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.
36 #ifdef HAVE_BYTESWAP_H
40 #define STREQ(a,b) (strcmp((a),(b)) == 0)
41 #define STRCASEEQ(a,b) (strcasecmp((a),(b)) == 0)
42 //#define STRNEQ(a,b) (strcmp((a),(b)) != 0)
43 //#define STRCASENEQ(a,b) (strcasecmp((a),(b)) != 0)
44 #define STREQLEN(a,b,n) (strncmp((a),(b),(n)) == 0)
45 //#define STRCASEEQLEN(a,b,n) (strncasecmp((a),(b),(n)) == 0)
46 //#define STRNEQLEN(a,b,n) (strncmp((a),(b),(n)) != 0)
47 //#define STRCASENEQLEN(a,b,n) (strncasecmp((a),(b),(n)) != 0)
48 //#define STRPREFIX(a,b) (strncmp((a),(b),strlen((b))) == 0)
50 #if __BYTE_ORDER == __LITTLE_ENDIAN
52 #define be32toh(x) __bswap_32 (x)
55 #define be64toh(x) __bswap_64 (x)
58 #define le16toh(x) (x)
61 #define le32toh(x) (x)
64 #define le64toh(x) (x)
68 #define be32toh(x) (x)
71 #define be64toh(x) (x)
74 #define le16toh(x) __bswap_16 (x)
77 #define le32toh(x) __bswap_32 (x)
80 #define le64toh(x) __bswap_64 (x)
91 /* Memory-mapped (readonly) registry file. */
94 struct ntreg_header *hdr;
97 /* Use a bitmap to store which file offsets are valid (point to a
98 * used block). We only need to store 1 bit per 32 bits of the file
99 * (because blocks are 4-byte aligned). We found that the average
100 * block size in a registry file is ~50 bytes. So roughly 1 in 12
101 * bits in the bitmap will be set, making it likely a more efficient
102 * structure than a hash table.
105 #define BITMAP_SET(bitmap,off) (bitmap[(off)>>5] |= 1 << (((off)>>2)&7))
106 #define BITMAP_CLR(bitmap,off) (bitmap[(off)>>5] &= ~ (1 << (((off)>>2)&7)))
107 #define BITMAP_TST(bitmap,off) (bitmap[(off)>>5] & (1 << (((off)>>2)&7)))
108 #define IS_VALID_BLOCK(h,off) \
109 (((off) & 3) == 0 && \
111 (off) < (h)->size && \
112 BITMAP_TST((h)->bitmap,(off)))
114 /* Fields from the header, extracted from little-endianness hell. */
115 size_t rootoffs; /* Root key offset (always an nk-block). */
118 size_t pages; /* Number of hbin pages read. */
119 size_t blocks; /* Total number of blocks found. */
120 size_t used_blocks; /* Total number of used blocks found. */
121 size_t used_size; /* Total size (bytes) of used blocks. */
124 /* NB. All fields are little endian. */
125 struct ntreg_header {
126 char magic[4]; /* "regf" */
129 char last_modified[8];
130 uint32_t unknown3; /* 1 */
131 uint32_t unknown4; /* 3 */
132 uint32_t unknown5; /* 0 */
133 uint32_t unknown6; /* 1 */
134 uint32_t offset; /* offset of root key record - 4KB */
135 uint32_t blocks; /* size in bytes of data (filesize - 4KB) */
136 uint32_t unknown7; /* 1 */
137 char name[0x1fc-0x2c];
138 uint32_t csum; /* checksum: sum of 32 bit words 0-0x1fb. */
139 } __attribute__((__packed__));
141 struct ntreg_hbin_page {
142 char magic[4]; /* "hbin" */
143 uint32_t offset_first; /* offset from 1st block */
144 uint32_t offset_next; /* offset of next (relative to this) */
146 /* Linked list of blocks follows here. */
147 } __attribute__((__packed__));
149 struct ntreg_hbin_block {
150 int32_t seg_len; /* length of this block (-ve for used block) */
151 char id[2]; /* the block type (eg. "nk" for nk record) */
152 /* Block data follows here. */
153 } __attribute__((__packed__));
155 #define BLOCK_ID_EQ(h,offs,eqid) \
156 (STREQLEN (((struct ntreg_hbin_block *)((h)->addr + (offs)))->id, (eqid), 2))
159 block_len (hive_h *h, size_t blkoff, int *used)
161 struct ntreg_hbin_block *block;
162 block = (struct ntreg_hbin_block *) (h->addr + blkoff);
164 int32_t len = le32toh (block->seg_len);
175 struct ntreg_nk_record {
176 int32_t seg_len; /* length (always -ve because used) */
177 char id[2]; /* "nk" */
180 uint32_t parent; /* offset of owner/parent */
181 uint32_t nr_subkeys; /* number of subkeys */
183 uint32_t subkey_lf; /* lf record containing list of subkeys */
185 uint32_t nr_values; /* number of values */
186 uint32_t vallist; /* value-list record */
187 uint32_t sk; /* offset of sk-record */
188 uint32_t classname; /* offset of classname record */
191 uint16_t name_len; /* length of name */
192 uint16_t classname_len; /* length of classname */
193 char name[1]; /* name follows here */
194 } __attribute__((__packed__));
196 struct ntreg_lf_record {
198 char id[2]; /* "lf" */
199 uint16_t nr_keys; /* number of keys in this record */
201 uint32_t offset; /* offset of nk-record for this subkey */
202 char name[4]; /* first 4 characters of subkey name */
204 } __attribute__((__packed__));
206 struct ntreg_ri_record {
208 char id[2]; /* "ri" */
209 uint16_t nr_offsets; /* number of pointers to lh records */
210 uint32_t offset[1]; /* list of pointers to lh records */
211 } __attribute__((__packed__));
213 /* This has no ID header. */
214 struct ntreg_value_list {
216 uint32_t offset[1]; /* list of pointers to vk records */
217 } __attribute__((__packed__));
219 struct ntreg_vk_record {
220 int32_t seg_len; /* length (always -ve because used) */
221 char id[2]; /* "vk" */
222 uint16_t name_len; /* length of name */
223 /* length of the data:
224 * If data_len is <= 4, then it's stored inline.
225 * If data_len is 0x80000000, then it's an inline dword.
226 * Top bit may be set or not set at random.
229 uint32_t data_offset; /* pointer to the data (or data if inline) */
230 hive_type data_type; /* type of the data */
231 uint16_t unknown1; /* possibly always 1 */
233 char name[1]; /* key name follows here */
234 } __attribute__((__packed__));
237 hivex_open (const char *filename, int flags)
241 h = calloc (1, sizeof *h);
245 h->msglvl = flags & HIVEX_OPEN_MSGLVL_MASK;
247 const char *debug = getenv ("HIVEX_DEBUG");
248 if (debug && STREQ (debug, "1"))
252 fprintf (stderr, "hivex_open: created handle %p\n", h);
254 h->fd = open (filename, O_RDONLY);
259 if (fstat (h->fd, &statbuf) == -1)
262 h->size = statbuf.st_size;
264 h->addr = mmap (NULL, h->size, PROT_READ, MAP_SHARED, h->fd, 0);
265 if (h->addr == MAP_FAILED)
269 fprintf (stderr, "hivex_open: mapped file at %p\n", h->addr);
272 if (h->hdr->magic[0] != 'r' ||
273 h->hdr->magic[1] != 'e' ||
274 h->hdr->magic[2] != 'g' ||
275 h->hdr->magic[3] != 'f') {
276 fprintf (stderr, "hivex: %s: not a Windows NT Registry hive file\n",
282 h->bitmap = calloc (1 + h->size / 32, 1);
283 if (h->bitmap == NULL)
286 /* Header checksum. */
287 uint32_t *daddr = (uint32_t *) h->addr;
290 for (i = 0; i < 0x1fc / 4; ++i) {
291 sum ^= le32toh (*daddr);
295 #if 0 /* Doesn't work. */
296 if (sum != le32toh (h->hdr->csum)) {
297 fprintf (stderr, "hivex: %s: bad checksum in hive header\n", filename);
305 "hivex_open: header fields:\n"
306 " root offset - 4KB 0x%x\n"
307 " blocks (file size - 4KB) 0x%x (real file size 0x%zx)\n"
308 " checksum 0x%x (calculated 0x%x)\n",
309 le32toh (h->hdr->offset),
310 le32toh (h->hdr->blocks), h->size,
311 le32toh (h->hdr->csum), sum);
313 h->rootoffs = le32toh (h->hdr->offset) + 0x1000;
316 fprintf (stderr, "hivex_open: root offset = 0x%zx\n", h->rootoffs);
318 /* We'll set this flag when we see a block with the root offset (ie.
321 int seen_root_block = 0, bad_root_block = 0;
323 /* Read the pages and blocks. The aim here is to be robust against
324 * corrupt or malicious registries. So we make sure the loops
325 * always make forward progress. We add the address of each block
326 * we read to a hash table so pointers will only reference the start
330 struct ntreg_hbin_page *page;
331 for (off = 0x1000; off < h->size; off += le32toh (page->offset_next)) {
334 page = (struct ntreg_hbin_page *) (h->addr + off);
335 if (page->magic[0] != 'h' ||
336 page->magic[1] != 'b' ||
337 page->magic[2] != 'i' ||
338 page->magic[3] != 'n') {
339 /* NB: This error is seemingly common in uncorrupt registry files. */
341 fprintf (stderr, "hivex: %s: ignoring trailing garbage at end of file (at 0x%zx, after %zu pages)\n",
342 filename, off, h->pages);
347 fprintf (stderr, "hivex_open: page at 0x%zx\n", off);
349 if (le32toh (page->offset_next) <= sizeof (struct ntreg_hbin_page) ||
350 (le32toh (page->offset_next) & 3) != 0) {
351 fprintf (stderr, "hivex: %s: pagesize %d at %zu, bad registry\n",
352 filename, le32toh (page->offset_next), off);
357 /* Read the blocks in this page. */
359 struct ntreg_hbin_block *block;
361 for (blkoff = off + 0x20;
362 blkoff < off + le32toh (page->offset_next);
366 int is_root = blkoff == h->rootoffs;
370 block = (struct ntreg_hbin_block *) (h->addr + blkoff);
372 seg_len = block_len (h, blkoff, &used);
373 if (seg_len <= 4 || (seg_len & 3) != 0) {
374 fprintf (stderr, "hivex: %s: block size %d at %zu, bad registry\n",
375 filename, le32toh (block->seg_len), blkoff);
381 fprintf (stderr, "hivex_open: %s block id %d,%d at 0x%zx%s\n",
382 used ? "used" : "free", block->id[0], block->id[1], blkoff,
383 is_root ? " (root)" : "");
385 if (is_root && !used)
390 h->used_size += seg_len;
392 /* Root block must be an nk-block. */
393 if (is_root && (block->id[0] != 'n' || block->id[1] != 'k'))
396 /* Note this blkoff is a valid address. */
397 BITMAP_SET (h->bitmap, blkoff);
402 if (!seen_root_block) {
403 fprintf (stderr, "hivex: %s: no root block found\n", filename);
408 if (bad_root_block) {
409 fprintf (stderr, "hivex: %s: bad root block (free or not nk)\n", filename);
416 "hivex_open: successfully read Windows Registry hive file:\n"
419 " blocks used: %zu\n"
420 " bytes used: %zu\n",
421 h->pages, h->blocks, h->used_blocks, h->used_size);
429 if (h->addr && h->size && h->addr != MAP_FAILED)
430 munmap (h->addr, h->size);
440 hivex_close (hive_h *h)
445 munmap (h->addr, h->size);
453 hivex_root (hive_h *h)
455 hive_node_h ret = h->rootoffs;
456 if (!IS_VALID_BLOCK (h, ret)) {
464 hivex_node_name (hive_h *h, hive_node_h node)
466 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
471 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
473 /* AFAIK the node name is always plain ASCII, so no conversion
474 * to UTF-8 is necessary. However we do need to nul-terminate
478 /* nk->name_len is unsigned, 16 bit, so this is safe ... However
479 * we have to make sure the length doesn't exceed the block length.
481 size_t len = le16toh (nk->name_len);
482 size_t seg_len = block_len (h, node, NULL);
483 if (sizeof (struct ntreg_nk_record) + len - 1 > seg_len) {
485 fprintf (stderr, "hivex_node_name: returning EFAULT because node name is too long (%zu, %zu)\n",
491 char *ret = malloc (len + 1);
494 memcpy (ret, nk->name, len);
500 /* I think the documentation for the sk and classname fields in the nk
501 * record is wrong, or else the offset field is in the wrong place.
502 * Otherwise this makes no sense. Disabled this for now -- it's not
503 * useful for reading the registry anyway.
507 hivex_node_security (hive_h *h, hive_node_h node)
509 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
514 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
516 hive_node_h ret = le32toh (nk->sk);
518 if (!IS_VALID_BLOCK (h, ret)) {
526 hivex_node_classname (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 hive_node_h ret = le32toh (nk->classname);
537 if (!IS_VALID_BLOCK (h, ret)) {
546 hivex_node_children (hive_h *h, hive_node_h node)
548 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
553 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
555 size_t nr_subkeys_in_nk = le32toh (nk->nr_subkeys);
557 /* Deal with the common "no subkeys" case quickly. */
559 if (nr_subkeys_in_nk == 0) {
560 ret = malloc (sizeof (hive_node_h));
567 /* Arbitrarily limit the number of subkeys we will ever deal with. */
568 if (nr_subkeys_in_nk > 1000000) {
573 /* The subkey_lf field can point either to an lf-record, which is
574 * the common case, or if there are lots of subkeys, to an
577 size_t subkey_lf = le32toh (nk->subkey_lf);
579 if (!IS_VALID_BLOCK (h, subkey_lf)) {
581 fprintf (stderr, "hivex_node_children: returning EFAULT because subkey_lf is not a valid block (%zu)\n",
587 struct ntreg_hbin_block *block =
588 (struct ntreg_hbin_block *) (h->addr + subkey_lf);
590 /* Points to lf-record? (Note, also "lh" but that is basically the
591 * same as "lf" as far as we are concerned here).
593 if (block->id[0] == 'l' && (block->id[1] == 'f' || block->id[1] == 'h')) {
594 struct ntreg_lf_record *lf = (struct ntreg_lf_record *) block;
596 /* Check number of subkeys in the nk-record matches number of subkeys
599 size_t nr_subkeys_in_lf = le16toh (lf->nr_keys);
602 fprintf (stderr, "hivex_node_children: nr_subkeys_in_nk = %zu, nr_subkeys_in_lf = %zu\n",
603 nr_subkeys_in_nk, nr_subkeys_in_lf);
605 if (nr_subkeys_in_nk != nr_subkeys_in_lf) {
610 size_t len = block_len (h, subkey_lf, NULL);
611 if (8 + nr_subkeys_in_lf * 8 > len) {
613 fprintf (stderr, "hivex_node_children: returning EFAULT because too many subkeys (%zu, %zu)\n",
614 nr_subkeys_in_lf, len);
619 /* Allocate space for the returned values. Note that
620 * nr_subkeys_in_lf is limited to a 16 bit value.
622 ret = malloc ((1 + nr_subkeys_in_lf) * sizeof (hive_node_h));
627 for (i = 0; i < nr_subkeys_in_lf; ++i) {
628 hive_node_h subkey = lf->keys[i].offset;
630 if (!IS_VALID_BLOCK (h, subkey)) {
632 fprintf (stderr, "hivex_node_children: returning EFAULT because subkey is not a valid block (0x%zx)\n",
643 /* Points to ri-record? */
644 else if (block->id[0] == 'r' && block->id[1] == 'i') {
645 struct ntreg_ri_record *ri = (struct ntreg_ri_record *) block;
647 size_t nr_offsets = le16toh (ri->nr_offsets);
649 /* Count total number of children. */
651 for (i = 0; i < nr_offsets; ++i) {
652 hive_node_h offset = ri->offset[i];
654 if (!IS_VALID_BLOCK (h, offset)) {
656 fprintf (stderr, "hivex_node_children: returning EFAULT because ri-offset is not a valid block (0x%zx)\n",
661 if (!BLOCK_ID_EQ (h, offset, "lf") && !BLOCK_ID_EQ (h, offset, "lh")) {
666 struct ntreg_lf_record *lf =
667 (struct ntreg_lf_record *) (h->addr + offset);
669 count += le16toh (lf->nr_keys);
673 fprintf (stderr, "hivex_node_children: nr_subkeys_in_nk = %zu, counted = %zu\n",
674 nr_subkeys_in_nk, count);
676 if (nr_subkeys_in_nk != count) {
681 /* Copy list of children. Note nr_subkeys_in_nk is limited to
682 * something reasonable above.
684 ret = malloc ((1 + nr_subkeys_in_nk) * sizeof (hive_node_h));
689 for (i = 0; i < nr_offsets; ++i) {
690 hive_node_h offset = ri->offset[i];
692 if (!IS_VALID_BLOCK (h, offset)) {
694 fprintf (stderr, "hivex_node_children: returning EFAULT because ri-offset is not a valid block (0x%zx)\n",
699 if (!BLOCK_ID_EQ (h, offset, "lf") && !BLOCK_ID_EQ (h, offset, "lh")) {
704 struct ntreg_lf_record *lf =
705 (struct ntreg_lf_record *) (h->addr + offset);
708 for (j = 0; j < le16toh (lf->nr_keys); ++j) {
709 hive_node_h subkey = lf->keys[j].offset;
711 if (!IS_VALID_BLOCK (h, subkey)) {
713 fprintf (stderr, "hivex_node_children: returning EFAULT because indirect subkey is not a valid block (0x%zx)\n",
719 ret[count++] = subkey;
732 /* Very inefficient, but at least having a separate API call
733 * allows us to make it more efficient in future.
736 hivex_node_get_child (hive_h *h, hive_node_h node, const char *nname)
738 hive_node_h *children = NULL;
742 children = hivex_node_children (h, node);
743 if (!children) goto error;
746 for (i = 0; children[i] != 0; ++i) {
747 name = hivex_node_name (h, children[i]);
748 if (!name) goto error;
749 if (STRCASEEQ (name, nname)) {
753 free (name); name = NULL;
763 hivex_node_parent (hive_h *h, hive_node_h node)
765 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
770 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
772 hive_node_h ret = le32toh (nk->parent);
774 if (!IS_VALID_BLOCK (h, ret)) {
776 fprintf (stderr, "hivex_node_parent: returning EFAULT because parent is not a valid block (0x%zx)\n",
785 hivex_node_values (hive_h *h, hive_node_h node)
787 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
792 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
794 size_t nr_values = le32toh (nk->nr_values);
797 fprintf (stderr, "hivex_node_values: nr_values = %zu\n", nr_values);
799 /* Deal with the common "no values" case quickly. */
801 if (nr_values == 0) {
802 ret = malloc (sizeof (hive_node_h));
809 /* Arbitrarily limit the number of values we will ever deal with. */
810 if (nr_values > 100000) {
815 /* Get the value list and check it looks reasonable. */
816 size_t vlist_offset = le32toh (nk->vallist);
817 vlist_offset += 0x1000;
818 if (!IS_VALID_BLOCK (h, vlist_offset)) {
820 fprintf (stderr, "hivex_node_values: returning EFAULT because value list is not a valid block (0x%zx)\n",
826 struct ntreg_value_list *vlist =
827 (struct ntreg_value_list *) (h->addr + vlist_offset);
829 size_t len = block_len (h, vlist_offset, NULL);
830 if (4 + nr_values * 4 > len) {
832 fprintf (stderr, "hivex_node_values: returning EFAULT because value list is too long (%zu, %zu)\n",
838 /* Allocate return array and copy values in. */
839 ret = malloc ((1 + nr_values) * sizeof (hive_node_h));
844 for (i = 0; i < nr_values; ++i) {
845 hive_node_h value = vlist->offset[i];
847 if (!IS_VALID_BLOCK (h, value)) {
849 fprintf (stderr, "hivex_node_values: returning EFAULT because value is not a valid block (0x%zx)\n",
862 /* Very inefficient, but at least having a separate API call
863 * allows us to make it more efficient in future.
866 hivex_node_get_value (hive_h *h, hive_node_h node, const char *key)
868 hive_value_h *values = NULL;
870 hive_value_h ret = 0;
872 values = hivex_node_values (h, node);
873 if (!values) goto error;
876 for (i = 0; values[i] != 0; ++i) {
877 name = hivex_value_key (h, values[i]);
878 if (!name) goto error;
879 if (STRCASEEQ (name, key)) {
883 free (name); name = NULL;
893 hivex_value_key (hive_h *h, hive_value_h value)
895 if (!IS_VALID_BLOCK (h, value) || !BLOCK_ID_EQ (h, value, "vk")) {
900 struct ntreg_vk_record *vk = (struct ntreg_vk_record *) (h->addr + value);
902 /* AFAIK the key is always plain ASCII, so no conversion to UTF-8 is
903 * necessary. However we do need to nul-terminate the string.
906 /* vk->name_len is unsigned, 16 bit, so this is safe ... However
907 * we have to make sure the length doesn't exceed the block length.
909 size_t len = le16toh (vk->name_len);
910 size_t seg_len = block_len (h, value, NULL);
911 if (sizeof (struct ntreg_vk_record) + len - 1 > seg_len) {
913 fprintf (stderr, "hivex_value_key: returning EFAULT because key length is too long (%zu, %zu)\n",
919 char *ret = malloc (len + 1);
922 memcpy (ret, vk->name, len);
928 hivex_value_type (hive_h *h, hive_value_h value, hive_type *t, size_t *len)
930 if (!IS_VALID_BLOCK (h, value) || !BLOCK_ID_EQ (h, value, "vk")) {
935 struct ntreg_vk_record *vk = (struct ntreg_vk_record *) (h->addr + value);
938 *t = le32toh (vk->data_type);
941 *len = le32toh (vk->data_len);
942 if (*len == 0x80000000) { /* special case */
944 if (t) *t = hive_t_dword;
953 hivex_value_value (hive_h *h, hive_value_h value,
954 hive_type *t_rtn, size_t *len_rtn)
956 if (!IS_VALID_BLOCK (h, value) || !BLOCK_ID_EQ (h, value, "vk")) {
961 struct ntreg_vk_record *vk = (struct ntreg_vk_record *) (h->addr + value);
966 t = le32toh (vk->data_type);
968 len = le32toh (vk->data_len);
969 if (len == 0x80000000) { /* special case */
976 fprintf (stderr, "hivex_value_value: value=0x%zx, t=%d, len=%zu\n",
984 /* Arbitrarily limit the length that we will read. */
990 char *ret = malloc (len);
994 /* If length is <= 4 it's always stored inline. */
996 memcpy (ret, (char *) &vk->data_offset, len);
1000 size_t data_offset = vk->data_offset;
1001 data_offset += 0x1000;
1002 if (!IS_VALID_BLOCK (h, data_offset)) {
1004 fprintf (stderr, "hivex_value_value: returning EFAULT because data offset is not a valid block (0x%zx)\n",
1011 /* Check that the declared size isn't larger than the block its in. */
1012 size_t blen = block_len (h, data_offset, NULL);
1015 fprintf (stderr, "hivex_value_value: returning EFAULT because data is longer than its block (%zu, %zu)\n",
1022 char *data = h->addr + data_offset + 4;
1023 memcpy (ret, data, len);
1028 windows_utf16_to_utf8 (/* const */ char *input, size_t len)
1030 iconv_t ic = iconv_open ("UTF-8", "UTF-16");
1031 if (ic == (iconv_t) -1)
1034 /* iconv(3) has an insane interface ... */
1036 /* Mostly UTF-8 will be smaller, so this is a good initial guess. */
1037 size_t outalloc = len;
1041 size_t outlen = outalloc;
1042 char *out = malloc (outlen + 1);
1052 size_t r = iconv (ic, &inp, &inlen, &outp, &outlen);
1053 if (r == (size_t) -1) {
1054 if (errno == E2BIG) {
1055 size_t prev = outalloc;
1056 /* Try again with a larger output buffer. */
1059 if (outalloc < prev)
1064 /* Else some conversion failure, eg. EILSEQ, EINVAL. */
1080 hivex_value_string (hive_h *h, hive_value_h value)
1084 char *data = hivex_value_value (h, value, &t, &len);
1089 if (t != hive_t_string && t != hive_t_expand_string && t != hive_t_link) {
1095 char *ret = windows_utf16_to_utf8 (data, len);
1104 free_strings (char **argv)
1109 for (i = 0; argv[i] != NULL; ++i)
1115 /* Get the length of a UTF-16 format string. Handle the string as
1116 * pairs of bytes, looking for the first \0\0 pair.
1119 utf16_string_len_in_bytes (const char *str)
1123 while (str[0] || str[1]) {
1131 /* http://blogs.msdn.com/oldnewthing/archive/2009/10/08/9904646.aspx */
1133 hivex_value_multiple_strings (hive_h *h, hive_value_h value)
1137 char *data = hivex_value_value (h, value, &t, &len);
1142 if (t != hive_t_multiple_strings) {
1148 size_t nr_strings = 0;
1149 char **ret = malloc ((1 + nr_strings) * sizeof (char *));
1159 while (p < data + len && (plen = utf16_string_len_in_bytes (p)) > 0) {
1161 char **ret2 = realloc (ret, (1 + nr_strings) * sizeof (char *));
1169 ret[nr_strings-1] = windows_utf16_to_utf8 (p, plen);
1170 ret[nr_strings] = NULL;
1171 if (ret[nr_strings-1] == NULL) {
1177 p += plen + 2 /* skip over UTF-16 \0\0 at the end of this string */;
1185 hivex_value_dword (hive_h *h, hive_value_h value)
1189 char *data = hivex_value_value (h, value, &t, &len);
1194 if ((t != hive_t_dword && t != hive_t_dword_be) || len != 4) {
1200 int32_t ret = *(int32_t*)data;
1202 if (t == hive_t_dword) /* little endian */
1203 ret = le32toh (ret);
1205 ret = be32toh (ret);
1211 hivex_value_qword (hive_h *h, hive_value_h value)
1215 char *data = hivex_value_value (h, value, &t, &len);
1220 if (t != hive_t_qword || len != 8) {
1226 int64_t ret = *(int64_t*)data;
1228 ret = le64toh (ret); /* always little endian */
1234 hivex_visit (hive_h *h, const struct hivex_visitor *visitor, size_t len,
1235 void *opaque, int flags)
1237 return hivex_visit_node (h, hivex_root (h), visitor, len, opaque, flags);
1240 static int hivex__visit_node (hive_h *h, hive_node_h node, const struct hivex_visitor *vtor, char *unvisited, void *opaque, int flags);
1243 hivex_visit_node (hive_h *h, hive_node_h node,
1244 const struct hivex_visitor *visitor, size_t len, void *opaque,
1247 struct hivex_visitor vtor;
1248 memset (&vtor, 0, sizeof vtor);
1250 /* Note that len might be larger *or smaller* than the expected size. */
1251 size_t copysize = len <= sizeof vtor ? len : sizeof vtor;
1252 memcpy (&vtor, visitor, copysize);
1254 /* This bitmap records unvisited nodes, so we don't loop if the
1255 * registry contains cycles.
1257 char *unvisited = malloc (1 + h->size / 32);
1258 if (unvisited == NULL)
1260 memcpy (unvisited, h->bitmap, 1 + h->size / 32);
1262 int r = hivex__visit_node (h, node, &vtor, unvisited, opaque, flags);
1268 hivex__visit_node (hive_h *h, hive_node_h node,
1269 const struct hivex_visitor *vtor, char *unvisited,
1270 void *opaque, int flags)
1272 int skip_bad = flags & HIVEX_VISIT_SKIP_BAD;
1274 hive_value_h *values = NULL;
1275 hive_node_h *children = NULL;
1281 /* Return -1 on all callback errors. However on internal errors,
1282 * check if skip_bad is set and suppress those errors if so.
1286 if (!BITMAP_TST (unvisited, node)) {
1288 fprintf (stderr, "hivex__visit_node: contains cycle: visited node 0x%zx already\n",
1292 return skip_bad ? 0 : -1;
1294 BITMAP_CLR (unvisited, node);
1296 name = hivex_node_name (h, node);
1297 if (!name) return skip_bad ? 0 : -1;
1298 if (vtor->node_start && vtor->node_start (h, opaque, node, name) == -1)
1301 values = hivex_node_values (h, node);
1303 ret = skip_bad ? 0 : -1;
1307 for (i = 0; values[i] != 0; ++i) {
1311 if (hivex_value_type (h, values[i], &t, &len) == -1) {
1312 ret = skip_bad ? 0 : -1;
1316 key = hivex_value_key (h, values[i]);
1318 ret = skip_bad ? 0 : -1;
1324 str = hivex_value_value (h, values[i], &t, &len);
1326 ret = skip_bad ? 0 : -1;
1329 if (t != hive_t_none) {
1330 ret = skip_bad ? 0 : -1;
1333 if (vtor->value_none &&
1334 vtor->value_none (h, opaque, node, values[i], t, len, key, str) == -1)
1336 free (str); str = NULL;
1340 case hive_t_expand_string:
1342 str = hivex_value_string (h, values[i]);
1344 if (errno != EILSEQ && errno != EINVAL) {
1345 ret = skip_bad ? 0 : -1;
1348 if (vtor->value_string_invalid_utf16) {
1349 str = hivex_value_value (h, values[i], &t, &len);
1350 if (vtor->value_string_invalid_utf16 (h, opaque, node, values[i], t, len, key, str) == -1)
1352 free (str); str = NULL;
1356 if (vtor->value_string &&
1357 vtor->value_string (h, opaque, node, values[i], t, len, key, str) == -1)
1359 free (str); str = NULL;
1363 case hive_t_dword_be: {
1364 int32_t i32 = hivex_value_dword (h, values[i]);
1365 if (vtor->value_dword &&
1366 vtor->value_dword (h, opaque, node, values[i], t, len, key, i32) == -1)
1371 case hive_t_qword: {
1372 int64_t i64 = hivex_value_qword (h, values[i]);
1373 if (vtor->value_qword &&
1374 vtor->value_qword (h, opaque, node, values[i], t, len, key, i64) == -1)
1380 str = hivex_value_value (h, values[i], &t, &len);
1382 ret = skip_bad ? 0 : -1;
1385 if (t != hive_t_binary) {
1386 ret = skip_bad ? 0 : -1;
1389 if (vtor->value_binary &&
1390 vtor->value_binary (h, opaque, node, values[i], t, len, key, str) == -1)
1392 free (str); str = NULL;
1395 case hive_t_multiple_strings:
1396 strs = hivex_value_multiple_strings (h, values[i]);
1398 if (errno != EILSEQ && errno != EINVAL) {
1399 ret = skip_bad ? 0 : -1;
1402 if (vtor->value_string_invalid_utf16) {
1403 str = hivex_value_value (h, values[i], &t, &len);
1404 if (vtor->value_string_invalid_utf16 (h, opaque, node, values[i], t, len, key, str) == -1)
1406 free (str); str = NULL;
1410 if (vtor->value_multiple_strings &&
1411 vtor->value_multiple_strings (h, opaque, node, values[i], t, len, key, strs) == -1)
1413 free_strings (strs); strs = NULL;
1416 case hive_t_resource_list:
1417 case hive_t_full_resource_description:
1418 case hive_t_resource_requirements_list:
1420 str = hivex_value_value (h, values[i], &t, &len);
1422 ret = skip_bad ? 0 : -1;
1425 if (vtor->value_other &&
1426 vtor->value_other (h, opaque, node, values[i], t, len, key, str) == -1)
1428 free (str); str = NULL;
1432 free (key); key = NULL;
1435 children = hivex_node_children (h, node);
1436 if (children == NULL) {
1437 ret = skip_bad ? 0 : -1;
1441 for (i = 0; children[i] != 0; ++i) {
1443 fprintf (stderr, "hivex__visit_node: %s: visiting subkey %d (0x%zx)\n",
1444 name, i, children[i]);
1446 if (hivex__visit_node (h, children[i], vtor, unvisited, opaque, flags) == -1)
1450 if (vtor->node_end && vtor->node_end (h, opaque, node, name) == -1)
1461 free_strings (strs);