1 /* hivex - Windows Registry "hive" extraction library.
2 * Copyright (C) 2009 Red Hat Inc.
3 * Derived from code by Petter Nordahl-Hagen under a compatible license:
4 * Copyright (c) 1997-2007 Petter Nordahl-Hagen.
5 * Derived from code by Markus Stephany under a compatible license:
6 * Copyright (c) 2000-2004, Markus Stephany.
8 * This library is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation;
11 * version 2.1 of the License.
13 * This library is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
18 * See file LICENSE for the full license.
39 #ifdef HAVE_BYTESWAP_H
43 #define STREQ(a,b) (strcmp((a),(b)) == 0)
44 #define STRCASEEQ(a,b) (strcasecmp((a),(b)) == 0)
45 //#define STRNEQ(a,b) (strcmp((a),(b)) != 0)
46 //#define STRCASENEQ(a,b) (strcasecmp((a),(b)) != 0)
47 #define STREQLEN(a,b,n) (strncmp((a),(b),(n)) == 0)
48 //#define STRCASEEQLEN(a,b,n) (strncasecmp((a),(b),(n)) == 0)
49 //#define STRNEQLEN(a,b,n) (strncmp((a),(b),(n)) != 0)
50 //#define STRCASENEQLEN(a,b,n) (strncasecmp((a),(b),(n)) != 0)
51 //#define STRPREFIX(a,b) (strncmp((a),(b),strlen((b))) == 0)
53 #if __BYTE_ORDER == __LITTLE_ENDIAN
55 #define be32toh(x) __bswap_32 (x)
58 #define be64toh(x) __bswap_64 (x)
61 #define le16toh(x) (x)
64 #define le32toh(x) (x)
67 #define le64toh(x) (x)
71 #define be32toh(x) (x)
74 #define be64toh(x) (x)
77 #define le16toh(x) __bswap_16 (x)
80 #define le32toh(x) __bswap_32 (x)
83 #define le64toh(x) __bswap_64 (x)
89 static char *windows_utf16_to_utf8 (/* const */ char *input, size_t len);
97 /* Memory-mapped (readonly) registry file. */
100 struct ntreg_header *hdr;
103 /* Use a bitmap to store which file offsets are valid (point to a
104 * used block). We only need to store 1 bit per 32 bits of the file
105 * (because blocks are 4-byte aligned). We found that the average
106 * block size in a registry file is ~50 bytes. So roughly 1 in 12
107 * bits in the bitmap will be set, making it likely a more efficient
108 * structure than a hash table.
111 #define BITMAP_SET(bitmap,off) (bitmap[(off)>>5] |= 1 << (((off)>>2)&7))
112 #define BITMAP_CLR(bitmap,off) (bitmap[(off)>>5] &= ~ (1 << (((off)>>2)&7)))
113 #define BITMAP_TST(bitmap,off) (bitmap[(off)>>5] & (1 << (((off)>>2)&7)))
114 #define IS_VALID_BLOCK(h,off) \
115 (((off) & 3) == 0 && \
117 (off) < (h)->size && \
118 BITMAP_TST((h)->bitmap,(off)))
120 /* Fields from the header, extracted from little-endianness hell. */
121 size_t rootoffs; /* Root key offset (always an nk-block). */
122 size_t endpages; /* Offset of end of pages. */
125 /* NB. All fields are little endian. */
126 struct ntreg_header {
127 char magic[4]; /* "regf" */
130 char last_modified[8];
131 uint32_t major_ver; /* 1 */
132 uint32_t minor_ver; /* 3 */
133 uint32_t unknown5; /* 0 */
134 uint32_t unknown6; /* 1 */
135 uint32_t offset; /* offset of root key record - 4KB */
136 uint32_t blocks; /* pointer AFTER last hbin in file - 4KB */
137 uint32_t unknown7; /* 1 */
139 char name[64]; /* original file name of hive */
140 char unknown_guid1[16];
141 char unknown_guid2[16];
144 char unknown_guid3[16];
149 uint32_t csum; /* checksum: xor of dwords 0-0x1fb. */
151 char unknown11[3528];
153 char unknown_guid4[16];
154 char unknown_guid5[16];
155 char unknown_guid6[16];
159 } __attribute__((__packed__));
161 struct ntreg_hbin_page {
162 char magic[4]; /* "hbin" */
163 uint32_t offset_first; /* offset from 1st block */
164 uint32_t page_size; /* size of this page (multiple of 4KB) */
166 /* Linked list of blocks follows here. */
167 } __attribute__((__packed__));
169 struct ntreg_hbin_block {
170 int32_t seg_len; /* length of this block (-ve for used block) */
171 char id[2]; /* the block type (eg. "nk" for nk record) */
172 /* Block data follows here. */
173 } __attribute__((__packed__));
175 #define BLOCK_ID_EQ(h,offs,eqid) \
176 (STREQLEN (((struct ntreg_hbin_block *)((h)->addr + (offs)))->id, (eqid), 2))
179 block_len (hive_h *h, size_t blkoff, int *used)
181 struct ntreg_hbin_block *block;
182 block = (struct ntreg_hbin_block *) (h->addr + blkoff);
184 int32_t len = le32toh (block->seg_len);
195 struct ntreg_nk_record {
196 int32_t seg_len; /* length (always -ve because used) */
197 char id[2]; /* "nk" */
200 uint32_t parent; /* offset of owner/parent */
201 uint32_t nr_subkeys; /* number of subkeys */
203 uint32_t subkey_lf; /* lf record containing list of subkeys */
205 uint32_t nr_values; /* number of values */
206 uint32_t vallist; /* value-list record */
207 uint32_t sk; /* offset of sk-record */
208 uint32_t classname; /* offset of classname record */
211 uint16_t name_len; /* length of name */
212 uint16_t classname_len; /* length of classname */
213 char name[1]; /* name follows here */
214 } __attribute__((__packed__));
216 struct ntreg_lf_record {
218 char id[2]; /* "lf" */
219 uint16_t nr_keys; /* number of keys in this record */
221 uint32_t offset; /* offset of nk-record for this subkey */
222 char name[4]; /* first 4 characters of subkey name */
224 } __attribute__((__packed__));
226 struct ntreg_ri_record {
228 char id[2]; /* "ri" */
229 uint16_t nr_offsets; /* number of pointers to lh records */
230 uint32_t offset[1]; /* list of pointers to lh records */
231 } __attribute__((__packed__));
233 /* This has no ID header. */
234 struct ntreg_value_list {
236 uint32_t offset[1]; /* list of pointers to vk records */
237 } __attribute__((__packed__));
239 struct ntreg_vk_record {
240 int32_t seg_len; /* length (always -ve because used) */
241 char id[2]; /* "vk" */
242 uint16_t name_len; /* length of name */
243 /* length of the data:
244 * If data_len is <= 4, then it's stored inline.
245 * If data_len is 0x80000000, then it's an inline dword.
246 * Top bit may be set or not set at random.
249 uint32_t data_offset; /* pointer to the data (or data if inline) */
250 hive_type data_type; /* type of the data */
251 uint16_t flags; /* bit 0 set => key name ASCII,
252 bit 0 clr => key name UTF-16.
253 Only seen ASCII here in the wild. */
255 char name[1]; /* key name follows here */
256 } __attribute__((__packed__));
259 header_checksum (hive_h *h)
261 uint32_t *daddr = (uint32_t *) h->addr;
265 for (i = 0; i < 0x1fc / 4; ++i) {
266 sum ^= le32toh (*daddr);
274 hivex_open (const char *filename, int flags)
278 assert (sizeof (struct ntreg_header) == 0x1000);
279 assert (offsetof (struct ntreg_header, csum) == 0x1fc);
281 h = calloc (1, sizeof *h);
285 h->msglvl = flags & HIVEX_OPEN_MSGLVL_MASK;
287 const char *debug = getenv ("HIVEX_DEBUG");
288 if (debug && STREQ (debug, "1"))
292 fprintf (stderr, "hivex_open: created handle %p\n", h);
294 h->filename = strdup (filename);
295 if (h->filename == NULL)
298 h->fd = open (filename, O_RDONLY);
303 if (fstat (h->fd, &statbuf) == -1)
306 h->size = statbuf.st_size;
308 h->addr = mmap (NULL, h->size, PROT_READ, MAP_SHARED, h->fd, 0);
309 if (h->addr == MAP_FAILED)
313 fprintf (stderr, "hivex_open: mapped file at %p\n", h->addr);
316 if (h->hdr->magic[0] != 'r' ||
317 h->hdr->magic[1] != 'e' ||
318 h->hdr->magic[2] != 'g' ||
319 h->hdr->magic[3] != 'f') {
320 fprintf (stderr, "hivex: %s: not a Windows NT Registry hive file\n",
326 /* Check major version. */
327 uint32_t major_ver = le32toh (h->hdr->major_ver);
328 if (major_ver != 1) {
330 "hivex: %s: hive file major version %" PRIu32 " (expected 1)\n",
331 filename, major_ver);
336 h->bitmap = calloc (1 + h->size / 32, 1);
337 if (h->bitmap == NULL)
340 /* Header checksum. */
341 uint32_t sum = header_checksum (h);
342 if (sum != le32toh (h->hdr->csum)) {
343 fprintf (stderr, "hivex: %s: bad checksum in hive header\n", filename);
348 if (h->msglvl >= 2) {
349 char *name = windows_utf16_to_utf8 (h->hdr->name, 64);
352 "hivex_open: header fields:\n"
353 " file version %" PRIu32 ".%" PRIu32 "\n"
354 " sequence nos %" PRIu32 " %" PRIu32 "\n"
355 " (sequences nos should match if hive was synched at shutdown)\n"
356 " original file name %s\n"
357 " (only 32 chars are stored, name is probably truncated)\n"
358 " root offset 0x%x + 0x1000\n"
359 " end of last page 0x%x + 0x1000 (total file size 0x%zx)\n"
360 " checksum 0x%x (calculated 0x%x)\n",
361 major_ver, le32toh (h->hdr->minor_ver),
362 le32toh (h->hdr->sequence1), le32toh (h->hdr->sequence2),
363 name ? name : "(conversion failed)",
364 le32toh (h->hdr->offset),
365 le32toh (h->hdr->blocks), h->size,
366 le32toh (h->hdr->csum), sum);
370 h->rootoffs = le32toh (h->hdr->offset) + 0x1000;
371 h->endpages = le32toh (h->hdr->blocks) + 0x1000;
374 fprintf (stderr, "hivex_open: root offset = 0x%zx\n", h->rootoffs);
376 /* We'll set this flag when we see a block with the root offset (ie.
379 int seen_root_block = 0, bad_root_block = 0;
381 /* Collect some stats. */
382 size_t pages = 0; /* Number of hbin pages read. */
383 size_t smallest_page = SIZE_MAX, largest_page = 0;
384 size_t blocks = 0; /* Total number of blocks found. */
385 size_t smallest_block = SIZE_MAX, largest_block = 0, blocks_bytes = 0;
386 size_t used_blocks = 0; /* Total number of used blocks found. */
387 size_t used_size = 0; /* Total size (bytes) of used blocks. */
389 /* Read the pages and blocks. The aim here is to be robust against
390 * corrupt or malicious registries. So we make sure the loops
391 * always make forward progress. We add the address of each block
392 * we read to a hash table so pointers will only reference the start
396 struct ntreg_hbin_page *page;
397 for (off = 0x1000; off < h->size; off += le32toh (page->page_size)) {
398 if (off >= h->endpages)
401 page = (struct ntreg_hbin_page *) (h->addr + off);
402 if (page->magic[0] != 'h' ||
403 page->magic[1] != 'b' ||
404 page->magic[2] != 'i' ||
405 page->magic[3] != 'n') {
406 fprintf (stderr, "hivex: %s: trailing garbage at end of file (at 0x%zx, after %zu pages)\n",
407 filename, off, pages);
412 size_t page_size = le32toh (page->page_size);
414 fprintf (stderr, "hivex_open: page at 0x%zx, size %zu\n", off, page_size);
416 if (page_size < smallest_page) smallest_page = page_size;
417 if (page_size > largest_page) largest_page = page_size;
419 if (page_size <= sizeof (struct ntreg_hbin_page) ||
420 (page_size & 0x0fff) != 0) {
421 fprintf (stderr, "hivex: %s: page size %zu at 0x%zx, bad registry\n",
422 filename, page_size, off);
427 /* Read the blocks in this page. */
429 struct ntreg_hbin_block *block;
431 for (blkoff = off + 0x20;
432 blkoff < off + page_size;
436 int is_root = blkoff == h->rootoffs;
440 block = (struct ntreg_hbin_block *) (h->addr + blkoff);
442 seg_len = block_len (h, blkoff, &used);
443 if (seg_len <= 4 || (seg_len & 3) != 0) {
444 fprintf (stderr, "hivex: %s: block size %d at %zu, bad registry\n",
445 filename, le32toh (block->seg_len), blkoff);
451 fprintf (stderr, "hivex_open: %s block id %d,%d at 0x%zx size %zu%s\n",
452 used ? "used" : "free", block->id[0], block->id[1], blkoff,
453 seg_len, is_root ? " (root)" : "");
455 blocks_bytes += seg_len;
456 if (seg_len < smallest_block) smallest_block = seg_len;
457 if (seg_len > largest_block) largest_block = seg_len;
459 if (is_root && !used)
464 used_size += seg_len;
466 /* Root block must be an nk-block. */
467 if (is_root && (block->id[0] != 'n' || block->id[1] != 'k'))
470 /* Note this blkoff is a valid address. */
471 BITMAP_SET (h->bitmap, blkoff);
476 if (!seen_root_block) {
477 fprintf (stderr, "hivex: %s: no root block found\n", filename);
482 if (bad_root_block) {
483 fprintf (stderr, "hivex: %s: bad root block (free or not nk)\n", filename);
490 "hivex_open: successfully read Windows Registry hive file:\n"
491 " pages: %zu [sml: %zu, lge: %zu]\n"
492 " blocks: %zu [sml: %zu, avg: %zu, lge: %zu]\n"
493 " blocks used: %zu\n"
494 " bytes used: %zu\n",
495 pages, smallest_page, largest_page,
496 blocks, smallest_block, blocks_bytes / blocks, largest_block,
497 used_blocks, used_size);
505 if (h->addr && h->size && h->addr != MAP_FAILED)
506 munmap (h->addr, h->size);
517 hivex_close (hive_h *h)
522 munmap (h->addr, h->size);
531 hivex_root (hive_h *h)
533 hive_node_h ret = h->rootoffs;
534 if (!IS_VALID_BLOCK (h, ret)) {
542 hivex_node_name (hive_h *h, hive_node_h node)
544 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
549 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
551 /* AFAIK the node name is always plain ASCII, so no conversion
552 * to UTF-8 is necessary. However we do need to nul-terminate
556 /* nk->name_len is unsigned, 16 bit, so this is safe ... However
557 * we have to make sure the length doesn't exceed the block length.
559 size_t len = le16toh (nk->name_len);
560 size_t seg_len = block_len (h, node, NULL);
561 if (sizeof (struct ntreg_nk_record) + len - 1 > seg_len) {
563 fprintf (stderr, "hivex_node_name: returning EFAULT because node name is too long (%zu, %zu)\n",
569 char *ret = malloc (len + 1);
572 memcpy (ret, nk->name, len);
578 /* I think the documentation for the sk and classname fields in the nk
579 * record is wrong, or else the offset field is in the wrong place.
580 * Otherwise this makes no sense. Disabled this for now -- it's not
581 * useful for reading the registry anyway.
585 hivex_node_security (hive_h *h, hive_node_h node)
587 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
592 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
594 hive_node_h ret = le32toh (nk->sk);
596 if (!IS_VALID_BLOCK (h, ret)) {
604 hivex_node_classname (hive_h *h, hive_node_h node)
606 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
611 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
613 hive_node_h ret = le32toh (nk->classname);
615 if (!IS_VALID_BLOCK (h, ret)) {
623 /* Structure for returning 0-terminated lists of offsets (nodes,
633 init_offset_list (struct offset_list *list)
637 list->offsets = NULL;
640 #define INIT_OFFSET_LIST(name) \
641 struct offset_list name; \
642 init_offset_list (&name)
644 /* Preallocates the offset_list, but doesn't make the contents longer. */
646 grow_offset_list (struct offset_list *list, size_t alloc)
648 assert (alloc >= list->len);
649 size_t *p = realloc (list->offsets, alloc * sizeof (size_t));
658 add_to_offset_list (struct offset_list *list, size_t offset)
660 if (list->len >= list->alloc) {
661 if (grow_offset_list (list, list->alloc ? list->alloc * 2 : 4) == -1)
664 list->offsets[list->len] = offset;
670 free_offset_list (struct offset_list *list)
672 free (list->offsets);
676 return_offset_list (struct offset_list *list)
678 if (add_to_offset_list (list, 0) == -1)
680 return list->offsets; /* caller frees */
683 /* Iterate over children, returning child nodes and intermediate blocks. */
685 get_children (hive_h *h, hive_node_h node,
686 hive_node_h **children_ret, size_t **blocks_ret)
688 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
693 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
695 size_t nr_subkeys_in_nk = le32toh (nk->nr_subkeys);
697 INIT_OFFSET_LIST (children);
698 INIT_OFFSET_LIST (blocks);
700 /* Deal with the common "no subkeys" case quickly. */
701 if (nr_subkeys_in_nk == 0)
704 /* Arbitrarily limit the number of subkeys we will ever deal with. */
705 if (nr_subkeys_in_nk > 1000000) {
710 /* Preallocate space for the children. */
711 if (grow_offset_list (&children, nr_subkeys_in_nk) == -1)
714 /* The subkey_lf field can point either to an lf-record, which is
715 * the common case, or if there are lots of subkeys, to an
718 size_t subkey_lf = le32toh (nk->subkey_lf);
720 if (!IS_VALID_BLOCK (h, subkey_lf)) {
722 fprintf (stderr, "hivex_node_children: returning EFAULT because subkey_lf is not a valid block (%zu)\n",
728 if (add_to_offset_list (&blocks, subkey_lf) == -1)
731 struct ntreg_hbin_block *block =
732 (struct ntreg_hbin_block *) (h->addr + subkey_lf);
734 /* Points to lf-record? (Note, also "lh" but that is basically the
735 * same as "lf" as far as we are concerned here).
737 if (block->id[0] == 'l' && (block->id[1] == 'f' || block->id[1] == 'h')) {
738 struct ntreg_lf_record *lf = (struct ntreg_lf_record *) block;
740 /* Check number of subkeys in the nk-record matches number of subkeys
743 size_t nr_subkeys_in_lf = le16toh (lf->nr_keys);
746 fprintf (stderr, "hivex_node_children: nr_subkeys_in_nk = %zu, nr_subkeys_in_lf = %zu\n",
747 nr_subkeys_in_nk, nr_subkeys_in_lf);
749 if (nr_subkeys_in_nk != nr_subkeys_in_lf) {
754 size_t len = block_len (h, subkey_lf, NULL);
755 if (8 + nr_subkeys_in_lf * 8 > len) {
757 fprintf (stderr, "hivex_node_children: returning EFAULT because too many subkeys (%zu, %zu)\n",
758 nr_subkeys_in_lf, len);
764 for (i = 0; i < nr_subkeys_in_lf; ++i) {
765 hive_node_h subkey = lf->keys[i].offset;
767 if (!IS_VALID_BLOCK (h, subkey)) {
769 fprintf (stderr, "hivex_node_children: returning EFAULT because subkey is not a valid block (0x%zx)\n",
774 if (add_to_offset_list (&children, subkey) == -1)
779 /* Points to ri-record? */
780 else if (block->id[0] == 'r' && block->id[1] == 'i') {
781 struct ntreg_ri_record *ri = (struct ntreg_ri_record *) block;
783 size_t nr_offsets = le16toh (ri->nr_offsets);
785 /* Count total number of children. */
787 for (i = 0; i < nr_offsets; ++i) {
788 hive_node_h offset = ri->offset[i];
790 if (!IS_VALID_BLOCK (h, offset)) {
792 fprintf (stderr, "hivex_node_children: returning EFAULT because ri-offset is not a valid block (0x%zx)\n",
797 if (!BLOCK_ID_EQ (h, offset, "lf") && !BLOCK_ID_EQ (h, offset, "lh")) {
802 if (add_to_offset_list (&blocks, offset) == -1)
805 struct ntreg_lf_record *lf =
806 (struct ntreg_lf_record *) (h->addr + offset);
808 count += le16toh (lf->nr_keys);
812 fprintf (stderr, "hivex_node_children: nr_subkeys_in_nk = %zu, counted = %zu\n",
813 nr_subkeys_in_nk, count);
815 if (nr_subkeys_in_nk != count) {
820 /* Copy list of children. Note nr_subkeys_in_nk is limited to
821 * something reasonable above.
823 for (i = 0; i < nr_offsets; ++i) {
824 hive_node_h offset = ri->offset[i];
826 if (!IS_VALID_BLOCK (h, offset)) {
828 fprintf (stderr, "hivex_node_children: returning EFAULT because ri-offset is not a valid block (0x%zx)\n",
833 if (!BLOCK_ID_EQ (h, offset, "lf") && !BLOCK_ID_EQ (h, offset, "lh")) {
838 struct ntreg_lf_record *lf =
839 (struct ntreg_lf_record *) (h->addr + offset);
842 for (j = 0; j < le16toh (lf->nr_keys); ++j) {
843 hive_node_h subkey = lf->keys[j].offset;
845 if (!IS_VALID_BLOCK (h, subkey)) {
847 fprintf (stderr, "hivex_node_children: returning EFAULT because indirect subkey is not a valid block (0x%zx)\n",
852 if (add_to_offset_list (&children, subkey) == -1)
858 /* else not supported, set errno and fall through */
861 free_offset_list (&children);
862 free_offset_list (&blocks);
866 *children_ret = return_offset_list (&children);
867 *blocks_ret = return_offset_list (&blocks);
868 if (!*children_ret || !*blocks_ret)
874 hivex_node_children (hive_h *h, hive_node_h node)
876 hive_node_h *children;
879 if (get_children (h, node, &children, &blocks) == -1)
886 /* Very inefficient, but at least having a separate API call
887 * allows us to make it more efficient in future.
890 hivex_node_get_child (hive_h *h, hive_node_h node, const char *nname)
892 hive_node_h *children = NULL;
896 children = hivex_node_children (h, node);
897 if (!children) goto error;
900 for (i = 0; children[i] != 0; ++i) {
901 name = hivex_node_name (h, children[i]);
902 if (!name) goto error;
903 if (STRCASEEQ (name, nname)) {
907 free (name); name = NULL;
917 hivex_node_parent (hive_h *h, hive_node_h node)
919 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
924 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
926 hive_node_h ret = le32toh (nk->parent);
928 if (!IS_VALID_BLOCK (h, ret)) {
930 fprintf (stderr, "hivex_node_parent: returning EFAULT because parent is not a valid block (0x%zx)\n",
939 get_values (hive_h *h, hive_node_h node,
940 hive_value_h **values_ret, size_t **blocks_ret)
942 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
947 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
949 size_t nr_values = le32toh (nk->nr_values);
952 fprintf (stderr, "hivex_node_values: nr_values = %zu\n", nr_values);
954 INIT_OFFSET_LIST (values);
955 INIT_OFFSET_LIST (blocks);
957 /* Deal with the common "no values" case quickly. */
961 /* Arbitrarily limit the number of values we will ever deal with. */
962 if (nr_values > 100000) {
967 /* Preallocate space for the values. */
968 if (grow_offset_list (&values, nr_values) == -1)
971 /* Get the value list and check it looks reasonable. */
972 size_t vlist_offset = le32toh (nk->vallist);
973 vlist_offset += 0x1000;
974 if (!IS_VALID_BLOCK (h, vlist_offset)) {
976 fprintf (stderr, "hivex_node_values: returning EFAULT because value list is not a valid block (0x%zx)\n",
982 if (add_to_offset_list (&blocks, vlist_offset) == -1)
985 struct ntreg_value_list *vlist =
986 (struct ntreg_value_list *) (h->addr + vlist_offset);
988 size_t len = block_len (h, vlist_offset, NULL);
989 if (4 + nr_values * 4 > len) {
991 fprintf (stderr, "hivex_node_values: returning EFAULT because value list is too long (%zu, %zu)\n",
998 for (i = 0; i < nr_values; ++i) {
999 hive_node_h value = vlist->offset[i];
1001 if (!IS_VALID_BLOCK (h, value)) {
1003 fprintf (stderr, "hivex_node_values: returning EFAULT because value is not a valid block (0x%zx)\n",
1008 if (add_to_offset_list (&values, value) == -1)
1013 *values_ret = return_offset_list (&values);
1014 *blocks_ret = return_offset_list (&blocks);
1015 if (!*values_ret || !*blocks_ret)
1020 free_offset_list (&values);
1021 free_offset_list (&blocks);
1026 hivex_node_values (hive_h *h, hive_node_h node)
1028 hive_value_h *values;
1031 if (get_values (h, node, &values, &blocks) == -1)
1038 /* Very inefficient, but at least having a separate API call
1039 * allows us to make it more efficient in future.
1042 hivex_node_get_value (hive_h *h, hive_node_h node, const char *key)
1044 hive_value_h *values = NULL;
1046 hive_value_h ret = 0;
1048 values = hivex_node_values (h, node);
1049 if (!values) goto error;
1052 for (i = 0; values[i] != 0; ++i) {
1053 name = hivex_value_key (h, values[i]);
1054 if (!name) goto error;
1055 if (STRCASEEQ (name, key)) {
1059 free (name); name = NULL;
1069 hivex_value_key (hive_h *h, hive_value_h value)
1071 if (!IS_VALID_BLOCK (h, value) || !BLOCK_ID_EQ (h, value, "vk")) {
1076 struct ntreg_vk_record *vk = (struct ntreg_vk_record *) (h->addr + value);
1078 /* AFAIK the key is always plain ASCII, so no conversion to UTF-8 is
1079 * necessary. However we do need to nul-terminate the string.
1082 /* vk->name_len is unsigned, 16 bit, so this is safe ... However
1083 * we have to make sure the length doesn't exceed the block length.
1085 size_t len = le16toh (vk->name_len);
1086 size_t seg_len = block_len (h, value, NULL);
1087 if (sizeof (struct ntreg_vk_record) + len - 1 > seg_len) {
1089 fprintf (stderr, "hivex_value_key: returning EFAULT because key length is too long (%zu, %zu)\n",
1095 char *ret = malloc (len + 1);
1098 memcpy (ret, vk->name, len);
1104 hivex_value_type (hive_h *h, hive_value_h value, hive_type *t, size_t *len)
1106 if (!IS_VALID_BLOCK (h, value) || !BLOCK_ID_EQ (h, value, "vk")) {
1111 struct ntreg_vk_record *vk = (struct ntreg_vk_record *) (h->addr + value);
1114 *t = le32toh (vk->data_type);
1117 *len = le32toh (vk->data_len);
1118 if (*len == 0x80000000) { /* special case */
1120 if (t) *t = hive_t_dword;
1129 hivex_value_value (hive_h *h, hive_value_h value,
1130 hive_type *t_rtn, size_t *len_rtn)
1132 if (!IS_VALID_BLOCK (h, value) || !BLOCK_ID_EQ (h, value, "vk")) {
1137 struct ntreg_vk_record *vk = (struct ntreg_vk_record *) (h->addr + value);
1142 t = le32toh (vk->data_type);
1144 len = le32toh (vk->data_len);
1145 if (len == 0x80000000) { /* special case */
1152 fprintf (stderr, "hivex_value_value: value=0x%zx, t=%d, len=%zu\n",
1160 /* Arbitrarily limit the length that we will read. */
1161 if (len > 1000000) {
1166 char *ret = malloc (len);
1170 /* If length is <= 4 it's always stored inline. */
1172 memcpy (ret, (char *) &vk->data_offset, len);
1176 size_t data_offset = vk->data_offset;
1177 data_offset += 0x1000;
1178 if (!IS_VALID_BLOCK (h, data_offset)) {
1180 fprintf (stderr, "hivex_value_value: returning EFAULT because data offset is not a valid block (0x%zx)\n",
1187 /* Check that the declared size isn't larger than the block its in. */
1188 size_t blen = block_len (h, data_offset, NULL);
1191 fprintf (stderr, "hivex_value_value: returning EFAULT because data is longer than its block (data 0x%zx, data len %zu, block len %zu)\n",
1192 data_offset, len, blen);
1198 char *data = h->addr + data_offset + 4;
1199 memcpy (ret, data, len);
1204 windows_utf16_to_utf8 (/* const */ char *input, size_t len)
1206 iconv_t ic = iconv_open ("UTF-8", "UTF-16");
1207 if (ic == (iconv_t) -1)
1210 /* iconv(3) has an insane interface ... */
1212 /* Mostly UTF-8 will be smaller, so this is a good initial guess. */
1213 size_t outalloc = len;
1217 size_t outlen = outalloc;
1218 char *out = malloc (outlen + 1);
1228 size_t r = iconv (ic, &inp, &inlen, &outp, &outlen);
1229 if (r == (size_t) -1) {
1230 if (errno == E2BIG) {
1231 size_t prev = outalloc;
1232 /* Try again with a larger output buffer. */
1235 if (outalloc < prev)
1240 /* Else some conversion failure, eg. EILSEQ, EINVAL. */
1256 hivex_value_string (hive_h *h, hive_value_h value)
1260 char *data = hivex_value_value (h, value, &t, &len);
1265 if (t != hive_t_string && t != hive_t_expand_string && t != hive_t_link) {
1271 char *ret = windows_utf16_to_utf8 (data, len);
1280 free_strings (char **argv)
1285 for (i = 0; argv[i] != NULL; ++i)
1291 /* Get the length of a UTF-16 format string. Handle the string as
1292 * pairs of bytes, looking for the first \0\0 pair.
1295 utf16_string_len_in_bytes (const char *str)
1299 while (str[0] || str[1]) {
1307 /* http://blogs.msdn.com/oldnewthing/archive/2009/10/08/9904646.aspx */
1309 hivex_value_multiple_strings (hive_h *h, hive_value_h value)
1313 char *data = hivex_value_value (h, value, &t, &len);
1318 if (t != hive_t_multiple_strings) {
1324 size_t nr_strings = 0;
1325 char **ret = malloc ((1 + nr_strings) * sizeof (char *));
1335 while (p < data + len && (plen = utf16_string_len_in_bytes (p)) > 0) {
1337 char **ret2 = realloc (ret, (1 + nr_strings) * sizeof (char *));
1345 ret[nr_strings-1] = windows_utf16_to_utf8 (p, plen);
1346 ret[nr_strings] = NULL;
1347 if (ret[nr_strings-1] == NULL) {
1353 p += plen + 2 /* skip over UTF-16 \0\0 at the end of this string */;
1361 hivex_value_dword (hive_h *h, hive_value_h value)
1365 char *data = hivex_value_value (h, value, &t, &len);
1370 if ((t != hive_t_dword && t != hive_t_dword_be) || len != 4) {
1376 int32_t ret = *(int32_t*)data;
1378 if (t == hive_t_dword) /* little endian */
1379 ret = le32toh (ret);
1381 ret = be32toh (ret);
1387 hivex_value_qword (hive_h *h, hive_value_h value)
1391 char *data = hivex_value_value (h, value, &t, &len);
1396 if (t != hive_t_qword || len != 8) {
1402 int64_t ret = *(int64_t*)data;
1404 ret = le64toh (ret); /* always little endian */
1410 hivex_visit (hive_h *h, const struct hivex_visitor *visitor, size_t len,
1411 void *opaque, int flags)
1413 return hivex_visit_node (h, hivex_root (h), visitor, len, opaque, flags);
1416 static int hivex__visit_node (hive_h *h, hive_node_h node, const struct hivex_visitor *vtor, char *unvisited, void *opaque, int flags);
1419 hivex_visit_node (hive_h *h, hive_node_h node,
1420 const struct hivex_visitor *visitor, size_t len, void *opaque,
1423 struct hivex_visitor vtor;
1424 memset (&vtor, 0, sizeof vtor);
1426 /* Note that len might be larger *or smaller* than the expected size. */
1427 size_t copysize = len <= sizeof vtor ? len : sizeof vtor;
1428 memcpy (&vtor, visitor, copysize);
1430 /* This bitmap records unvisited nodes, so we don't loop if the
1431 * registry contains cycles.
1433 char *unvisited = malloc (1 + h->size / 32);
1434 if (unvisited == NULL)
1436 memcpy (unvisited, h->bitmap, 1 + h->size / 32);
1438 int r = hivex__visit_node (h, node, &vtor, unvisited, opaque, flags);
1444 hivex__visit_node (hive_h *h, hive_node_h node,
1445 const struct hivex_visitor *vtor, char *unvisited,
1446 void *opaque, int flags)
1448 int skip_bad = flags & HIVEX_VISIT_SKIP_BAD;
1450 hive_value_h *values = NULL;
1451 hive_node_h *children = NULL;
1457 /* Return -1 on all callback errors. However on internal errors,
1458 * check if skip_bad is set and suppress those errors if so.
1462 if (!BITMAP_TST (unvisited, node)) {
1464 fprintf (stderr, "hivex__visit_node: contains cycle: visited node 0x%zx already\n",
1468 return skip_bad ? 0 : -1;
1470 BITMAP_CLR (unvisited, node);
1472 name = hivex_node_name (h, node);
1473 if (!name) return skip_bad ? 0 : -1;
1474 if (vtor->node_start && vtor->node_start (h, opaque, node, name) == -1)
1477 values = hivex_node_values (h, node);
1479 ret = skip_bad ? 0 : -1;
1483 for (i = 0; values[i] != 0; ++i) {
1487 if (hivex_value_type (h, values[i], &t, &len) == -1) {
1488 ret = skip_bad ? 0 : -1;
1492 key = hivex_value_key (h, values[i]);
1494 ret = skip_bad ? 0 : -1;
1498 if (vtor->value_any) {
1499 str = hivex_value_value (h, values[i], &t, &len);
1501 ret = skip_bad ? 0 : -1;
1504 if (vtor->value_any (h, opaque, node, values[i], t, len, key, str) == -1)
1506 free (str); str = NULL;
1511 str = hivex_value_value (h, values[i], &t, &len);
1513 ret = skip_bad ? 0 : -1;
1516 if (t != hive_t_none) {
1517 ret = skip_bad ? 0 : -1;
1520 if (vtor->value_none &&
1521 vtor->value_none (h, opaque, node, values[i], t, len, key, str) == -1)
1523 free (str); str = NULL;
1527 case hive_t_expand_string:
1529 str = hivex_value_string (h, values[i]);
1531 if (errno != EILSEQ && errno != EINVAL) {
1532 ret = skip_bad ? 0 : -1;
1535 if (vtor->value_string_invalid_utf16) {
1536 str = hivex_value_value (h, values[i], &t, &len);
1537 if (vtor->value_string_invalid_utf16 (h, opaque, node, values[i], t, len, key, str) == -1)
1539 free (str); str = NULL;
1543 if (vtor->value_string &&
1544 vtor->value_string (h, opaque, node, values[i], t, len, key, str) == -1)
1546 free (str); str = NULL;
1550 case hive_t_dword_be: {
1551 int32_t i32 = hivex_value_dword (h, values[i]);
1552 if (vtor->value_dword &&
1553 vtor->value_dword (h, opaque, node, values[i], t, len, key, i32) == -1)
1558 case hive_t_qword: {
1559 int64_t i64 = hivex_value_qword (h, values[i]);
1560 if (vtor->value_qword &&
1561 vtor->value_qword (h, opaque, node, values[i], t, len, key, i64) == -1)
1567 str = hivex_value_value (h, values[i], &t, &len);
1569 ret = skip_bad ? 0 : -1;
1572 if (t != hive_t_binary) {
1573 ret = skip_bad ? 0 : -1;
1576 if (vtor->value_binary &&
1577 vtor->value_binary (h, opaque, node, values[i], t, len, key, str) == -1)
1579 free (str); str = NULL;
1582 case hive_t_multiple_strings:
1583 strs = hivex_value_multiple_strings (h, values[i]);
1585 if (errno != EILSEQ && errno != EINVAL) {
1586 ret = skip_bad ? 0 : -1;
1589 if (vtor->value_string_invalid_utf16) {
1590 str = hivex_value_value (h, values[i], &t, &len);
1591 if (vtor->value_string_invalid_utf16 (h, opaque, node, values[i], t, len, key, str) == -1)
1593 free (str); str = NULL;
1597 if (vtor->value_multiple_strings &&
1598 vtor->value_multiple_strings (h, opaque, node, values[i], t, len, key, strs) == -1)
1600 free_strings (strs); strs = NULL;
1603 case hive_t_resource_list:
1604 case hive_t_full_resource_description:
1605 case hive_t_resource_requirements_list:
1607 str = hivex_value_value (h, values[i], &t, &len);
1609 ret = skip_bad ? 0 : -1;
1612 if (vtor->value_other &&
1613 vtor->value_other (h, opaque, node, values[i], t, len, key, str) == -1)
1615 free (str); str = NULL;
1620 free (key); key = NULL;
1623 children = hivex_node_children (h, node);
1624 if (children == NULL) {
1625 ret = skip_bad ? 0 : -1;
1629 for (i = 0; children[i] != 0; ++i) {
1631 fprintf (stderr, "hivex__visit_node: %s: visiting subkey %d (0x%zx)\n",
1632 name, i, children[i]);
1634 if (hivex__visit_node (h, children[i], vtor, unvisited, opaque, flags) == -1)
1638 if (vtor->node_end && vtor->node_end (h, opaque, node, name) == -1)
1649 free_strings (strs);