1 /* hivex - Windows Registry "hive" extraction library.
2 * Copyright (C) 2009 Red Hat Inc.
3 * Derived from code by Petter Nordahl-Hagen under a compatible license:
4 * Copyright (c) 1997-2007 Petter Nordahl-Hagen.
5 * Derived from code by Markus Stephany under a compatible license:
6 * Copyright (c) 2000-2004, Markus Stephany.
8 * This library is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation;
11 * version 2.1 of the License.
13 * This library is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
18 * See file LICENSE for the full license.
39 #ifdef HAVE_BYTESWAP_H
43 #define STREQ(a,b) (strcmp((a),(b)) == 0)
44 #define STRCASEEQ(a,b) (strcasecmp((a),(b)) == 0)
45 //#define STRNEQ(a,b) (strcmp((a),(b)) != 0)
46 //#define STRCASENEQ(a,b) (strcasecmp((a),(b)) != 0)
47 #define STREQLEN(a,b,n) (strncmp((a),(b),(n)) == 0)
48 //#define STRCASEEQLEN(a,b,n) (strncasecmp((a),(b),(n)) == 0)
49 //#define STRNEQLEN(a,b,n) (strncmp((a),(b),(n)) != 0)
50 //#define STRCASENEQLEN(a,b,n) (strncasecmp((a),(b),(n)) != 0)
51 //#define STRPREFIX(a,b) (strncmp((a),(b),strlen((b))) == 0)
53 #if __BYTE_ORDER == __LITTLE_ENDIAN
55 #define be32toh(x) __bswap_32 (x)
58 #define be64toh(x) __bswap_64 (x)
61 #define le16toh(x) (x)
64 #define le32toh(x) (x)
67 #define le64toh(x) (x)
71 #define be32toh(x) (x)
74 #define be64toh(x) (x)
77 #define le16toh(x) __bswap_16 (x)
80 #define le32toh(x) __bswap_32 (x)
83 #define le64toh(x) __bswap_64 (x)
89 static char *windows_utf16_to_utf8 (/* const */ char *input, size_t len);
97 /* Memory-mapped (readonly) registry file. */
100 struct ntreg_header *hdr;
103 /* Use a bitmap to store which file offsets are valid (point to a
104 * used block). We only need to store 1 bit per 32 bits of the file
105 * (because blocks are 4-byte aligned). We found that the average
106 * block size in a registry file is ~50 bytes. So roughly 1 in 12
107 * bits in the bitmap will be set, making it likely a more efficient
108 * structure than a hash table.
111 #define BITMAP_SET(bitmap,off) (bitmap[(off)>>5] |= 1 << (((off)>>2)&7))
112 #define BITMAP_CLR(bitmap,off) (bitmap[(off)>>5] &= ~ (1 << (((off)>>2)&7)))
113 #define BITMAP_TST(bitmap,off) (bitmap[(off)>>5] & (1 << (((off)>>2)&7)))
114 #define IS_VALID_BLOCK(h,off) \
115 (((off) & 3) == 0 && \
117 (off) < (h)->size && \
118 BITMAP_TST((h)->bitmap,(off)))
120 /* Fields from the header, extracted from little-endianness hell. */
121 size_t rootoffs; /* Root key offset (always an nk-block). */
122 size_t endpages; /* Offset of end of pages. */
125 size_t pages; /* Number of hbin pages read. */
126 size_t blocks; /* Total number of blocks found. */
127 size_t used_blocks; /* Total number of used blocks found. */
128 size_t used_size; /* Total size (bytes) of used blocks. */
131 /* NB. All fields are little endian. */
132 struct ntreg_header {
133 char magic[4]; /* "regf" */
136 char last_modified[8];
137 uint32_t major_ver; /* 1 */
138 uint32_t minor_ver; /* 3 */
139 uint32_t unknown5; /* 0 */
140 uint32_t unknown6; /* 1 */
141 uint32_t offset; /* offset of root key record - 4KB */
142 uint32_t blocks; /* pointer AFTER last hbin in file - 4KB */
143 uint32_t unknown7; /* 1 */
145 char name[64]; /* original file name of hive */
146 char unknown_guid1[16];
147 char unknown_guid2[16];
150 char unknown_guid3[16];
155 uint32_t csum; /* checksum: xor of dwords 0-0x1fb. */
157 char unknown11[3528];
159 char unknown_guid4[16];
160 char unknown_guid5[16];
161 char unknown_guid6[16];
165 } __attribute__((__packed__));
167 struct ntreg_hbin_page {
168 char magic[4]; /* "hbin" */
169 uint32_t offset_first; /* offset from 1st block */
170 uint32_t offset_next; /* offset of next (relative to this) */
172 /* Linked list of blocks follows here. */
173 } __attribute__((__packed__));
175 struct ntreg_hbin_block {
176 int32_t seg_len; /* length of this block (-ve for used block) */
177 char id[2]; /* the block type (eg. "nk" for nk record) */
178 /* Block data follows here. */
179 } __attribute__((__packed__));
181 #define BLOCK_ID_EQ(h,offs,eqid) \
182 (STREQLEN (((struct ntreg_hbin_block *)((h)->addr + (offs)))->id, (eqid), 2))
185 block_len (hive_h *h, size_t blkoff, int *used)
187 struct ntreg_hbin_block *block;
188 block = (struct ntreg_hbin_block *) (h->addr + blkoff);
190 int32_t len = le32toh (block->seg_len);
201 struct ntreg_nk_record {
202 int32_t seg_len; /* length (always -ve because used) */
203 char id[2]; /* "nk" */
206 uint32_t parent; /* offset of owner/parent */
207 uint32_t nr_subkeys; /* number of subkeys */
209 uint32_t subkey_lf; /* lf record containing list of subkeys */
211 uint32_t nr_values; /* number of values */
212 uint32_t vallist; /* value-list record */
213 uint32_t sk; /* offset of sk-record */
214 uint32_t classname; /* offset of classname record */
217 uint16_t name_len; /* length of name */
218 uint16_t classname_len; /* length of classname */
219 char name[1]; /* name follows here */
220 } __attribute__((__packed__));
222 struct ntreg_lf_record {
224 char id[2]; /* "lf" */
225 uint16_t nr_keys; /* number of keys in this record */
227 uint32_t offset; /* offset of nk-record for this subkey */
228 char name[4]; /* first 4 characters of subkey name */
230 } __attribute__((__packed__));
232 struct ntreg_ri_record {
234 char id[2]; /* "ri" */
235 uint16_t nr_offsets; /* number of pointers to lh records */
236 uint32_t offset[1]; /* list of pointers to lh records */
237 } __attribute__((__packed__));
239 /* This has no ID header. */
240 struct ntreg_value_list {
242 uint32_t offset[1]; /* list of pointers to vk records */
243 } __attribute__((__packed__));
245 struct ntreg_vk_record {
246 int32_t seg_len; /* length (always -ve because used) */
247 char id[2]; /* "vk" */
248 uint16_t name_len; /* length of name */
249 /* length of the data:
250 * If data_len is <= 4, then it's stored inline.
251 * If data_len is 0x80000000, then it's an inline dword.
252 * Top bit may be set or not set at random.
255 uint32_t data_offset; /* pointer to the data (or data if inline) */
256 hive_type data_type; /* type of the data */
257 uint16_t flags; /* bit 0 set => key name ASCII,
258 bit 0 clr => key name UTF-16.
259 Only seen ASCII here in the wild. */
261 char name[1]; /* key name follows here */
262 } __attribute__((__packed__));
265 hivex_open (const char *filename, int flags)
269 assert (sizeof (struct ntreg_header) == 0x1000);
270 assert (offsetof (struct ntreg_header, csum) == 0x1fc);
272 h = calloc (1, sizeof *h);
276 h->msglvl = flags & HIVEX_OPEN_MSGLVL_MASK;
278 const char *debug = getenv ("HIVEX_DEBUG");
279 if (debug && STREQ (debug, "1"))
283 fprintf (stderr, "hivex_open: created handle %p\n", h);
285 h->filename = strdup (filename);
286 if (h->filename == NULL)
289 h->fd = open (filename, O_RDONLY);
294 if (fstat (h->fd, &statbuf) == -1)
297 h->size = statbuf.st_size;
299 h->addr = mmap (NULL, h->size, PROT_READ, MAP_SHARED, h->fd, 0);
300 if (h->addr == MAP_FAILED)
304 fprintf (stderr, "hivex_open: mapped file at %p\n", h->addr);
307 if (h->hdr->magic[0] != 'r' ||
308 h->hdr->magic[1] != 'e' ||
309 h->hdr->magic[2] != 'g' ||
310 h->hdr->magic[3] != 'f') {
311 fprintf (stderr, "hivex: %s: not a Windows NT Registry hive file\n",
317 /* Check major version. */
318 uint32_t major_ver = le32toh (h->hdr->major_ver);
319 if (major_ver != 1) {
321 "hivex: %s: hive file major version %" PRIu32 " (expected 1)\n",
322 filename, major_ver);
327 h->bitmap = calloc (1 + h->size / 32, 1);
328 if (h->bitmap == NULL)
331 /* Header checksum. */
332 uint32_t *daddr = (uint32_t *) h->addr;
335 for (i = 0; i < 0x1fc / 4; ++i) {
336 sum ^= le32toh (*daddr);
340 if (sum != le32toh (h->hdr->csum)) {
341 fprintf (stderr, "hivex: %s: bad checksum in hive header\n", filename);
346 if (h->msglvl >= 2) {
347 char *name = windows_utf16_to_utf8 (h->hdr->name, 64);
350 "hivex_open: header fields:\n"
351 " file version %" PRIu32 ".%" PRIu32 "\n"
352 " sequence nos %" PRIu32 " %" PRIu32 "\n"
353 " (sequences nos should match if hive was synched at shutdown)\n"
354 " original file name %s\n"
355 " (only 32 chars are stored, name is probably truncated)\n"
356 " root offset 0x%x + 0x1000\n"
357 " end of last page 0x%x + 0x1000 (total file size 0x%zx)\n"
358 " checksum 0x%x (calculated 0x%x)\n",
359 major_ver, le32toh (h->hdr->minor_ver),
360 le32toh (h->hdr->sequence1), le32toh (h->hdr->sequence2),
361 name ? name : "(conversion failed)",
362 le32toh (h->hdr->offset),
363 le32toh (h->hdr->blocks), h->size,
364 le32toh (h->hdr->csum), sum);
368 h->rootoffs = le32toh (h->hdr->offset) + 0x1000;
369 h->endpages = le32toh (h->hdr->blocks) + 0x1000;
372 fprintf (stderr, "hivex_open: root offset = 0x%zx\n", h->rootoffs);
374 /* We'll set this flag when we see a block with the root offset (ie.
377 int seen_root_block = 0, bad_root_block = 0;
379 /* Read the pages and blocks. The aim here is to be robust against
380 * corrupt or malicious registries. So we make sure the loops
381 * always make forward progress. We add the address of each block
382 * we read to a hash table so pointers will only reference the start
386 struct ntreg_hbin_page *page;
387 for (off = 0x1000; off < h->size; off += le32toh (page->offset_next)) {
388 if (off >= h->endpages)
391 page = (struct ntreg_hbin_page *) (h->addr + off);
392 if (page->magic[0] != 'h' ||
393 page->magic[1] != 'b' ||
394 page->magic[2] != 'i' ||
395 page->magic[3] != 'n') {
396 fprintf (stderr, "hivex: %s: trailing garbage at end of file (at 0x%zx, after %zu pages)\n",
397 filename, off, h->pages);
403 fprintf (stderr, "hivex_open: page at 0x%zx\n", off);
406 if (le32toh (page->offset_next) <= sizeof (struct ntreg_hbin_page) ||
407 (le32toh (page->offset_next) & 3) != 0) {
408 fprintf (stderr, "hivex: %s: pagesize %d at %zu, bad registry\n",
409 filename, le32toh (page->offset_next), off);
414 /* Read the blocks in this page. */
416 struct ntreg_hbin_block *block;
418 for (blkoff = off + 0x20;
419 blkoff < off + le32toh (page->offset_next);
423 int is_root = blkoff == h->rootoffs;
427 block = (struct ntreg_hbin_block *) (h->addr + blkoff);
429 seg_len = block_len (h, blkoff, &used);
430 if (seg_len <= 4 || (seg_len & 3) != 0) {
431 fprintf (stderr, "hivex: %s: block size %d at %zu, bad registry\n",
432 filename, le32toh (block->seg_len), blkoff);
438 fprintf (stderr, "hivex_open: %s block id %d,%d at 0x%zx%s\n",
439 used ? "used" : "free", block->id[0], block->id[1], blkoff,
440 is_root ? " (root)" : "");
442 if (is_root && !used)
447 h->used_size += seg_len;
449 /* Root block must be an nk-block. */
450 if (is_root && (block->id[0] != 'n' || block->id[1] != 'k'))
453 /* Note this blkoff is a valid address. */
454 BITMAP_SET (h->bitmap, blkoff);
459 if (!seen_root_block) {
460 fprintf (stderr, "hivex: %s: no root block found\n", filename);
465 if (bad_root_block) {
466 fprintf (stderr, "hivex: %s: bad root block (free or not nk)\n", filename);
473 "hivex_open: successfully read Windows Registry hive file:\n"
476 " blocks used: %zu\n"
477 " bytes used: %zu\n",
478 h->pages, h->blocks, h->used_blocks, h->used_size);
486 if (h->addr && h->size && h->addr != MAP_FAILED)
487 munmap (h->addr, h->size);
498 hivex_close (hive_h *h)
503 munmap (h->addr, h->size);
512 hivex_root (hive_h *h)
514 hive_node_h ret = h->rootoffs;
515 if (!IS_VALID_BLOCK (h, ret)) {
523 hivex_node_name (hive_h *h, hive_node_h node)
525 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
530 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
532 /* AFAIK the node name is always plain ASCII, so no conversion
533 * to UTF-8 is necessary. However we do need to nul-terminate
537 /* nk->name_len is unsigned, 16 bit, so this is safe ... However
538 * we have to make sure the length doesn't exceed the block length.
540 size_t len = le16toh (nk->name_len);
541 size_t seg_len = block_len (h, node, NULL);
542 if (sizeof (struct ntreg_nk_record) + len - 1 > seg_len) {
544 fprintf (stderr, "hivex_node_name: returning EFAULT because node name is too long (%zu, %zu)\n",
550 char *ret = malloc (len + 1);
553 memcpy (ret, nk->name, len);
559 /* I think the documentation for the sk and classname fields in the nk
560 * record is wrong, or else the offset field is in the wrong place.
561 * Otherwise this makes no sense. Disabled this for now -- it's not
562 * useful for reading the registry anyway.
566 hivex_node_security (hive_h *h, hive_node_h node)
568 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
573 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
575 hive_node_h ret = le32toh (nk->sk);
577 if (!IS_VALID_BLOCK (h, ret)) {
585 hivex_node_classname (hive_h *h, hive_node_h node)
587 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
592 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
594 hive_node_h ret = le32toh (nk->classname);
596 if (!IS_VALID_BLOCK (h, ret)) {
605 hivex_node_children (hive_h *h, hive_node_h node)
607 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
612 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
614 size_t nr_subkeys_in_nk = le32toh (nk->nr_subkeys);
616 /* Deal with the common "no subkeys" case quickly. */
618 if (nr_subkeys_in_nk == 0) {
619 ret = malloc (sizeof (hive_node_h));
626 /* Arbitrarily limit the number of subkeys we will ever deal with. */
627 if (nr_subkeys_in_nk > 1000000) {
632 /* The subkey_lf field can point either to an lf-record, which is
633 * the common case, or if there are lots of subkeys, to an
636 size_t subkey_lf = le32toh (nk->subkey_lf);
638 if (!IS_VALID_BLOCK (h, subkey_lf)) {
640 fprintf (stderr, "hivex_node_children: returning EFAULT because subkey_lf is not a valid block (%zu)\n",
646 struct ntreg_hbin_block *block =
647 (struct ntreg_hbin_block *) (h->addr + subkey_lf);
649 /* Points to lf-record? (Note, also "lh" but that is basically the
650 * same as "lf" as far as we are concerned here).
652 if (block->id[0] == 'l' && (block->id[1] == 'f' || block->id[1] == 'h')) {
653 struct ntreg_lf_record *lf = (struct ntreg_lf_record *) block;
655 /* Check number of subkeys in the nk-record matches number of subkeys
658 size_t nr_subkeys_in_lf = le16toh (lf->nr_keys);
661 fprintf (stderr, "hivex_node_children: nr_subkeys_in_nk = %zu, nr_subkeys_in_lf = %zu\n",
662 nr_subkeys_in_nk, nr_subkeys_in_lf);
664 if (nr_subkeys_in_nk != nr_subkeys_in_lf) {
669 size_t len = block_len (h, subkey_lf, NULL);
670 if (8 + nr_subkeys_in_lf * 8 > len) {
672 fprintf (stderr, "hivex_node_children: returning EFAULT because too many subkeys (%zu, %zu)\n",
673 nr_subkeys_in_lf, len);
678 /* Allocate space for the returned values. Note that
679 * nr_subkeys_in_lf is limited to a 16 bit value.
681 ret = malloc ((1 + nr_subkeys_in_lf) * sizeof (hive_node_h));
686 for (i = 0; i < nr_subkeys_in_lf; ++i) {
687 hive_node_h subkey = lf->keys[i].offset;
689 if (!IS_VALID_BLOCK (h, subkey)) {
691 fprintf (stderr, "hivex_node_children: returning EFAULT because subkey is not a valid block (0x%zx)\n",
702 /* Points to ri-record? */
703 else if (block->id[0] == 'r' && block->id[1] == 'i') {
704 struct ntreg_ri_record *ri = (struct ntreg_ri_record *) block;
706 size_t nr_offsets = le16toh (ri->nr_offsets);
708 /* Count total number of children. */
710 for (i = 0; i < nr_offsets; ++i) {
711 hive_node_h offset = ri->offset[i];
713 if (!IS_VALID_BLOCK (h, offset)) {
715 fprintf (stderr, "hivex_node_children: returning EFAULT because ri-offset is not a valid block (0x%zx)\n",
720 if (!BLOCK_ID_EQ (h, offset, "lf") && !BLOCK_ID_EQ (h, offset, "lh")) {
725 struct ntreg_lf_record *lf =
726 (struct ntreg_lf_record *) (h->addr + offset);
728 count += le16toh (lf->nr_keys);
732 fprintf (stderr, "hivex_node_children: nr_subkeys_in_nk = %zu, counted = %zu\n",
733 nr_subkeys_in_nk, count);
735 if (nr_subkeys_in_nk != count) {
740 /* Copy list of children. Note nr_subkeys_in_nk is limited to
741 * something reasonable above.
743 ret = malloc ((1 + nr_subkeys_in_nk) * sizeof (hive_node_h));
748 for (i = 0; i < nr_offsets; ++i) {
749 hive_node_h offset = ri->offset[i];
751 if (!IS_VALID_BLOCK (h, offset)) {
753 fprintf (stderr, "hivex_node_children: returning EFAULT because ri-offset is not a valid block (0x%zx)\n",
758 if (!BLOCK_ID_EQ (h, offset, "lf") && !BLOCK_ID_EQ (h, offset, "lh")) {
763 struct ntreg_lf_record *lf =
764 (struct ntreg_lf_record *) (h->addr + offset);
767 for (j = 0; j < le16toh (lf->nr_keys); ++j) {
768 hive_node_h subkey = lf->keys[j].offset;
770 if (!IS_VALID_BLOCK (h, subkey)) {
772 fprintf (stderr, "hivex_node_children: returning EFAULT because indirect subkey is not a valid block (0x%zx)\n",
778 ret[count++] = subkey;
791 /* Very inefficient, but at least having a separate API call
792 * allows us to make it more efficient in future.
795 hivex_node_get_child (hive_h *h, hive_node_h node, const char *nname)
797 hive_node_h *children = NULL;
801 children = hivex_node_children (h, node);
802 if (!children) goto error;
805 for (i = 0; children[i] != 0; ++i) {
806 name = hivex_node_name (h, children[i]);
807 if (!name) goto error;
808 if (STRCASEEQ (name, nname)) {
812 free (name); name = NULL;
822 hivex_node_parent (hive_h *h, hive_node_h node)
824 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
829 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
831 hive_node_h ret = le32toh (nk->parent);
833 if (!IS_VALID_BLOCK (h, ret)) {
835 fprintf (stderr, "hivex_node_parent: returning EFAULT because parent is not a valid block (0x%zx)\n",
844 hivex_node_values (hive_h *h, hive_node_h node)
846 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
851 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
853 size_t nr_values = le32toh (nk->nr_values);
856 fprintf (stderr, "hivex_node_values: nr_values = %zu\n", nr_values);
858 /* Deal with the common "no values" case quickly. */
860 if (nr_values == 0) {
861 ret = malloc (sizeof (hive_node_h));
868 /* Arbitrarily limit the number of values we will ever deal with. */
869 if (nr_values > 100000) {
874 /* Get the value list and check it looks reasonable. */
875 size_t vlist_offset = le32toh (nk->vallist);
876 vlist_offset += 0x1000;
877 if (!IS_VALID_BLOCK (h, vlist_offset)) {
879 fprintf (stderr, "hivex_node_values: returning EFAULT because value list is not a valid block (0x%zx)\n",
885 struct ntreg_value_list *vlist =
886 (struct ntreg_value_list *) (h->addr + vlist_offset);
888 size_t len = block_len (h, vlist_offset, NULL);
889 if (4 + nr_values * 4 > len) {
891 fprintf (stderr, "hivex_node_values: returning EFAULT because value list is too long (%zu, %zu)\n",
897 /* Allocate return array and copy values in. */
898 ret = malloc ((1 + nr_values) * sizeof (hive_node_h));
903 for (i = 0; i < nr_values; ++i) {
904 hive_node_h value = vlist->offset[i];
906 if (!IS_VALID_BLOCK (h, value)) {
908 fprintf (stderr, "hivex_node_values: returning EFAULT because value is not a valid block (0x%zx)\n",
921 /* Very inefficient, but at least having a separate API call
922 * allows us to make it more efficient in future.
925 hivex_node_get_value (hive_h *h, hive_node_h node, const char *key)
927 hive_value_h *values = NULL;
929 hive_value_h ret = 0;
931 values = hivex_node_values (h, node);
932 if (!values) goto error;
935 for (i = 0; values[i] != 0; ++i) {
936 name = hivex_value_key (h, values[i]);
937 if (!name) goto error;
938 if (STRCASEEQ (name, key)) {
942 free (name); name = NULL;
952 hivex_value_key (hive_h *h, hive_value_h value)
954 if (!IS_VALID_BLOCK (h, value) || !BLOCK_ID_EQ (h, value, "vk")) {
959 struct ntreg_vk_record *vk = (struct ntreg_vk_record *) (h->addr + value);
961 /* AFAIK the key is always plain ASCII, so no conversion to UTF-8 is
962 * necessary. However we do need to nul-terminate the string.
965 /* vk->name_len is unsigned, 16 bit, so this is safe ... However
966 * we have to make sure the length doesn't exceed the block length.
968 size_t len = le16toh (vk->name_len);
969 size_t seg_len = block_len (h, value, NULL);
970 if (sizeof (struct ntreg_vk_record) + len - 1 > seg_len) {
972 fprintf (stderr, "hivex_value_key: returning EFAULT because key length is too long (%zu, %zu)\n",
978 char *ret = malloc (len + 1);
981 memcpy (ret, vk->name, len);
987 hivex_value_type (hive_h *h, hive_value_h value, hive_type *t, size_t *len)
989 if (!IS_VALID_BLOCK (h, value) || !BLOCK_ID_EQ (h, value, "vk")) {
994 struct ntreg_vk_record *vk = (struct ntreg_vk_record *) (h->addr + value);
997 *t = le32toh (vk->data_type);
1000 *len = le32toh (vk->data_len);
1001 if (*len == 0x80000000) { /* special case */
1003 if (t) *t = hive_t_dword;
1012 hivex_value_value (hive_h *h, hive_value_h value,
1013 hive_type *t_rtn, size_t *len_rtn)
1015 if (!IS_VALID_BLOCK (h, value) || !BLOCK_ID_EQ (h, value, "vk")) {
1020 struct ntreg_vk_record *vk = (struct ntreg_vk_record *) (h->addr + value);
1025 t = le32toh (vk->data_type);
1027 len = le32toh (vk->data_len);
1028 if (len == 0x80000000) { /* special case */
1035 fprintf (stderr, "hivex_value_value: value=0x%zx, t=%d, len=%zu\n",
1043 /* Arbitrarily limit the length that we will read. */
1044 if (len > 1000000) {
1049 char *ret = malloc (len);
1053 /* If length is <= 4 it's always stored inline. */
1055 memcpy (ret, (char *) &vk->data_offset, len);
1059 size_t data_offset = vk->data_offset;
1060 data_offset += 0x1000;
1061 if (!IS_VALID_BLOCK (h, data_offset)) {
1063 fprintf (stderr, "hivex_value_value: returning EFAULT because data offset is not a valid block (0x%zx)\n",
1070 /* Check that the declared size isn't larger than the block its in. */
1071 size_t blen = block_len (h, data_offset, NULL);
1074 fprintf (stderr, "hivex_value_value: returning EFAULT because data is longer than its block (data 0x%zx, data len %zu, block len %zu)\n",
1075 data_offset, len, blen);
1081 char *data = h->addr + data_offset + 4;
1082 memcpy (ret, data, len);
1087 windows_utf16_to_utf8 (/* const */ char *input, size_t len)
1089 iconv_t ic = iconv_open ("UTF-8", "UTF-16");
1090 if (ic == (iconv_t) -1)
1093 /* iconv(3) has an insane interface ... */
1095 /* Mostly UTF-8 will be smaller, so this is a good initial guess. */
1096 size_t outalloc = len;
1100 size_t outlen = outalloc;
1101 char *out = malloc (outlen + 1);
1111 size_t r = iconv (ic, &inp, &inlen, &outp, &outlen);
1112 if (r == (size_t) -1) {
1113 if (errno == E2BIG) {
1114 size_t prev = outalloc;
1115 /* Try again with a larger output buffer. */
1118 if (outalloc < prev)
1123 /* Else some conversion failure, eg. EILSEQ, EINVAL. */
1139 hivex_value_string (hive_h *h, hive_value_h value)
1143 char *data = hivex_value_value (h, value, &t, &len);
1148 if (t != hive_t_string && t != hive_t_expand_string && t != hive_t_link) {
1154 char *ret = windows_utf16_to_utf8 (data, len);
1163 free_strings (char **argv)
1168 for (i = 0; argv[i] != NULL; ++i)
1174 /* Get the length of a UTF-16 format string. Handle the string as
1175 * pairs of bytes, looking for the first \0\0 pair.
1178 utf16_string_len_in_bytes (const char *str)
1182 while (str[0] || str[1]) {
1190 /* http://blogs.msdn.com/oldnewthing/archive/2009/10/08/9904646.aspx */
1192 hivex_value_multiple_strings (hive_h *h, hive_value_h value)
1196 char *data = hivex_value_value (h, value, &t, &len);
1201 if (t != hive_t_multiple_strings) {
1207 size_t nr_strings = 0;
1208 char **ret = malloc ((1 + nr_strings) * sizeof (char *));
1218 while (p < data + len && (plen = utf16_string_len_in_bytes (p)) > 0) {
1220 char **ret2 = realloc (ret, (1 + nr_strings) * sizeof (char *));
1228 ret[nr_strings-1] = windows_utf16_to_utf8 (p, plen);
1229 ret[nr_strings] = NULL;
1230 if (ret[nr_strings-1] == NULL) {
1236 p += plen + 2 /* skip over UTF-16 \0\0 at the end of this string */;
1244 hivex_value_dword (hive_h *h, hive_value_h value)
1248 char *data = hivex_value_value (h, value, &t, &len);
1253 if ((t != hive_t_dword && t != hive_t_dword_be) || len != 4) {
1259 int32_t ret = *(int32_t*)data;
1261 if (t == hive_t_dword) /* little endian */
1262 ret = le32toh (ret);
1264 ret = be32toh (ret);
1270 hivex_value_qword (hive_h *h, hive_value_h value)
1274 char *data = hivex_value_value (h, value, &t, &len);
1279 if (t != hive_t_qword || len != 8) {
1285 int64_t ret = *(int64_t*)data;
1287 ret = le64toh (ret); /* always little endian */
1293 hivex_visit (hive_h *h, const struct hivex_visitor *visitor, size_t len,
1294 void *opaque, int flags)
1296 return hivex_visit_node (h, hivex_root (h), visitor, len, opaque, flags);
1299 static int hivex__visit_node (hive_h *h, hive_node_h node, const struct hivex_visitor *vtor, char *unvisited, void *opaque, int flags);
1302 hivex_visit_node (hive_h *h, hive_node_h node,
1303 const struct hivex_visitor *visitor, size_t len, void *opaque,
1306 struct hivex_visitor vtor;
1307 memset (&vtor, 0, sizeof vtor);
1309 /* Note that len might be larger *or smaller* than the expected size. */
1310 size_t copysize = len <= sizeof vtor ? len : sizeof vtor;
1311 memcpy (&vtor, visitor, copysize);
1313 /* This bitmap records unvisited nodes, so we don't loop if the
1314 * registry contains cycles.
1316 char *unvisited = malloc (1 + h->size / 32);
1317 if (unvisited == NULL)
1319 memcpy (unvisited, h->bitmap, 1 + h->size / 32);
1321 int r = hivex__visit_node (h, node, &vtor, unvisited, opaque, flags);
1327 hivex__visit_node (hive_h *h, hive_node_h node,
1328 const struct hivex_visitor *vtor, char *unvisited,
1329 void *opaque, int flags)
1331 int skip_bad = flags & HIVEX_VISIT_SKIP_BAD;
1333 hive_value_h *values = NULL;
1334 hive_node_h *children = NULL;
1340 /* Return -1 on all callback errors. However on internal errors,
1341 * check if skip_bad is set and suppress those errors if so.
1345 if (!BITMAP_TST (unvisited, node)) {
1347 fprintf (stderr, "hivex__visit_node: contains cycle: visited node 0x%zx already\n",
1351 return skip_bad ? 0 : -1;
1353 BITMAP_CLR (unvisited, node);
1355 name = hivex_node_name (h, node);
1356 if (!name) return skip_bad ? 0 : -1;
1357 if (vtor->node_start && vtor->node_start (h, opaque, node, name) == -1)
1360 values = hivex_node_values (h, node);
1362 ret = skip_bad ? 0 : -1;
1366 for (i = 0; values[i] != 0; ++i) {
1370 if (hivex_value_type (h, values[i], &t, &len) == -1) {
1371 ret = skip_bad ? 0 : -1;
1375 key = hivex_value_key (h, values[i]);
1377 ret = skip_bad ? 0 : -1;
1383 str = hivex_value_value (h, values[i], &t, &len);
1385 ret = skip_bad ? 0 : -1;
1388 if (t != hive_t_none) {
1389 ret = skip_bad ? 0 : -1;
1392 if (vtor->value_none &&
1393 vtor->value_none (h, opaque, node, values[i], t, len, key, str) == -1)
1395 free (str); str = NULL;
1399 case hive_t_expand_string:
1401 str = hivex_value_string (h, values[i]);
1403 if (errno != EILSEQ && errno != EINVAL) {
1404 ret = skip_bad ? 0 : -1;
1407 if (vtor->value_string_invalid_utf16) {
1408 str = hivex_value_value (h, values[i], &t, &len);
1409 if (vtor->value_string_invalid_utf16 (h, opaque, node, values[i], t, len, key, str) == -1)
1411 free (str); str = NULL;
1415 if (vtor->value_string &&
1416 vtor->value_string (h, opaque, node, values[i], t, len, key, str) == -1)
1418 free (str); str = NULL;
1422 case hive_t_dword_be: {
1423 int32_t i32 = hivex_value_dword (h, values[i]);
1424 if (vtor->value_dword &&
1425 vtor->value_dword (h, opaque, node, values[i], t, len, key, i32) == -1)
1430 case hive_t_qword: {
1431 int64_t i64 = hivex_value_qword (h, values[i]);
1432 if (vtor->value_qword &&
1433 vtor->value_qword (h, opaque, node, values[i], t, len, key, i64) == -1)
1439 str = hivex_value_value (h, values[i], &t, &len);
1441 ret = skip_bad ? 0 : -1;
1444 if (t != hive_t_binary) {
1445 ret = skip_bad ? 0 : -1;
1448 if (vtor->value_binary &&
1449 vtor->value_binary (h, opaque, node, values[i], t, len, key, str) == -1)
1451 free (str); str = NULL;
1454 case hive_t_multiple_strings:
1455 strs = hivex_value_multiple_strings (h, values[i]);
1457 if (errno != EILSEQ && errno != EINVAL) {
1458 ret = skip_bad ? 0 : -1;
1461 if (vtor->value_string_invalid_utf16) {
1462 str = hivex_value_value (h, values[i], &t, &len);
1463 if (vtor->value_string_invalid_utf16 (h, opaque, node, values[i], t, len, key, str) == -1)
1465 free (str); str = NULL;
1469 if (vtor->value_multiple_strings &&
1470 vtor->value_multiple_strings (h, opaque, node, values[i], t, len, key, strs) == -1)
1472 free_strings (strs); strs = NULL;
1475 case hive_t_resource_list:
1476 case hive_t_full_resource_description:
1477 case hive_t_resource_requirements_list:
1479 str = hivex_value_value (h, values[i], &t, &len);
1481 ret = skip_bad ? 0 : -1;
1484 if (vtor->value_other &&
1485 vtor->value_other (h, opaque, node, values[i], t, len, key, str) == -1)
1487 free (str); str = NULL;
1491 free (key); key = NULL;
1494 children = hivex_node_children (h, node);
1495 if (children == NULL) {
1496 ret = skip_bad ? 0 : -1;
1500 for (i = 0; children[i] != 0; ++i) {
1502 fprintf (stderr, "hivex__visit_node: %s: visiting subkey %d (0x%zx)\n",
1503 name, i, children[i]);
1505 if (hivex__visit_node (h, children[i], vtor, unvisited, opaque, flags) == -1)
1509 if (vtor->node_end && vtor->node_end (h, opaque, node, name) == -1)
1520 free_strings (strs);