inode() executed on it.
1167 static int ntfs_read_locked_attr_inode(struct inode *base_vi, struct inode *vi)
1169 ntfs_volume *vol = NTFS_SB(vi->i_sb);
1170 ntfs_inode *ni, *base_ni;
1173 ntfs_attr_search_ctx *ctx;
1176 ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino);
1178 ntfs_init_big_inode(vi);
1181 base_ni = NTFS_I(base_vi);
1183 /* Just mirror the values from the base inode. */
1184 vi->i_blksize = base_vi->i_blksize;
1185 vi->i_version = base_vi->i_version;
1186 vi->i_uid = base_vi->i_uid;
1187 vi->i_gid = base_vi->i_gid;
1188 vi->i_nlink = base_vi->i_nlink;
1189 vi->i_mtime = base_vi->i_mtime;
1190 vi->i_ctime = base_vi->i_ctime;
1191 vi->i_atime = base_vi->i_atime;
1192 vi->i_generation = ni->seq_no = base_ni->seq_no;
1194 /* Set inode type to zero but preserve permissions. */
1195 vi->i_mode = base_vi->i_mode & ~S_IFMT;
1197 m = map_mft_record(base_ni);
1202 ctx = ntfs_attr_get_search_ctx(base_ni, m);
1207 /* Find the attribute. */
1208 err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
1209 CASE_SENSITIVE, 0, NULL, 0, ctx);
1213 if (!a->non_resident) {
1214 /* Ensure the attribute name is placed before the value. */
1215 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1216 le16_to_cpu(a->data.resident.value_offset)))) {
1217 ntfs_error(vol->sb, "Attribute name is placed after "
1218 "the attribute value.");
1221 if (NInoMstProtected(ni) || a->flags) {
1222 ntfs_error(vi->i_sb, "Found mst protected attribute "
1223 "or attribute with non-zero flags but "
1224 "the attribute is resident. Please "
1225 "report you saw this message to "
1226 "linux-ntfs-dev@lists.sourceforge.net");
1229 /* Resident attribute. Setup all the sizes. */
1230 vi->i_size = ni->initialized_size = le32_to_cpu(
1231 a->data.resident.value_length);
1232 ni->allocated_size = le32_to_cpu(a->length) -
1233 le16_to_cpu(a->data.resident.value_offset);
1234 if (vi->i_size > ni->allocated_size) {
1235 ntfs_error(vi->i_sb, "Resident data attribute is "
1236 "corrupt (size exceeds allocation).");
1240 NInoSetNonResident(ni);
1242 * Ensure the attribute name is placed before the mapping pairs
1245 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1247 a->data.non_resident.mapping_pairs_offset)))) {
1248 ntfs_error(vol->sb, "Attribute name is placed after "
1249 "the mapping pairs array.");
1252 if (a->flags & ATTR_COMPRESSION_MASK) {
1253 if (NInoMstProtected(ni)) {
1254 ntfs_error(vi->i_sb, "Found mst protected "
1255 "attribute but the attribute "
1256 "is compressed. Please report "
1257 "you saw this message to "
1258 "linux-ntfs-dev@lists."
1262 NInoSetCompressed(ni);
1263 if ((ni->type != AT_DATA) || (ni->type == AT_DATA &&
1265 ntfs_error(vi->i_sb, "Found compressed "
1266 "non-data or named data "
1267 "attribute. Please report "
1268 "you saw this message to "
1269 "linux-ntfs-dev@lists."
1273 if (vol->cluster_size > 4096) {
1274 ntfs_error(vi->i_sb, "Found compressed "
1275 "attribute but compression is "
1276 "disabled due to cluster size "
1281 if ((a->flags & ATTR_COMPRESSION_MASK) !=
1282 ATTR_IS_COMPRESSED) {
1283 ntfs_error(vi->i_sb, "Found unknown "
1284 "compression method.");
1287 ni->itype.compressed.block_clusters = 1U <<
1288 a->data.non_resident.compression_unit;
1289 if (a->data.non_resident.compression_unit != 4) {
1290 ntfs_error(vi->i_sb, "Found nonstandard "
1291 "compression unit (%u instead "
1292 "of 4). Cannot handle this.",
1293 a->data.non_resident.
1298 ni->itype.compressed.block_size = 1U << (
1299 a->data.non_resident.compression_unit +
1300 vol->cluster_size_bits);
1301 ni->itype.compressed.block_size_bits = ffs(
1302 ni->itype.compressed.block_size) - 1;
1304 if (a->flags & ATTR_IS_ENCRYPTED) {
1305 if (a->flags & ATTR_COMPRESSION_MASK) {
1306 ntfs_error(vi->i_sb, "Found encrypted and "
1307 "compressed data.");
1310 if (NInoMstProtected(ni)) {
1311 ntfs_error(vi->i_sb, "Found mst protected "
1312 "attribute but the attribute "
1313 "is encrypted. Please report "
1314 "you saw this message to "
1315 "linux-ntfs-dev@lists."
1319 NInoSetEncrypted(ni);
1321 if (a->flags & ATTR_IS_SPARSE) {
1322 if (NInoMstProtected(ni)) {
1323 ntfs_error(vi->i_sb, "Found mst protected "
1324 "attribute but the attribute "
1325 "is sparse. Please report "
1326 "you saw this message to "
1327 "linux-ntfs-dev@lists."
1333 if (a->data.non_resident.lowest_vcn) {
1334 ntfs_error(vi->i_sb, "First extent of attribute has "
1335 "non-zero lowest_vcn.");
1338 /* Setup all the sizes. */
1339 vi->i_size = sle64_to_cpu(a->data.non_resident.data_size);
1340 ni->initialized_size = sle64_to_cpu(
1341 a->data.non_resident.initialized_size);
1342 ni->allocated_size = sle64_to_cpu(
1343 a->data.non_resident.allocated_size);
1344 if (NInoCompressed(ni))
1345 ni->itype.compressed.size = sle64_to_cpu(
1346 a->data.non_resident.compressed_size);
1349 /* Setup the operations for this attribute inode. */
1352 if (NInoMstProtected(ni))
1353 vi->i_mapping->a_ops = &ntfs_mst_aops;
1355 vi->i_mapping->a_ops = &ntfs_aops;
1357 if (!NInoCompressed(ni))
1358 vi->i_blocks = ni->allocated_size >> 9;
1360 vi->i_blocks = ni->itype.compressed.size >> 9;
1363 * Make sure the base inode doesn't go away and attach it to the
1367 ni->ext.base_ntfs_ino = base_ni;
1368 ni->nr_extents = -1;
1370 ntfs_attr_put_search_ctx(ctx);
1371 unmap_mft_record(base_ni);
1373 ntfs_debug("Done.");
1380 ntfs_attr_put_search_ctx(ctx);
1381 unmap_mft_record(base_ni);
1383 ntfs_error(vol->sb, "Failed with error code %i while reading attribute "
1384 "inode (mft_no 0x%lx, type 0x%x, name_len %i). "
1385 "Marking corrupt inode and base inode 0x%lx as bad. "
1386 "Run chkdsk.", err, vi->i_ino, ni->type, ni->name_len,
1389 make_bad_inode(base_vi);
1396 * ntfs_read_locked_index_inode - read an index inode from its base inode
1397 * @base_vi: base inode
1398 * @vi: index inode to read
1400 * ntfs_read_locked_index_inode() is called from ntfs_index_iget() to read the
1401 * index inode described by @vi into memory from the base mft record described
1404 * ntfs_read_locked_index_inode() maps, pins and locks the base inode for
1405 * reading and looks up the attributes relating to the index described by @vi
1406 * before setting up the necessary fields in @vi as well as initializing the
1409 * Note, index inodes are essentially attribute inodes (NInoAttr() is true)
1410 * with the attribute type set to AT_INDEX_ALLOCATION. Apart from that, they
1411 * are setup like directory inodes since directories are a special case of
1412 * indices ao they need to be treated in much the same way. Most importantly,
1413 * for small indices the index allocation attribute might not actually exist.
1414 * However, the index root attribute always exists but this does not need to
1415 * have an inode associated with it and this is why we define a new inode type
1416 * index. Also, like for directories, we need to have an attribute inode for
1417 * the bitmap attribute corresponding to the index allocation attribute and we
1418 * can store this in the appropriate field of the inode, just like we do for
1419 * normal directory inodes.
1421 * Q: What locks are held when the function is called?
1422 * A: i_state has I_LOCK set, hence the inode is locked, also
1423 * i_count is set to 1, so it is not going to go away
1425 * Return 0 on success and -errno on error. In the error case, the inode will
1426 * have had make_bad_inode() executed on it.
1428 static int ntfs_read_locked_index_inode(struct inode *base_vi, struct inode *vi)
1431 ntfs_volume *vol = NTFS_SB(vi->i_sb);
1432 ntfs_inode *ni, *base_ni, *bni;
1436 ntfs_attr_search_ctx *ctx;
1438 u8 *ir_end, *index_end;
1441 ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino);
1442 ntfs_init_big_inode(vi);
1444 base_ni = NTFS_I(base_vi);
1445 /* Just mirror the values from the base inode. */
1446 vi->i_blksize = base_vi->i_blksize;
1447 vi->i_version = base_vi->i_version;
1448 vi->i_uid = base_vi->i_uid;
1449 vi->i_gid = base_vi->i_gid;
1450 vi->i_nlink = base_vi->i_nlink;
1451 vi->i_mtime = base_vi->i_mtime;
1452 vi->i_ctime = base_vi->i_ctime;
1453 vi->i_atime = base_vi->i_atime;
1454 vi->i_generation = ni->seq_no = base_ni->seq_no;
1455 /* Set inode type to zero but preserve permissions. */
1456 vi->i_mode = base_vi->i_mode & ~S_IFMT;
1457 /* Map the mft record for the base inode. */
1458 m = map_mft_record(base_ni);
1463 ctx = ntfs_attr_get_search_ctx(base_ni, m);
1468 /* Find the index root attribute. */
1469 err = ntfs_attr_lookup(AT_INDEX_ROOT, ni->name, ni->name_len,
1470 CASE_SENSITIVE, 0, NULL, 0, ctx);
1471 if (unlikely(err)) {
1473 ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is "
1478 /* Set up the state. */
1479 if (unlikely(a->non_resident)) {
1480 ntfs_error(vol->sb, "$INDEX_ROOT attribute is not resident.");
1483 /* Ensure the attribute name is placed before the value. */
1484 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1485 le16_to_cpu(a->data.resident.value_offset)))) {
1486 ntfs_error(vol->sb, "$INDEX_ROOT attribute name is placed "
1487 "after the attribute value.");
1490 /* Compressed/encrypted/sparse index root is not allowed. */
1491 if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_ENCRYPTED |
1493 ntfs_error(vi->i_sb, "Found compressed/encrypted/sparse index "
1497 ir = (INDEX_ROOT*)((u8*)a + le16_to_cpu(a->data.resident.value_offset));
1498 ir_end = (u8*)ir + le32_to_cpu(a->data.resident.value_length);
1499 if (ir_end > (u8*)ctx->mrec + vol->mft_record_size) {
1500 ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is corrupt.");
1503 index_end = (u8*)&ir->index + le32_to_cpu(ir->index.index_length);
1504 if (index_end > ir_end) {
1505 ntfs_error(vi->i_sb, "Index is corrupt.");
1509 ntfs_error(vi->i_sb, "Index type is not 0 (type is 0x%x).",
1510 le32_to_cpu(ir->type));
1513 ni->itype.index.collation_rule = ir->collation_rule;
1514 ntfs_debug("Index collation rule is 0x%x.",
1515 le32_to_cpu(ir->collation_rule));
1516 ni->itype.index.block_size = le32_to_cpu(ir->index_block_size);
1517 if (ni->itype.index.block_size & (ni->itype.index.block_size - 1)) {
1518 ntfs_error(vi->i_sb, "Index block size (%u) is not a power of "
1519 "two.", ni->itype.index.block_size);
1522 if (ni->itype.index.block_size > PAGE_CACHE_SIZE) {
1523 ntfs_error(vi->i_sb, "Index block size (%u) > PAGE_CACHE_SIZE "
1524 "(%ld) is not supported. Sorry.",
1525 ni->itype.index.block_size, PAGE_CACHE_SIZE);
1529 if (ni->itype.index.block_size < NTFS_BLOCK_SIZE) {
1530 ntfs_error(vi->i_sb, "Index block size (%u) < NTFS_BLOCK_SIZE "
1531 "(%i) is not supported. Sorry.",
1532 ni->itype.index.block_size, NTFS_BLOCK_SIZE);
1536 ni->itype.index.block_size_bits = ffs(ni->itype.index.block_size) - 1;
1537 /* Determine the size of a vcn in the index. */
1538 if (vol->cluster_size <= ni->itype.index.block_size) {
1539 ni->itype.index.vcn_size = vol->cluster_size;
1540 ni->itype.index.vcn_size_bits = vol->cluster_size_bits;
1542 ni->itype.index.vcn_size = vol->sector_size;
1543 ni->itype.index.vcn_size_bits = vol->sector_size_bits;
1545 /* Check for presence of index allocation attribute. */
1546 if (!(ir->index.flags & LARGE_INDEX)) {
1547 /* No index allocation. */
1548 vi->i_size = ni->initialized_size = ni->allocated_size = 0;
1549 /* We are done with the mft record, so we release it. */
1550 ntfs_attr_put_search_ctx(ctx);
1551 unmap_mft_record(base_ni);
1554 goto skip_large_index_stuff;
1555 } /* LARGE_INDEX: Index allocation present. Setup state. */
1556 NInoSetIndexAllocPresent(ni);
1557 /* Find index allocation attribute. */
1558 ntfs_attr_reinit_search_ctx(ctx);
1559 err = ntfs_attr_lookup(AT_INDEX_ALLOCATION, ni->name, ni->name_len,
1560 CASE_SENSITIVE, 0, NULL, 0, ctx);
1561 if (unlikely(err)) {
1563 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1564 "not present but $INDEX_ROOT "
1565 "indicated it is.");
1567 ntfs_error(vi->i_sb, "Failed to lookup "
1568 "$INDEX_ALLOCATION attribute.");
1571 if (!a->non_resident) {
1572 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1577 * Ensure the attribute name is placed before the mapping pairs array.
1579 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1581 a->data.non_resident.mapping_pairs_offset)))) {
1582 ntfs_error(vol->sb, "$INDEX_ALLOCATION attribute name is "
1583 "placed after the mapping pairs array.");
1586 if (a->flags & ATTR_IS_ENCRYPTED) {
1587 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1591 if (a->flags & ATTR_IS_SPARSE) {
1592 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is sparse.");
1595 if (a->flags & ATTR_COMPRESSION_MASK) {
1596 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1600 if (a->data.non_resident.lowest_vcn) {
1601 ntfs_error(vi->i_sb, "First extent of $INDEX_ALLOCATION "
1602 "attribute has non zero lowest_vcn.");
1605 vi->i_size = sle64_to_cpu(a->data.non_resident.data_size);
1606 ni->initialized_size = sle64_to_cpu(
1607 a->data.non_resident.initialized_size);
1608 ni->allocated_size = sle64_to_cpu(a->data.non_resident.allocated_size);
1610 * We are done with the mft record, so we release it. Otherwise
1611 * we would deadlock in ntfs_attr_iget().
1613 ntfs_attr_put_search_ctx(ctx);
1614 unmap_mft_record(base_ni);
1617 /* Get the index bitmap attribute inode. */
1618 bvi = ntfs_attr_iget(base_vi, AT_BITMAP, ni->name, ni->name_len);
1620 ntfs_error(vi->i_sb, "Failed to get bitmap attribute.");
1625 if (NInoCompressed(bni) || NInoEncrypted(bni) ||
1627 ntfs_error(vi->i_sb, "$BITMAP attribute is compressed and/or "
1628 "encrypted and/or sparse.");
1629 goto iput_unm_err_out;
1631 /* Consistency check bitmap size vs. index allocation size. */
1632 bvi_size = i_size_read(bvi);
1633 if ((bvi_size << 3) < (vi->i_size >> ni->itype.index.block_size_bits)) {
1634 ntfs_error(vi->i_sb, "Index bitmap too small (0x%llx) for "
1635 "index allocation (0x%llx).", bvi_size << 3,
1637 goto iput_unm_err_out;
1639 ni->itype.index.bmp_ino = bvi;
1640 skip_large_index_stuff:
1641 /* Setup the operations for this index inode. */
1644 vi->i_mapping->a_ops = &ntfs_mst_aops;
1645 vi->i_blocks = ni->allocated_size >> 9;
1648 * Make sure the base inode doesn't go away and attach it to the
1652 ni->ext.base_ntfs_ino = base_ni;
1653 ni->nr_extents = -1;
1655 ntfs_debug("Done.");
1664 ntfs_attr_put_search_ctx(ctx);
1666 unmap_mft_record(base_ni);
1668 ntfs_error(vi->i_sb, "Failed with error code %i while reading index "
1669 "inode (mft_no 0x%lx, name_len %i.", err, vi->i_ino,
1672 if (err != -EOPNOTSUPP && err != -ENOMEM)
1678 * ntfs_read_inode_mount - special read_inode for mount time use only
1679 * @vi: inode to read
1681 * Read inode FILE_MFT at mount time, only called with super_block lock
1682 * held from within the read_super() code path.
1684 * This function exists because when it is called the page cache for $MFT/$DATA
1685 * is not initialized and hence we cannot get at the contents of mft records
1686 * by calling map_mft_record*().
1688 * Further it needs to cope with the circular references problem, i.e. cannot
1689 * load any attributes other than $ATTRIBUTE_LIST until $DATA is loaded, because
1690 * we do not know where the other extent mft records are yet and again, because
1691 * we cannot call map_mft_record*() yet. Obviously this applies only when an
1692 * attribute list is actually present in $MFT inode.
1694 * We solve these problems by starting with the $DATA attribute before anything
1695 * else and iterating using ntfs_attr_lookup($DATA) over all extents. As each
1696 * extent is found, we ntfs_mapping_pairs_decompress() including the implied
1697 * ntfs_runlists_merge(). Each step of the iteration necessarily provides
1698 * sufficient information for the next step to complete.
1700 * This should work but there are two possible pit falls (see inline comments
1701 * below), but only time will tell if they are real pits or just smoke...
1703 int ntfs_read_inode_mount(struct inode *vi)
1705 VCN next_vcn, last_vcn, highest_vcn;
1707 struct super_block *sb = vi->i_sb;
1708 ntfs_volume *vol = NTFS_SB(sb);
1709 struct buffer_head *bh;
1711 MFT_RECORD *m = NULL;
1713 ntfs_attr_search_ctx *ctx;
1714 unsigned int i, nr_blocks;
1717 ntfs_debug("Entering.");
1719 /* Initialize the ntfs specific part of @vi. */
1720 ntfs_init_big_inode(vi);
1724 /* Setup the data attribute. It is special as it is mst protected. */
1725 NInoSetNonResident(ni);
1726 NInoSetMstProtected(ni);
1727 NInoSetSparseDisabled(ni);
1733 * This sets up our little cheat allowing us to reuse the async read io
1734 * completion handler for directories.
1736 ni->itype.index.block_size = vol->mft_record_size;
1737 ni->itype.index.block_size_bits = vol->mft_record_size_bits;
1739 /* Very important! Needed to be able to call map_mft_record*(). */
1742 /* Allocate enough memory to read the first mft record. */
1743 if (vol->mft_record_size > 64 * 1024) {
1744 ntfs_error(sb, "Unsupported mft record size %i (max 64kiB).",
1745 vol->mft_record_size);
1748 i = vol->mft_record_size;
1749 if (i < sb->s_blocksize)
1750 i = sb->s_blocksize;
1751 m = (MFT_RECORD*)ntfs_malloc_nofs(i);
1753 ntfs_error(sb, "Failed to allocate buffer for $MFT record 0.");
1757 /* Determine the first block of the $MFT/$DATA attribute. */
1758 block = vol->mft_lcn << vol->cluster_size_bits >>
1759 sb->s_blocksize_bits;
1760 nr_blocks = vol->mft_record_size >> sb->s_blocksize_bits;
1764 /* Load $MFT/$DATA's first mft record. */
1765 for (i = 0; i < nr_blocks; i++) {
1766 bh = sb_bread(sb, block++);
1768 ntfs_error(sb, "Device read failed.");
1771 memcpy((char*)m + (i << sb->s_blocksize_bits), bh->b_data,
1776 /* Apply the mst fixups. */
1777 if (post_read_mst_fixup((NTFS_RECORD*)m, vol->mft_record_size)) {
1778 /* FIXME: Try to use the $MFTMirr now. */
1779 ntfs_error(sb, "MST fixup failed. $MFT is corrupt.");
1783 /* Need this to sanity check attribute list references to $MFT. */
1784 vi->i_generation = ni->seq_no = le16_to_cpu(m->sequence_number);
1786 /* Provides readpage() and sync_page() for map_mft_record(). */
1787 vi->i_mapping->a_ops = &ntfs_mst_aops;
1789 ctx = ntfs_attr_get_search_ctx(ni, m);
1795 /* Find the attribute list attribute if present. */
1796 err = ntfs_attr_lookup(AT_ATTRIBUTE_LIST, NULL, 0, 0, 0, NULL, 0, ctx);
1798 if (unlikely(err != -ENOENT)) {
1799 ntfs_error(sb, "Failed to lookup attribute list "
1800 "attribute. You should run chkdsk.");
1803 } else /* if (!err) */ {
1804 ATTR_LIST_ENTRY *al_entry, *next_al_entry;
1807 ntfs_debug("Attribute list attribute found in $MFT.");
1808 NInoSetAttrList(ni);
1810 if (a->flags & ATTR_IS_ENCRYPTED ||
1811 a->flags & ATTR_COMPRESSION_MASK ||
1812 a->flags & ATTR_IS_SPARSE) {
1813 ntfs_error(sb, "Attribute list attribute is "
1814 "compressed/encrypted/sparse. Not "
1815 "allowed. $MFT is corrupt. You should "
1819 /* Now allocate memory for the attribute list. */
1820 ni->attr_list_size = (u32)ntfs_attr_size(a);
1821 ni->attr_list = ntfs_malloc_nofs(ni->attr_list_size);
1822 if (!ni->attr_list) {
1823 ntfs_error(sb, "Not enough memory to allocate buffer "
1824 "for attribute list.");
1827 if (a->non_resident) {
1828 NInoSetAttrListNonResident(ni);
1829 if (a->data.non_resident.lowest_vcn) {
1830 ntfs_error(sb, "Attribute list has non zero "
1831 "lowest_vcn. $MFT is corrupt. "
1832 "You should run chkdsk.");
1835 /* Setup the runlist. */
1836 ni->attr_list_rl.rl = ntfs_mapping_pairs_decompress(vol,
1838 if (IS_ERR(ni->attr_list_rl.rl)) {
1839 err = PTR_ERR(ni->attr_list_rl.rl);
1840 ni->attr_list_rl.rl = NULL;
1841 ntfs_error(sb, "Mapping pairs decompression "
1842 "failed with error code %i.",
1846 /* Now load the attribute list. */
1847 if ((err = load_attribute_list(vol, &ni->attr_list_rl,
1848 ni->attr_list, ni->attr_list_size,
1849 sle64_to_cpu(a->data.
1850 non_resident.initialized_size)))) {
1851 ntfs_error(sb, "Failed to load attribute list "
1852 "attribute with error code %i.",
1856 } else /* if (!ctx.attr->non_resident) */ {
1857 if ((u8*)a + le16_to_cpu(
1858 a->data.resident.value_offset) +
1860 a->data.resident.value_length) >
1861 (u8*)ctx->mrec + vol->mft_record_size) {
1862 ntfs_error(sb, "Corrupt attribute list "
1866 /* Now copy the attribute list. */
1867 memcpy(ni->attr_list, (u8*)a + le16_to_cpu(
1868 a->data.resident.value_offset),
1870 a->data.resident.value_length));
1872 /* The attribute list is now setup in memory. */
1874 * FIXME: I don't know if this case is actually possible.
1875 * According to logic it is not possible but I have seen too
1876 * many weird things in MS software to rely on logic... Thus we
1877 * perform a manual search and make sure the first $MFT/$DATA
1878 * extent is in the base inode. If it is not we abort with an
1879 * error and if we ever see a report of this error we will need
1880 * to do some magic in order to have the necessary mft record
1881 * loaded and in the right place in the page cache. But
1882 * hopefully logic will prevail and this never happens...
1884 al_entry = (ATTR_LIST_ENTRY*)ni->attr_list;
1885 al_end = (u8*)al_entry + ni->attr_list_size;
1886 for (;; al_entry = next_al_entry) {
1887 /* Out of bounds check. */
1888 if ((u8*)al_entry < ni->attr_list ||
1889 (u8*)al_entry > al_end)
1890 goto em_put_err_out;
1891 /* Catch the end of the attribute list. */
1892 if ((u8*)al_entry == al_end)
1893 goto em_put_err_out;
1894 if (!al_entry->length)
1895 goto em_put_err_out;
1896 if ((u8*)al_entry + 6 > al_end || (u8*)al_entry +
1897 le16_to_cpu(al_entry->length) > al_end)
1898 goto em_put_err_out;
1899 next_al_entry = (ATTR_LIST_ENTRY*)((u8*)al_entry +
1900 le16_to_cpu(al_entry->length));
1901 if (le32_to_cpu(al_entry->type) >
1902 const_le32_to_cpu(AT_DATA))
1903 goto em_put_err_out;
1904 if (AT_DATA != al_entry->type)
1906 /* We want an unnamed attribute. */
1907 if (al_entry->name_length)
1908 goto em_put_err_out;
1909 /* Want the first entry, i.e. lowest_vcn == 0. */
1910 if (al_entry->lowest_vcn)
1911 goto em_put_err_out;
1912 /* First entry has to be in the base mft record. */
1913 if (MREF_LE(al_entry->mft_reference) != vi->i_ino) {
1914 /* MFT references do not match, logic fails. */
1915 ntfs_error(sb, "BUG: The first $DATA extent "
1916 "of $MFT is not in the base "
1917 "mft record. Please report "
1918 "you saw this message to "
1919 "linux-ntfs-dev@lists."
1923 /* Sequence numbers must match. */
1924 if (MSEQNO_LE(al_entry->mft_reference) !=
1926 goto em_put_err_out;
1927 /* Got it. All is ok. We can stop now. */
1933 ntfs_attr_reinit_search_ctx(ctx);
1935 /* Now load all attribute extents. */
1937 next_vcn = last_vcn = highest_vcn = 0;
1938 while (!(err = ntfs_attr_lookup(AT_DATA, NULL, 0, 0, next_vcn, NULL, 0,
1940 runlist_element *nrl;
1942 /* Cache the current attribute. */
1944 /* $MFT must be non-resident. */
1945 if (!a->non_resident) {
1946 ntfs_error(sb, "$MFT must be non-resident but a "
1947 "resident extent was found. $MFT is "
1948 "corrupt. Run chkdsk.");
1951 /* $MFT must be uncompressed and unencrypted. */
1952 if (a->flags & ATTR_COMPRESSION_MASK ||
1953 a->flags & ATTR_IS_ENCRYPTED ||
1954 a->flags & ATTR_IS_SPARSE) {
1955 ntfs_error(sb, "$MFT must be uncompressed, "
1956 "non-sparse, and unencrypted but a "
1957 "compressed/sparse/encrypted extent "
1958 "was found. $MFT is corrupt. Run "
1963 * Decompress the mapping pairs array of this extent and merge
1964 * the result into the existing runlist. No need for locking
1965 * as we have exclusive access to the inode at this time and we
1966 * are a mount in progress task, too.
1968 nrl = ntfs_mapping_pairs_decompress(vol, a, ni->runlist.rl);
1970 ntfs_error(sb, "ntfs_mapping_pairs_decompress() "
1971 "failed with error code %ld. $MFT is "
1972 "corrupt.", PTR_ERR(nrl));
1975 ni->runlist.rl = nrl;
1977 /* Are we in the first extent? */
1979 if (a->data.non_resident.lowest_vcn) {
1980 ntfs_error(sb, "First extent of $DATA "
1981 "attribute has non zero "
1982 "lowest_vcn. $MFT is corrupt. "
1983 "You should run chkdsk.");
1986 /* Get the last vcn in the $DATA attribute. */
1987 last_vcn = sle64_to_cpu(
1988 a->data.non_resident.allocated_size)
1989 >> vol->cluster_size_bits;
1990 /* Fill in the inode size. */
1991 vi->i_size = sle64_to_cpu(
1992 a->data.non_resident.data_size);
1993 ni->initialized_size = sle64_to_cpu(
1994 a->data.non_resident.initialized_size);
1995 ni->allocated_size = sle64_to_cpu(
1996 a->data.non_resident.allocated_size);
1998 * Verify the number of mft records does not exceed
2001 if ((vi->i_size >> vol->mft_record_size_bits) >=
2003 ntfs_error(sb, "$MFT is too big! Aborting.");
2007 * We have got the first extent of the runlist for
2008 * $MFT which means it is now relatively safe to call
2009 * the normal ntfs_read_inode() function.
2010 * Complete reading the inode, this will actually
2011 * re-read the mft record for $MFT, this time entering
2012 * it into the page cache with which we complete the
2013 * kick start of the volume. It should be safe to do
2014 * this now as the first extent of $MFT/$DATA is
2015 * already known and we would hope that we don't need
2016 * further extents in order to find the other
2017 * attributes belonging to $MFT. Only time will tell if
2018 * this is really the case. If not we will have to play
2019 * magic at this point, possibly duplicating a lot of
2020 * ntfs_read_inode() at this point. We will need to
2021 * ensure we do enough of its work to be able to call
2022 * ntfs_read_inode() on extents of $MFT/$DATA. But lets
2023 * hope this never happens...
2025 ntfs_read_locked_inode(vi);
2026 if (is_bad_inode(vi)) {
2027 ntfs_error(sb, "ntfs_read_inode() of $MFT "
2028 "failed. BUG or corrupt $MFT. "
2029 "Run chkdsk and if no errors "
2030 "are found, please report you "
2031 "saw this message to "
2032 "linux-ntfs-dev@lists."
2034 ntfs_attr_put_search_ctx(ctx);
2035 /* Revert to the safe super operations. */
2040 * Re-initialize some specifics about $MFT's inode as
2041 * ntfs_read_inode() will have set up the default ones.
2043 /* Set uid and gid to root. */
2044 vi->i_uid = vi->i_gid = 0;
2045 /* Regular file. No access for anyone. */
2046 vi->i_mode = S_IFREG;
2047 /* No VFS initiated operations allowed for $MFT. */
2048 vi->i_op = &ntfs_empty_inode_ops;
2049 vi->i_fop = &ntfs_empty_file_ops;
2052 /* Get the lowest vcn for the next extent. */
2053 highest_vcn = sle64_to_cpu(a->data.non_resident.highest_vcn);
2054 next_vcn = highest_vcn + 1;
2056 /* Only one extent or error, which we catch below. */
2060 /* Avoid endless loops due to corruption. */
2061 if (next_vcn < sle64_to_cpu(
2062 a->data.non_resident.lowest_vcn)) {
2063 ntfs_error(sb, "$MFT has corrupt attribute list "
2064 "attribute. Run chkdsk.");
2068 if (err != -ENOENT) {
2069 ntfs_error(sb, "Failed to lookup $MFT/$DATA attribute extent. "
2070 "$MFT is corrupt. Run chkdsk.");
2074 ntfs_error(sb, "$MFT/$DATA attribute not found. $MFT is "
2075 "corrupt. Run chkdsk.");
2078 if (highest_vcn && highest_vcn != last_vcn - 1) {
2079 ntfs_error(sb, "Failed to load the complete runlist for "
2080 "$MFT/$DATA. Driver bug or corrupt $MFT. "
2082 ntfs_debug("highest_vcn = 0x%llx, last_vcn - 1 = 0x%llx",
2083 (unsigned long long)highest_vcn,
2084 (unsigned long long)last_vcn - 1);
2087 ntfs_attr_put_search_ctx(ctx);
2088 ntfs_debug("Done.");
2093 ntfs_error(sb, "Couldn't find first extent of $DATA attribute in "
2094 "attribute list. $MFT is corrupt. Run chkdsk.");
2096 ntfs_attr_put_search_ctx(ctx);
2098 ntfs_error(sb, "Failed. Marking inode as bad.");
2105 * ntfs_put_inode - handler for when the inode reference count is decremented
2108 * The VFS calls ntfs_put_inode() every time the inode reference count (i_count)
2109 * is about to be decremented (but before the decrement itself.
2111 * If the inode @vi is a directory with two references, one of which is being
2112 * dropped, we need to put the attribute inode for the directory index bitmap,
2113 * if it is present, otherwise the directory inode would remain pinned for
2116 void ntfs_put_inode(struct inode *vi)
2118 if (S_ISDIR(vi->i_mode) && atomic_read(&vi->i_count) == 2) {
2119 ntfs_inode *ni = NTFS_I(vi);
2120 if (NInoIndexAllocPresent(ni)) {
2121 struct inode *bvi = NULL;
2123 if (atomic_read(&vi->i_count) == 2) {
2124 bvi = ni->itype.index.bmp_ino;
2126 ni->itype.index.bmp_ino = NULL;
2135 static void __ntfs_clear_inode(ntfs_inode *ni)
2137 /* Free all alocated memory. */
2138 down_write(&ni->runlist.lock);
2139 if (ni->runlist.rl) {
2140 ntfs_free(ni->runlist.rl);
2141 ni->runlist.rl = NULL;
2143 up_write(&ni->runlist.lock);
2145 if (ni->attr_list) {
2146 ntfs_free(ni->attr_list);
2147 ni->attr_list = NULL;
2150 down_write(&ni->attr_list_rl.lock);
2151 if (ni->attr_list_rl.rl) {
2152 ntfs_free(ni->attr_list_rl.rl);
2153 ni->attr_list_rl.rl = NULL;
2155 up_write(&ni->attr_list_rl.lock);
2157 if (ni->name_len && ni->name != I30) {
2164 void ntfs_clear_extent_inode(ntfs_inode *ni)
2166 ntfs_debug("Entering for inode 0x%lx.", ni->mft_no);
2168 BUG_ON(NInoAttr(ni));
2169 BUG_ON(ni->nr_extents != -1);
2172 if (NInoDirty(ni)) {
2173 if (!is_bad_inode(VFS_I(ni->ext.base_ntfs_ino)))
2174 ntfs_error(ni->vol->sb, "Clearing dirty extent inode! "
2175 "Losing data! This is a BUG!!!");
2176 // FIXME: Do something!!!
2178 #endif /* NTFS_RW */
2180 __ntfs_clear_inode(ni);
2183 ntfs_destroy_extent_inode(ni);
2187 * ntfs_clear_big_inode - clean up the ntfs specific part of an inode
2188 * @vi: vfs inode pending annihilation
2190 * When the VFS is going to remove an inode from memory, ntfs_clear_big_inode()
2191 * is called, which deallocates all memory belonging to the NTFS specific part
2192 * of the inode and returns.
2194 * If the MFT record is dirty, we commit it before doing anything else.
2196 void ntfs_clear_big_inode(struct inode *vi)
2198 ntfs_inode *ni = NTFS_I(vi);
2201 * If the inode @vi is an index inode we need to put the attribute
2202 * inode for the index bitmap, if it is present, otherwise the index
2203 * inode would disappear and the attribute inode for the index bitmap
2204 * would no longer be referenced from anywhere and thus it would remain
2207 if (NInoAttr(ni) && (ni->type == AT_INDEX_ALLOCATION) &&
2208 NInoIndexAllocPresent(ni) && ni->itype.index.bmp_ino) {
2209 iput(ni->itype.index.bmp_ino);
2210 ni->itype.index.bmp_ino = NULL;
2213 if (NInoDirty(ni)) {
2214 BOOL was_bad = (is_bad_inode(vi));
2216 /* Committing the inode also commits all extent inodes. */
2217 ntfs_commit_inode(vi);
2219 if (!was_bad && (is_bad_inode(vi) || NInoDirty(ni))) {
2220 ntfs_error(vi->i_sb, "Failed to commit dirty inode "
2221 "0x%lx. Losing data!", vi->i_ino);
2222 // FIXME: Do something!!!
2225 #endif /* NTFS_RW */
2227 /* No need to lock at this stage as no one else has a reference. */
2228 if (ni->nr_extents > 0) {
2231 for (i = 0; i < ni->nr_extents; i++)
2232 ntfs_clear_extent_inode(ni->ext.extent_ntfs_inos[i]);
2233 kfree(ni->ext.extent_ntfs_inos);
2236 __ntfs_clear_inode(ni);
2239 /* Release the base inode if we are holding it. */
2240 if (ni->nr_extents == -1) {
2241 iput(VFS_I(ni->ext.base_ntfs_ino));
2243 ni->ext.base_ntfs_ino = NULL;
2250 * ntfs_show_options - show mount options in /proc/mounts
2251 * @sf: seq_file in which to write our mount options
2252 * @mnt: vfs mount whose mount options to display
2254 * Called by the VFS once for each mounted ntfs volume when someone reads
2255 * /proc/mounts in order to display the NTFS specific mount options of each
2256 * mount. The mount options of the vfs mount @mnt are written to the seq file
2257 * @sf and success is returned.
2259 int ntfs_show_options(struct seq_file *sf, struct vfsmount *mnt)
2261 ntfs_volume *vol = NTFS_SB(mnt->mnt_sb);
2264 seq_printf(sf, ",uid=%i", vol->uid);
2265 seq_printf(sf, ",gid=%i", vol->gid);
2266 if (vol->fmask == vol->dmask)
2267 seq_printf(sf, ",umask=0%o", vol->fmask);
2269 seq_printf(sf, ",fmask=0%o", vol->fmask);
2270 seq_printf(sf, ",dmask=0%o", vol->dmask);
2272 seq_printf(sf, ",nls=%s", vol->nls_map->charset);
2273 if (NVolCaseSensitive(vol))
2274 seq_printf(sf, ",case_sensitive");
2275 if (NVolShowSystemFiles(vol))
2276 seq_printf(sf, ",show_sys_files");
2277 if (!NVolSparseEnabled(vol))
2278 seq_printf(sf, ",disable_sparse");
2279 for (i = 0; on_errors_arr[i].val; i++) {
2280 if (on_errors_arr[i].val & vol->on_errors)
2281 seq_printf(sf, ",errors=%s", on_errors_arr[i].str);
2283 seq_printf(sf, ",mft_zone_multiplier=%i", vol->mft_zone_multiplier);
2290 * ntfs_truncate - called when the i_size of an ntfs inode is changed
2291 * @vi: inode for which the i_size was changed
2293 * We do not support i_size changes yet.
2295 * The kernel guarantees that @vi is a regular file (S_ISREG() is true) and
2296 * that the change is allowed.
2298 * This implies for us that @vi is a file inode rather than a directory, index,
2299 * or attribute inode as well as that @vi is a base inode.
2301 * Returns 0 on success or -errno on error.
2303 * Called with ->i_sem held. In all but one case ->i_alloc_sem is held for
2304 * writing. The only case where ->i_alloc_sem is not held is
2305 * mm/filemap.c::generic_file_buffered_write() where vmtruncate() is called
2306 * with the current i_size as the offset which means that it is a noop as far
2307 * as ntfs_truncate() is concerned.
2309 int ntfs_truncate(struct inode *vi)
2311 ntfs_inode *ni = NTFS_I(vi);
2312 ntfs_volume *vol = ni->vol;
2313 ntfs_attr_search_ctx *ctx;
2316 const char *te = " Leaving file length out of sync with i_size.";
2319 ntfs_debug("Entering for inode 0x%lx.", vi->i_ino);
2320 BUG_ON(NInoAttr(ni));
2321 BUG_ON(ni->nr_extents < 0);
2322 m = map_mft_record(ni);
2325 ntfs_error(vi->i_sb, "Failed to map mft record for inode 0x%lx "
2326 "(error code %d).%s", vi->i_ino, err, te);
2331 ctx = ntfs_attr_get_search_ctx(ni, m);
2332 if (unlikely(!ctx)) {
2333 ntfs_error(vi->i_sb, "Failed to allocate a search context for "
2334 "inode 0x%lx (not enough memory).%s",
2339 err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
2340 CASE_SENSITIVE, 0, NULL, 0, ctx);
2341 if (unlikely(err)) {
2343 ntfs_error(vi->i_sb, "Open attribute is missing from "
2344 "mft record. Inode 0x%lx is corrupt. "
2345 "Run chkdsk.", vi->i_ino);
2347 ntfs_error(vi->i_sb, "Failed to lookup attribute in "
2348 "inode 0x%lx (error code %d).",
2353 /* If the size has not changed there is nothing to do. */
2354 if (ntfs_attr_size(a) == i_size_read(vi))
2356 // TODO: Implement the truncate...
2357 ntfs_error(vi->i_sb, "Inode size has changed but this is not "
2358 "implemented yet. Resetting inode size to old value. "
2359 " This is most likely a bug in the ntfs driver!");
2360 i_size_write(vi, ntfs_attr_size(a));
2362 ntfs_attr_put_search_ctx(ctx);
2363 unmap_mft_record(ni);
2364 NInoClearTruncateFailed(ni);
2365 ntfs_debug("Done.");
2368 if (err != -ENOMEM) {
2373 ntfs_attr_put_search_ctx(ctx);
2375 unmap_mft_record(ni);
2376 NInoSetTruncateFailed(ni);
2381 * ntfs_truncate_vfs - wrapper for ntfs_truncate() that has no return value
2382 * @vi: inode for which the i_size was changed
2384 * Wrapper for ntfs_truncate() that has no return value.
2386 * See ntfs_truncate() description above for details.
2388 void ntfs_truncate_vfs(struct inode *vi) {
2393 * ntfs_setattr - called from notify_change() when an attribute is being changed
2394 * @dentry: dentry whose attributes to change
2395 * @attr: structure describing the attributes and the changes
2397 * We have to trap VFS attempts to truncate the file described by @dentry as
2398 * soon as possible, because we do not implement changes in i_size yet. So we
2399 * abort all i_size changes here.
2401 * We also abort all changes of user, group, and mode as we do not implement
2402 * the NTFS ACLs yet.
2404 * Called with ->i_sem held. For the ATTR_SIZE (i.e. ->truncate) case, also
2405 * called with ->i_alloc_sem held for writing.
2407 * Basically this is a copy of generic notify_change() and inode_setattr()
2408 * functionality, except we intercept and abort changes in i_size.
2410 int ntfs_setattr(struct dentry *dentry, struct iattr *attr)
2412 struct inode *vi = dentry->d_inode;
2414 unsigned int ia_valid = attr->ia_valid;
2416 err = inode_change_ok(vi, attr);
2420 /* We do not support NTFS ACLs yet. */
2421 if (ia_valid & (ATTR_UID | ATTR_GID | ATTR_MODE)) {
2422 ntfs_warning(vi->i_sb, "Changes in user/group/mode are not "
2423 "supported yet, ignoring.");
2428 if (ia_valid & ATTR_SIZE) {
2429 if (attr->ia_size != i_size_read(vi)) {
2430 ntfs_warning(vi->i_sb, "Changes in inode size are not "
2431 "supported yet, ignoring.");
2433 // TODO: Implement...
2434 // err = vmtruncate(vi, attr->ia_size);
2435 if (err || ia_valid == ATTR_SIZE)
2439 * We skipped the truncate but must still update
2442 ia_valid |= ATTR_MTIME|ATTR_CTIME;
2446 if (ia_valid & ATTR_ATIME)
2447 vi->i_atime = attr->ia_atime;
2448 if (ia_valid & ATTR_MTIME)
2449 vi->i_mtime = attr->ia_mtime;
2450 if (ia_valid & ATTR_CTIME)
2451 vi->i_ctime = attr->ia_ctime;
2452 mark_inode_dirty(vi);
2458 * ntfs_write_inode - write out a dirty inode
2459 * @vi: inode to write out
2460 * @sync: if true, write out synchronously
2462 * Write out a dirty inode to disk including any extent inodes if present.
2464 * If @sync is true, commit the inode to disk and wait for io completion. This
2465 * is done using write_mft_record().
2467 * If @sync is false, just schedule the write to happen but do not wait for i/o
2468 * completion. In 2.6 kernels, scheduling usually happens just by virtue of
2469 * marking the page (and in this case mft record) dirty but we do not implement
2470 * this yet as write_mft_record() largely ignores the @sync parameter and
2471 * always performs synchronous writes.
2473 * Return 0 on success and -errno on error.
2475 int ntfs_write_inode(struct inode *vi, int sync)
2478 ntfs_inode *ni = NTFS_I(vi);
2479 ntfs_attr_search_ctx *ctx;
2481 STANDARD_INFORMATION *si;
2483 BOOL modified = FALSE;
2485 ntfs_debug("Entering for %sinode 0x%lx.", NInoAttr(ni) ? "attr " : "",
2488 * Dirty attribute inodes are written via their real inodes so just
2489 * clean them here. Access time updates are taken care off when the
2490 * real inode is written.
2494 ntfs_debug("Done.");
2497 /* Map, pin, and lock the mft record belonging to the inode. */
2498 m = map_mft_record(ni);
2503 /* Update the access times in the standard information attribute. */
2504 ctx = ntfs_attr_get_search_ctx(ni, m);
2505 if (unlikely(!ctx)) {
2509 err = ntfs_attr_lookup(AT_STANDARD_INFORMATION, NULL, 0,
2510 CASE_SENSITIVE, 0, NULL, 0, ctx);
2511 if (unlikely(err)) {
2512 ntfs_attr_put_search_ctx(ctx);
2515 si = (STANDARD_INFORMATION*)((u8*)ctx->attr +
2516 le16_to_cpu(ctx->attr->data.resident.value_offset));
2517 /* Update the access times if they have changed. */
2518 nt = utc2ntfs(vi->i_mtime);
2519 if (si->last_data_change_time != nt) {
2520 ntfs_debug("Updating mtime for inode 0x%lx: old = 0x%llx, "
2521 "new = 0x%llx", vi->i_ino, (long long)
2522 sle64_to_cpu(si->last_data_change_time),
2523 (long long)sle64_to_cpu(nt));
2524 si->last_data_change_time = nt;
2527 nt = utc2ntfs(vi->i_ctime);
2528 if (si->last_mft_change_time != nt) {
2529 ntfs_debug("Updating ctime for inode 0x%lx: old = 0x%llx, "
2530 "new = 0x%llx", vi->i_ino, (long long)
2531 sle64_to_cpu(si->last_mft_change_time),
2532 (long long)sle64_to_cpu(nt));
2533 si->last_mft_change_time = nt;
2536 nt = utc2ntfs(vi->i_atime);
2537 if (si->last_access_time != nt) {
2538 ntfs_debug("Updating atime for inode 0x%lx: old = 0x%llx, "
2539 "new = 0x%llx", vi->i_ino,
2540 (long long)sle64_to_cpu(si->last_access_time),
2541 (long long)sle64_to_cpu(nt));
2542 si->last_access_time = nt;
2546 * If we just modified the standard information attribute we need to
2547 * mark the mft record it is in dirty. We do this manually so that
2548 * mark_inode_dirty() is not called which would redirty the inode and
2549 * hence result in an infinite loop of trying to write the inode.
2550 * There is no need to mark the base inode nor the base mft record
2551 * dirty, since we are going to write this mft record below in any case
2552 * and the base mft record may actually not have been modified so it
2553 * might not need to be written out.
2554 * NOTE: It is not a problem when the inode for $MFT itself is being
2555 * written out as mark_ntfs_record_dirty() will only set I_DIRTY_PAGES
2556 * on the $MFT inode and hence ntfs_write_inode() will not be
2557 * re-invoked because of it which in turn is ok since the dirtied mft
2558 * record will be cleaned and written out to disk below, i.e. before
2559 * this function returns.
2561 if (modified && !NInoTestSetDirty(ctx->ntfs_ino))
2562 mark_ntfs_record_dirty(ctx->ntfs_ino->page,
2563 ctx->ntfs_ino->page_ofs);
2564 ntfs_attr_put_search_ctx(ctx);
2565 /* Now the access times are updated, write the base mft record. */
2567 err = write_mft_record(ni, m, sync);
2568 /* Write all attached extent mft records. */
2569 down(&ni->extent_lock);
2570 if (ni->nr_extents > 0) {
2571 ntfs_inode **extent_nis = ni->ext.extent_ntfs_inos;
2574 ntfs_debug("Writing %i extent inodes.", ni->nr_extents);
2575 for (i = 0; i < ni->nr_extents; i++) {
2576 ntfs_inode *tni = extent_nis[i];
2578 if (NInoDirty(tni)) {
2579 MFT_RECORD *tm = map_mft_record(tni);
2583 if (!err || err == -ENOMEM)
2587 ret = write_mft_record(tni, tm, sync);
2588 unmap_mft_record(tni);
2589 if (unlikely(ret)) {
2590 if (!err || err == -ENOMEM)
2596 up(&ni->extent_lock);
2597 unmap_mft_record(ni);
2600 ntfs_debug("Done.");
2603 unmap_mft_record(ni);
2605 if (err == -ENOMEM) {
2606 ntfs_warning(vi->i_sb, "Not enough memory to write inode. "
2607 "Marking the inode dirty again, so the VFS "
2609 mark_inode_dirty(vi);
2611 ntfs_error(vi->i_sb, "Failed (error code %i): Marking inode "
2612 "as bad. You should run chkdsk.", -err);
2614 NVolSetErrors(ni->vol);
2619 #endif /* NTFS_RW */