2 * attrib.c - NTFS attribute operations. Part of the Linux-NTFS project.
4 * Copyright (c) 2001-2012 Anton Altaparmakov and Tuxera Inc.
5 * Copyright (c) 2002 Richard Russon
7 * This program/include file is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU General Public License as published
9 * by the Free Software Foundation; either version 2 of the License, or
10 * (at your option) any later version.
12 * This program/include file is distributed in the hope that it will be
13 * useful, but WITHOUT ANY WARRANTY; without even the implied warranty
14 * of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
17 * You should have received a copy of the GNU General Public License
18 * along with this program (in the main directory of the Linux-NTFS
19 * distribution in the file COPYING); if not, write to the Free Software
20 * Foundation,Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
23 #include <linux/buffer_head.h>
24 #include <linux/sched.h>
25 #include <linux/slab.h>
26 #include <linux/swap.h>
27 #include <linux/writeback.h>
39 * ntfs_map_runlist_nolock - map (a part of) a runlist of an ntfs inode
40 * @ni: ntfs inode for which to map (part of) a runlist
41 * @vcn: map runlist part containing this vcn
42 * @ctx: active attribute search context if present or NULL if not
44 * Map the part of a runlist containing the @vcn of the ntfs inode @ni.
46 * If @ctx is specified, it is an active search context of @ni and its base mft
47 * record. This is needed when ntfs_map_runlist_nolock() encounters unmapped
48 * runlist fragments and allows their mapping. If you do not have the mft
49 * record mapped, you can specify @ctx as NULL and ntfs_map_runlist_nolock()
50 * will perform the necessary mapping and unmapping.
52 * Note, ntfs_map_runlist_nolock() saves the state of @ctx on entry and
53 * restores it before returning. Thus, @ctx will be left pointing to the same
54 * attribute on return as on entry. However, the actual pointers in @ctx may
55 * point to different memory locations on return, so you must remember to reset
56 * any cached pointers from the @ctx, i.e. after the call to
57 * ntfs_map_runlist_nolock(), you will probably want to do:
60 * Assuming you cache ctx->attr in a variable @a of type ATTR_RECORD * and that
61 * you cache ctx->mrec in a variable @m of type MFT_RECORD *.
63 * Return 0 on success and -errno on error. There is one special error code
64 * which is not an error as such. This is -ENOENT. It means that @vcn is out
65 * of bounds of the runlist.
67 * Note the runlist can be NULL after this function returns if @vcn is zero and
68 * the attribute has zero allocated size, i.e. there simply is no runlist.
70 * WARNING: If @ctx is supplied, regardless of whether success or failure is
71 * returned, you need to check IS_ERR(@ctx->mrec) and if 'true' the @ctx
72 * is no longer valid, i.e. you need to either call
73 * ntfs_attr_reinit_search_ctx() or ntfs_attr_put_search_ctx() on it.
74 * In that case PTR_ERR(@ctx->mrec) will give you the error code for
75 * why the mapping of the old inode failed.
77 * Locking: - The runlist described by @ni must be locked for writing on entry
78 * and is locked on return. Note the runlist will be modified.
79 * - If @ctx is NULL, the base mft record of @ni must not be mapped on
80 * entry and it will be left unmapped on return.
81 * - If @ctx is not NULL, the base mft record must be mapped on entry
82 * and it will be left mapped on return.
84 int ntfs_map_runlist_nolock(ntfs_inode *ni, VCN vcn, ntfs_attr_search_ctx *ctx)
92 struct page *put_this_page = NULL;
94 bool ctx_is_temporary, ctx_needs_reset;
95 ntfs_attr_search_ctx old_ctx = { NULL, };
97 ntfs_debug("Mapping runlist part containing vcn 0x%llx.",
98 (unsigned long long)vcn);
102 base_ni = ni->ext.base_ntfs_ino;
104 ctx_is_temporary = ctx_needs_reset = true;
105 m = map_mft_record(base_ni);
108 ctx = ntfs_attr_get_search_ctx(base_ni, m);
109 if (unlikely(!ctx)) {
114 VCN allocated_size_vcn;
116 BUG_ON(IS_ERR(ctx->mrec));
118 BUG_ON(!a->non_resident);
119 ctx_is_temporary = false;
120 end_vcn = sle64_to_cpu(a->data.non_resident.highest_vcn);
121 read_lock_irqsave(&ni->size_lock, flags);
122 allocated_size_vcn = ni->allocated_size >>
123 ni->vol->cluster_size_bits;
124 read_unlock_irqrestore(&ni->size_lock, flags);
125 if (!a->data.non_resident.lowest_vcn && end_vcn <= 0)
126 end_vcn = allocated_size_vcn - 1;
128 * If we already have the attribute extent containing @vcn in
129 * @ctx, no need to look it up again. We slightly cheat in
130 * that if vcn exceeds the allocated size, we will refuse to
131 * map the runlist below, so there is definitely no need to get
132 * the right attribute extent.
134 if (vcn >= allocated_size_vcn || (a->type == ni->type &&
135 a->name_length == ni->name_len &&
136 !memcmp((u8*)a + le16_to_cpu(a->name_offset),
137 ni->name, ni->name_len) &&
138 sle64_to_cpu(a->data.non_resident.lowest_vcn)
139 <= vcn && end_vcn >= vcn))
140 ctx_needs_reset = false;
142 /* Save the old search context. */
145 * If the currently mapped (extent) inode is not the
146 * base inode we will unmap it when we reinitialize the
147 * search context which means we need to get a
148 * reference to the page containing the mapped mft
149 * record so we do not accidentally drop changes to the
150 * mft record when it has not been marked dirty yet.
152 if (old_ctx.base_ntfs_ino && old_ctx.ntfs_ino !=
153 old_ctx.base_ntfs_ino) {
154 put_this_page = old_ctx.ntfs_ino->page;
155 get_page(put_this_page);
158 * Reinitialize the search context so we can lookup the
159 * needed attribute extent.
161 ntfs_attr_reinit_search_ctx(ctx);
162 ctx_needs_reset = true;
165 if (ctx_needs_reset) {
166 err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
167 CASE_SENSITIVE, vcn, NULL, 0, ctx);
173 BUG_ON(!ctx->attr->non_resident);
177 * Only decompress the mapping pairs if @vcn is inside it. Otherwise
178 * we get into problems when we try to map an out of bounds vcn because
179 * we then try to map the already mapped runlist fragment and
180 * ntfs_mapping_pairs_decompress() fails.
182 end_vcn = sle64_to_cpu(a->data.non_resident.highest_vcn) + 1;
183 if (unlikely(vcn && vcn >= end_vcn)) {
187 rl = ntfs_mapping_pairs_decompress(ni->vol, a, ni->runlist.rl);
193 if (ctx_is_temporary) {
195 ntfs_attr_put_search_ctx(ctx);
196 unmap_mft_record(base_ni);
197 } else if (ctx_needs_reset) {
199 * If there is no attribute list, restoring the search context
200 * is accomplished simply by copying the saved context back over
201 * the caller supplied context. If there is an attribute list,
202 * things are more complicated as we need to deal with mapping
203 * of mft records and resulting potential changes in pointers.
205 if (NInoAttrList(base_ni)) {
207 * If the currently mapped (extent) inode is not the
208 * one we had before, we need to unmap it and map the
211 if (ctx->ntfs_ino != old_ctx.ntfs_ino) {
213 * If the currently mapped inode is not the
214 * base inode, unmap it.
216 if (ctx->base_ntfs_ino && ctx->ntfs_ino !=
217 ctx->base_ntfs_ino) {
218 unmap_extent_mft_record(ctx->ntfs_ino);
219 ctx->mrec = ctx->base_mrec;
223 * If the old mapped inode is not the base
226 if (old_ctx.base_ntfs_ino &&
228 old_ctx.base_ntfs_ino) {
230 ctx->mrec = map_mft_record(
233 * Something bad has happened. If out
234 * of memory retry till it succeeds.
235 * Any other errors are fatal and we
236 * return the error code in ctx->mrec.
237 * Let the caller deal with it... We
238 * just need to fudge things so the
239 * caller can reinit and/or put the
240 * search context safely.
242 if (IS_ERR(ctx->mrec)) {
243 if (PTR_ERR(ctx->mrec) ==
254 /* Update the changed pointers in the saved context. */
255 if (ctx->mrec != old_ctx.mrec) {
256 if (!IS_ERR(ctx->mrec))
257 old_ctx.attr = (ATTR_RECORD*)(
261 old_ctx.mrec = ctx->mrec;
264 /* Restore the search context to the saved one. */
267 * We drop the reference on the page we took earlier. In the
268 * case that IS_ERR(ctx->mrec) is true this means we might lose
269 * some changes to the mft record that had been made between
270 * the last time it was marked dirty/written out and now. This
271 * at this stage is not a problem as the mapping error is fatal
272 * enough that the mft record cannot be written out anyway and
273 * the caller is very likely to shutdown the whole inode
274 * immediately and mark the volume dirty for chkdsk to pick up
278 put_page(put_this_page);
284 * ntfs_map_runlist - map (a part of) a runlist of an ntfs inode
285 * @ni: ntfs inode for which to map (part of) a runlist
286 * @vcn: map runlist part containing this vcn
288 * Map the part of a runlist containing the @vcn of the ntfs inode @ni.
290 * Return 0 on success and -errno on error. There is one special error code
291 * which is not an error as such. This is -ENOENT. It means that @vcn is out
292 * of bounds of the runlist.
294 * Locking: - The runlist must be unlocked on entry and is unlocked on return.
295 * - This function takes the runlist lock for writing and may modify
298 int ntfs_map_runlist(ntfs_inode *ni, VCN vcn)
302 down_write(&ni->runlist.lock);
303 /* Make sure someone else didn't do the work while we were sleeping. */
304 if (likely(ntfs_rl_vcn_to_lcn(ni->runlist.rl, vcn) <=
306 err = ntfs_map_runlist_nolock(ni, vcn, NULL);
307 up_write(&ni->runlist.lock);
312 * ntfs_attr_vcn_to_lcn_nolock - convert a vcn into a lcn given an ntfs inode
313 * @ni: ntfs inode of the attribute whose runlist to search
314 * @vcn: vcn to convert
315 * @write_locked: true if the runlist is locked for writing
317 * Find the virtual cluster number @vcn in the runlist of the ntfs attribute
318 * described by the ntfs inode @ni and return the corresponding logical cluster
321 * If the @vcn is not mapped yet, the attempt is made to map the attribute
322 * extent containing the @vcn and the vcn to lcn conversion is retried.
324 * If @write_locked is true the caller has locked the runlist for writing and
325 * if false for reading.
327 * Since lcns must be >= 0, we use negative return codes with special meaning:
329 * Return code Meaning / Description
330 * ==========================================
331 * LCN_HOLE Hole / not allocated on disk.
332 * LCN_ENOENT There is no such vcn in the runlist, i.e. @vcn is out of bounds.
333 * LCN_ENOMEM Not enough memory to map runlist.
334 * LCN_EIO Critical error (runlist/file is corrupt, i/o error, etc).
336 * Locking: - The runlist must be locked on entry and is left locked on return.
337 * - If @write_locked is 'false', i.e. the runlist is locked for reading,
338 * the lock may be dropped inside the function so you cannot rely on
339 * the runlist still being the same when this function returns.
341 LCN ntfs_attr_vcn_to_lcn_nolock(ntfs_inode *ni, const VCN vcn,
342 const bool write_locked)
346 bool is_retry = false;
349 ntfs_debug("Entering for i_ino 0x%lx, vcn 0x%llx, %s_locked.",
350 ni->mft_no, (unsigned long long)vcn,
351 write_locked ? "write" : "read");
352 BUG_ON(!NInoNonResident(ni));
354 if (!ni->runlist.rl) {
355 read_lock_irqsave(&ni->size_lock, flags);
356 if (!ni->allocated_size) {
357 read_unlock_irqrestore(&ni->size_lock, flags);
360 read_unlock_irqrestore(&ni->size_lock, flags);
363 /* Convert vcn to lcn. If that fails map the runlist and retry once. */
364 lcn = ntfs_rl_vcn_to_lcn(ni->runlist.rl, vcn);
365 if (likely(lcn >= LCN_HOLE)) {
366 ntfs_debug("Done, lcn 0x%llx.", (long long)lcn);
369 if (lcn != LCN_RL_NOT_MAPPED) {
370 if (lcn != LCN_ENOENT)
372 } else if (!is_retry) {
376 up_read(&ni->runlist.lock);
377 down_write(&ni->runlist.lock);
378 if (unlikely(ntfs_rl_vcn_to_lcn(ni->runlist.rl, vcn) !=
379 LCN_RL_NOT_MAPPED)) {
380 up_write(&ni->runlist.lock);
381 down_read(&ni->runlist.lock);
385 err = ntfs_map_runlist_nolock(ni, vcn, NULL);
387 up_write(&ni->runlist.lock);
388 down_read(&ni->runlist.lock);
396 else if (err == -ENOMEM)
401 if (lcn != LCN_ENOENT)
402 ntfs_error(ni->vol->sb, "Failed with error code %lli.",
408 * ntfs_attr_find_vcn_nolock - find a vcn in the runlist of an ntfs inode
409 * @ni: ntfs inode describing the runlist to search
411 * @ctx: active attribute search context if present or NULL if not
413 * Find the virtual cluster number @vcn in the runlist described by the ntfs
414 * inode @ni and return the address of the runlist element containing the @vcn.
416 * If the @vcn is not mapped yet, the attempt is made to map the attribute
417 * extent containing the @vcn and the vcn to lcn conversion is retried.
419 * If @ctx is specified, it is an active search context of @ni and its base mft
420 * record. This is needed when ntfs_attr_find_vcn_nolock() encounters unmapped
421 * runlist fragments and allows their mapping. If you do not have the mft
422 * record mapped, you can specify @ctx as NULL and ntfs_attr_find_vcn_nolock()
423 * will perform the necessary mapping and unmapping.
425 * Note, ntfs_attr_find_vcn_nolock() saves the state of @ctx on entry and
426 * restores it before returning. Thus, @ctx will be left pointing to the same
427 * attribute on return as on entry. However, the actual pointers in @ctx may
428 * point to different memory locations on return, so you must remember to reset
429 * any cached pointers from the @ctx, i.e. after the call to
430 * ntfs_attr_find_vcn_nolock(), you will probably want to do:
433 * Assuming you cache ctx->attr in a variable @a of type ATTR_RECORD * and that
434 * you cache ctx->mrec in a variable @m of type MFT_RECORD *.
435 * Note you need to distinguish between the lcn of the returned runlist element
436 * being >= 0 and LCN_HOLE. In the later case you have to return zeroes on
437 * read and allocate clusters on write.
439 * Return the runlist element containing the @vcn on success and
440 * ERR_PTR(-errno) on error. You need to test the return value with IS_ERR()
441 * to decide if the return is success or failure and PTR_ERR() to get to the
442 * error code if IS_ERR() is true.
444 * The possible error return codes are:
445 * -ENOENT - No such vcn in the runlist, i.e. @vcn is out of bounds.
446 * -ENOMEM - Not enough memory to map runlist.
447 * -EIO - Critical error (runlist/file is corrupt, i/o error, etc).
449 * WARNING: If @ctx is supplied, regardless of whether success or failure is
450 * returned, you need to check IS_ERR(@ctx->mrec) and if 'true' the @ctx
451 * is no longer valid, i.e. you need to either call
452 * ntfs_attr_reinit_search_ctx() or ntfs_attr_put_search_ctx() on it.
453 * In that case PTR_ERR(@ctx->mrec) will give you the error code for
454 * why the mapping of the old inode failed.
456 * Locking: - The runlist described by @ni must be locked for writing on entry
457 * and is locked on return. Note the runlist may be modified when
458 * needed runlist fragments need to be mapped.
459 * - If @ctx is NULL, the base mft record of @ni must not be mapped on
460 * entry and it will be left unmapped on return.
461 * - If @ctx is not NULL, the base mft record must be mapped on entry
462 * and it will be left mapped on return.
464 runlist_element *ntfs_attr_find_vcn_nolock(ntfs_inode *ni, const VCN vcn,
465 ntfs_attr_search_ctx *ctx)
470 bool is_retry = false;
473 ntfs_debug("Entering for i_ino 0x%lx, vcn 0x%llx, with%s ctx.",
474 ni->mft_no, (unsigned long long)vcn, ctx ? "" : "out");
475 BUG_ON(!NInoNonResident(ni));
477 if (!ni->runlist.rl) {
478 read_lock_irqsave(&ni->size_lock, flags);
479 if (!ni->allocated_size) {
480 read_unlock_irqrestore(&ni->size_lock, flags);
481 return ERR_PTR(-ENOENT);
483 read_unlock_irqrestore(&ni->size_lock, flags);
487 if (likely(rl && vcn >= rl[0].vcn)) {
488 while (likely(rl->length)) {
489 if (unlikely(vcn < rl[1].vcn)) {
490 if (likely(rl->lcn >= LCN_HOLE)) {
498 if (likely(rl->lcn != LCN_RL_NOT_MAPPED)) {
499 if (likely(rl->lcn == LCN_ENOENT))
505 if (!err && !is_retry) {
507 * If the search context is invalid we cannot map the unmapped
510 if (IS_ERR(ctx->mrec))
511 err = PTR_ERR(ctx->mrec);
514 * The @vcn is in an unmapped region, map the runlist
517 err = ntfs_map_runlist_nolock(ni, vcn, ctx);
528 ntfs_error(ni->vol->sb, "Failed with error code %i.", err);
533 * ntfs_attr_find - find (next) attribute in mft record
534 * @type: attribute type to find
535 * @name: attribute name to find (optional, i.e. NULL means don't care)
536 * @name_len: attribute name length (only needed if @name present)
537 * @ic: IGNORE_CASE or CASE_SENSITIVE (ignored if @name not present)
538 * @val: attribute value to find (optional, resident attributes only)
539 * @val_len: attribute value length
540 * @ctx: search context with mft record and attribute to search from
542 * You should not need to call this function directly. Use ntfs_attr_lookup()
545 * ntfs_attr_find() takes a search context @ctx as parameter and searches the
546 * mft record specified by @ctx->mrec, beginning at @ctx->attr, for an
547 * attribute of @type, optionally @name and @val.
549 * If the attribute is found, ntfs_attr_find() returns 0 and @ctx->attr will
550 * point to the found attribute.
552 * If the attribute is not found, ntfs_attr_find() returns -ENOENT and
553 * @ctx->attr will point to the attribute before which the attribute being
554 * searched for would need to be inserted if such an action were to be desired.
556 * On actual error, ntfs_attr_find() returns -EIO. In this case @ctx->attr is
557 * undefined and in particular do not rely on it not changing.
559 * If @ctx->is_first is 'true', the search begins with @ctx->attr itself. If it
560 * is 'false', the search begins after @ctx->attr.
562 * If @ic is IGNORE_CASE, the @name comparisson is not case sensitive and
563 * @ctx->ntfs_ino must be set to the ntfs inode to which the mft record
564 * @ctx->mrec belongs. This is so we can get at the ntfs volume and hence at
565 * the upcase table. If @ic is CASE_SENSITIVE, the comparison is case
566 * sensitive. When @name is present, @name_len is the @name length in Unicode
569 * If @name is not present (NULL), we assume that the unnamed attribute is
570 * being searched for.
572 * Finally, the resident attribute value @val is looked for, if present. If
573 * @val is not present (NULL), @val_len is ignored.
575 * ntfs_attr_find() only searches the specified mft record and it ignores the
576 * presence of an attribute list attribute (unless it is the one being searched
577 * for, obviously). If you need to take attribute lists into consideration,
578 * use ntfs_attr_lookup() instead (see below). This also means that you cannot
579 * use ntfs_attr_find() to search for extent records of non-resident
580 * attributes, as extents with lowest_vcn != 0 are usually described by the
581 * attribute list attribute only. - Note that it is possible that the first
582 * extent is only in the attribute list while the last extent is in the base
583 * mft record, so do not rely on being able to find the first extent in the
586 * Warning: Never use @val when looking for attribute types which can be
587 * non-resident as this most likely will result in a crash!
589 static int ntfs_attr_find(const ATTR_TYPE type, const ntfschar *name,
590 const u32 name_len, const IGNORE_CASE_BOOL ic,
591 const u8 *val, const u32 val_len, ntfs_attr_search_ctx *ctx)
594 ntfs_volume *vol = ctx->ntfs_ino->vol;
595 ntfschar *upcase = vol->upcase;
596 u32 upcase_len = vol->upcase_len;
599 * Iterate over attributes in mft record starting at @ctx->attr, or the
600 * attribute following that, if @ctx->is_first is 'true'.
604 ctx->is_first = false;
606 a = (ATTR_RECORD*)((u8*)ctx->attr +
607 le32_to_cpu(ctx->attr->length));
608 for (;; a = (ATTR_RECORD*)((u8*)a + le32_to_cpu(a->length))) {
609 u8 *mrec_end = (u8 *)ctx->mrec +
610 le32_to_cpu(ctx->mrec->bytes_allocated);
611 u8 *name_end = (u8 *)a + le16_to_cpu(a->name_offset) +
612 a->name_length * sizeof(ntfschar);
613 if ((u8*)a < (u8*)ctx->mrec || (u8*)a > mrec_end ||
617 if (unlikely(le32_to_cpu(a->type) > le32_to_cpu(type) ||
620 if (unlikely(!a->length))
625 * If @name is present, compare the two names. If @name is
626 * missing, assume we want an unnamed attribute.
629 /* The search failed if the found attribute is named. */
632 } else if (!ntfs_are_names_equal(name, name_len,
633 (ntfschar*)((u8*)a + le16_to_cpu(a->name_offset)),
634 a->name_length, ic, upcase, upcase_len)) {
637 rc = ntfs_collate_names(name, name_len,
639 le16_to_cpu(a->name_offset)),
640 a->name_length, 1, IGNORE_CASE,
643 * If @name collates before a->name, there is no
644 * matching attribute.
648 /* If the strings are not equal, continue search. */
651 rc = ntfs_collate_names(name, name_len,
653 le16_to_cpu(a->name_offset)),
654 a->name_length, 1, CASE_SENSITIVE,
662 * The names match or @name not present and attribute is
663 * unnamed. If no @val specified, we have found the attribute
668 /* @val is present; compare values. */
672 rc = memcmp(val, (u8*)a + le16_to_cpu(
673 a->data.resident.value_offset),
674 min_t(u32, val_len, le32_to_cpu(
675 a->data.resident.value_length)));
677 * If @val collates before the current attribute's
678 * value, there is no matching attribute.
684 a->data.resident.value_length);
693 ntfs_error(vol->sb, "Inode is corrupt. Run chkdsk.");
699 * load_attribute_list - load an attribute list into memory
700 * @vol: ntfs volume from which to read
701 * @runlist: runlist of the attribute list
702 * @al_start: destination buffer
703 * @size: size of the destination buffer in bytes
704 * @initialized_size: initialized size of the attribute list
706 * Walk the runlist @runlist and load all clusters from it copying them into
707 * the linear buffer @al. The maximum number of bytes copied to @al is @size
708 * bytes. Note, @size does not need to be a multiple of the cluster size. If
709 * @initialized_size is less than @size, the region in @al between
710 * @initialized_size and @size will be zeroed and not read from disk.
712 * Return 0 on success or -errno on error.
714 int load_attribute_list(ntfs_volume *vol, runlist *runlist, u8 *al_start,
715 const s64 size, const s64 initialized_size)
719 u8 *al_end = al + initialized_size;
721 struct buffer_head *bh;
722 struct super_block *sb;
723 unsigned long block_size;
724 unsigned long block, max_block;
726 unsigned char block_size_bits;
728 ntfs_debug("Entering.");
729 if (!vol || !runlist || !al || size <= 0 || initialized_size < 0 ||
730 initialized_size > size)
732 if (!initialized_size) {
737 block_size = sb->s_blocksize;
738 block_size_bits = sb->s_blocksize_bits;
739 down_read(&runlist->lock);
742 ntfs_error(sb, "Cannot read attribute list since runlist is "
746 /* Read all clusters specified by the runlist one run at a time. */
748 lcn = ntfs_rl_vcn_to_lcn(rl, rl->vcn);
749 ntfs_debug("Reading vcn = 0x%llx, lcn = 0x%llx.",
750 (unsigned long long)rl->vcn,
751 (unsigned long long)lcn);
752 /* The attribute list cannot be sparse. */
754 ntfs_error(sb, "ntfs_rl_vcn_to_lcn() failed. Cannot "
755 "read attribute list.");
758 block = lcn << vol->cluster_size_bits >> block_size_bits;
759 /* Read the run from device in chunks of block_size bytes. */
760 max_block = block + (rl->length << vol->cluster_size_bits >>
762 ntfs_debug("max_block = 0x%lx.", max_block);
764 ntfs_debug("Reading block = 0x%lx.", block);
765 bh = sb_bread(sb, block);
767 ntfs_error(sb, "sb_bread() failed. Cannot "
768 "read attribute list.");
771 if (al + block_size >= al_end)
773 memcpy(al, bh->b_data, block_size);
776 } while (++block < max_block);
779 if (initialized_size < size) {
781 memset(al_start + initialized_size, 0, size - initialized_size);
784 up_read(&runlist->lock);
791 * Note: The attribute list can be smaller than its allocation
792 * by multiple clusters. This has been encountered by at least
793 * two people running Windows XP, thus we cannot do any
794 * truncation sanity checking here. (AIA)
796 memcpy(al, bh->b_data, al_end - al);
798 if (initialized_size < size)
804 ntfs_error(sb, "Attribute list buffer overflow. Read attribute list "
812 * ntfs_external_attr_find - find an attribute in the attribute list of an inode
813 * @type: attribute type to find
814 * @name: attribute name to find (optional, i.e. NULL means don't care)
815 * @name_len: attribute name length (only needed if @name present)
816 * @ic: IGNORE_CASE or CASE_SENSITIVE (ignored if @name not present)
817 * @lowest_vcn: lowest vcn to find (optional, non-resident attributes only)
818 * @val: attribute value to find (optional, resident attributes only)
819 * @val_len: attribute value length
820 * @ctx: search context with mft record and attribute to search from
822 * You should not need to call this function directly. Use ntfs_attr_lookup()
825 * Find an attribute by searching the attribute list for the corresponding
826 * attribute list entry. Having found the entry, map the mft record if the
827 * attribute is in a different mft record/inode, ntfs_attr_find() the attribute
828 * in there and return it.
830 * On first search @ctx->ntfs_ino must be the base mft record and @ctx must
831 * have been obtained from a call to ntfs_attr_get_search_ctx(). On subsequent
832 * calls @ctx->ntfs_ino can be any extent inode, too (@ctx->base_ntfs_ino is
833 * then the base inode).
835 * After finishing with the attribute/mft record you need to call
836 * ntfs_attr_put_search_ctx() to cleanup the search context (unmapping any
837 * mapped inodes, etc).
839 * If the attribute is found, ntfs_external_attr_find() returns 0 and
840 * @ctx->attr will point to the found attribute. @ctx->mrec will point to the
841 * mft record in which @ctx->attr is located and @ctx->al_entry will point to
842 * the attribute list entry for the attribute.
844 * If the attribute is not found, ntfs_external_attr_find() returns -ENOENT and
845 * @ctx->attr will point to the attribute in the base mft record before which
846 * the attribute being searched for would need to be inserted if such an action
847 * were to be desired. @ctx->mrec will point to the mft record in which
848 * @ctx->attr is located and @ctx->al_entry will point to the attribute list
849 * entry of the attribute before which the attribute being searched for would
850 * need to be inserted if such an action were to be desired.
852 * Thus to insert the not found attribute, one wants to add the attribute to
853 * @ctx->mrec (the base mft record) and if there is not enough space, the
854 * attribute should be placed in a newly allocated extent mft record. The
855 * attribute list entry for the inserted attribute should be inserted in the
856 * attribute list attribute at @ctx->al_entry.
858 * On actual error, ntfs_external_attr_find() returns -EIO. In this case
859 * @ctx->attr is undefined and in particular do not rely on it not changing.
861 static int ntfs_external_attr_find(const ATTR_TYPE type,
862 const ntfschar *name, const u32 name_len,
863 const IGNORE_CASE_BOOL ic, const VCN lowest_vcn,
864 const u8 *val, const u32 val_len, ntfs_attr_search_ctx *ctx)
866 ntfs_inode *base_ni, *ni;
868 ATTR_LIST_ENTRY *al_entry, *next_al_entry;
869 u8 *al_start, *al_end;
874 static const char *es = " Unmount and run chkdsk.";
877 base_ni = ctx->base_ntfs_ino;
878 ntfs_debug("Entering for inode 0x%lx, type 0x%x.", ni->mft_no, type);
880 /* First call happens with the base mft record. */
881 base_ni = ctx->base_ntfs_ino = ctx->ntfs_ino;
882 ctx->base_mrec = ctx->mrec;
885 ctx->base_attr = ctx->attr;
889 al_start = base_ni->attr_list;
890 al_end = al_start + base_ni->attr_list_size;
892 ctx->al_entry = (ATTR_LIST_ENTRY*)al_start;
894 * Iterate over entries in attribute list starting at @ctx->al_entry,
895 * or the entry following that, if @ctx->is_first is 'true'.
898 al_entry = ctx->al_entry;
899 ctx->is_first = false;
901 al_entry = (ATTR_LIST_ENTRY*)((u8*)ctx->al_entry +
902 le16_to_cpu(ctx->al_entry->length));
903 for (;; al_entry = next_al_entry) {
904 /* Out of bounds check. */
905 if ((u8*)al_entry < base_ni->attr_list ||
906 (u8*)al_entry > al_end)
907 break; /* Inode is corrupt. */
908 ctx->al_entry = al_entry;
909 /* Catch the end of the attribute list. */
910 if ((u8*)al_entry == al_end)
912 if (!al_entry->length)
914 if ((u8*)al_entry + 6 > al_end || (u8*)al_entry +
915 le16_to_cpu(al_entry->length) > al_end)
917 next_al_entry = (ATTR_LIST_ENTRY*)((u8*)al_entry +
918 le16_to_cpu(al_entry->length));
919 if (le32_to_cpu(al_entry->type) > le32_to_cpu(type))
921 if (type != al_entry->type)
924 * If @name is present, compare the two names. If @name is
925 * missing, assume we want an unnamed attribute.
927 al_name_len = al_entry->name_length;
928 al_name = (ntfschar*)((u8*)al_entry + al_entry->name_offset);
932 } else if (!ntfs_are_names_equal(al_name, al_name_len, name,
933 name_len, ic, vol->upcase, vol->upcase_len)) {
936 rc = ntfs_collate_names(name, name_len, al_name,
937 al_name_len, 1, IGNORE_CASE,
938 vol->upcase, vol->upcase_len);
940 * If @name collates before al_name, there is no
941 * matching attribute.
945 /* If the strings are not equal, continue search. */
949 * FIXME: Reverse engineering showed 0, IGNORE_CASE but
950 * that is inconsistent with ntfs_attr_find(). The
951 * subsequent rc checks were also different. Perhaps I
952 * made a mistake in one of the two. Need to recheck
953 * which is correct or at least see what is going on...
956 rc = ntfs_collate_names(name, name_len, al_name,
957 al_name_len, 1, CASE_SENSITIVE,
958 vol->upcase, vol->upcase_len);
965 * The names match or @name not present and attribute is
966 * unnamed. Now check @lowest_vcn. Continue search if the
967 * next attribute list entry still fits @lowest_vcn. Otherwise
968 * we have reached the right one or the search has failed.
970 if (lowest_vcn && (u8*)next_al_entry >= al_start &&
971 (u8*)next_al_entry + 6 < al_end &&
972 (u8*)next_al_entry + le16_to_cpu(
973 next_al_entry->length) <= al_end &&
974 sle64_to_cpu(next_al_entry->lowest_vcn) <=
976 next_al_entry->type == al_entry->type &&
977 next_al_entry->name_length == al_name_len &&
978 ntfs_are_names_equal((ntfschar*)((u8*)
980 next_al_entry->name_offset),
981 next_al_entry->name_length,
982 al_name, al_name_len, CASE_SENSITIVE,
983 vol->upcase, vol->upcase_len))
985 if (MREF_LE(al_entry->mft_reference) == ni->mft_no) {
986 if (MSEQNO_LE(al_entry->mft_reference) != ni->seq_no) {
987 ntfs_error(vol->sb, "Found stale mft "
988 "reference in attribute list "
989 "of base inode 0x%lx.%s",
990 base_ni->mft_no, es);
994 } else { /* Mft references do not match. */
995 /* If there is a mapped record unmap it first. */
997 unmap_extent_mft_record(ni);
998 /* Do we want the base record back? */
999 if (MREF_LE(al_entry->mft_reference) ==
1001 ni = ctx->ntfs_ino = base_ni;
1002 ctx->mrec = ctx->base_mrec;
1004 /* We want an extent record. */
1005 ctx->mrec = map_extent_mft_record(base_ni,
1007 al_entry->mft_reference), &ni);
1008 if (IS_ERR(ctx->mrec)) {
1009 ntfs_error(vol->sb, "Failed to map "
1010 "extent mft record "
1011 "0x%lx of base inode "
1015 base_ni->mft_no, es);
1016 err = PTR_ERR(ctx->mrec);
1019 /* Cause @ctx to be sanitized below. */
1025 ctx->attr = (ATTR_RECORD*)((u8*)ctx->mrec +
1026 le16_to_cpu(ctx->mrec->attrs_offset));
1029 * ctx->vfs_ino, ctx->mrec, and ctx->attr now point to the
1030 * mft record containing the attribute represented by the
1034 * We could call into ntfs_attr_find() to find the right
1035 * attribute in this mft record but this would be less
1036 * efficient and not quite accurate as ntfs_attr_find() ignores
1037 * the attribute instance numbers for example which become
1038 * important when one plays with attribute lists. Also,
1039 * because a proper match has been found in the attribute list
1040 * entry above, the comparison can now be optimized. So it is
1041 * worth re-implementing a simplified ntfs_attr_find() here.
1045 * Use a manual loop so we can still use break and continue
1046 * with the same meanings as above.
1049 if ((u8*)a < (u8*)ctx->mrec || (u8*)a > (u8*)ctx->mrec +
1050 le32_to_cpu(ctx->mrec->bytes_allocated))
1052 if (a->type == AT_END)
1056 if (al_entry->instance != a->instance)
1059 * If the type and/or the name are mismatched between the
1060 * attribute list entry and the attribute record, there is
1061 * corruption so we break and return error EIO.
1063 if (al_entry->type != a->type)
1065 if (!ntfs_are_names_equal((ntfschar*)((u8*)a +
1066 le16_to_cpu(a->name_offset)), a->name_length,
1067 al_name, al_name_len, CASE_SENSITIVE,
1068 vol->upcase, vol->upcase_len))
1072 * If no @val specified or @val specified and it matches, we
1075 if (!val || (!a->non_resident && le32_to_cpu(
1076 a->data.resident.value_length) == val_len &&
1078 le16_to_cpu(a->data.resident.value_offset),
1080 ntfs_debug("Done, found.");
1084 /* Proceed to the next attribute in the current mft record. */
1085 a = (ATTR_RECORD*)((u8*)a + le32_to_cpu(a->length));
1086 goto do_next_attr_loop;
1089 ntfs_error(vol->sb, "Base inode 0x%lx contains corrupt "
1090 "attribute list attribute.%s", base_ni->mft_no,
1094 if (ni != base_ni) {
1096 unmap_extent_mft_record(ni);
1097 ctx->ntfs_ino = base_ni;
1098 ctx->mrec = ctx->base_mrec;
1099 ctx->attr = ctx->base_attr;
1106 * If we were looking for AT_END, we reset the search context @ctx and
1107 * use ntfs_attr_find() to seek to the end of the base mft record.
1109 if (type == AT_END) {
1110 ntfs_attr_reinit_search_ctx(ctx);
1111 return ntfs_attr_find(AT_END, name, name_len, ic, val, val_len,
1115 * The attribute was not found. Before we return, we want to ensure
1116 * @ctx->mrec and @ctx->attr indicate the position at which the
1117 * attribute should be inserted in the base mft record. Since we also
1118 * want to preserve @ctx->al_entry we cannot reinitialize the search
1119 * context using ntfs_attr_reinit_search_ctx() as this would set
1120 * @ctx->al_entry to NULL. Thus we do the necessary bits manually (see
1121 * ntfs_attr_init_search_ctx() below). Note, we _only_ preserve
1122 * @ctx->al_entry as the remaining fields (base_*) are identical to
1123 * their non base_ counterparts and we cannot set @ctx->base_attr
1124 * correctly yet as we do not know what @ctx->attr will be set to by
1125 * the call to ntfs_attr_find() below.
1128 unmap_extent_mft_record(ni);
1129 ctx->mrec = ctx->base_mrec;
1130 ctx->attr = (ATTR_RECORD*)((u8*)ctx->mrec +
1131 le16_to_cpu(ctx->mrec->attrs_offset));
1132 ctx->is_first = true;
1133 ctx->ntfs_ino = base_ni;
1134 ctx->base_ntfs_ino = NULL;
1135 ctx->base_mrec = NULL;
1136 ctx->base_attr = NULL;
1138 * In case there are multiple matches in the base mft record, need to
1139 * keep enumerating until we get an attribute not found response (or
1140 * another error), otherwise we would keep returning the same attribute
1141 * over and over again and all programs using us for enumeration would
1142 * lock up in a tight loop.
1145 err = ntfs_attr_find(type, name, name_len, ic, val, val_len,
1148 ntfs_debug("Done, not found.");
1153 * ntfs_attr_lookup - find an attribute in an ntfs inode
1154 * @type: attribute type to find
1155 * @name: attribute name to find (optional, i.e. NULL means don't care)
1156 * @name_len: attribute name length (only needed if @name present)
1157 * @ic: IGNORE_CASE or CASE_SENSITIVE (ignored if @name not present)
1158 * @lowest_vcn: lowest vcn to find (optional, non-resident attributes only)
1159 * @val: attribute value to find (optional, resident attributes only)
1160 * @val_len: attribute value length
1161 * @ctx: search context with mft record and attribute to search from
1163 * Find an attribute in an ntfs inode. On first search @ctx->ntfs_ino must
1164 * be the base mft record and @ctx must have been obtained from a call to
1165 * ntfs_attr_get_search_ctx().
1167 * This function transparently handles attribute lists and @ctx is used to
1168 * continue searches where they were left off at.
1170 * After finishing with the attribute/mft record you need to call
1171 * ntfs_attr_put_search_ctx() to cleanup the search context (unmapping any
1172 * mapped inodes, etc).
1174 * Return 0 if the search was successful and -errno if not.
1176 * When 0, @ctx->attr is the found attribute and it is in mft record
1177 * @ctx->mrec. If an attribute list attribute is present, @ctx->al_entry is
1178 * the attribute list entry of the found attribute.
1180 * When -ENOENT, @ctx->attr is the attribute which collates just after the
1181 * attribute being searched for, i.e. if one wants to add the attribute to the
1182 * mft record this is the correct place to insert it into. If an attribute
1183 * list attribute is present, @ctx->al_entry is the attribute list entry which
1184 * collates just after the attribute list entry of the attribute being searched
1185 * for, i.e. if one wants to add the attribute to the mft record this is the
1186 * correct place to insert its attribute list entry into.
1188 * When -errno != -ENOENT, an error occurred during the lookup. @ctx->attr is
1189 * then undefined and in particular you should not rely on it not changing.
1191 int ntfs_attr_lookup(const ATTR_TYPE type, const ntfschar *name,
1192 const u32 name_len, const IGNORE_CASE_BOOL ic,
1193 const VCN lowest_vcn, const u8 *val, const u32 val_len,
1194 ntfs_attr_search_ctx *ctx)
1196 ntfs_inode *base_ni;
1198 ntfs_debug("Entering.");
1199 BUG_ON(IS_ERR(ctx->mrec));
1200 if (ctx->base_ntfs_ino)
1201 base_ni = ctx->base_ntfs_ino;
1203 base_ni = ctx->ntfs_ino;
1204 /* Sanity check, just for debugging really. */
1206 if (!NInoAttrList(base_ni) || type == AT_ATTRIBUTE_LIST)
1207 return ntfs_attr_find(type, name, name_len, ic, val, val_len,
1209 return ntfs_external_attr_find(type, name, name_len, ic, lowest_vcn,
1214 * ntfs_attr_init_search_ctx - initialize an attribute search context
1215 * @ctx: attribute search context to initialize
1216 * @ni: ntfs inode with which to initialize the search context
1217 * @mrec: mft record with which to initialize the search context
1219 * Initialize the attribute search context @ctx with @ni and @mrec.
1221 static inline void ntfs_attr_init_search_ctx(ntfs_attr_search_ctx *ctx,
1222 ntfs_inode *ni, MFT_RECORD *mrec)
1224 *ctx = (ntfs_attr_search_ctx) {
1226 /* Sanity checks are performed elsewhere. */
1227 .attr = (ATTR_RECORD*)((u8*)mrec +
1228 le16_to_cpu(mrec->attrs_offset)),
1235 * ntfs_attr_reinit_search_ctx - reinitialize an attribute search context
1236 * @ctx: attribute search context to reinitialize
1238 * Reinitialize the attribute search context @ctx, unmapping an associated
1239 * extent mft record if present, and initialize the search context again.
1241 * This is used when a search for a new attribute is being started to reset
1242 * the search context to the beginning.
1244 void ntfs_attr_reinit_search_ctx(ntfs_attr_search_ctx *ctx)
1246 if (likely(!ctx->base_ntfs_ino)) {
1247 /* No attribute list. */
1248 ctx->is_first = true;
1249 /* Sanity checks are performed elsewhere. */
1250 ctx->attr = (ATTR_RECORD*)((u8*)ctx->mrec +
1251 le16_to_cpu(ctx->mrec->attrs_offset));
1253 * This needs resetting due to ntfs_external_attr_find() which
1254 * can leave it set despite having zeroed ctx->base_ntfs_ino.
1256 ctx->al_entry = NULL;
1258 } /* Attribute list. */
1259 if (ctx->ntfs_ino != ctx->base_ntfs_ino)
1260 unmap_extent_mft_record(ctx->ntfs_ino);
1261 ntfs_attr_init_search_ctx(ctx, ctx->base_ntfs_ino, ctx->base_mrec);
1266 * ntfs_attr_get_search_ctx - allocate/initialize a new attribute search context
1267 * @ni: ntfs inode with which to initialize the search context
1268 * @mrec: mft record with which to initialize the search context
1270 * Allocate a new attribute search context, initialize it with @ni and @mrec,
1271 * and return it. Return NULL if allocation failed.
1273 ntfs_attr_search_ctx *ntfs_attr_get_search_ctx(ntfs_inode *ni, MFT_RECORD *mrec)
1275 ntfs_attr_search_ctx *ctx;
1277 ctx = kmem_cache_alloc(ntfs_attr_ctx_cache, GFP_NOFS);
1279 ntfs_attr_init_search_ctx(ctx, ni, mrec);
1284 * ntfs_attr_put_search_ctx - release an attribute search context
1285 * @ctx: attribute search context to free
1287 * Release the attribute search context @ctx, unmapping an associated extent
1288 * mft record if present.
1290 void ntfs_attr_put_search_ctx(ntfs_attr_search_ctx *ctx)
1292 if (ctx->base_ntfs_ino && ctx->ntfs_ino != ctx->base_ntfs_ino)
1293 unmap_extent_mft_record(ctx->ntfs_ino);
1294 kmem_cache_free(ntfs_attr_ctx_cache, ctx);
1301 * ntfs_attr_find_in_attrdef - find an attribute in the $AttrDef system file
1302 * @vol: ntfs volume to which the attribute belongs
1303 * @type: attribute type which to find
1305 * Search for the attribute definition record corresponding to the attribute
1306 * @type in the $AttrDef system file.
1308 * Return the attribute type definition record if found and NULL if not found.
1310 static ATTR_DEF *ntfs_attr_find_in_attrdef(const ntfs_volume *vol,
1311 const ATTR_TYPE type)
1315 BUG_ON(!vol->attrdef);
1317 for (ad = vol->attrdef; (u8*)ad - (u8*)vol->attrdef <
1318 vol->attrdef_size && ad->type; ++ad) {
1319 /* We have not found it yet, carry on searching. */
1320 if (likely(le32_to_cpu(ad->type) < le32_to_cpu(type)))
1322 /* We found the attribute; return it. */
1323 if (likely(ad->type == type))
1325 /* We have gone too far already. No point in continuing. */
1328 /* Attribute not found. */
1329 ntfs_debug("Attribute type 0x%x not found in $AttrDef.",
1335 * ntfs_attr_size_bounds_check - check a size of an attribute type for validity
1336 * @vol: ntfs volume to which the attribute belongs
1337 * @type: attribute type which to check
1338 * @size: size which to check
1340 * Check whether the @size in bytes is valid for an attribute of @type on the
1341 * ntfs volume @vol. This information is obtained from $AttrDef system file.
1343 * Return 0 if valid, -ERANGE if not valid, or -ENOENT if the attribute is not
1344 * listed in $AttrDef.
1346 int ntfs_attr_size_bounds_check(const ntfs_volume *vol, const ATTR_TYPE type,
1353 * $ATTRIBUTE_LIST has a maximum size of 256kiB, but this is not
1354 * listed in $AttrDef.
1356 if (unlikely(type == AT_ATTRIBUTE_LIST && size > 256 * 1024))
1358 /* Get the $AttrDef entry for the attribute @type. */
1359 ad = ntfs_attr_find_in_attrdef(vol, type);
1362 /* Do the bounds check. */
1363 if (((sle64_to_cpu(ad->min_size) > 0) &&
1364 size < sle64_to_cpu(ad->min_size)) ||
1365 ((sle64_to_cpu(ad->max_size) > 0) && size >
1366 sle64_to_cpu(ad->max_size)))
1372 * ntfs_attr_can_be_non_resident - check if an attribute can be non-resident
1373 * @vol: ntfs volume to which the attribute belongs
1374 * @type: attribute type which to check
1376 * Check whether the attribute of @type on the ntfs volume @vol is allowed to
1377 * be non-resident. This information is obtained from $AttrDef system file.
1379 * Return 0 if the attribute is allowed to be non-resident, -EPERM if not, and
1380 * -ENOENT if the attribute is not listed in $AttrDef.
1382 int ntfs_attr_can_be_non_resident(const ntfs_volume *vol, const ATTR_TYPE type)
1386 /* Find the attribute definition record in $AttrDef. */
1387 ad = ntfs_attr_find_in_attrdef(vol, type);
1390 /* Check the flags and return the result. */
1391 if (ad->flags & ATTR_DEF_RESIDENT)
1397 * ntfs_attr_can_be_resident - check if an attribute can be resident
1398 * @vol: ntfs volume to which the attribute belongs
1399 * @type: attribute type which to check
1401 * Check whether the attribute of @type on the ntfs volume @vol is allowed to
1402 * be resident. This information is derived from our ntfs knowledge and may
1403 * not be completely accurate, especially when user defined attributes are
1404 * present. Basically we allow everything to be resident except for index
1405 * allocation and $EA attributes.
1407 * Return 0 if the attribute is allowed to be non-resident and -EPERM if not.
1409 * Warning: In the system file $MFT the attribute $Bitmap must be non-resident
1410 * otherwise windows will not boot (blue screen of death)! We cannot
1411 * check for this here as we do not know which inode's $Bitmap is
1412 * being asked about so the caller needs to special case this.
1414 int ntfs_attr_can_be_resident(const ntfs_volume *vol, const ATTR_TYPE type)
1416 if (type == AT_INDEX_ALLOCATION)
1422 * ntfs_attr_record_resize - resize an attribute record
1423 * @m: mft record containing attribute record
1424 * @a: attribute record to resize
1425 * @new_size: new size in bytes to which to resize the attribute record @a
1427 * Resize the attribute record @a, i.e. the resident part of the attribute, in
1428 * the mft record @m to @new_size bytes.
1430 * Return 0 on success and -errno on error. The following error codes are
1432 * -ENOSPC - Not enough space in the mft record @m to perform the resize.
1434 * Note: On error, no modifications have been performed whatsoever.
1436 * Warning: If you make a record smaller without having copied all the data you
1437 * are interested in the data may be overwritten.
1439 int ntfs_attr_record_resize(MFT_RECORD *m, ATTR_RECORD *a, u32 new_size)
1441 ntfs_debug("Entering for new_size %u.", new_size);
1442 /* Align to 8 bytes if it is not already done. */
1444 new_size = (new_size + 7) & ~7;
1445 /* If the actual attribute length has changed, move things around. */
1446 if (new_size != le32_to_cpu(a->length)) {
1447 u32 new_muse = le32_to_cpu(m->bytes_in_use) -
1448 le32_to_cpu(a->length) + new_size;
1449 /* Not enough space in this mft record. */
1450 if (new_muse > le32_to_cpu(m->bytes_allocated))
1452 /* Move attributes following @a to their new location. */
1453 memmove((u8*)a + new_size, (u8*)a + le32_to_cpu(a->length),
1454 le32_to_cpu(m->bytes_in_use) - ((u8*)a -
1455 (u8*)m) - le32_to_cpu(a->length));
1456 /* Adjust @m to reflect the change in used space. */
1457 m->bytes_in_use = cpu_to_le32(new_muse);
1458 /* Adjust @a to reflect the new size. */
1459 if (new_size >= offsetof(ATTR_REC, length) + sizeof(a->length))
1460 a->length = cpu_to_le32(new_size);
1466 * ntfs_resident_attr_value_resize - resize the value of a resident attribute
1467 * @m: mft record containing attribute record
1468 * @a: attribute record whose value to resize
1469 * @new_size: new size in bytes to which to resize the attribute value of @a
1471 * Resize the value of the attribute @a in the mft record @m to @new_size bytes.
1472 * If the value is made bigger, the newly allocated space is cleared.
1474 * Return 0 on success and -errno on error. The following error codes are
1476 * -ENOSPC - Not enough space in the mft record @m to perform the resize.
1478 * Note: On error, no modifications have been performed whatsoever.
1480 * Warning: If you make a record smaller without having copied all the data you
1481 * are interested in the data may be overwritten.
1483 int ntfs_resident_attr_value_resize(MFT_RECORD *m, ATTR_RECORD *a,
1488 /* Resize the resident part of the attribute record. */
1489 if (ntfs_attr_record_resize(m, a,
1490 le16_to_cpu(a->data.resident.value_offset) + new_size))
1493 * The resize succeeded! If we made the attribute value bigger, clear
1494 * the area between the old size and @new_size.
1496 old_size = le32_to_cpu(a->data.resident.value_length);
1497 if (new_size > old_size)
1498 memset((u8*)a + le16_to_cpu(a->data.resident.value_offset) +
1499 old_size, 0, new_size - old_size);
1500 /* Finally update the length of the attribute value. */
1501 a->data.resident.value_length = cpu_to_le32(new_size);
1506 * ntfs_attr_make_non_resident - convert a resident to a non-resident attribute
1507 * @ni: ntfs inode describing the attribute to convert
1508 * @data_size: size of the resident data to copy to the non-resident attribute
1510 * Convert the resident ntfs attribute described by the ntfs inode @ni to a
1513 * @data_size must be equal to the attribute value size. This is needed since
1514 * we need to know the size before we can map the mft record and our callers
1515 * always know it. The reason we cannot simply read the size from the vfs
1516 * inode i_size is that this is not necessarily uptodate. This happens when
1517 * ntfs_attr_make_non_resident() is called in the ->truncate call path(s).
1519 * Return 0 on success and -errno on error. The following error return codes
1521 * -EPERM - The attribute is not allowed to be non-resident.
1522 * -ENOMEM - Not enough memory.
1523 * -ENOSPC - Not enough disk space.
1524 * -EINVAL - Attribute not defined on the volume.
1525 * -EIO - I/o error or other error.
1526 * Note that -ENOSPC is also returned in the case that there is not enough
1527 * space in the mft record to do the conversion. This can happen when the mft
1528 * record is already very full. The caller is responsible for trying to make
1529 * space in the mft record and trying again. FIXME: Do we need a separate
1530 * error return code for this kind of -ENOSPC or is it always worth trying
1531 * again in case the attribute may then fit in a resident state so no need to
1532 * make it non-resident at all? Ho-hum... (AIA)
1534 * NOTE to self: No changes in the attribute list are required to move from
1535 * a resident to a non-resident attribute.
1537 * Locking: - The caller must hold i_mutex on the inode.
1539 int ntfs_attr_make_non_resident(ntfs_inode *ni, const u32 data_size)
1542 struct inode *vi = VFS_I(ni);
1543 ntfs_volume *vol = ni->vol;
1544 ntfs_inode *base_ni;
1547 ntfs_attr_search_ctx *ctx;
1549 runlist_element *rl;
1551 unsigned long flags;
1552 int mp_size, mp_ofs, name_ofs, arec_size, err, err2;
1554 u8 old_res_attr_flags;
1556 /* Check that the attribute is allowed to be non-resident. */
1557 err = ntfs_attr_can_be_non_resident(vol, ni->type);
1558 if (unlikely(err)) {
1560 ntfs_debug("Attribute is not allowed to be "
1563 ntfs_debug("Attribute not defined on the NTFS "
1568 * FIXME: Compressed and encrypted attributes are not supported when
1569 * writing and we should never have gotten here for them.
1571 BUG_ON(NInoCompressed(ni));
1572 BUG_ON(NInoEncrypted(ni));
1574 * The size needs to be aligned to a cluster boundary for allocation
1577 new_size = (data_size + vol->cluster_size - 1) &
1578 ~(vol->cluster_size - 1);
1581 * Will need the page later and since the page lock nests
1582 * outside all ntfs locks, we need to get the page now.
1584 page = find_or_create_page(vi->i_mapping, 0,
1585 mapping_gfp_mask(vi->i_mapping));
1586 if (unlikely(!page))
1588 /* Start by allocating clusters to hold the attribute value. */
1589 rl = ntfs_cluster_alloc(vol, 0, new_size >>
1590 vol->cluster_size_bits, -1, DATA_ZONE, true);
1593 ntfs_debug("Failed to allocate cluster%s, error code "
1595 vol->cluster_size_bits) > 1 ? "s" : "",
1603 /* Determine the size of the mapping pairs array. */
1604 mp_size = ntfs_get_size_for_mapping_pairs(vol, rl, 0, -1);
1605 if (unlikely(mp_size < 0)) {
1607 ntfs_debug("Failed to get size for mapping pairs array, error "
1611 down_write(&ni->runlist.lock);
1615 base_ni = ni->ext.base_ntfs_ino;
1616 m = map_mft_record(base_ni);
1623 ctx = ntfs_attr_get_search_ctx(base_ni, m);
1624 if (unlikely(!ctx)) {
1628 err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
1629 CASE_SENSITIVE, 0, NULL, 0, ctx);
1630 if (unlikely(err)) {
1637 BUG_ON(NInoNonResident(ni));
1638 BUG_ON(a->non_resident);
1640 * Calculate new offsets for the name and the mapping pairs array.
1642 if (NInoSparse(ni) || NInoCompressed(ni))
1643 name_ofs = (offsetof(ATTR_REC,
1644 data.non_resident.compressed_size) +
1645 sizeof(a->data.non_resident.compressed_size) +
1648 name_ofs = (offsetof(ATTR_REC,
1649 data.non_resident.compressed_size) + 7) & ~7;
1650 mp_ofs = (name_ofs + a->name_length * sizeof(ntfschar) + 7) & ~7;
1652 * Determine the size of the resident part of the now non-resident
1655 arec_size = (mp_ofs + mp_size + 7) & ~7;
1657 * If the page is not uptodate bring it uptodate by copying from the
1660 attr_size = le32_to_cpu(a->data.resident.value_length);
1661 BUG_ON(attr_size != data_size);
1662 if (page && !PageUptodate(page)) {
1663 kaddr = kmap_atomic(page);
1664 memcpy(kaddr, (u8*)a +
1665 le16_to_cpu(a->data.resident.value_offset),
1667 memset(kaddr + attr_size, 0, PAGE_SIZE - attr_size);
1668 kunmap_atomic(kaddr);
1669 flush_dcache_page(page);
1670 SetPageUptodate(page);
1672 /* Backup the attribute flag. */
1673 old_res_attr_flags = a->data.resident.flags;
1674 /* Resize the resident part of the attribute record. */
1675 err = ntfs_attr_record_resize(m, a, arec_size);
1679 * Convert the resident part of the attribute record to describe a
1680 * non-resident attribute.
1682 a->non_resident = 1;
1683 /* Move the attribute name if it exists and update the offset. */
1685 memmove((u8*)a + name_ofs, (u8*)a + le16_to_cpu(a->name_offset),
1686 a->name_length * sizeof(ntfschar));
1687 a->name_offset = cpu_to_le16(name_ofs);
1688 /* Setup the fields specific to non-resident attributes. */
1689 a->data.non_resident.lowest_vcn = 0;
1690 a->data.non_resident.highest_vcn = cpu_to_sle64((new_size - 1) >>
1691 vol->cluster_size_bits);
1692 a->data.non_resident.mapping_pairs_offset = cpu_to_le16(mp_ofs);
1693 memset(&a->data.non_resident.reserved, 0,
1694 sizeof(a->data.non_resident.reserved));
1695 a->data.non_resident.allocated_size = cpu_to_sle64(new_size);
1696 a->data.non_resident.data_size =
1697 a->data.non_resident.initialized_size =
1698 cpu_to_sle64(attr_size);
1699 if (NInoSparse(ni) || NInoCompressed(ni)) {
1700 a->data.non_resident.compression_unit = 0;
1701 if (NInoCompressed(ni) || vol->major_ver < 3)
1702 a->data.non_resident.compression_unit = 4;
1703 a->data.non_resident.compressed_size =
1704 a->data.non_resident.allocated_size;
1706 a->data.non_resident.compression_unit = 0;
1707 /* Generate the mapping pairs array into the attribute record. */
1708 err = ntfs_mapping_pairs_build(vol, (u8*)a + mp_ofs,
1709 arec_size - mp_ofs, rl, 0, -1, NULL);
1710 if (unlikely(err)) {
1711 ntfs_debug("Failed to build mapping pairs, error code %i.",
1715 /* Setup the in-memory attribute structure to be non-resident. */
1716 ni->runlist.rl = rl;
1717 write_lock_irqsave(&ni->size_lock, flags);
1718 ni->allocated_size = new_size;
1719 if (NInoSparse(ni) || NInoCompressed(ni)) {
1720 ni->itype.compressed.size = ni->allocated_size;
1721 if (a->data.non_resident.compression_unit) {
1722 ni->itype.compressed.block_size = 1U << (a->data.
1723 non_resident.compression_unit +
1724 vol->cluster_size_bits);
1725 ni->itype.compressed.block_size_bits =
1726 ffs(ni->itype.compressed.block_size) -
1728 ni->itype.compressed.block_clusters = 1U <<
1729 a->data.non_resident.compression_unit;
1731 ni->itype.compressed.block_size = 0;
1732 ni->itype.compressed.block_size_bits = 0;
1733 ni->itype.compressed.block_clusters = 0;
1735 vi->i_blocks = ni->itype.compressed.size >> 9;
1737 vi->i_blocks = ni->allocated_size >> 9;
1738 write_unlock_irqrestore(&ni->size_lock, flags);
1740 * This needs to be last since the address space operations ->readpage
1741 * and ->writepage can run concurrently with us as they are not
1742 * serialized on i_mutex. Note, we are not allowed to fail once we flip
1743 * this switch, which is another reason to do this last.
1745 NInoSetNonResident(ni);
1746 /* Mark the mft record dirty, so it gets written back. */
1747 flush_dcache_mft_record_page(ctx->ntfs_ino);
1748 mark_mft_record_dirty(ctx->ntfs_ino);
1749 ntfs_attr_put_search_ctx(ctx);
1750 unmap_mft_record(base_ni);
1751 up_write(&ni->runlist.lock);
1753 set_page_dirty(page);
1757 ntfs_debug("Done.");
1760 /* Convert the attribute back into a resident attribute. */
1761 a->non_resident = 0;
1762 /* Move the attribute name if it exists and update the offset. */
1763 name_ofs = (offsetof(ATTR_RECORD, data.resident.reserved) +
1764 sizeof(a->data.resident.reserved) + 7) & ~7;
1766 memmove((u8*)a + name_ofs, (u8*)a + le16_to_cpu(a->name_offset),
1767 a->name_length * sizeof(ntfschar));
1768 mp_ofs = (name_ofs + a->name_length * sizeof(ntfschar) + 7) & ~7;
1769 a->name_offset = cpu_to_le16(name_ofs);
1770 arec_size = (mp_ofs + attr_size + 7) & ~7;
1771 /* Resize the resident part of the attribute record. */
1772 err2 = ntfs_attr_record_resize(m, a, arec_size);
1773 if (unlikely(err2)) {
1775 * This cannot happen (well if memory corruption is at work it
1776 * could happen in theory), but deal with it as well as we can.
1777 * If the old size is too small, truncate the attribute,
1778 * otherwise simply give it a larger allocated size.
1779 * FIXME: Should check whether chkdsk complains when the
1780 * allocated size is much bigger than the resident value size.
1782 arec_size = le32_to_cpu(a->length);
1783 if ((mp_ofs + attr_size) > arec_size) {
1785 attr_size = arec_size - mp_ofs;
1786 ntfs_error(vol->sb, "Failed to undo partial resident "
1787 "to non-resident attribute "
1788 "conversion. Truncating inode 0x%lx, "
1789 "attribute type 0x%x from %i bytes to "
1790 "%i bytes to maintain metadata "
1791 "consistency. THIS MEANS YOU ARE "
1792 "LOSING %i BYTES DATA FROM THIS %s.",
1794 (unsigned)le32_to_cpu(ni->type),
1795 err2, attr_size, err2 - attr_size,
1796 ((ni->type == AT_DATA) &&
1797 !ni->name_len) ? "FILE": "ATTRIBUTE");
1798 write_lock_irqsave(&ni->size_lock, flags);
1799 ni->initialized_size = attr_size;
1800 i_size_write(vi, attr_size);
1801 write_unlock_irqrestore(&ni->size_lock, flags);
1804 /* Setup the fields specific to resident attributes. */
1805 a->data.resident.value_length = cpu_to_le32(attr_size);
1806 a->data.resident.value_offset = cpu_to_le16(mp_ofs);
1807 a->data.resident.flags = old_res_attr_flags;
1808 memset(&a->data.resident.reserved, 0,
1809 sizeof(a->data.resident.reserved));
1810 /* Copy the data from the page back to the attribute value. */
1812 kaddr = kmap_atomic(page);
1813 memcpy((u8*)a + mp_ofs, kaddr, attr_size);
1814 kunmap_atomic(kaddr);
1816 /* Setup the allocated size in the ntfs inode in case it changed. */
1817 write_lock_irqsave(&ni->size_lock, flags);
1818 ni->allocated_size = arec_size - mp_ofs;
1819 write_unlock_irqrestore(&ni->size_lock, flags);
1820 /* Mark the mft record dirty, so it gets written back. */
1821 flush_dcache_mft_record_page(ctx->ntfs_ino);
1822 mark_mft_record_dirty(ctx->ntfs_ino);
1825 ntfs_attr_put_search_ctx(ctx);
1827 unmap_mft_record(base_ni);
1828 ni->runlist.rl = NULL;
1829 up_write(&ni->runlist.lock);
1832 if (ntfs_cluster_free_from_rl(vol, rl) < 0) {
1833 ntfs_error(vol->sb, "Failed to release allocated "
1834 "cluster(s) in error code path. Run "
1835 "chkdsk to recover the lost "
1850 * ntfs_attr_extend_allocation - extend the allocated space of an attribute
1851 * @ni: ntfs inode of the attribute whose allocation to extend
1852 * @new_alloc_size: new size in bytes to which to extend the allocation to
1853 * @new_data_size: new size in bytes to which to extend the data to
1854 * @data_start: beginning of region which is required to be non-sparse
1856 * Extend the allocated space of an attribute described by the ntfs inode @ni
1857 * to @new_alloc_size bytes. If @data_start is -1, the whole extension may be
1858 * implemented as a hole in the file (as long as both the volume and the ntfs
1859 * inode @ni have sparse support enabled). If @data_start is >= 0, then the
1860 * region between the old allocated size and @data_start - 1 may be made sparse
1861 * but the regions between @data_start and @new_alloc_size must be backed by
1864 * If @new_data_size is -1, it is ignored. If it is >= 0, then the data size
1865 * of the attribute is extended to @new_data_size. Note that the i_size of the
1866 * vfs inode is not updated. Only the data size in the base attribute record
1867 * is updated. The caller has to update i_size separately if this is required.
1868 * WARNING: It is a BUG() for @new_data_size to be smaller than the old data
1869 * size as well as for @new_data_size to be greater than @new_alloc_size.
1871 * For resident attributes this involves resizing the attribute record and if
1872 * necessary moving it and/or other attributes into extent mft records and/or
1873 * converting the attribute to a non-resident attribute which in turn involves
1874 * extending the allocation of a non-resident attribute as described below.
1876 * For non-resident attributes this involves allocating clusters in the data
1877 * zone on the volume (except for regions that are being made sparse) and
1878 * extending the run list to describe the allocated clusters as well as
1879 * updating the mapping pairs array of the attribute. This in turn involves
1880 * resizing the attribute record and if necessary moving it and/or other
1881 * attributes into extent mft records and/or splitting the attribute record
1882 * into multiple extent attribute records.
1884 * Also, the attribute list attribute is updated if present and in some of the
1885 * above cases (the ones where extent mft records/attributes come into play),
1886 * an attribute list attribute is created if not already present.
1888 * Return the new allocated size on success and -errno on error. In the case
1889 * that an error is encountered but a partial extension at least up to
1890 * @data_start (if present) is possible, the allocation is partially extended
1891 * and this is returned. This means the caller must check the returned size to
1892 * determine if the extension was partial. If @data_start is -1 then partial
1893 * allocations are not performed.
1895 * WARNING: Do not call ntfs_attr_extend_allocation() for $MFT/$DATA.
1897 * Locking: This function takes the runlist lock of @ni for writing as well as
1898 * locking the mft record of the base ntfs inode. These locks are maintained
1899 * throughout execution of the function. These locks are required so that the
1900 * attribute can be resized safely and so that it can for example be converted
1901 * from resident to non-resident safely.
1903 * TODO: At present attribute list attribute handling is not implemented.
1905 * TODO: At present it is not safe to call this function for anything other
1906 * than the $DATA attribute(s) of an uncompressed and unencrypted file.
1908 s64 ntfs_attr_extend_allocation(ntfs_inode *ni, s64 new_alloc_size,
1909 const s64 new_data_size, const s64 data_start)
1912 s64 ll, allocated_size, start = data_start;
1913 struct inode *vi = VFS_I(ni);
1914 ntfs_volume *vol = ni->vol;
1915 ntfs_inode *base_ni;
1918 ntfs_attr_search_ctx *ctx;
1919 runlist_element *rl, *rl2;
1920 unsigned long flags;
1922 u32 attr_len = 0; /* Silence stupid gcc warning. */
1926 read_lock_irqsave(&ni->size_lock, flags);
1927 allocated_size = ni->allocated_size;
1928 read_unlock_irqrestore(&ni->size_lock, flags);
1929 ntfs_debug("Entering for i_ino 0x%lx, attribute type 0x%x, "
1930 "old_allocated_size 0x%llx, "
1931 "new_allocated_size 0x%llx, new_data_size 0x%llx, "
1932 "data_start 0x%llx.", vi->i_ino,
1933 (unsigned)le32_to_cpu(ni->type),
1934 (unsigned long long)allocated_size,
1935 (unsigned long long)new_alloc_size,
1936 (unsigned long long)new_data_size,
1937 (unsigned long long)start);
1941 * For non-resident attributes, @start and @new_size need to be aligned
1942 * to cluster boundaries for allocation purposes.
1944 if (NInoNonResident(ni)) {
1946 start &= ~(s64)vol->cluster_size_mask;
1947 new_alloc_size = (new_alloc_size + vol->cluster_size - 1) &
1948 ~(s64)vol->cluster_size_mask;
1950 BUG_ON(new_data_size >= 0 && new_data_size > new_alloc_size);
1951 /* Check if new size is allowed in $AttrDef. */
1952 err = ntfs_attr_size_bounds_check(vol, ni->type, new_alloc_size);
1953 if (unlikely(err)) {
1954 /* Only emit errors when the write will fail completely. */
1955 read_lock_irqsave(&ni->size_lock, flags);
1956 allocated_size = ni->allocated_size;
1957 read_unlock_irqrestore(&ni->size_lock, flags);
1958 if (start < 0 || start >= allocated_size) {
1959 if (err == -ERANGE) {
1960 ntfs_error(vol->sb, "Cannot extend allocation "
1961 "of inode 0x%lx, attribute "
1962 "type 0x%x, because the new "
1963 "allocation would exceed the "
1964 "maximum allowed size for "
1965 "this attribute type.",
1966 vi->i_ino, (unsigned)
1967 le32_to_cpu(ni->type));
1969 ntfs_error(vol->sb, "Cannot extend allocation "
1970 "of inode 0x%lx, attribute "
1971 "type 0x%x, because this "
1972 "attribute type is not "
1973 "defined on the NTFS volume. "
1974 "Possible corruption! You "
1975 "should run chkdsk!",
1976 vi->i_ino, (unsigned)
1977 le32_to_cpu(ni->type));
1980 /* Translate error code to be POSIX conformant for write(2). */
1990 base_ni = ni->ext.base_ntfs_ino;
1992 * We will be modifying both the runlist (if non-resident) and the mft
1993 * record so lock them both down.
1995 down_write(&ni->runlist.lock);
1996 m = map_mft_record(base_ni);
2003 ctx = ntfs_attr_get_search_ctx(base_ni, m);
2004 if (unlikely(!ctx)) {
2008 read_lock_irqsave(&ni->size_lock, flags);
2009 allocated_size = ni->allocated_size;
2010 read_unlock_irqrestore(&ni->size_lock, flags);
2012 * If non-resident, seek to the last extent. If resident, there is
2013 * only one extent, so seek to that.
2015 vcn = NInoNonResident(ni) ? allocated_size >> vol->cluster_size_bits :
2018 * Abort if someone did the work whilst we waited for the locks. If we
2019 * just converted the attribute from resident to non-resident it is
2020 * likely that exactly this has happened already. We cannot quite
2021 * abort if we need to update the data size.
2023 if (unlikely(new_alloc_size <= allocated_size)) {
2024 ntfs_debug("Allocated size already exceeds requested size.");
2025 new_alloc_size = allocated_size;
2026 if (new_data_size < 0)
2029 * We want the first attribute extent so that we can update the
2034 err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
2035 CASE_SENSITIVE, vcn, NULL, 0, ctx);
2036 if (unlikely(err)) {
2043 /* Use goto to reduce indentation. */
2044 if (a->non_resident)
2045 goto do_non_resident_extend;
2046 BUG_ON(NInoNonResident(ni));
2047 /* The total length of the attribute value. */
2048 attr_len = le32_to_cpu(a->data.resident.value_length);
2050 * Extend the attribute record to be able to store the new attribute
2051 * size. ntfs_attr_record_resize() will not do anything if the size is
2054 if (new_alloc_size < vol->mft_record_size &&
2055 !ntfs_attr_record_resize(m, a,
2056 le16_to_cpu(a->data.resident.value_offset) +
2058 /* The resize succeeded! */
2059 write_lock_irqsave(&ni->size_lock, flags);
2060 ni->allocated_size = le32_to_cpu(a->length) -
2061 le16_to_cpu(a->data.resident.value_offset);
2062 write_unlock_irqrestore(&ni->size_lock, flags);
2063 if (new_data_size >= 0) {
2064 BUG_ON(new_data_size < attr_len);
2065 a->data.resident.value_length =
2066 cpu_to_le32((u32)new_data_size);
2071 * We have to drop all the locks so we can call
2072 * ntfs_attr_make_non_resident(). This could be optimised by try-
2073 * locking the first page cache page and only if that fails dropping
2074 * the locks, locking the page, and redoing all the locking and
2075 * lookups. While this would be a huge optimisation, it is not worth
2076 * it as this is definitely a slow code path.
2078 ntfs_attr_put_search_ctx(ctx);
2079 unmap_mft_record(base_ni);
2080 up_write(&ni->runlist.lock);
2082 * Not enough space in the mft record, try to make the attribute
2083 * non-resident and if successful restart the extension process.
2085 err = ntfs_attr_make_non_resident(ni, attr_len);
2089 * Could not make non-resident. If this is due to this not being
2090 * permitted for this attribute type or there not being enough space,
2091 * try to make other attributes non-resident. Otherwise fail.
2093 if (unlikely(err != -EPERM && err != -ENOSPC)) {
2094 /* Only emit errors when the write will fail completely. */
2095 read_lock_irqsave(&ni->size_lock, flags);
2096 allocated_size = ni->allocated_size;
2097 read_unlock_irqrestore(&ni->size_lock, flags);
2098 if (start < 0 || start >= allocated_size)
2099 ntfs_error(vol->sb, "Cannot extend allocation of "
2100 "inode 0x%lx, attribute type 0x%x, "
2101 "because the conversion from resident "
2102 "to non-resident attribute failed "
2103 "with error code %i.", vi->i_ino,
2104 (unsigned)le32_to_cpu(ni->type), err);
2109 /* TODO: Not implemented from here, abort. */
2110 read_lock_irqsave(&ni->size_lock, flags);
2111 allocated_size = ni->allocated_size;
2112 read_unlock_irqrestore(&ni->size_lock, flags);
2113 if (start < 0 || start >= allocated_size) {
2115 ntfs_error(vol->sb, "Not enough space in the mft "
2116 "record/on disk for the non-resident "
2117 "attribute value. This case is not "
2118 "implemented yet.");
2119 else /* if (err == -EPERM) */
2120 ntfs_error(vol->sb, "This attribute type may not be "
2121 "non-resident. This case is not "
2122 "implemented yet.");
2127 // TODO: Attempt to make other attributes non-resident.
2129 goto do_resident_extend;
2131 * Both the attribute list attribute and the standard information
2132 * attribute must remain in the base inode. Thus, if this is one of
2133 * these attributes, we have to try to move other attributes out into
2134 * extent mft records instead.
2136 if (ni->type == AT_ATTRIBUTE_LIST ||
2137 ni->type == AT_STANDARD_INFORMATION) {
2138 // TODO: Attempt to move other attributes into extent mft
2142 goto do_resident_extend;
2145 // TODO: Attempt to move this attribute to an extent mft record, but
2146 // only if it is not already the only attribute in an mft record in
2147 // which case there would be nothing to gain.
2150 goto do_resident_extend;
2151 /* There is nothing we can do to make enough space. )-: */
2154 do_non_resident_extend:
2155 BUG_ON(!NInoNonResident(ni));
2156 if (new_alloc_size == allocated_size) {
2161 * If the data starts after the end of the old allocation, this is a
2162 * $DATA attribute and sparse attributes are enabled on the volume and
2163 * for this inode, then create a sparse region between the old
2164 * allocated size and the start of the data. Otherwise simply proceed
2165 * with filling the whole space between the old allocated size and the
2166 * new allocated size with clusters.
2168 if ((start >= 0 && start <= allocated_size) || ni->type != AT_DATA ||
2169 !NVolSparseEnabled(vol) || NInoSparseDisabled(ni))
2171 // TODO: This is not implemented yet. We just fill in with real
2172 // clusters for now...
2173 ntfs_debug("Inserting holes is not-implemented yet. Falling back to "
2174 "allocating real clusters instead.");
2176 rl = ni->runlist.rl;
2178 /* Seek to the end of the runlist. */
2182 /* If this attribute extent is not mapped, map it now. */
2183 if (unlikely(!rl || rl->lcn == LCN_RL_NOT_MAPPED ||
2184 (rl->lcn == LCN_ENOENT && rl > ni->runlist.rl &&
2185 (rl-1)->lcn == LCN_RL_NOT_MAPPED))) {
2186 if (!rl && !allocated_size)
2188 rl = ntfs_mapping_pairs_decompress(vol, a, ni->runlist.rl);
2191 if (start < 0 || start >= allocated_size)
2192 ntfs_error(vol->sb, "Cannot extend allocation "
2193 "of inode 0x%lx, attribute "
2194 "type 0x%x, because the "
2195 "mapping of a runlist "
2196 "fragment failed with error "
2197 "code %i.", vi->i_ino,
2198 (unsigned)le32_to_cpu(ni->type),
2204 ni->runlist.rl = rl;
2205 /* Seek to the end of the runlist. */
2210 * We now know the runlist of the last extent is mapped and @rl is at
2211 * the end of the runlist. We want to begin allocating clusters
2212 * starting at the last allocated cluster to reduce fragmentation. If
2213 * there are no valid LCNs in the attribute we let the cluster
2214 * allocator choose the starting cluster.
2216 /* If the last LCN is a hole or simillar seek back to last real LCN. */
2217 while (rl->lcn < 0 && rl > ni->runlist.rl)
2220 // FIXME: Need to implement partial allocations so at least part of the
2221 // write can be performed when start >= 0. (Needed for POSIX write(2)
2223 rl2 = ntfs_cluster_alloc(vol, allocated_size >> vol->cluster_size_bits,
2224 (new_alloc_size - allocated_size) >>
2225 vol->cluster_size_bits, (rl && (rl->lcn >= 0)) ?
2226 rl->lcn + rl->length : -1, DATA_ZONE, true);
2229 if (start < 0 || start >= allocated_size)
2230 ntfs_error(vol->sb, "Cannot extend allocation of "
2231 "inode 0x%lx, attribute type 0x%x, "
2232 "because the allocation of clusters "
2233 "failed with error code %i.", vi->i_ino,
2234 (unsigned)le32_to_cpu(ni->type), err);
2235 if (err != -ENOMEM && err != -ENOSPC)
2239 rl = ntfs_runlists_merge(ni->runlist.rl, rl2);
2242 if (start < 0 || start >= allocated_size)
2243 ntfs_error(vol->sb, "Cannot extend allocation of "
2244 "inode 0x%lx, attribute type 0x%x, "
2245 "because the runlist merge failed "
2246 "with error code %i.", vi->i_ino,
2247 (unsigned)le32_to_cpu(ni->type), err);
2250 if (ntfs_cluster_free_from_rl(vol, rl2)) {
2251 ntfs_error(vol->sb, "Failed to release allocated "
2252 "cluster(s) in error code path. Run "
2253 "chkdsk to recover the lost "
2260 ni->runlist.rl = rl;
2261 ntfs_debug("Allocated 0x%llx clusters.", (long long)(new_alloc_size -
2262 allocated_size) >> vol->cluster_size_bits);
2263 /* Find the runlist element with which the attribute extent starts. */
2264 ll = sle64_to_cpu(a->data.non_resident.lowest_vcn);
2265 rl2 = ntfs_rl_find_vcn_nolock(rl, ll);
2267 BUG_ON(!rl2->length);
2268 BUG_ON(rl2->lcn < LCN_HOLE);
2270 /* Get the size for the new mapping pairs array for this extent. */
2271 mp_size = ntfs_get_size_for_mapping_pairs(vol, rl2, ll, -1);
2272 if (unlikely(mp_size <= 0)) {
2274 if (start < 0 || start >= allocated_size)
2275 ntfs_error(vol->sb, "Cannot extend allocation of "
2276 "inode 0x%lx, attribute type 0x%x, "
2277 "because determining the size for the "
2278 "mapping pairs failed with error code "
2280 (unsigned)le32_to_cpu(ni->type), err);
2284 /* Extend the attribute record to fit the bigger mapping pairs array. */
2285 attr_len = le32_to_cpu(a->length);
2286 err = ntfs_attr_record_resize(m, a, mp_size +
2287 le16_to_cpu(a->data.non_resident.mapping_pairs_offset));
2288 if (unlikely(err)) {
2289 BUG_ON(err != -ENOSPC);
2290 // TODO: Deal with this by moving this extent to a new mft
2291 // record or by starting a new extent in a new mft record,
2292 // possibly by extending this extent partially and filling it
2293 // and creating a new extent for the remainder, or by making
2294 // other attributes non-resident and/or by moving other
2295 // attributes out of this mft record.
2296 if (start < 0 || start >= allocated_size)
2297 ntfs_error(vol->sb, "Not enough space in the mft "
2298 "record for the extended attribute "
2299 "record. This case is not "
2300 "implemented yet.");
2305 /* Generate the mapping pairs array directly into the attr record. */
2306 err = ntfs_mapping_pairs_build(vol, (u8*)a +
2307 le16_to_cpu(a->data.non_resident.mapping_pairs_offset),
2308 mp_size, rl2, ll, -1, NULL);
2309 if (unlikely(err)) {
2310 if (start < 0 || start >= allocated_size)
2311 ntfs_error(vol->sb, "Cannot extend allocation of "
2312 "inode 0x%lx, attribute type 0x%x, "
2313 "because building the mapping pairs "
2314 "failed with error code %i.", vi->i_ino,
2315 (unsigned)le32_to_cpu(ni->type), err);
2319 /* Update the highest_vcn. */
2320 a->data.non_resident.highest_vcn = cpu_to_sle64((new_alloc_size >>
2321 vol->cluster_size_bits) - 1);
2323 * We now have extended the allocated size of the attribute. Reflect
2324 * this in the ntfs_inode structure and the attribute record.
2326 if (a->data.non_resident.lowest_vcn) {
2328 * We are not in the first attribute extent, switch to it, but
2329 * first ensure the changes will make it to disk later.
2331 flush_dcache_mft_record_page(ctx->ntfs_ino);
2332 mark_mft_record_dirty(ctx->ntfs_ino);
2333 ntfs_attr_reinit_search_ctx(ctx);
2334 err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
2335 CASE_SENSITIVE, 0, NULL, 0, ctx);
2337 goto restore_undo_alloc;
2338 /* @m is not used any more so no need to set it. */
2341 write_lock_irqsave(&ni->size_lock, flags);
2342 ni->allocated_size = new_alloc_size;
2343 a->data.non_resident.allocated_size = cpu_to_sle64(new_alloc_size);
2345 * FIXME: This would fail if @ni is a directory, $MFT, or an index,
2346 * since those can have sparse/compressed set. For example can be
2347 * set compressed even though it is not compressed itself and in that
2348 * case the bit means that files are to be created compressed in the
2349 * directory... At present this is ok as this code is only called for
2350 * regular files, and only for their $DATA attribute(s).
2351 * FIXME: The calculation is wrong if we created a hole above. For now
2352 * it does not matter as we never create holes.
2354 if (NInoSparse(ni) || NInoCompressed(ni)) {
2355 ni->itype.compressed.size += new_alloc_size - allocated_size;
2356 a->data.non_resident.compressed_size =
2357 cpu_to_sle64(ni->itype.compressed.size);
2358 vi->i_blocks = ni->itype.compressed.size >> 9;
2360 vi->i_blocks = new_alloc_size >> 9;
2361 write_unlock_irqrestore(&ni->size_lock, flags);
2363 if (new_data_size >= 0) {
2364 BUG_ON(new_data_size <
2365 sle64_to_cpu(a->data.non_resident.data_size));
2366 a->data.non_resident.data_size = cpu_to_sle64(new_data_size);
2369 /* Ensure the changes make it to disk. */
2370 flush_dcache_mft_record_page(ctx->ntfs_ino);
2371 mark_mft_record_dirty(ctx->ntfs_ino);
2373 ntfs_attr_put_search_ctx(ctx);
2374 unmap_mft_record(base_ni);
2375 up_write(&ni->runlist.lock);
2376 ntfs_debug("Done, new_allocated_size 0x%llx.",
2377 (unsigned long long)new_alloc_size);
2378 return new_alloc_size;
2380 if (start < 0 || start >= allocated_size)
2381 ntfs_error(vol->sb, "Cannot complete extension of allocation "
2382 "of inode 0x%lx, attribute type 0x%x, because "
2383 "lookup of first attribute extent failed with "
2384 "error code %i.", vi->i_ino,
2385 (unsigned)le32_to_cpu(ni->type), err);
2388 ntfs_attr_reinit_search_ctx(ctx);
2389 if (ntfs_attr_lookup(ni->type, ni->name, ni->name_len, CASE_SENSITIVE,
2390 allocated_size >> vol->cluster_size_bits, NULL, 0,
2392 ntfs_error(vol->sb, "Failed to find last attribute extent of "
2393 "attribute in error code path. Run chkdsk to "
2395 write_lock_irqsave(&ni->size_lock, flags);
2396 ni->allocated_size = new_alloc_size;
2398 * FIXME: This would fail if @ni is a directory... See above.
2399 * FIXME: The calculation is wrong if we created a hole above.
2400 * For now it does not matter as we never create holes.
2402 if (NInoSparse(ni) || NInoCompressed(ni)) {
2403 ni->itype.compressed.size += new_alloc_size -
2405 vi->i_blocks = ni->itype.compressed.size >> 9;
2407 vi->i_blocks = new_alloc_size >> 9;
2408 write_unlock_irqrestore(&ni->size_lock, flags);
2409 ntfs_attr_put_search_ctx(ctx);
2410 unmap_mft_record(base_ni);
2411 up_write(&ni->runlist.lock);
2413 * The only thing that is now wrong is the allocated size of the
2414 * base attribute extent which chkdsk should be able to fix.
2419 ctx->attr->data.non_resident.highest_vcn = cpu_to_sle64(
2420 (allocated_size >> vol->cluster_size_bits) - 1);
2422 ll = allocated_size >> vol->cluster_size_bits;
2423 if (ntfs_cluster_free(ni, ll, -1, ctx) < 0) {
2424 ntfs_error(vol->sb, "Failed to release allocated cluster(s) "
2425 "in error code path. Run chkdsk to recover "
2426 "the lost cluster(s).");
2432 * If the runlist truncation fails and/or the search context is no
2433 * longer valid, we cannot resize the attribute record or build the
2434 * mapping pairs array thus we mark the inode bad so that no access to
2435 * the freed clusters can happen.
2437 if (ntfs_rl_truncate_nolock(vol, &ni->runlist, ll) || IS_ERR(m)) {
2438 ntfs_error(vol->sb, "Failed to %s in error code path. Run "
2439 "chkdsk to recover.", IS_ERR(m) ?
2440 "restore attribute search context" :
2441 "truncate attribute runlist");
2443 } else if (mp_rebuilt) {
2444 if (ntfs_attr_record_resize(m, a, attr_len)) {
2445 ntfs_error(vol->sb, "Failed to restore attribute "
2446 "record in error code path. Run "
2447 "chkdsk to recover.");
2449 } else /* if (success) */ {
2450 if (ntfs_mapping_pairs_build(vol, (u8*)a + le16_to_cpu(
2451 a->data.non_resident.
2452 mapping_pairs_offset), attr_len -
2453 le16_to_cpu(a->data.non_resident.
2454 mapping_pairs_offset), rl2, ll, -1,
2456 ntfs_error(vol->sb, "Failed to restore "
2457 "mapping pairs array in error "
2458 "code path. Run chkdsk to "
2462 flush_dcache_mft_record_page(ctx->ntfs_ino);
2463 mark_mft_record_dirty(ctx->ntfs_ino);
2468 ntfs_attr_put_search_ctx(ctx);
2470 unmap_mft_record(base_ni);
2471 up_write(&ni->runlist.lock);
2473 ntfs_debug("Failed. Returning error code %i.", err);
2478 * ntfs_attr_set - fill (a part of) an attribute with a byte
2479 * @ni: ntfs inode describing the attribute to fill
2480 * @ofs: offset inside the attribute at which to start to fill
2481 * @cnt: number of bytes to fill
2482 * @val: the unsigned 8-bit value with which to fill the attribute
2484 * Fill @cnt bytes of the attribute described by the ntfs inode @ni starting at
2485 * byte offset @ofs inside the attribute with the constant byte @val.
2487 * This function is effectively like memset() applied to an ntfs attribute.
2488 * Note thie function actually only operates on the page cache pages belonging
2489 * to the ntfs attribute and it marks them dirty after doing the memset().
2490 * Thus it relies on the vm dirty page write code paths to cause the modified
2491 * pages to be written to the mft record/disk.
2493 * Return 0 on success and -errno on error. An error code of -ESPIPE means
2494 * that @ofs + @cnt were outside the end of the attribute and no write was
2497 int ntfs_attr_set(ntfs_inode *ni, const s64 ofs, const s64 cnt, const u8 val)
2499 ntfs_volume *vol = ni->vol;
2500 struct address_space *mapping;
2504 unsigned start_ofs, end_ofs, size;
2506 ntfs_debug("Entering for ofs 0x%llx, cnt 0x%llx, val 0x%hx.",
2507 (long long)ofs, (long long)cnt, val);
2513 * FIXME: Compressed and encrypted attributes are not supported when
2514 * writing and we should never have gotten here for them.
2516 BUG_ON(NInoCompressed(ni));
2517 BUG_ON(NInoEncrypted(ni));
2518 mapping = VFS_I(ni)->i_mapping;
2519 /* Work out the starting index and page offset. */
2520 idx = ofs >> PAGE_SHIFT;
2521 start_ofs = ofs & ~PAGE_MASK;
2522 /* Work out the ending index and page offset. */
2524 end_ofs = end & ~PAGE_MASK;
2525 /* If the end is outside the inode size return -ESPIPE. */
2526 if (unlikely(end > i_size_read(VFS_I(ni)))) {
2527 ntfs_error(vol->sb, "Request exceeds end of attribute.");
2531 /* If there is a first partial page, need to do it the slow way. */
2533 page = read_mapping_page(mapping, idx, NULL);
2535 ntfs_error(vol->sb, "Failed to read first partial "
2536 "page (error, index 0x%lx).", idx);
2537 return PTR_ERR(page);
2540 * If the last page is the same as the first page, need to
2541 * limit the write to the end offset.
2546 kaddr = kmap_atomic(page);
2547 memset(kaddr + start_ofs, val, size - start_ofs);
2548 flush_dcache_page(page);
2549 kunmap_atomic(kaddr);
2550 set_page_dirty(page);
2552 balance_dirty_pages_ratelimited(mapping);
2558 /* Do the whole pages the fast way. */
2559 for (; idx < end; idx++) {
2560 /* Find or create the current page. (The page is locked.) */
2561 page = grab_cache_page(mapping, idx);
2562 if (unlikely(!page)) {
2563 ntfs_error(vol->sb, "Insufficient memory to grab "
2564 "page (index 0x%lx).", idx);
2567 kaddr = kmap_atomic(page);
2568 memset(kaddr, val, PAGE_SIZE);
2569 flush_dcache_page(page);
2570 kunmap_atomic(kaddr);
2572 * If the page has buffers, mark them uptodate since buffer
2573 * state and not page state is definitive in 2.6 kernels.
2575 if (page_has_buffers(page)) {
2576 struct buffer_head *bh, *head;
2578 bh = head = page_buffers(page);
2580 set_buffer_uptodate(bh);
2581 } while ((bh = bh->b_this_page) != head);
2583 /* Now that buffers are uptodate, set the page uptodate, too. */
2584 SetPageUptodate(page);
2586 * Set the page and all its buffers dirty and mark the inode
2587 * dirty, too. The VM will write the page later on.
2589 set_page_dirty(page);
2590 /* Finally unlock and release the page. */
2593 balance_dirty_pages_ratelimited(mapping);
2596 /* If there is a last partial page, need to do it the slow way. */
2598 page = read_mapping_page(mapping, idx, NULL);
2600 ntfs_error(vol->sb, "Failed to read last partial page "
2601 "(error, index 0x%lx).", idx);
2602 return PTR_ERR(page);
2604 kaddr = kmap_atomic(page);
2605 memset(kaddr, val, end_ofs);
2606 flush_dcache_page(page);
2607 kunmap_atomic(kaddr);
2608 set_page_dirty(page);
2610 balance_dirty_pages_ratelimited(mapping);
2614 ntfs_debug("Done.");
2618 #endif /* NTFS_RW */