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 le32toh(x) (x)
52 #define le32toh(x) (x)
55 #define le32toh(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);
275 #if 0 /* Doesn't work. */
276 /* Header checksum. */
277 uint32_t *daddr = h->addr;
280 for (i = 0; i < 0x1fc / 4; ++i) {
281 sum += le32toh (*daddr);
284 if (sum != le32toh (h->hdr->csum)) {
285 fprintf (stderr, "hivex: %s: bad checksum in hive header\n", filename);
291 h->rootoffs = le32toh (h->hdr->offset) + 0x1000;
294 printf ("hivex_open: root offset = %zu\n", h->rootoffs);
296 /* We'll set this flag when we see a block with the root offset (ie.
299 int seen_root_block = 0, bad_root_block = 0;
301 /* Read the pages and blocks. The aim here is to be robust against
302 * corrupt or malicious registries. So we make sure the loops
303 * always make forward progress. We add the address of each block
304 * we read to a hash table so pointers will only reference the start
308 struct ntreg_hbin_page *page;
309 for (off = 0x1000; off < h->size; off += le32toh (page->offset_next)) {
312 page = (struct ntreg_hbin_page *) (h->addr + off);
313 if (page->magic[0] != 'h' ||
314 page->magic[1] != 'b' ||
315 page->magic[2] != 'i' ||
316 page->magic[3] != 'n') {
317 /* This error is seemingly common in uncorrupt registry files. */
319 fprintf (stderr, "hivex: %s: ignoring trailing garbage at end of file (at %zu, after %zu pages)\n",
320 filename, off, h->pages);
326 printf ("hivex_open: page at %zu\n", off);
328 if (le32toh (page->offset_next) <= sizeof (struct ntreg_hbin_page) ||
329 (le32toh (page->offset_next) & 3) != 0) {
330 fprintf (stderr, "hivex: %s: pagesize %d at %zu, bad registry\n",
331 filename, le32toh (page->offset_next), off);
336 /* Read the blocks in this page. */
338 struct ntreg_hbin_block *block;
340 for (blkoff = off + 0x20;
341 blkoff < off + le32toh (page->offset_next);
345 int is_root = blkoff == h->rootoffs;
349 block = (struct ntreg_hbin_block *) (h->addr + blkoff);
351 seg_len = block_len (h, blkoff, &used);
352 if (seg_len <= 4 || (seg_len & 3) != 0) {
353 fprintf (stderr, "hivex: %s: block size %d at %zu, bad registry\n",
354 filename, le32toh (block->seg_len), blkoff);
360 printf ("hivex_open: %s block id %d,%d at %zu%s\n",
361 used ? "used" : "free", block->id[0], block->id[1], blkoff,
362 is_root ? " (root)" : "");
364 if (is_root && !used)
369 h->used_size += seg_len;
371 /* Root block must be an nk-block. */
372 if (is_root && (block->id[0] != 'n' || block->id[1] != 'k'))
375 /* Note this blkoff is a valid address. */
376 BITMAP_SET (h->bitmap, blkoff);
381 if (!seen_root_block) {
382 fprintf (stderr, "hivex: %s: no root block found\n", filename);
387 if (bad_root_block) {
388 fprintf (stderr, "hivex: %s: bad root block (free or not nk)\n", filename);
394 printf ("hivex_open: successfully read Windows Registry hive file:\n"
397 " blocks used: %zu\n"
398 " bytes used: %zu\n",
399 h->pages, h->blocks, h->used_blocks, h->used_size);
407 if (h->addr && h->size && h->addr != MAP_FAILED)
408 munmap (h->addr, h->size);
418 hivex_close (hive_h *h)
423 munmap (h->addr, h->size);
431 hivex_root (hive_h *h)
433 hive_node_h ret = h->rootoffs;
434 if (!IS_VALID_BLOCK (h, ret)) {
442 hivex_node_name (hive_h *h, hive_node_h node)
444 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
449 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
451 /* AFAIK the node name is always plain ASCII, so no conversion
452 * to UTF-8 is necessary. However we do need to nul-terminate
456 /* nk->name_len is unsigned, 16 bit, so this is safe ... However
457 * we have to make sure the length doesn't exceed the block length.
459 size_t len = le16toh (nk->name_len);
460 size_t seg_len = block_len (h, node, NULL);
461 if (sizeof (struct ntreg_nk_record) + len - 1 > seg_len) {
463 printf ("hivex_node_name: returning EFAULT because node name is too long (%zu, %zu)\n",
469 char *ret = malloc (len + 1);
472 memcpy (ret, nk->name, len);
478 /* I think the documentation for the sk and classname fields in the nk
479 * record is wrong, or else the offset field is in the wrong place.
480 * Otherwise this makes no sense. Disabled this for now -- it's not
481 * useful for reading the registry anyway.
485 hivex_node_security (hive_h *h, hive_node_h node)
487 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
492 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
494 hive_node_h ret = le32toh (nk->sk);
496 if (!IS_VALID_BLOCK (h, ret)) {
504 hivex_node_classname (hive_h *h, hive_node_h node)
506 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
511 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
513 hive_node_h ret = le32toh (nk->classname);
515 if (!IS_VALID_BLOCK (h, ret)) {
524 hivex_node_children (hive_h *h, hive_node_h node)
526 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
531 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
533 size_t nr_subkeys_in_nk = le32toh (nk->nr_subkeys);
535 /* Deal with the common "no subkeys" case quickly. */
537 if (nr_subkeys_in_nk == 0) {
538 ret = malloc (sizeof (hive_node_h));
545 /* Arbitrarily limit the number of subkeys we will ever deal with. */
546 if (nr_subkeys_in_nk > 1000000) {
551 /* The subkey_lf field can point either to an lf-record, which is
552 * the common case, or if there are lots of subkeys, to an
555 size_t subkey_lf = le32toh (nk->subkey_lf);
557 if (!IS_VALID_BLOCK (h, subkey_lf)) {
559 printf ("hivex_node_children: returning EFAULT because subkey_lf is not a valid block (%zu)\n",
565 struct ntreg_hbin_block *block =
566 (struct ntreg_hbin_block *) (h->addr + subkey_lf);
568 /* Points to lf-record? (Note, also "lh" but that is basically the
569 * same as "lf" as far as we are concerned here).
571 if (block->id[0] == 'l' && (block->id[1] == 'f' || block->id[1] == 'h')) {
572 struct ntreg_lf_record *lf = (struct ntreg_lf_record *) block;
574 /* Check number of subkeys in the nk-record matches number of subkeys
577 size_t nr_subkeys_in_lf = le16toh (lf->nr_keys);
580 printf ("hivex_node_children: nr_subkeys_in_nk = %zu, nr_subkeys_in_lf = %zu\n",
581 nr_subkeys_in_nk, nr_subkeys_in_lf);
583 if (nr_subkeys_in_nk != nr_subkeys_in_lf) {
588 size_t len = block_len (h, subkey_lf, NULL);
589 if (8 + nr_subkeys_in_lf * 8 > len) {
591 printf ("hivex_node_children: returning EFAULT because too many subkeys (%zu, %zu)\n",
592 nr_subkeys_in_lf, len);
597 /* Allocate space for the returned values. Note that
598 * nr_subkeys_in_lf is limited to a 16 bit value.
600 ret = malloc ((1 + nr_subkeys_in_lf) * sizeof (hive_node_h));
605 for (i = 0; i < nr_subkeys_in_lf; ++i) {
606 hive_node_h subkey = lf->keys[i].offset;
608 if (!IS_VALID_BLOCK (h, subkey)) {
610 printf ("hivex_node_children: returning EFAULT because subkey is not a valid block (%zu)\n",
621 /* Points to ri-record? */
622 else if (block->id[0] == 'r' && block->id[1] == 'i') {
623 struct ntreg_ri_record *ri = (struct ntreg_ri_record *) block;
625 size_t nr_offsets = le16toh (ri->nr_offsets);
627 /* Count total number of children. */
629 for (i = 0; i < nr_offsets; ++i) {
630 hive_node_h offset = ri->offset[i];
632 if (!IS_VALID_BLOCK (h, offset)) {
634 printf ("hivex_node_children: returning EFAULT because ri-offset is not a valid block (%zu)\n",
639 if (!BLOCK_ID_EQ (h, offset, "lf") && !BLOCK_ID_EQ (h, offset, "lh")) {
644 struct ntreg_lf_record *lf =
645 (struct ntreg_lf_record *) (h->addr + offset);
647 count += le16toh (lf->nr_keys);
651 printf ("hivex_node_children: nr_subkeys_in_nk = %zu, counted = %zu\n",
652 nr_subkeys_in_nk, count);
654 if (nr_subkeys_in_nk != count) {
659 /* Copy list of children. Note nr_subkeys_in_nk is limited to
660 * something reasonable above.
662 ret = malloc ((1 + nr_subkeys_in_nk) * sizeof (hive_node_h));
667 for (i = 0; i < nr_offsets; ++i) {
668 hive_node_h offset = ri->offset[i];
670 if (!IS_VALID_BLOCK (h, offset)) {
672 printf ("hivex_node_children: returning EFAULT because ri-offset is not a valid block (%zu)\n",
677 if (!BLOCK_ID_EQ (h, offset, "lf") && !BLOCK_ID_EQ (h, offset, "lh")) {
682 struct ntreg_lf_record *lf =
683 (struct ntreg_lf_record *) (h->addr + offset);
686 for (j = 0; j < le16toh (lf->nr_keys); ++j) {
687 hive_node_h subkey = lf->keys[j].offset;
689 if (!IS_VALID_BLOCK (h, subkey)) {
691 printf ("hivex_node_children: returning EFAULT because indirect subkey is not a valid block (%zu)\n",
697 ret[count++] = subkey;
710 /* Very inefficient, but at least having a separate API call
711 * allows us to make it more efficient in future.
714 hivex_node_get_child (hive_h *h, hive_node_h node, const char *nname)
716 hive_node_h *children = NULL;
720 children = hivex_node_children (h, node);
721 if (!children) goto error;
724 for (i = 0; children[i] != 0; ++i) {
725 name = hivex_node_name (h, children[i]);
726 if (!name) goto error;
727 if (strcasecmp (name, nname) == 0) {
731 free (name); name = NULL;
741 hivex_node_parent (hive_h *h, hive_node_h node)
743 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
748 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
750 hive_node_h ret = le32toh (nk->parent);
752 printf ("parent = %zu\n", ret);
753 if (!IS_VALID_BLOCK (h, ret)) {
755 printf ("hivex_node_parent: returning EFAULT because parent is not a valid block (%zu)\n",
764 hivex_node_values (hive_h *h, hive_node_h node)
766 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
771 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
773 size_t nr_values = le32toh (nk->nr_values);
776 printf ("hivex_node_values: nr_values = %zu\n", nr_values);
778 /* Deal with the common "no values" case quickly. */
780 if (nr_values == 0) {
781 ret = malloc (sizeof (hive_node_h));
788 /* Arbitrarily limit the number of values we will ever deal with. */
789 if (nr_values > 100000) {
794 /* Get the value list and check it looks reasonable. */
795 size_t vlist_offset = le32toh (nk->vallist);
796 vlist_offset += 0x1000;
797 if (!IS_VALID_BLOCK (h, vlist_offset)) {
799 printf ("hivex_node_values: returning EFAULT because value list is not a valid block (%zu)\n",
805 struct ntreg_value_list *vlist =
806 (struct ntreg_value_list *) (h->addr + vlist_offset);
808 size_t len = block_len (h, vlist_offset, NULL);
809 if (4 + nr_values * 4 > len) {
811 printf ("hivex_node_values: returning EFAULT because value list is too long (%zu, %zu)\n",
817 /* Allocate return array and copy values in. */
818 ret = malloc ((1 + nr_values) * sizeof (hive_node_h));
823 for (i = 0; i < nr_values; ++i) {
824 hive_node_h value = vlist->offset[i];
826 if (!IS_VALID_BLOCK (h, value)) {
828 printf ("hivex_node_values: returning EFAULT because value is not a valid block (%zu)\n",
841 /* Very inefficient, but at least having a separate API call
842 * allows us to make it more efficient in future.
845 hivex_node_get_value (hive_h *h, hive_node_h node, const char *key)
847 hive_value_h *values = NULL;
849 hive_value_h ret = 0;
851 values = hivex_node_values (h, node);
852 if (!values) goto error;
855 for (i = 0; values[i] != 0; ++i) {
856 name = hivex_value_key (h, values[i]);
857 if (!name) goto error;
858 if (strcasecmp (name, key) == 0) {
862 free (name); name = NULL;
872 hivex_value_key (hive_h *h, hive_value_h value)
874 if (!IS_VALID_BLOCK (h, value) || !BLOCK_ID_EQ (h, value, "vk")) {
879 struct ntreg_vk_record *vk = (struct ntreg_vk_record *) (h->addr + value);
881 /* AFAIK the key is always plain ASCII, so no conversion to UTF-8 is
882 * necessary. However we do need to nul-terminate the string.
885 /* vk->name_len is unsigned, 16 bit, so this is safe ... However
886 * we have to make sure the length doesn't exceed the block length.
888 size_t len = le16toh (vk->name_len);
889 size_t seg_len = block_len (h, value, NULL);
890 if (sizeof (struct ntreg_vk_record) + len - 1 > seg_len) {
892 printf ("hivex_value_key: returning EFAULT because key length is too long (%zu, %zu)\n",
898 char *ret = malloc (len + 1);
901 memcpy (ret, vk->name, len);
907 hivex_value_type (hive_h *h, hive_value_h value, hive_type *t, size_t *len)
909 if (!IS_VALID_BLOCK (h, value) || !BLOCK_ID_EQ (h, value, "vk")) {
914 struct ntreg_vk_record *vk = (struct ntreg_vk_record *) (h->addr + value);
917 *t = le32toh (vk->data_type);
920 *len = le32toh (vk->data_len);
921 if (*len == 0x80000000) { /* special case */
923 if (t) *t = hive_t_dword;
932 hivex_value_value (hive_h *h, hive_value_h value,
933 hive_type *t_rtn, size_t *len_rtn)
935 if (!IS_VALID_BLOCK (h, value) || !BLOCK_ID_EQ (h, value, "vk")) {
940 struct ntreg_vk_record *vk = (struct ntreg_vk_record *) (h->addr + value);
945 t = le32toh (vk->data_type);
947 len = le32toh (vk->data_len);
948 if (len == 0x80000000) { /* special case */
955 printf ("hivex_value_value: value=%zu, t=%d, len=%zu\n",
963 /* Arbitrarily limit the length that we will read. */
969 char *ret = malloc (len);
973 /* If length is <= 4 it's always stored inline. */
975 memcpy (ret, (char *) &vk->data_offset, len);
979 size_t data_offset = vk->data_offset;
980 data_offset += 0x1000;
981 if (!IS_VALID_BLOCK (h, data_offset)) {
983 printf ("hivex_value_value: returning EFAULT because data offset is not a valid block (%zu)\n",
990 /* Check that the declared size isn't larger than the block its in. */
991 size_t blen = block_len (h, data_offset, NULL);
994 printf ("hivex_value_value: returning EFAULT because data is longer than its block (%zu, %zu)\n",
1001 char *data = h->addr + data_offset + 4;
1002 memcpy (ret, data, len);
1007 windows_utf16_to_utf8 (/* const */ char *input, size_t len)
1009 iconv_t ic = iconv_open ("UTF-8", "UTF-16");
1010 if (ic == (iconv_t) -1)
1013 /* iconv(3) has an insane interface ... */
1015 /* Mostly UTF-8 will be smaller, so this is a good initial guess. */
1016 size_t outalloc = len;
1020 size_t outlen = outalloc;
1021 char *out = malloc (outlen + 1);
1031 size_t r = iconv (ic, &inp, &inlen, &outp, &outlen);
1032 if (r == (size_t) -1) {
1033 if (errno == E2BIG) {
1034 /* Try again with a larger output buffer. */
1040 /* Else some conversion failure, eg. EILSEQ, EINVAL. */
1056 hivex_value_string (hive_h *h, hive_value_h value)
1060 char *data = hivex_value_value (h, value, &t, &len);
1065 if (t != hive_t_string && t != hive_t_expand_string && t != hive_t_link) {
1071 char *ret = windows_utf16_to_utf8 (data, len);
1080 free_strings (char **argv)
1085 for (i = 0; argv[i] != NULL; ++i)
1091 /* Get the length of a UTF-16 format string. Handle the string as
1092 * pairs of bytes, looking for the first \0\0 pair.
1095 utf16_string_len_in_bytes (const char *str)
1099 while (str[0] || str[1]) {
1107 /* http://blogs.msdn.com/oldnewthing/archive/2009/10/08/9904646.aspx */
1109 hivex_value_multiple_strings (hive_h *h, hive_value_h value)
1113 char *data = hivex_value_value (h, value, &t, &len);
1118 if (t != hive_t_multiple_strings) {
1124 size_t nr_strings = 0;
1125 char **ret = malloc ((1 + nr_strings) * sizeof (char *));
1135 while (p < data + len && (plen = utf16_string_len_in_bytes (p)) > 0) {
1137 char **ret2 = realloc (ret, (1 + nr_strings) * sizeof (char *));
1145 ret[nr_strings-1] = windows_utf16_to_utf8 (p, plen);
1146 ret[nr_strings] = NULL;
1147 if (ret[nr_strings-1] == NULL) {
1153 p += plen + 2 /* skip over UTF-16 \0\0 at the end of this string */;
1161 hivex_value_dword (hive_h *h, hive_value_h value)
1165 char *data = hivex_value_value (h, value, &t, &len);
1170 if ((t != hive_t_dword && t != hive_t_dword_be) || len != 4) {
1176 int32_t ret = *(int32_t*)data;
1178 if (t == hive_t_dword) /* little endian */
1179 ret = le32toh (ret);
1181 ret = be32toh (ret);
1187 hivex_value_qword (hive_h *h, hive_value_h value)
1191 char *data = hivex_value_value (h, value, &t, &len);
1196 if (t != hive_t_qword || len != 8) {
1202 int64_t ret = *(int64_t*)data;
1204 ret = le64toh (ret); /* always little endian */
1210 hivex_visit (hive_h *h, const struct hivex_visitor *visitor, size_t len,
1211 void *opaque, int flags)
1213 return hivex_visit_node (h, hivex_root (h), visitor, len, opaque, flags);
1216 static int hivex__visit_node (hive_h *h, hive_node_h node, const struct hivex_visitor *vtor, char *unvisited, void *opaque, int flags);
1219 hivex_visit_node (hive_h *h, hive_node_h node,
1220 const struct hivex_visitor *visitor, size_t len, void *opaque,
1223 struct hivex_visitor vtor;
1224 memset (&vtor, 0, sizeof vtor);
1226 /* Note that len might be larger *or smaller* than the expected size. */
1227 size_t copysize = len <= sizeof vtor ? len : sizeof vtor;
1228 memcpy (&vtor, visitor, copysize);
1230 /* This bitmap records unvisited nodes, so we don't loop if the
1231 * registry contains cycles.
1233 char *unvisited = malloc (1 + h->size / 32);
1234 if (unvisited == NULL)
1236 memcpy (unvisited, h->bitmap, 1 + h->size / 32);
1238 int r = hivex__visit_node (h, node, &vtor, unvisited, opaque, flags);
1244 hivex__visit_node (hive_h *h, hive_node_h node,
1245 const struct hivex_visitor *vtor, char *unvisited,
1246 void *opaque, int flags)
1248 int skip_bad = flags & HIVEX_VISIT_SKIP_BAD;
1250 hive_value_h *values = NULL;
1251 hive_node_h *children = NULL;
1257 /* Return -1 on all callback errors. However on internal errors,
1258 * check if skip_bad is set and suppress those errors if so.
1262 if (!BITMAP_TST (unvisited, node)) {
1264 printf ("hivex__visit_node: contains cycle: visited node %zu already\n",
1268 return skip_bad ? 0 : -1;
1270 BITMAP_CLR (unvisited, node);
1272 name = hivex_node_name (h, node);
1273 if (!name) return skip_bad ? 0 : -1;
1274 if (vtor->node_start && vtor->node_start (h, opaque, node, name) == -1)
1277 values = hivex_node_values (h, node);
1279 ret = skip_bad ? 0 : -1;
1283 for (i = 0; values[i] != 0; ++i) {
1287 if (hivex_value_type (h, values[i], &t, &len) == -1) {
1288 ret = skip_bad ? 0 : -1;
1292 key = hivex_value_key (h, values[i]);
1294 ret = skip_bad ? 0 : -1;
1300 str = hivex_value_value (h, values[i], &t, &len);
1302 ret = skip_bad ? 0 : -1;
1305 if (t != hive_t_none) {
1306 ret = skip_bad ? 0 : -1;
1309 if (vtor->value_none &&
1310 vtor->value_none (h, opaque, node, values[i], t, len, key, str) == -1)
1312 free (str); str = NULL;
1316 case hive_t_expand_string:
1318 str = hivex_value_string (h, values[i]);
1320 if (errno != EILSEQ && errno != EINVAL) {
1321 ret = skip_bad ? 0 : -1;
1324 if (vtor->value_string_invalid_utf16) {
1325 str = hivex_value_value (h, values[i], &t, &len);
1326 if (vtor->value_string_invalid_utf16 (h, opaque, node, values[i], t, len, key, str) == -1)
1328 free (str); str = NULL;
1332 if (vtor->value_string &&
1333 vtor->value_string (h, opaque, node, values[i], t, len, key, str) == -1)
1335 free (str); str = NULL;
1339 case hive_t_dword_be: {
1340 int32_t i32 = hivex_value_dword (h, values[i]);
1341 if (vtor->value_dword &&
1342 vtor->value_dword (h, opaque, node, values[i], t, len, key, i32) == -1)
1347 case hive_t_qword: {
1348 int64_t i64 = hivex_value_qword (h, values[i]);
1349 if (vtor->value_qword &&
1350 vtor->value_qword (h, opaque, node, values[i], t, len, key, i64) == -1)
1356 str = hivex_value_value (h, values[i], &t, &len);
1358 ret = skip_bad ? 0 : -1;
1361 if (t != hive_t_binary) {
1362 ret = skip_bad ? 0 : -1;
1365 if (vtor->value_binary &&
1366 vtor->value_binary (h, opaque, node, values[i], t, len, key, str) == -1)
1368 free (str); str = NULL;
1371 case hive_t_multiple_strings:
1372 strs = hivex_value_multiple_strings (h, values[i]);
1374 if (errno != EILSEQ && errno != EINVAL) {
1375 ret = skip_bad ? 0 : -1;
1378 if (vtor->value_string_invalid_utf16) {
1379 str = hivex_value_value (h, values[i], &t, &len);
1380 if (vtor->value_string_invalid_utf16 (h, opaque, node, values[i], t, len, key, str) == -1)
1382 free (str); str = NULL;
1386 if (vtor->value_multiple_strings &&
1387 vtor->value_multiple_strings (h, opaque, node, values[i], t, len, key, strs) == -1)
1389 free_strings (strs); strs = NULL;
1392 case hive_t_resource_list:
1393 case hive_t_full_resource_description:
1394 case hive_t_resource_requirements_list:
1396 str = hivex_value_value (h, values[i], &t, &len);
1398 ret = skip_bad ? 0 : -1;
1401 if (vtor->value_other &&
1402 vtor->value_other (h, opaque, node, values[i], t, len, key, str) == -1)
1404 free (str); str = NULL;
1408 free (key); key = NULL;
1411 children = hivex_node_children (h, node);
1412 if (children == NULL) {
1413 ret = skip_bad ? 0 : -1;
1417 for (i = 0; children[i] != 0; ++i) {
1419 printf ("hivex__visit_node: %s: visiting subkey %d (%zu)\n",
1420 name, i, children[i]);
1422 if (hivex__visit_node (h, children[i], vtor, unvisited, opaque, flags) == -1)
1426 if (vtor->node_end && vtor->node_end (h, opaque, node, name) == -1)
1437 free_strings (strs);