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. */
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 #if 0 /* Doesn't work. */
287 /* Header checksum. */
288 uint32_t *daddr = h->addr;
291 for (i = 0; i < 0x1fc / 4; ++i) {
292 sum += le32toh (*daddr);
295 if (sum != le32toh (h->hdr->csum)) {
296 fprintf (stderr, "hivex: %s: bad checksum in hive header\n", filename);
302 h->rootoffs = le32toh (h->hdr->offset) + 0x1000;
305 fprintf (stderr, "hivex_open: root offset = %zu\n", h->rootoffs);
307 /* We'll set this flag when we see a block with the root offset (ie.
310 int seen_root_block = 0, bad_root_block = 0;
312 /* Read the pages and blocks. The aim here is to be robust against
313 * corrupt or malicious registries. So we make sure the loops
314 * always make forward progress. We add the address of each block
315 * we read to a hash table so pointers will only reference the start
319 struct ntreg_hbin_page *page;
320 for (off = 0x1000; off < h->size; off += le32toh (page->offset_next)) {
323 page = (struct ntreg_hbin_page *) (h->addr + off);
324 if (page->magic[0] != 'h' ||
325 page->magic[1] != 'b' ||
326 page->magic[2] != 'i' ||
327 page->magic[3] != 'n') {
328 /* NB: This error is seemingly common in uncorrupt registry files. */
330 fprintf (stderr, "hivex: %s: ignoring trailing garbage at end of file (at %zu, after %zu pages)\n",
331 filename, off, h->pages);
336 fprintf (stderr, "hivex_open: page at %zu\n", off);
338 if (le32toh (page->offset_next) <= sizeof (struct ntreg_hbin_page) ||
339 (le32toh (page->offset_next) & 3) != 0) {
340 fprintf (stderr, "hivex: %s: pagesize %d at %zu, bad registry\n",
341 filename, le32toh (page->offset_next), off);
346 /* Read the blocks in this page. */
348 struct ntreg_hbin_block *block;
350 for (blkoff = off + 0x20;
351 blkoff < off + le32toh (page->offset_next);
355 int is_root = blkoff == h->rootoffs;
359 block = (struct ntreg_hbin_block *) (h->addr + blkoff);
361 seg_len = block_len (h, blkoff, &used);
362 if (seg_len <= 4 || (seg_len & 3) != 0) {
363 fprintf (stderr, "hivex: %s: block size %d at %zu, bad registry\n",
364 filename, le32toh (block->seg_len), blkoff);
370 fprintf (stderr, "hivex_open: %s block id %d,%d at %zu%s\n",
371 used ? "used" : "free", block->id[0], block->id[1], blkoff,
372 is_root ? " (root)" : "");
374 if (is_root && !used)
379 h->used_size += seg_len;
381 /* Root block must be an nk-block. */
382 if (is_root && (block->id[0] != 'n' || block->id[1] != 'k'))
385 /* Note this blkoff is a valid address. */
386 BITMAP_SET (h->bitmap, blkoff);
391 if (!seen_root_block) {
392 fprintf (stderr, "hivex: %s: no root block found\n", filename);
397 if (bad_root_block) {
398 fprintf (stderr, "hivex: %s: bad root block (free or not nk)\n", filename);
405 "hivex_open: successfully read Windows Registry hive file:\n"
408 " blocks used: %zu\n"
409 " bytes used: %zu\n",
410 h->pages, h->blocks, h->used_blocks, h->used_size);
418 if (h->addr && h->size && h->addr != MAP_FAILED)
419 munmap (h->addr, h->size);
429 hivex_close (hive_h *h)
434 munmap (h->addr, h->size);
442 hivex_root (hive_h *h)
444 hive_node_h ret = h->rootoffs;
445 if (!IS_VALID_BLOCK (h, ret)) {
453 hivex_node_name (hive_h *h, hive_node_h node)
455 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
460 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
462 /* AFAIK the node name is always plain ASCII, so no conversion
463 * to UTF-8 is necessary. However we do need to nul-terminate
467 /* nk->name_len is unsigned, 16 bit, so this is safe ... However
468 * we have to make sure the length doesn't exceed the block length.
470 size_t len = le16toh (nk->name_len);
471 size_t seg_len = block_len (h, node, NULL);
472 if (sizeof (struct ntreg_nk_record) + len - 1 > seg_len) {
474 fprintf (stderr, "hivex_node_name: returning EFAULT because node name is too long (%zu, %zu)\n",
480 char *ret = malloc (len + 1);
483 memcpy (ret, nk->name, len);
489 /* I think the documentation for the sk and classname fields in the nk
490 * record is wrong, or else the offset field is in the wrong place.
491 * Otherwise this makes no sense. Disabled this for now -- it's not
492 * useful for reading the registry anyway.
496 hivex_node_security (hive_h *h, hive_node_h node)
498 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
503 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
505 hive_node_h ret = le32toh (nk->sk);
507 if (!IS_VALID_BLOCK (h, ret)) {
515 hivex_node_classname (hive_h *h, hive_node_h node)
517 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
522 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
524 hive_node_h ret = le32toh (nk->classname);
526 if (!IS_VALID_BLOCK (h, ret)) {
535 hivex_node_children (hive_h *h, hive_node_h node)
537 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
542 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
544 size_t nr_subkeys_in_nk = le32toh (nk->nr_subkeys);
546 /* Deal with the common "no subkeys" case quickly. */
548 if (nr_subkeys_in_nk == 0) {
549 ret = malloc (sizeof (hive_node_h));
556 /* Arbitrarily limit the number of subkeys we will ever deal with. */
557 if (nr_subkeys_in_nk > 1000000) {
562 /* The subkey_lf field can point either to an lf-record, which is
563 * the common case, or if there are lots of subkeys, to an
566 size_t subkey_lf = le32toh (nk->subkey_lf);
568 if (!IS_VALID_BLOCK (h, subkey_lf)) {
570 fprintf (stderr, "hivex_node_children: returning EFAULT because subkey_lf is not a valid block (%zu)\n",
576 struct ntreg_hbin_block *block =
577 (struct ntreg_hbin_block *) (h->addr + subkey_lf);
579 /* Points to lf-record? (Note, also "lh" but that is basically the
580 * same as "lf" as far as we are concerned here).
582 if (block->id[0] == 'l' && (block->id[1] == 'f' || block->id[1] == 'h')) {
583 struct ntreg_lf_record *lf = (struct ntreg_lf_record *) block;
585 /* Check number of subkeys in the nk-record matches number of subkeys
588 size_t nr_subkeys_in_lf = le16toh (lf->nr_keys);
591 fprintf (stderr, "hivex_node_children: nr_subkeys_in_nk = %zu, nr_subkeys_in_lf = %zu\n",
592 nr_subkeys_in_nk, nr_subkeys_in_lf);
594 if (nr_subkeys_in_nk != nr_subkeys_in_lf) {
599 size_t len = block_len (h, subkey_lf, NULL);
600 if (8 + nr_subkeys_in_lf * 8 > len) {
602 fprintf (stderr, "hivex_node_children: returning EFAULT because too many subkeys (%zu, %zu)\n",
603 nr_subkeys_in_lf, len);
608 /* Allocate space for the returned values. Note that
609 * nr_subkeys_in_lf is limited to a 16 bit value.
611 ret = malloc ((1 + nr_subkeys_in_lf) * sizeof (hive_node_h));
616 for (i = 0; i < nr_subkeys_in_lf; ++i) {
617 hive_node_h subkey = lf->keys[i].offset;
619 if (!IS_VALID_BLOCK (h, subkey)) {
621 fprintf (stderr, "hivex_node_children: returning EFAULT because subkey is not a valid block (%zu)\n",
632 /* Points to ri-record? */
633 else if (block->id[0] == 'r' && block->id[1] == 'i') {
634 struct ntreg_ri_record *ri = (struct ntreg_ri_record *) block;
636 size_t nr_offsets = le16toh (ri->nr_offsets);
638 /* Count total number of children. */
640 for (i = 0; i < nr_offsets; ++i) {
641 hive_node_h offset = ri->offset[i];
643 if (!IS_VALID_BLOCK (h, offset)) {
645 fprintf (stderr, "hivex_node_children: returning EFAULT because ri-offset is not a valid block (%zu)\n",
650 if (!BLOCK_ID_EQ (h, offset, "lf") && !BLOCK_ID_EQ (h, offset, "lh")) {
655 struct ntreg_lf_record *lf =
656 (struct ntreg_lf_record *) (h->addr + offset);
658 count += le16toh (lf->nr_keys);
662 fprintf (stderr, "hivex_node_children: nr_subkeys_in_nk = %zu, counted = %zu\n",
663 nr_subkeys_in_nk, count);
665 if (nr_subkeys_in_nk != count) {
670 /* Copy list of children. Note nr_subkeys_in_nk is limited to
671 * something reasonable above.
673 ret = malloc ((1 + nr_subkeys_in_nk) * sizeof (hive_node_h));
678 for (i = 0; i < nr_offsets; ++i) {
679 hive_node_h offset = ri->offset[i];
681 if (!IS_VALID_BLOCK (h, offset)) {
683 fprintf (stderr, "hivex_node_children: returning EFAULT because ri-offset is not a valid block (%zu)\n",
688 if (!BLOCK_ID_EQ (h, offset, "lf") && !BLOCK_ID_EQ (h, offset, "lh")) {
693 struct ntreg_lf_record *lf =
694 (struct ntreg_lf_record *) (h->addr + offset);
697 for (j = 0; j < le16toh (lf->nr_keys); ++j) {
698 hive_node_h subkey = lf->keys[j].offset;
700 if (!IS_VALID_BLOCK (h, subkey)) {
702 fprintf (stderr, "hivex_node_children: returning EFAULT because indirect subkey is not a valid block (%zu)\n",
708 ret[count++] = subkey;
721 /* Very inefficient, but at least having a separate API call
722 * allows us to make it more efficient in future.
725 hivex_node_get_child (hive_h *h, hive_node_h node, const char *nname)
727 hive_node_h *children = NULL;
731 children = hivex_node_children (h, node);
732 if (!children) goto error;
735 for (i = 0; children[i] != 0; ++i) {
736 name = hivex_node_name (h, children[i]);
737 if (!name) goto error;
738 if (STRCASEEQ (name, nname)) {
742 free (name); name = NULL;
752 hivex_node_parent (hive_h *h, hive_node_h node)
754 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
759 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
761 hive_node_h ret = le32toh (nk->parent);
763 if (!IS_VALID_BLOCK (h, ret)) {
765 fprintf (stderr, "hivex_node_parent: returning EFAULT because parent is not a valid block (%zu)\n",
774 hivex_node_values (hive_h *h, hive_node_h node)
776 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
781 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
783 size_t nr_values = le32toh (nk->nr_values);
786 fprintf (stderr, "hivex_node_values: nr_values = %zu\n", nr_values);
788 /* Deal with the common "no values" case quickly. */
790 if (nr_values == 0) {
791 ret = malloc (sizeof (hive_node_h));
798 /* Arbitrarily limit the number of values we will ever deal with. */
799 if (nr_values > 100000) {
804 /* Get the value list and check it looks reasonable. */
805 size_t vlist_offset = le32toh (nk->vallist);
806 vlist_offset += 0x1000;
807 if (!IS_VALID_BLOCK (h, vlist_offset)) {
809 fprintf (stderr, "hivex_node_values: returning EFAULT because value list is not a valid block (%zu)\n",
815 struct ntreg_value_list *vlist =
816 (struct ntreg_value_list *) (h->addr + vlist_offset);
818 size_t len = block_len (h, vlist_offset, NULL);
819 if (4 + nr_values * 4 > len) {
821 fprintf (stderr, "hivex_node_values: returning EFAULT because value list is too long (%zu, %zu)\n",
827 /* Allocate return array and copy values in. */
828 ret = malloc ((1 + nr_values) * sizeof (hive_node_h));
833 for (i = 0; i < nr_values; ++i) {
834 hive_node_h value = vlist->offset[i];
836 if (!IS_VALID_BLOCK (h, value)) {
838 fprintf (stderr, "hivex_node_values: returning EFAULT because value is not a valid block (%zu)\n",
851 /* Very inefficient, but at least having a separate API call
852 * allows us to make it more efficient in future.
855 hivex_node_get_value (hive_h *h, hive_node_h node, const char *key)
857 hive_value_h *values = NULL;
859 hive_value_h ret = 0;
861 values = hivex_node_values (h, node);
862 if (!values) goto error;
865 for (i = 0; values[i] != 0; ++i) {
866 name = hivex_value_key (h, values[i]);
867 if (!name) goto error;
868 if (STRCASEEQ (name, key)) {
872 free (name); name = NULL;
882 hivex_value_key (hive_h *h, hive_value_h value)
884 if (!IS_VALID_BLOCK (h, value) || !BLOCK_ID_EQ (h, value, "vk")) {
889 struct ntreg_vk_record *vk = (struct ntreg_vk_record *) (h->addr + value);
891 /* AFAIK the key is always plain ASCII, so no conversion to UTF-8 is
892 * necessary. However we do need to nul-terminate the string.
895 /* vk->name_len is unsigned, 16 bit, so this is safe ... However
896 * we have to make sure the length doesn't exceed the block length.
898 size_t len = le16toh (vk->name_len);
899 size_t seg_len = block_len (h, value, NULL);
900 if (sizeof (struct ntreg_vk_record) + len - 1 > seg_len) {
902 fprintf (stderr, "hivex_value_key: returning EFAULT because key length is too long (%zu, %zu)\n",
908 char *ret = malloc (len + 1);
911 memcpy (ret, vk->name, len);
917 hivex_value_type (hive_h *h, hive_value_h value, hive_type *t, size_t *len)
919 if (!IS_VALID_BLOCK (h, value) || !BLOCK_ID_EQ (h, value, "vk")) {
924 struct ntreg_vk_record *vk = (struct ntreg_vk_record *) (h->addr + value);
927 *t = le32toh (vk->data_type);
930 *len = le32toh (vk->data_len);
931 if (*len == 0x80000000) { /* special case */
933 if (t) *t = hive_t_dword;
942 hivex_value_value (hive_h *h, hive_value_h value,
943 hive_type *t_rtn, size_t *len_rtn)
945 if (!IS_VALID_BLOCK (h, value) || !BLOCK_ID_EQ (h, value, "vk")) {
950 struct ntreg_vk_record *vk = (struct ntreg_vk_record *) (h->addr + value);
955 t = le32toh (vk->data_type);
957 len = le32toh (vk->data_len);
958 if (len == 0x80000000) { /* special case */
965 fprintf (stderr, "hivex_value_value: value=%zu, t=%d, len=%zu\n",
973 /* Arbitrarily limit the length that we will read. */
979 char *ret = malloc (len);
983 /* If length is <= 4 it's always stored inline. */
985 memcpy (ret, (char *) &vk->data_offset, len);
989 size_t data_offset = vk->data_offset;
990 data_offset += 0x1000;
991 if (!IS_VALID_BLOCK (h, data_offset)) {
993 fprintf (stderr, "hivex_value_value: returning EFAULT because data offset is not a valid block (%zu)\n",
1000 /* Check that the declared size isn't larger than the block its in. */
1001 size_t blen = block_len (h, data_offset, NULL);
1004 fprintf (stderr, "hivex_value_value: returning EFAULT because data is longer than its block (%zu, %zu)\n",
1011 char *data = h->addr + data_offset + 4;
1012 memcpy (ret, data, len);
1017 windows_utf16_to_utf8 (/* const */ char *input, size_t len)
1019 iconv_t ic = iconv_open ("UTF-8", "UTF-16");
1020 if (ic == (iconv_t) -1)
1023 /* iconv(3) has an insane interface ... */
1025 /* Mostly UTF-8 will be smaller, so this is a good initial guess. */
1026 size_t outalloc = len;
1030 size_t outlen = outalloc;
1031 char *out = malloc (outlen + 1);
1041 size_t r = iconv (ic, &inp, &inlen, &outp, &outlen);
1042 if (r == (size_t) -1) {
1043 if (errno == E2BIG) {
1044 size_t prev = outalloc;
1045 /* Try again with a larger output buffer. */
1048 if (outalloc < prev)
1053 /* Else some conversion failure, eg. EILSEQ, EINVAL. */
1069 hivex_value_string (hive_h *h, hive_value_h value)
1073 char *data = hivex_value_value (h, value, &t, &len);
1078 if (t != hive_t_string && t != hive_t_expand_string && t != hive_t_link) {
1084 char *ret = windows_utf16_to_utf8 (data, len);
1093 free_strings (char **argv)
1098 for (i = 0; argv[i] != NULL; ++i)
1104 /* Get the length of a UTF-16 format string. Handle the string as
1105 * pairs of bytes, looking for the first \0\0 pair.
1108 utf16_string_len_in_bytes (const char *str)
1112 while (str[0] || str[1]) {
1120 /* http://blogs.msdn.com/oldnewthing/archive/2009/10/08/9904646.aspx */
1122 hivex_value_multiple_strings (hive_h *h, hive_value_h value)
1126 char *data = hivex_value_value (h, value, &t, &len);
1131 if (t != hive_t_multiple_strings) {
1137 size_t nr_strings = 0;
1138 char **ret = malloc ((1 + nr_strings) * sizeof (char *));
1148 while (p < data + len && (plen = utf16_string_len_in_bytes (p)) > 0) {
1150 char **ret2 = realloc (ret, (1 + nr_strings) * sizeof (char *));
1158 ret[nr_strings-1] = windows_utf16_to_utf8 (p, plen);
1159 ret[nr_strings] = NULL;
1160 if (ret[nr_strings-1] == NULL) {
1166 p += plen + 2 /* skip over UTF-16 \0\0 at the end of this string */;
1174 hivex_value_dword (hive_h *h, hive_value_h value)
1178 char *data = hivex_value_value (h, value, &t, &len);
1183 if ((t != hive_t_dword && t != hive_t_dword_be) || len != 4) {
1189 int32_t ret = *(int32_t*)data;
1191 if (t == hive_t_dword) /* little endian */
1192 ret = le32toh (ret);
1194 ret = be32toh (ret);
1200 hivex_value_qword (hive_h *h, hive_value_h value)
1204 char *data = hivex_value_value (h, value, &t, &len);
1209 if (t != hive_t_qword || len != 8) {
1215 int64_t ret = *(int64_t*)data;
1217 ret = le64toh (ret); /* always little endian */
1223 hivex_visit (hive_h *h, const struct hivex_visitor *visitor, size_t len,
1224 void *opaque, int flags)
1226 return hivex_visit_node (h, hivex_root (h), visitor, len, opaque, flags);
1229 static int hivex__visit_node (hive_h *h, hive_node_h node, const struct hivex_visitor *vtor, char *unvisited, void *opaque, int flags);
1232 hivex_visit_node (hive_h *h, hive_node_h node,
1233 const struct hivex_visitor *visitor, size_t len, void *opaque,
1236 struct hivex_visitor vtor;
1237 memset (&vtor, 0, sizeof vtor);
1239 /* Note that len might be larger *or smaller* than the expected size. */
1240 size_t copysize = len <= sizeof vtor ? len : sizeof vtor;
1241 memcpy (&vtor, visitor, copysize);
1243 /* This bitmap records unvisited nodes, so we don't loop if the
1244 * registry contains cycles.
1246 char *unvisited = malloc (1 + h->size / 32);
1247 if (unvisited == NULL)
1249 memcpy (unvisited, h->bitmap, 1 + h->size / 32);
1251 int r = hivex__visit_node (h, node, &vtor, unvisited, opaque, flags);
1257 hivex__visit_node (hive_h *h, hive_node_h node,
1258 const struct hivex_visitor *vtor, char *unvisited,
1259 void *opaque, int flags)
1261 int skip_bad = flags & HIVEX_VISIT_SKIP_BAD;
1263 hive_value_h *values = NULL;
1264 hive_node_h *children = NULL;
1270 /* Return -1 on all callback errors. However on internal errors,
1271 * check if skip_bad is set and suppress those errors if so.
1275 if (!BITMAP_TST (unvisited, node)) {
1277 fprintf (stderr, "hivex__visit_node: contains cycle: visited node %zu already\n",
1281 return skip_bad ? 0 : -1;
1283 BITMAP_CLR (unvisited, node);
1285 name = hivex_node_name (h, node);
1286 if (!name) return skip_bad ? 0 : -1;
1287 if (vtor->node_start && vtor->node_start (h, opaque, node, name) == -1)
1290 values = hivex_node_values (h, node);
1292 ret = skip_bad ? 0 : -1;
1296 for (i = 0; values[i] != 0; ++i) {
1300 if (hivex_value_type (h, values[i], &t, &len) == -1) {
1301 ret = skip_bad ? 0 : -1;
1305 key = hivex_value_key (h, values[i]);
1307 ret = skip_bad ? 0 : -1;
1313 str = hivex_value_value (h, values[i], &t, &len);
1315 ret = skip_bad ? 0 : -1;
1318 if (t != hive_t_none) {
1319 ret = skip_bad ? 0 : -1;
1322 if (vtor->value_none &&
1323 vtor->value_none (h, opaque, node, values[i], t, len, key, str) == -1)
1325 free (str); str = NULL;
1329 case hive_t_expand_string:
1331 str = hivex_value_string (h, values[i]);
1333 if (errno != EILSEQ && errno != EINVAL) {
1334 ret = skip_bad ? 0 : -1;
1337 if (vtor->value_string_invalid_utf16) {
1338 str = hivex_value_value (h, values[i], &t, &len);
1339 if (vtor->value_string_invalid_utf16 (h, opaque, node, values[i], t, len, key, str) == -1)
1341 free (str); str = NULL;
1345 if (vtor->value_string &&
1346 vtor->value_string (h, opaque, node, values[i], t, len, key, str) == -1)
1348 free (str); str = NULL;
1352 case hive_t_dword_be: {
1353 int32_t i32 = hivex_value_dword (h, values[i]);
1354 if (vtor->value_dword &&
1355 vtor->value_dword (h, opaque, node, values[i], t, len, key, i32) == -1)
1360 case hive_t_qword: {
1361 int64_t i64 = hivex_value_qword (h, values[i]);
1362 if (vtor->value_qword &&
1363 vtor->value_qword (h, opaque, node, values[i], t, len, key, i64) == -1)
1369 str = hivex_value_value (h, values[i], &t, &len);
1371 ret = skip_bad ? 0 : -1;
1374 if (t != hive_t_binary) {
1375 ret = skip_bad ? 0 : -1;
1378 if (vtor->value_binary &&
1379 vtor->value_binary (h, opaque, node, values[i], t, len, key, str) == -1)
1381 free (str); str = NULL;
1384 case hive_t_multiple_strings:
1385 strs = hivex_value_multiple_strings (h, values[i]);
1387 if (errno != EILSEQ && errno != EINVAL) {
1388 ret = skip_bad ? 0 : -1;
1391 if (vtor->value_string_invalid_utf16) {
1392 str = hivex_value_value (h, values[i], &t, &len);
1393 if (vtor->value_string_invalid_utf16 (h, opaque, node, values[i], t, len, key, str) == -1)
1395 free (str); str = NULL;
1399 if (vtor->value_multiple_strings &&
1400 vtor->value_multiple_strings (h, opaque, node, values[i], t, len, key, strs) == -1)
1402 free_strings (strs); strs = NULL;
1405 case hive_t_resource_list:
1406 case hive_t_full_resource_description:
1407 case hive_t_resource_requirements_list:
1409 str = hivex_value_value (h, values[i], &t, &len);
1411 ret = skip_bad ? 0 : -1;
1414 if (vtor->value_other &&
1415 vtor->value_other (h, opaque, node, values[i], t, len, key, str) == -1)
1417 free (str); str = NULL;
1421 free (key); key = NULL;
1424 children = hivex_node_children (h, node);
1425 if (children == NULL) {
1426 ret = skip_bad ? 0 : -1;
1430 for (i = 0; children[i] != 0; ++i) {
1432 fprintf (stderr, "hivex__visit_node: %s: visiting subkey %d (%zu)\n",
1433 name, i, children[i]);
1435 if (hivex__visit_node (h, children[i], vtor, unvisited, opaque, flags) == -1)
1439 if (vtor->node_end && vtor->node_end (h, opaque, node, name) == -1)
1450 free_strings (strs);