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.
37 #ifdef HAVE_BYTESWAP_H
41 #if __BYTE_ORDER == __LITTLE_ENDIAN
43 #define be32toh(x) __bswap_32 (x)
46 #define be64toh(x) __bswap_64 (x)
49 #define le16toh(x) (x)
52 #define le32toh(x) (x)
55 #define le64toh(x) (x)
59 #define be32toh(x) (x)
62 #define be64toh(x) (x)
65 #define le16toh(x) __bswap_16 (x)
68 #define le32toh(x) __bswap_32 (x)
71 #define le64toh(x) __bswap_64 (x)
82 /* Memory-mapped (readonly) registry file. */
85 struct ntreg_header *hdr;
88 /* Use a bitmap to store which file offsets are valid (point to a
89 * used block). We only need to store 1 bit per 32 bits of the file
90 * (because blocks are 4-byte aligned). We found that the average
91 * block size in a registry file is ~50 bytes. So roughly 1 in 12
92 * bits in the bitmap will be set, making it likely a more efficient
93 * structure than a hash table.
96 #define BITMAP_SET(bitmap,off) (bitmap[(off)>>5] |= 1 << (((off)>>2)&7))
97 #define BITMAP_CLR(bitmap,off) (bitmap[(off)>>5] &= ~ (1 << (((off)>>2)&7)))
98 #define BITMAP_TST(bitmap,off) (bitmap[(off)>>5] & (1 << (((off)>>2)&7)))
99 #define IS_VALID_BLOCK(h,off) \
100 (((off) & 3) == 0 && \
102 (off) < (h)->size && \
103 BITMAP_TST((h)->bitmap,(off)))
105 /* Fields from the header, extracted from little-endianness. */
106 size_t rootoffs; /* Root key offset (always an nk-block). */
109 size_t pages; /* Number of hbin pages read. */
110 size_t blocks; /* Total number of blocks found. */
111 size_t used_blocks; /* Total number of used blocks found. */
112 size_t used_size; /* Total size (bytes) of used blocks. */
115 /* NB. All fields are little endian. */
116 struct ntreg_header {
117 char magic[4]; /* "regf" */
120 char last_modified[8];
121 uint32_t unknown3; /* 1 */
122 uint32_t unknown4; /* 3 */
123 uint32_t unknown5; /* 0 */
124 uint32_t unknown6; /* 1 */
125 uint32_t offset; /* offset of root key record - 4KB */
126 uint32_t blocks; /* size in bytes of data (filesize - 4KB) */
127 uint32_t unknown7; /* 1 */
128 char name[0x1fc-0x2c];
129 uint32_t csum; /* checksum: sum of 32 bit words 0-0x1fb. */
130 } __attribute__((__packed__));
132 struct ntreg_hbin_page {
133 char magic[4]; /* "hbin" */
134 uint32_t offset_first; /* offset from 1st block */
135 uint32_t offset_next; /* offset of next (relative to this) */
137 /* Linked list of blocks follows here. */
138 } __attribute__((__packed__));
140 struct ntreg_hbin_block {
141 int32_t seg_len; /* length of this block (-ve for used block) */
142 char id[2]; /* the block type (eg. "nk" for nk record) */
143 /* Block data follows here. */
144 } __attribute__((__packed__));
146 #define BLOCK_ID_EQ(h,offs,eqid) \
147 (strncmp (((struct ntreg_hbin_block *)((h)->addr + (offs)))->id, (eqid), 2) == 0)
150 block_len (hive_h *h, size_t blkoff, int *used)
152 struct ntreg_hbin_block *block;
153 block = (struct ntreg_hbin_block *) (h->addr + blkoff);
155 int32_t len = le32toh (block->seg_len);
166 struct ntreg_nk_record {
167 int32_t seg_len; /* length (always -ve because used) */
168 char id[2]; /* "nk" */
171 uint32_t parent; /* offset of owner/parent */
172 uint32_t nr_subkeys; /* number of subkeys */
174 uint32_t subkey_lf; /* lf record containing list of subkeys */
176 uint32_t nr_values; /* number of values */
177 uint32_t vallist; /* value-list record */
178 uint32_t sk; /* offset of sk-record */
179 uint32_t classname; /* offset of classname record */
182 uint16_t name_len; /* length of name */
183 uint16_t classname_len; /* length of classname */
184 char name[1]; /* name follows here */
185 } __attribute__((__packed__));
187 struct ntreg_lf_record {
189 char id[2]; /* "lf" */
190 uint16_t nr_keys; /* number of keys in this record */
192 uint32_t offset; /* offset of nk-record for this subkey */
193 char name[4]; /* first 4 characters of subkey name */
195 } __attribute__((__packed__));
197 struct ntreg_ri_record {
199 char id[2]; /* "ri" */
200 uint16_t nr_offsets; /* number of pointers to lh records */
201 uint32_t offset[1]; /* list of pointers to lh records */
202 } __attribute__((__packed__));
204 /* This has no ID header. */
205 struct ntreg_value_list {
207 uint32_t offset[1]; /* list of pointers to vk records */
208 } __attribute__((__packed__));
210 struct ntreg_vk_record {
211 int32_t seg_len; /* length (always -ve because used) */
212 char id[2]; /* "vk" */
213 uint16_t name_len; /* length of name */
214 /* length of the data:
215 * If data_len is <= 4, then it's stored inline.
216 * If data_len is 0x80000000, then it's an inline dword.
217 * Top bit may be set or not set at random.
220 uint32_t data_offset; /* pointer to the data (or data if inline) */
221 hive_type data_type; /* type of the data */
222 uint16_t unknown1; /* possibly always 1 */
224 char name[1]; /* key name follows here */
225 } __attribute__((__packed__));
228 hivex_open (const char *filename, int flags)
232 h = calloc (1, sizeof *h);
236 h->msglvl = flags & HIVEX_OPEN_MSGLVL_MASK;
238 const char *debug = getenv ("HIVEX_DEBUG");
239 if (debug && strcmp (debug, "1") == 0)
243 printf ("hivex_open: created handle %p\n", h);
245 h->fd = open (filename, O_RDONLY);
250 if (fstat (h->fd, &statbuf) == -1)
253 h->size = statbuf.st_size;
255 h->addr = mmap (NULL, h->size, PROT_READ, MAP_SHARED, h->fd, 0);
256 if (h->addr == MAP_FAILED)
260 printf ("hivex_open: mapped file at %p\n", h->addr);
263 if (h->hdr->magic[0] != 'r' ||
264 h->hdr->magic[1] != 'e' ||
265 h->hdr->magic[2] != 'g' ||
266 h->hdr->magic[3] != 'f') {
267 fprintf (stderr, "hivex: %s: not a Windows NT Registry hive file\n",
273 h->bitmap = calloc (1 + h->size / 32, 1);
274 if (h->bitmap == NULL)
277 #if 0 /* Doesn't work. */
278 /* Header checksum. */
279 uint32_t *daddr = h->addr;
282 for (i = 0; i < 0x1fc / 4; ++i) {
283 sum += le32toh (*daddr);
286 if (sum != le32toh (h->hdr->csum)) {
287 fprintf (stderr, "hivex: %s: bad checksum in hive header\n", filename);
293 h->rootoffs = le32toh (h->hdr->offset) + 0x1000;
296 printf ("hivex_open: root offset = %zu\n", h->rootoffs);
298 /* We'll set this flag when we see a block with the root offset (ie.
301 int seen_root_block = 0, bad_root_block = 0;
303 /* Read the pages and blocks. The aim here is to be robust against
304 * corrupt or malicious registries. So we make sure the loops
305 * always make forward progress. We add the address of each block
306 * we read to a hash table so pointers will only reference the start
310 struct ntreg_hbin_page *page;
311 for (off = 0x1000; off < h->size; off += le32toh (page->offset_next)) {
314 page = (struct ntreg_hbin_page *) (h->addr + off);
315 if (page->magic[0] != 'h' ||
316 page->magic[1] != 'b' ||
317 page->magic[2] != 'i' ||
318 page->magic[3] != 'n') {
319 /* This error is seemingly common in uncorrupt registry files. */
321 fprintf (stderr, "hivex: %s: ignoring trailing garbage at end of file (at %zu, after %zu pages)\n",
322 filename, off, h->pages);
328 printf ("hivex_open: page at %zu\n", off);
330 if (le32toh (page->offset_next) <= sizeof (struct ntreg_hbin_page) ||
331 (le32toh (page->offset_next) & 3) != 0) {
332 fprintf (stderr, "hivex: %s: pagesize %d at %zu, bad registry\n",
333 filename, le32toh (page->offset_next), off);
338 /* Read the blocks in this page. */
340 struct ntreg_hbin_block *block;
342 for (blkoff = off + 0x20;
343 blkoff < off + le32toh (page->offset_next);
347 int is_root = blkoff == h->rootoffs;
351 block = (struct ntreg_hbin_block *) (h->addr + blkoff);
353 seg_len = block_len (h, blkoff, &used);
354 if (seg_len <= 4 || (seg_len & 3) != 0) {
355 fprintf (stderr, "hivex: %s: block size %d at %zu, bad registry\n",
356 filename, le32toh (block->seg_len), blkoff);
362 printf ("hivex_open: %s block id %d,%d at %zu%s\n",
363 used ? "used" : "free", block->id[0], block->id[1], blkoff,
364 is_root ? " (root)" : "");
366 if (is_root && !used)
371 h->used_size += seg_len;
373 /* Root block must be an nk-block. */
374 if (is_root && (block->id[0] != 'n' || block->id[1] != 'k'))
377 /* Note this blkoff is a valid address. */
378 BITMAP_SET (h->bitmap, blkoff);
383 if (!seen_root_block) {
384 fprintf (stderr, "hivex: %s: no root block found\n", filename);
389 if (bad_root_block) {
390 fprintf (stderr, "hivex: %s: bad root block (free or not nk)\n", filename);
396 printf ("hivex_open: successfully read Windows Registry hive file:\n"
399 " blocks used: %zu\n"
400 " bytes used: %zu\n",
401 h->pages, h->blocks, h->used_blocks, h->used_size);
409 if (h->addr && h->size && h->addr != MAP_FAILED)
410 munmap (h->addr, h->size);
420 hivex_close (hive_h *h)
425 munmap (h->addr, h->size);
433 hivex_root (hive_h *h)
435 hive_node_h ret = h->rootoffs;
436 if (!IS_VALID_BLOCK (h, ret)) {
444 hivex_node_name (hive_h *h, hive_node_h node)
446 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
451 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
453 /* AFAIK the node name is always plain ASCII, so no conversion
454 * to UTF-8 is necessary. However we do need to nul-terminate
458 /* nk->name_len is unsigned, 16 bit, so this is safe ... However
459 * we have to make sure the length doesn't exceed the block length.
461 size_t len = le16toh (nk->name_len);
462 size_t seg_len = block_len (h, node, NULL);
463 if (sizeof (struct ntreg_nk_record) + len - 1 > seg_len) {
465 printf ("hivex_node_name: returning EFAULT because node name is too long (%zu, %zu)\n",
471 char *ret = malloc (len + 1);
474 memcpy (ret, nk->name, len);
480 /* I think the documentation for the sk and classname fields in the nk
481 * record is wrong, or else the offset field is in the wrong place.
482 * Otherwise this makes no sense. Disabled this for now -- it's not
483 * useful for reading the registry anyway.
487 hivex_node_security (hive_h *h, hive_node_h node)
489 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
494 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
496 hive_node_h ret = le32toh (nk->sk);
498 if (!IS_VALID_BLOCK (h, ret)) {
506 hivex_node_classname (hive_h *h, hive_node_h node)
508 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
513 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
515 hive_node_h ret = le32toh (nk->classname);
517 if (!IS_VALID_BLOCK (h, ret)) {
526 hivex_node_children (hive_h *h, hive_node_h node)
528 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
533 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
535 size_t nr_subkeys_in_nk = le32toh (nk->nr_subkeys);
537 /* Deal with the common "no subkeys" case quickly. */
539 if (nr_subkeys_in_nk == 0) {
540 ret = malloc (sizeof (hive_node_h));
547 /* Arbitrarily limit the number of subkeys we will ever deal with. */
548 if (nr_subkeys_in_nk > 1000000) {
553 /* The subkey_lf field can point either to an lf-record, which is
554 * the common case, or if there are lots of subkeys, to an
557 size_t subkey_lf = le32toh (nk->subkey_lf);
559 if (!IS_VALID_BLOCK (h, subkey_lf)) {
561 printf ("hivex_node_children: returning EFAULT because subkey_lf is not a valid block (%zu)\n",
567 struct ntreg_hbin_block *block =
568 (struct ntreg_hbin_block *) (h->addr + subkey_lf);
570 /* Points to lf-record? (Note, also "lh" but that is basically the
571 * same as "lf" as far as we are concerned here).
573 if (block->id[0] == 'l' && (block->id[1] == 'f' || block->id[1] == 'h')) {
574 struct ntreg_lf_record *lf = (struct ntreg_lf_record *) block;
576 /* Check number of subkeys in the nk-record matches number of subkeys
579 size_t nr_subkeys_in_lf = le16toh (lf->nr_keys);
582 printf ("hivex_node_children: nr_subkeys_in_nk = %zu, nr_subkeys_in_lf = %zu\n",
583 nr_subkeys_in_nk, nr_subkeys_in_lf);
585 if (nr_subkeys_in_nk != nr_subkeys_in_lf) {
590 size_t len = block_len (h, subkey_lf, NULL);
591 if (8 + nr_subkeys_in_lf * 8 > len) {
593 printf ("hivex_node_children: returning EFAULT because too many subkeys (%zu, %zu)\n",
594 nr_subkeys_in_lf, len);
599 /* Allocate space for the returned values. Note that
600 * nr_subkeys_in_lf is limited to a 16 bit value.
602 ret = malloc ((1 + nr_subkeys_in_lf) * sizeof (hive_node_h));
607 for (i = 0; i < nr_subkeys_in_lf; ++i) {
608 hive_node_h subkey = lf->keys[i].offset;
610 if (!IS_VALID_BLOCK (h, subkey)) {
612 printf ("hivex_node_children: returning EFAULT because subkey is not a valid block (%zu)\n",
623 /* Points to ri-record? */
624 else if (block->id[0] == 'r' && block->id[1] == 'i') {
625 struct ntreg_ri_record *ri = (struct ntreg_ri_record *) block;
627 size_t nr_offsets = le16toh (ri->nr_offsets);
629 /* Count total number of children. */
631 for (i = 0; i < nr_offsets; ++i) {
632 hive_node_h offset = ri->offset[i];
634 if (!IS_VALID_BLOCK (h, offset)) {
636 printf ("hivex_node_children: returning EFAULT because ri-offset is not a valid block (%zu)\n",
641 if (!BLOCK_ID_EQ (h, offset, "lf") && !BLOCK_ID_EQ (h, offset, "lh")) {
646 struct ntreg_lf_record *lf =
647 (struct ntreg_lf_record *) (h->addr + offset);
649 count += le16toh (lf->nr_keys);
653 printf ("hivex_node_children: nr_subkeys_in_nk = %zu, counted = %zu\n",
654 nr_subkeys_in_nk, count);
656 if (nr_subkeys_in_nk != count) {
661 /* Copy list of children. Note nr_subkeys_in_nk is limited to
662 * something reasonable above.
664 ret = malloc ((1 + nr_subkeys_in_nk) * sizeof (hive_node_h));
669 for (i = 0; i < nr_offsets; ++i) {
670 hive_node_h offset = ri->offset[i];
672 if (!IS_VALID_BLOCK (h, offset)) {
674 printf ("hivex_node_children: returning EFAULT because ri-offset is not a valid block (%zu)\n",
679 if (!BLOCK_ID_EQ (h, offset, "lf") && !BLOCK_ID_EQ (h, offset, "lh")) {
684 struct ntreg_lf_record *lf =
685 (struct ntreg_lf_record *) (h->addr + offset);
688 for (j = 0; j < le16toh (lf->nr_keys); ++j) {
689 hive_node_h subkey = lf->keys[j].offset;
691 if (!IS_VALID_BLOCK (h, subkey)) {
693 printf ("hivex_node_children: returning EFAULT because indirect subkey is not a valid block (%zu)\n",
699 ret[count++] = subkey;
712 /* Very inefficient, but at least having a separate API call
713 * allows us to make it more efficient in future.
716 hivex_node_get_child (hive_h *h, hive_node_h node, const char *nname)
718 hive_node_h *children = NULL;
722 children = hivex_node_children (h, node);
723 if (!children) goto error;
726 for (i = 0; children[i] != 0; ++i) {
727 name = hivex_node_name (h, children[i]);
728 if (!name) goto error;
729 if (strcasecmp (name, nname) == 0) {
733 free (name); name = NULL;
743 hivex_node_parent (hive_h *h, hive_node_h node)
745 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
750 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
752 hive_node_h ret = le32toh (nk->parent);
754 printf ("parent = %zu\n", ret);
755 if (!IS_VALID_BLOCK (h, ret)) {
757 printf ("hivex_node_parent: returning EFAULT because parent is not a valid block (%zu)\n",
766 hivex_node_values (hive_h *h, hive_node_h node)
768 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
773 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
775 size_t nr_values = le32toh (nk->nr_values);
778 printf ("hivex_node_values: nr_values = %zu\n", nr_values);
780 /* Deal with the common "no values" case quickly. */
782 if (nr_values == 0) {
783 ret = malloc (sizeof (hive_node_h));
790 /* Arbitrarily limit the number of values we will ever deal with. */
791 if (nr_values > 100000) {
796 /* Get the value list and check it looks reasonable. */
797 size_t vlist_offset = le32toh (nk->vallist);
798 vlist_offset += 0x1000;
799 if (!IS_VALID_BLOCK (h, vlist_offset)) {
801 printf ("hivex_node_values: returning EFAULT because value list is not a valid block (%zu)\n",
807 struct ntreg_value_list *vlist =
808 (struct ntreg_value_list *) (h->addr + vlist_offset);
810 size_t len = block_len (h, vlist_offset, NULL);
811 if (4 + nr_values * 4 > len) {
813 printf ("hivex_node_values: returning EFAULT because value list is too long (%zu, %zu)\n",
819 /* Allocate return array and copy values in. */
820 ret = malloc ((1 + nr_values) * sizeof (hive_node_h));
825 for (i = 0; i < nr_values; ++i) {
826 hive_node_h value = vlist->offset[i];
828 if (!IS_VALID_BLOCK (h, value)) {
830 printf ("hivex_node_values: returning EFAULT because value is not a valid block (%zu)\n",
843 /* Very inefficient, but at least having a separate API call
844 * allows us to make it more efficient in future.
847 hivex_node_get_value (hive_h *h, hive_node_h node, const char *key)
849 hive_value_h *values = NULL;
851 hive_value_h ret = 0;
853 values = hivex_node_values (h, node);
854 if (!values) goto error;
857 for (i = 0; values[i] != 0; ++i) {
858 name = hivex_value_key (h, values[i]);
859 if (!name) goto error;
860 if (strcasecmp (name, key) == 0) {
864 free (name); name = NULL;
874 hivex_value_key (hive_h *h, hive_value_h value)
876 if (!IS_VALID_BLOCK (h, value) || !BLOCK_ID_EQ (h, value, "vk")) {
881 struct ntreg_vk_record *vk = (struct ntreg_vk_record *) (h->addr + value);
883 /* AFAIK the key is always plain ASCII, so no conversion to UTF-8 is
884 * necessary. However we do need to nul-terminate the string.
887 /* vk->name_len is unsigned, 16 bit, so this is safe ... However
888 * we have to make sure the length doesn't exceed the block length.
890 size_t len = le16toh (vk->name_len);
891 size_t seg_len = block_len (h, value, NULL);
892 if (sizeof (struct ntreg_vk_record) + len - 1 > seg_len) {
894 printf ("hivex_value_key: returning EFAULT because key length is too long (%zu, %zu)\n",
900 char *ret = malloc (len + 1);
903 memcpy (ret, vk->name, len);
909 hivex_value_type (hive_h *h, hive_value_h value, hive_type *t, size_t *len)
911 if (!IS_VALID_BLOCK (h, value) || !BLOCK_ID_EQ (h, value, "vk")) {
916 struct ntreg_vk_record *vk = (struct ntreg_vk_record *) (h->addr + value);
919 *t = le32toh (vk->data_type);
922 *len = le32toh (vk->data_len);
923 if (*len == 0x80000000) { /* special case */
925 if (t) *t = hive_t_dword;
934 hivex_value_value (hive_h *h, hive_value_h value,
935 hive_type *t_rtn, size_t *len_rtn)
937 if (!IS_VALID_BLOCK (h, value) || !BLOCK_ID_EQ (h, value, "vk")) {
942 struct ntreg_vk_record *vk = (struct ntreg_vk_record *) (h->addr + value);
947 t = le32toh (vk->data_type);
949 len = le32toh (vk->data_len);
950 if (len == 0x80000000) { /* special case */
957 printf ("hivex_value_value: value=%zu, t=%d, len=%zu\n",
965 /* Arbitrarily limit the length that we will read. */
971 char *ret = malloc (len);
975 /* If length is <= 4 it's always stored inline. */
977 memcpy (ret, (char *) &vk->data_offset, len);
981 size_t data_offset = vk->data_offset;
982 data_offset += 0x1000;
983 if (!IS_VALID_BLOCK (h, data_offset)) {
985 printf ("hivex_value_value: returning EFAULT because data offset is not a valid block (%zu)\n",
992 /* Check that the declared size isn't larger than the block its in. */
993 size_t blen = block_len (h, data_offset, NULL);
996 printf ("hivex_value_value: returning EFAULT because data is longer than its block (%zu, %zu)\n",
1003 char *data = h->addr + data_offset + 4;
1004 memcpy (ret, data, len);
1009 windows_utf16_to_utf8 (/* const */ char *input, size_t len)
1011 iconv_t ic = iconv_open ("UTF-8", "UTF-16");
1012 if (ic == (iconv_t) -1)
1015 /* iconv(3) has an insane interface ... */
1017 /* Mostly UTF-8 will be smaller, so this is a good initial guess. */
1018 size_t outalloc = len;
1022 size_t outlen = outalloc;
1023 char *out = malloc (outlen + 1);
1033 size_t r = iconv (ic, &inp, &inlen, &outp, &outlen);
1034 if (r == (size_t) -1) {
1035 if (errno == E2BIG) {
1036 /* Try again with a larger output buffer. */
1042 /* Else some conversion failure, eg. EILSEQ, EINVAL. */
1058 hivex_value_string (hive_h *h, hive_value_h value)
1062 char *data = hivex_value_value (h, value, &t, &len);
1067 if (t != hive_t_string && t != hive_t_expand_string && t != hive_t_link) {
1073 char *ret = windows_utf16_to_utf8 (data, len);
1082 free_strings (char **argv)
1087 for (i = 0; argv[i] != NULL; ++i)
1093 /* Get the length of a UTF-16 format string. Handle the string as
1094 * pairs of bytes, looking for the first \0\0 pair.
1097 utf16_string_len_in_bytes (const char *str)
1101 while (str[0] || str[1]) {
1109 /* http://blogs.msdn.com/oldnewthing/archive/2009/10/08/9904646.aspx */
1111 hivex_value_multiple_strings (hive_h *h, hive_value_h value)
1115 char *data = hivex_value_value (h, value, &t, &len);
1120 if (t != hive_t_multiple_strings) {
1126 size_t nr_strings = 0;
1127 char **ret = malloc ((1 + nr_strings) * sizeof (char *));
1137 while (p < data + len && (plen = utf16_string_len_in_bytes (p)) > 0) {
1139 char **ret2 = realloc (ret, (1 + nr_strings) * sizeof (char *));
1147 ret[nr_strings-1] = windows_utf16_to_utf8 (p, plen);
1148 ret[nr_strings] = NULL;
1149 if (ret[nr_strings-1] == NULL) {
1155 p += plen + 2 /* skip over UTF-16 \0\0 at the end of this string */;
1163 hivex_value_dword (hive_h *h, hive_value_h value)
1167 char *data = hivex_value_value (h, value, &t, &len);
1172 if ((t != hive_t_dword && t != hive_t_dword_be) || len != 4) {
1178 int32_t ret = *(int32_t*)data;
1180 if (t == hive_t_dword) /* little endian */
1181 ret = le32toh (ret);
1183 ret = be32toh (ret);
1189 hivex_value_qword (hive_h *h, hive_value_h value)
1193 char *data = hivex_value_value (h, value, &t, &len);
1198 if (t != hive_t_qword || len != 8) {
1204 int64_t ret = *(int64_t*)data;
1206 ret = le64toh (ret); /* always little endian */
1212 hivex_visit (hive_h *h, const struct hivex_visitor *visitor, size_t len,
1213 void *opaque, int flags)
1215 return hivex_visit_node (h, hivex_root (h), visitor, len, opaque, flags);
1218 static int hivex__visit_node (hive_h *h, hive_node_h node, const struct hivex_visitor *vtor, char *unvisited, void *opaque, int flags);
1221 hivex_visit_node (hive_h *h, hive_node_h node,
1222 const struct hivex_visitor *visitor, size_t len, void *opaque,
1225 struct hivex_visitor vtor;
1226 memset (&vtor, 0, sizeof vtor);
1228 /* Note that len might be larger *or smaller* than the expected size. */
1229 size_t copysize = len <= sizeof vtor ? len : sizeof vtor;
1230 memcpy (&vtor, visitor, copysize);
1232 /* This bitmap records unvisited nodes, so we don't loop if the
1233 * registry contains cycles.
1235 char *unvisited = malloc (1 + h->size / 32);
1236 if (unvisited == NULL)
1238 memcpy (unvisited, h->bitmap, 1 + h->size / 32);
1240 int r = hivex__visit_node (h, node, &vtor, unvisited, opaque, flags);
1246 hivex__visit_node (hive_h *h, hive_node_h node,
1247 const struct hivex_visitor *vtor, char *unvisited,
1248 void *opaque, int flags)
1250 int skip_bad = flags & HIVEX_VISIT_SKIP_BAD;
1252 hive_value_h *values = NULL;
1253 hive_node_h *children = NULL;
1259 /* Return -1 on all callback errors. However on internal errors,
1260 * check if skip_bad is set and suppress those errors if so.
1264 if (!BITMAP_TST (unvisited, node)) {
1266 printf ("hivex__visit_node: contains cycle: visited node %zu already\n",
1270 return skip_bad ? 0 : -1;
1272 BITMAP_CLR (unvisited, node);
1274 name = hivex_node_name (h, node);
1275 if (!name) return skip_bad ? 0 : -1;
1276 if (vtor->node_start && vtor->node_start (h, opaque, node, name) == -1)
1279 values = hivex_node_values (h, node);
1281 ret = skip_bad ? 0 : -1;
1285 for (i = 0; values[i] != 0; ++i) {
1289 if (hivex_value_type (h, values[i], &t, &len) == -1) {
1290 ret = skip_bad ? 0 : -1;
1294 key = hivex_value_key (h, values[i]);
1296 ret = skip_bad ? 0 : -1;
1302 str = hivex_value_value (h, values[i], &t, &len);
1304 ret = skip_bad ? 0 : -1;
1307 if (t != hive_t_none) {
1308 ret = skip_bad ? 0 : -1;
1311 if (vtor->value_none &&
1312 vtor->value_none (h, opaque, node, values[i], t, len, key, str) == -1)
1314 free (str); str = NULL;
1318 case hive_t_expand_string:
1320 str = hivex_value_string (h, values[i]);
1322 if (errno != EILSEQ && errno != EINVAL) {
1323 ret = skip_bad ? 0 : -1;
1326 if (vtor->value_string_invalid_utf16) {
1327 str = hivex_value_value (h, values[i], &t, &len);
1328 if (vtor->value_string_invalid_utf16 (h, opaque, node, values[i], t, len, key, str) == -1)
1330 free (str); str = NULL;
1334 if (vtor->value_string &&
1335 vtor->value_string (h, opaque, node, values[i], t, len, key, str) == -1)
1337 free (str); str = NULL;
1341 case hive_t_dword_be: {
1342 int32_t i32 = hivex_value_dword (h, values[i]);
1343 if (vtor->value_dword &&
1344 vtor->value_dword (h, opaque, node, values[i], t, len, key, i32) == -1)
1349 case hive_t_qword: {
1350 int64_t i64 = hivex_value_qword (h, values[i]);
1351 if (vtor->value_qword &&
1352 vtor->value_qword (h, opaque, node, values[i], t, len, key, i64) == -1)
1358 str = hivex_value_value (h, values[i], &t, &len);
1360 ret = skip_bad ? 0 : -1;
1363 if (t != hive_t_binary) {
1364 ret = skip_bad ? 0 : -1;
1367 if (vtor->value_binary &&
1368 vtor->value_binary (h, opaque, node, values[i], t, len, key, str) == -1)
1370 free (str); str = NULL;
1373 case hive_t_multiple_strings:
1374 strs = hivex_value_multiple_strings (h, values[i]);
1376 if (errno != EILSEQ && errno != EINVAL) {
1377 ret = skip_bad ? 0 : -1;
1380 if (vtor->value_string_invalid_utf16) {
1381 str = hivex_value_value (h, values[i], &t, &len);
1382 if (vtor->value_string_invalid_utf16 (h, opaque, node, values[i], t, len, key, str) == -1)
1384 free (str); str = NULL;
1388 if (vtor->value_multiple_strings &&
1389 vtor->value_multiple_strings (h, opaque, node, values[i], t, len, key, strs) == -1)
1391 free_strings (strs); strs = NULL;
1394 case hive_t_resource_list:
1395 case hive_t_full_resource_description:
1396 case hive_t_resource_requirements_list:
1398 str = hivex_value_value (h, values[i], &t, &len);
1400 ret = skip_bad ? 0 : -1;
1403 if (vtor->value_other &&
1404 vtor->value_other (h, opaque, node, values[i], t, len, key, str) == -1)
1406 free (str); str = NULL;
1410 free (key); key = NULL;
1413 children = hivex_node_children (h, node);
1414 if (children == NULL) {
1415 ret = skip_bad ? 0 : -1;
1419 for (i = 0; children[i] != 0; ++i) {
1421 printf ("hivex__visit_node: %s: visiting subkey %d (%zu)\n",
1422 name, i, children[i]);
1424 if (hivex__visit_node (h, children[i], vtor, unvisited, opaque, flags) == -1)
1428 if (vtor->node_end && vtor->node_end (h, opaque, node, name) == -1)
1439 free_strings (strs);