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 /* NB. All fields are little endian. */
126 struct ntreg_header {
127 char magic[4]; /* "regf" */
130 char last_modified[8];
131 uint32_t major_ver; /* 1 */
132 uint32_t minor_ver; /* 3 */
133 uint32_t unknown5; /* 0 */
134 uint32_t unknown6; /* 1 */
135 uint32_t offset; /* offset of root key record - 4KB */
136 uint32_t blocks; /* pointer AFTER last hbin in file - 4KB */
137 uint32_t unknown7; /* 1 */
139 char name[64]; /* original file name of hive */
140 char unknown_guid1[16];
141 char unknown_guid2[16];
144 char unknown_guid3[16];
149 uint32_t csum; /* checksum: xor of dwords 0-0x1fb. */
151 char unknown11[3528];
153 char unknown_guid4[16];
154 char unknown_guid5[16];
155 char unknown_guid6[16];
159 } __attribute__((__packed__));
161 struct ntreg_hbin_page {
162 char magic[4]; /* "hbin" */
163 uint32_t offset_first; /* offset from 1st block */
164 uint32_t offset_next; /* offset of next (relative to this) */
166 /* Linked list of blocks follows here. */
167 } __attribute__((__packed__));
169 struct ntreg_hbin_block {
170 int32_t seg_len; /* length of this block (-ve for used block) */
171 char id[2]; /* the block type (eg. "nk" for nk record) */
172 /* Block data follows here. */
173 } __attribute__((__packed__));
175 #define BLOCK_ID_EQ(h,offs,eqid) \
176 (STREQLEN (((struct ntreg_hbin_block *)((h)->addr + (offs)))->id, (eqid), 2))
179 block_len (hive_h *h, size_t blkoff, int *used)
181 struct ntreg_hbin_block *block;
182 block = (struct ntreg_hbin_block *) (h->addr + blkoff);
184 int32_t len = le32toh (block->seg_len);
195 struct ntreg_nk_record {
196 int32_t seg_len; /* length (always -ve because used) */
197 char id[2]; /* "nk" */
200 uint32_t parent; /* offset of owner/parent */
201 uint32_t nr_subkeys; /* number of subkeys */
203 uint32_t subkey_lf; /* lf record containing list of subkeys */
205 uint32_t nr_values; /* number of values */
206 uint32_t vallist; /* value-list record */
207 uint32_t sk; /* offset of sk-record */
208 uint32_t classname; /* offset of classname record */
211 uint16_t name_len; /* length of name */
212 uint16_t classname_len; /* length of classname */
213 char name[1]; /* name follows here */
214 } __attribute__((__packed__));
216 struct ntreg_lf_record {
218 char id[2]; /* "lf" */
219 uint16_t nr_keys; /* number of keys in this record */
221 uint32_t offset; /* offset of nk-record for this subkey */
222 char name[4]; /* first 4 characters of subkey name */
224 } __attribute__((__packed__));
226 struct ntreg_ri_record {
228 char id[2]; /* "ri" */
229 uint16_t nr_offsets; /* number of pointers to lh records */
230 uint32_t offset[1]; /* list of pointers to lh records */
231 } __attribute__((__packed__));
233 /* This has no ID header. */
234 struct ntreg_value_list {
236 uint32_t offset[1]; /* list of pointers to vk records */
237 } __attribute__((__packed__));
239 struct ntreg_vk_record {
240 int32_t seg_len; /* length (always -ve because used) */
241 char id[2]; /* "vk" */
242 uint16_t name_len; /* length of name */
243 /* length of the data:
244 * If data_len is <= 4, then it's stored inline.
245 * If data_len is 0x80000000, then it's an inline dword.
246 * Top bit may be set or not set at random.
249 uint32_t data_offset; /* pointer to the data (or data if inline) */
250 hive_type data_type; /* type of the data */
251 uint16_t flags; /* bit 0 set => key name ASCII,
252 bit 0 clr => key name UTF-16.
253 Only seen ASCII here in the wild. */
255 char name[1]; /* key name follows here */
256 } __attribute__((__packed__));
259 hivex_open (const char *filename, int flags)
263 assert (sizeof (struct ntreg_header) == 0x1000);
264 assert (offsetof (struct ntreg_header, csum) == 0x1fc);
266 h = calloc (1, sizeof *h);
270 h->msglvl = flags & HIVEX_OPEN_MSGLVL_MASK;
272 const char *debug = getenv ("HIVEX_DEBUG");
273 if (debug && STREQ (debug, "1"))
277 fprintf (stderr, "hivex_open: created handle %p\n", h);
279 h->filename = strdup (filename);
280 if (h->filename == NULL)
283 h->fd = open (filename, O_RDONLY);
288 if (fstat (h->fd, &statbuf) == -1)
291 h->size = statbuf.st_size;
293 h->addr = mmap (NULL, h->size, PROT_READ, MAP_SHARED, h->fd, 0);
294 if (h->addr == MAP_FAILED)
298 fprintf (stderr, "hivex_open: mapped file at %p\n", h->addr);
301 if (h->hdr->magic[0] != 'r' ||
302 h->hdr->magic[1] != 'e' ||
303 h->hdr->magic[2] != 'g' ||
304 h->hdr->magic[3] != 'f') {
305 fprintf (stderr, "hivex: %s: not a Windows NT Registry hive file\n",
311 /* Check major version. */
312 uint32_t major_ver = le32toh (h->hdr->major_ver);
313 if (major_ver != 1) {
315 "hivex: %s: hive file major version %" PRIu32 " (expected 1)\n",
316 filename, major_ver);
321 h->bitmap = calloc (1 + h->size / 32, 1);
322 if (h->bitmap == NULL)
325 /* Header checksum. */
326 uint32_t *daddr = (uint32_t *) h->addr;
329 for (i = 0; i < 0x1fc / 4; ++i) {
330 sum ^= le32toh (*daddr);
334 if (sum != le32toh (h->hdr->csum)) {
335 fprintf (stderr, "hivex: %s: bad checksum in hive header\n", filename);
340 if (h->msglvl >= 2) {
341 char *name = windows_utf16_to_utf8 (h->hdr->name, 64);
344 "hivex_open: header fields:\n"
345 " file version %" PRIu32 ".%" PRIu32 "\n"
346 " sequence nos %" PRIu32 " %" PRIu32 "\n"
347 " (sequences nos should match if hive was synched at shutdown)\n"
348 " original file name %s\n"
349 " (only 32 chars are stored, name is probably truncated)\n"
350 " root offset 0x%x + 0x1000\n"
351 " end of last page 0x%x + 0x1000 (total file size 0x%zx)\n"
352 " checksum 0x%x (calculated 0x%x)\n",
353 major_ver, le32toh (h->hdr->minor_ver),
354 le32toh (h->hdr->sequence1), le32toh (h->hdr->sequence2),
355 name ? name : "(conversion failed)",
356 le32toh (h->hdr->offset),
357 le32toh (h->hdr->blocks), h->size,
358 le32toh (h->hdr->csum), sum);
362 h->rootoffs = le32toh (h->hdr->offset) + 0x1000;
363 h->endpages = le32toh (h->hdr->blocks) + 0x1000;
366 fprintf (stderr, "hivex_open: root offset = 0x%zx\n", h->rootoffs);
368 /* We'll set this flag when we see a block with the root offset (ie.
371 int seen_root_block = 0, bad_root_block = 0;
373 /* Collect some stats. */
374 size_t pages = 0; /* Number of hbin pages read. */
375 size_t smallest_page = SIZE_MAX, largest_page = 0;
376 size_t blocks = 0; /* Total number of blocks found. */
377 size_t smallest_block = SIZE_MAX, largest_block = 0, blocks_bytes = 0;
378 size_t used_blocks = 0; /* Total number of used blocks found. */
379 size_t used_size = 0; /* Total size (bytes) of used blocks. */
381 /* Read the pages and blocks. The aim here is to be robust against
382 * corrupt or malicious registries. So we make sure the loops
383 * always make forward progress. We add the address of each block
384 * we read to a hash table so pointers will only reference the start
388 struct ntreg_hbin_page *page;
389 for (off = 0x1000; off < h->size; off += le32toh (page->offset_next)) {
390 if (off >= h->endpages)
393 page = (struct ntreg_hbin_page *) (h->addr + off);
394 if (page->magic[0] != 'h' ||
395 page->magic[1] != 'b' ||
396 page->magic[2] != 'i' ||
397 page->magic[3] != 'n') {
398 fprintf (stderr, "hivex: %s: trailing garbage at end of file (at 0x%zx, after %zu pages)\n",
399 filename, off, pages);
404 size_t page_size = le32toh (page->offset_next);
406 fprintf (stderr, "hivex_open: page at 0x%zx, size %zu\n", off, page_size);
408 if (page_size < smallest_page) smallest_page = page_size;
409 if (page_size > largest_page) largest_page = page_size;
411 if (le32toh (page->offset_next) <= sizeof (struct ntreg_hbin_page) ||
412 (le32toh (page->offset_next) & 3) != 0) {
413 fprintf (stderr, "hivex: %s: pagesize %d at %zu, bad registry\n",
414 filename, le32toh (page->offset_next), off);
419 /* Read the blocks in this page. */
421 struct ntreg_hbin_block *block;
423 for (blkoff = off + 0x20;
424 blkoff < off + le32toh (page->offset_next);
428 int is_root = blkoff == h->rootoffs;
432 block = (struct ntreg_hbin_block *) (h->addr + blkoff);
434 seg_len = block_len (h, blkoff, &used);
435 if (seg_len <= 4 || (seg_len & 3) != 0) {
436 fprintf (stderr, "hivex: %s: block size %d at %zu, bad registry\n",
437 filename, le32toh (block->seg_len), blkoff);
443 fprintf (stderr, "hivex_open: %s block id %d,%d at 0x%zx size %zu%s\n",
444 used ? "used" : "free", block->id[0], block->id[1], blkoff,
445 seg_len, is_root ? " (root)" : "");
447 blocks_bytes += seg_len;
448 if (seg_len < smallest_block) smallest_block = seg_len;
449 if (seg_len > largest_block) largest_block = seg_len;
451 if (is_root && !used)
456 used_size += seg_len;
458 /* Root block must be an nk-block. */
459 if (is_root && (block->id[0] != 'n' || block->id[1] != 'k'))
462 /* Note this blkoff is a valid address. */
463 BITMAP_SET (h->bitmap, blkoff);
468 if (!seen_root_block) {
469 fprintf (stderr, "hivex: %s: no root block found\n", filename);
474 if (bad_root_block) {
475 fprintf (stderr, "hivex: %s: bad root block (free or not nk)\n", filename);
482 "hivex_open: successfully read Windows Registry hive file:\n"
483 " pages: %zu [sml: %zu, lge: %zu]\n"
484 " blocks: %zu [sml: %zu, avg: %zu, lge: %zu]\n"
485 " blocks used: %zu\n"
486 " bytes used: %zu\n",
487 pages, smallest_page, largest_page,
488 blocks, smallest_block, blocks_bytes / blocks, largest_block,
489 used_blocks, used_size);
497 if (h->addr && h->size && h->addr != MAP_FAILED)
498 munmap (h->addr, h->size);
509 hivex_close (hive_h *h)
514 munmap (h->addr, h->size);
523 hivex_root (hive_h *h)
525 hive_node_h ret = h->rootoffs;
526 if (!IS_VALID_BLOCK (h, ret)) {
534 hivex_node_name (hive_h *h, hive_node_h node)
536 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
541 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
543 /* AFAIK the node name is always plain ASCII, so no conversion
544 * to UTF-8 is necessary. However we do need to nul-terminate
548 /* nk->name_len is unsigned, 16 bit, so this is safe ... However
549 * we have to make sure the length doesn't exceed the block length.
551 size_t len = le16toh (nk->name_len);
552 size_t seg_len = block_len (h, node, NULL);
553 if (sizeof (struct ntreg_nk_record) + len - 1 > seg_len) {
555 fprintf (stderr, "hivex_node_name: returning EFAULT because node name is too long (%zu, %zu)\n",
561 char *ret = malloc (len + 1);
564 memcpy (ret, nk->name, len);
570 /* I think the documentation for the sk and classname fields in the nk
571 * record is wrong, or else the offset field is in the wrong place.
572 * Otherwise this makes no sense. Disabled this for now -- it's not
573 * useful for reading the registry anyway.
577 hivex_node_security (hive_h *h, hive_node_h node)
579 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
584 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
586 hive_node_h ret = le32toh (nk->sk);
588 if (!IS_VALID_BLOCK (h, ret)) {
596 hivex_node_classname (hive_h *h, hive_node_h node)
598 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
603 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
605 hive_node_h ret = le32toh (nk->classname);
607 if (!IS_VALID_BLOCK (h, ret)) {
616 hivex_node_children (hive_h *h, hive_node_h node)
618 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
623 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
625 size_t nr_subkeys_in_nk = le32toh (nk->nr_subkeys);
627 /* Deal with the common "no subkeys" case quickly. */
629 if (nr_subkeys_in_nk == 0) {
630 ret = malloc (sizeof (hive_node_h));
637 /* Arbitrarily limit the number of subkeys we will ever deal with. */
638 if (nr_subkeys_in_nk > 1000000) {
643 /* The subkey_lf field can point either to an lf-record, which is
644 * the common case, or if there are lots of subkeys, to an
647 size_t subkey_lf = le32toh (nk->subkey_lf);
649 if (!IS_VALID_BLOCK (h, subkey_lf)) {
651 fprintf (stderr, "hivex_node_children: returning EFAULT because subkey_lf is not a valid block (%zu)\n",
657 struct ntreg_hbin_block *block =
658 (struct ntreg_hbin_block *) (h->addr + subkey_lf);
660 /* Points to lf-record? (Note, also "lh" but that is basically the
661 * same as "lf" as far as we are concerned here).
663 if (block->id[0] == 'l' && (block->id[1] == 'f' || block->id[1] == 'h')) {
664 struct ntreg_lf_record *lf = (struct ntreg_lf_record *) block;
666 /* Check number of subkeys in the nk-record matches number of subkeys
669 size_t nr_subkeys_in_lf = le16toh (lf->nr_keys);
672 fprintf (stderr, "hivex_node_children: nr_subkeys_in_nk = %zu, nr_subkeys_in_lf = %zu\n",
673 nr_subkeys_in_nk, nr_subkeys_in_lf);
675 if (nr_subkeys_in_nk != nr_subkeys_in_lf) {
680 size_t len = block_len (h, subkey_lf, NULL);
681 if (8 + nr_subkeys_in_lf * 8 > len) {
683 fprintf (stderr, "hivex_node_children: returning EFAULT because too many subkeys (%zu, %zu)\n",
684 nr_subkeys_in_lf, len);
689 /* Allocate space for the returned values. Note that
690 * nr_subkeys_in_lf is limited to a 16 bit value.
692 ret = malloc ((1 + nr_subkeys_in_lf) * sizeof (hive_node_h));
697 for (i = 0; i < nr_subkeys_in_lf; ++i) {
698 hive_node_h subkey = lf->keys[i].offset;
700 if (!IS_VALID_BLOCK (h, subkey)) {
702 fprintf (stderr, "hivex_node_children: returning EFAULT because subkey is not a valid block (0x%zx)\n",
713 /* Points to ri-record? */
714 else if (block->id[0] == 'r' && block->id[1] == 'i') {
715 struct ntreg_ri_record *ri = (struct ntreg_ri_record *) block;
717 size_t nr_offsets = le16toh (ri->nr_offsets);
719 /* Count total number of children. */
721 for (i = 0; i < nr_offsets; ++i) {
722 hive_node_h offset = ri->offset[i];
724 if (!IS_VALID_BLOCK (h, offset)) {
726 fprintf (stderr, "hivex_node_children: returning EFAULT because ri-offset is not a valid block (0x%zx)\n",
731 if (!BLOCK_ID_EQ (h, offset, "lf") && !BLOCK_ID_EQ (h, offset, "lh")) {
736 struct ntreg_lf_record *lf =
737 (struct ntreg_lf_record *) (h->addr + offset);
739 count += le16toh (lf->nr_keys);
743 fprintf (stderr, "hivex_node_children: nr_subkeys_in_nk = %zu, counted = %zu\n",
744 nr_subkeys_in_nk, count);
746 if (nr_subkeys_in_nk != count) {
751 /* Copy list of children. Note nr_subkeys_in_nk is limited to
752 * something reasonable above.
754 ret = malloc ((1 + nr_subkeys_in_nk) * sizeof (hive_node_h));
759 for (i = 0; i < nr_offsets; ++i) {
760 hive_node_h offset = ri->offset[i];
762 if (!IS_VALID_BLOCK (h, offset)) {
764 fprintf (stderr, "hivex_node_children: returning EFAULT because ri-offset is not a valid block (0x%zx)\n",
769 if (!BLOCK_ID_EQ (h, offset, "lf") && !BLOCK_ID_EQ (h, offset, "lh")) {
774 struct ntreg_lf_record *lf =
775 (struct ntreg_lf_record *) (h->addr + offset);
778 for (j = 0; j < le16toh (lf->nr_keys); ++j) {
779 hive_node_h subkey = lf->keys[j].offset;
781 if (!IS_VALID_BLOCK (h, subkey)) {
783 fprintf (stderr, "hivex_node_children: returning EFAULT because indirect subkey is not a valid block (0x%zx)\n",
789 ret[count++] = subkey;
802 /* Very inefficient, but at least having a separate API call
803 * allows us to make it more efficient in future.
806 hivex_node_get_child (hive_h *h, hive_node_h node, const char *nname)
808 hive_node_h *children = NULL;
812 children = hivex_node_children (h, node);
813 if (!children) goto error;
816 for (i = 0; children[i] != 0; ++i) {
817 name = hivex_node_name (h, children[i]);
818 if (!name) goto error;
819 if (STRCASEEQ (name, nname)) {
823 free (name); name = NULL;
833 hivex_node_parent (hive_h *h, hive_node_h node)
835 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
840 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
842 hive_node_h ret = le32toh (nk->parent);
844 if (!IS_VALID_BLOCK (h, ret)) {
846 fprintf (stderr, "hivex_node_parent: returning EFAULT because parent is not a valid block (0x%zx)\n",
855 hivex_node_values (hive_h *h, hive_node_h node)
857 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
862 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
864 size_t nr_values = le32toh (nk->nr_values);
867 fprintf (stderr, "hivex_node_values: nr_values = %zu\n", nr_values);
869 /* Deal with the common "no values" case quickly. */
871 if (nr_values == 0) {
872 ret = malloc (sizeof (hive_node_h));
879 /* Arbitrarily limit the number of values we will ever deal with. */
880 if (nr_values > 100000) {
885 /* Get the value list and check it looks reasonable. */
886 size_t vlist_offset = le32toh (nk->vallist);
887 vlist_offset += 0x1000;
888 if (!IS_VALID_BLOCK (h, vlist_offset)) {
890 fprintf (stderr, "hivex_node_values: returning EFAULT because value list is not a valid block (0x%zx)\n",
896 struct ntreg_value_list *vlist =
897 (struct ntreg_value_list *) (h->addr + vlist_offset);
899 size_t len = block_len (h, vlist_offset, NULL);
900 if (4 + nr_values * 4 > len) {
902 fprintf (stderr, "hivex_node_values: returning EFAULT because value list is too long (%zu, %zu)\n",
908 /* Allocate return array and copy values in. */
909 ret = malloc ((1 + nr_values) * sizeof (hive_node_h));
914 for (i = 0; i < nr_values; ++i) {
915 hive_node_h value = vlist->offset[i];
917 if (!IS_VALID_BLOCK (h, value)) {
919 fprintf (stderr, "hivex_node_values: returning EFAULT because value is not a valid block (0x%zx)\n",
932 /* Very inefficient, but at least having a separate API call
933 * allows us to make it more efficient in future.
936 hivex_node_get_value (hive_h *h, hive_node_h node, const char *key)
938 hive_value_h *values = NULL;
940 hive_value_h ret = 0;
942 values = hivex_node_values (h, node);
943 if (!values) goto error;
946 for (i = 0; values[i] != 0; ++i) {
947 name = hivex_value_key (h, values[i]);
948 if (!name) goto error;
949 if (STRCASEEQ (name, key)) {
953 free (name); name = NULL;
963 hivex_value_key (hive_h *h, hive_value_h value)
965 if (!IS_VALID_BLOCK (h, value) || !BLOCK_ID_EQ (h, value, "vk")) {
970 struct ntreg_vk_record *vk = (struct ntreg_vk_record *) (h->addr + value);
972 /* AFAIK the key is always plain ASCII, so no conversion to UTF-8 is
973 * necessary. However we do need to nul-terminate the string.
976 /* vk->name_len is unsigned, 16 bit, so this is safe ... However
977 * we have to make sure the length doesn't exceed the block length.
979 size_t len = le16toh (vk->name_len);
980 size_t seg_len = block_len (h, value, NULL);
981 if (sizeof (struct ntreg_vk_record) + len - 1 > seg_len) {
983 fprintf (stderr, "hivex_value_key: returning EFAULT because key length is too long (%zu, %zu)\n",
989 char *ret = malloc (len + 1);
992 memcpy (ret, vk->name, len);
998 hivex_value_type (hive_h *h, hive_value_h value, hive_type *t, size_t *len)
1000 if (!IS_VALID_BLOCK (h, value) || !BLOCK_ID_EQ (h, value, "vk")) {
1005 struct ntreg_vk_record *vk = (struct ntreg_vk_record *) (h->addr + value);
1008 *t = le32toh (vk->data_type);
1011 *len = le32toh (vk->data_len);
1012 if (*len == 0x80000000) { /* special case */
1014 if (t) *t = hive_t_dword;
1023 hivex_value_value (hive_h *h, hive_value_h value,
1024 hive_type *t_rtn, size_t *len_rtn)
1026 if (!IS_VALID_BLOCK (h, value) || !BLOCK_ID_EQ (h, value, "vk")) {
1031 struct ntreg_vk_record *vk = (struct ntreg_vk_record *) (h->addr + value);
1036 t = le32toh (vk->data_type);
1038 len = le32toh (vk->data_len);
1039 if (len == 0x80000000) { /* special case */
1046 fprintf (stderr, "hivex_value_value: value=0x%zx, t=%d, len=%zu\n",
1054 /* Arbitrarily limit the length that we will read. */
1055 if (len > 1000000) {
1060 char *ret = malloc (len);
1064 /* If length is <= 4 it's always stored inline. */
1066 memcpy (ret, (char *) &vk->data_offset, len);
1070 size_t data_offset = vk->data_offset;
1071 data_offset += 0x1000;
1072 if (!IS_VALID_BLOCK (h, data_offset)) {
1074 fprintf (stderr, "hivex_value_value: returning EFAULT because data offset is not a valid block (0x%zx)\n",
1081 /* Check that the declared size isn't larger than the block its in. */
1082 size_t blen = block_len (h, data_offset, NULL);
1085 fprintf (stderr, "hivex_value_value: returning EFAULT because data is longer than its block (data 0x%zx, data len %zu, block len %zu)\n",
1086 data_offset, len, blen);
1092 char *data = h->addr + data_offset + 4;
1093 memcpy (ret, data, len);
1098 windows_utf16_to_utf8 (/* const */ char *input, size_t len)
1100 iconv_t ic = iconv_open ("UTF-8", "UTF-16");
1101 if (ic == (iconv_t) -1)
1104 /* iconv(3) has an insane interface ... */
1106 /* Mostly UTF-8 will be smaller, so this is a good initial guess. */
1107 size_t outalloc = len;
1111 size_t outlen = outalloc;
1112 char *out = malloc (outlen + 1);
1122 size_t r = iconv (ic, &inp, &inlen, &outp, &outlen);
1123 if (r == (size_t) -1) {
1124 if (errno == E2BIG) {
1125 size_t prev = outalloc;
1126 /* Try again with a larger output buffer. */
1129 if (outalloc < prev)
1134 /* Else some conversion failure, eg. EILSEQ, EINVAL. */
1150 hivex_value_string (hive_h *h, hive_value_h value)
1154 char *data = hivex_value_value (h, value, &t, &len);
1159 if (t != hive_t_string && t != hive_t_expand_string && t != hive_t_link) {
1165 char *ret = windows_utf16_to_utf8 (data, len);
1174 free_strings (char **argv)
1179 for (i = 0; argv[i] != NULL; ++i)
1185 /* Get the length of a UTF-16 format string. Handle the string as
1186 * pairs of bytes, looking for the first \0\0 pair.
1189 utf16_string_len_in_bytes (const char *str)
1193 while (str[0] || str[1]) {
1201 /* http://blogs.msdn.com/oldnewthing/archive/2009/10/08/9904646.aspx */
1203 hivex_value_multiple_strings (hive_h *h, hive_value_h value)
1207 char *data = hivex_value_value (h, value, &t, &len);
1212 if (t != hive_t_multiple_strings) {
1218 size_t nr_strings = 0;
1219 char **ret = malloc ((1 + nr_strings) * sizeof (char *));
1229 while (p < data + len && (plen = utf16_string_len_in_bytes (p)) > 0) {
1231 char **ret2 = realloc (ret, (1 + nr_strings) * sizeof (char *));
1239 ret[nr_strings-1] = windows_utf16_to_utf8 (p, plen);
1240 ret[nr_strings] = NULL;
1241 if (ret[nr_strings-1] == NULL) {
1247 p += plen + 2 /* skip over UTF-16 \0\0 at the end of this string */;
1255 hivex_value_dword (hive_h *h, hive_value_h value)
1259 char *data = hivex_value_value (h, value, &t, &len);
1264 if ((t != hive_t_dword && t != hive_t_dword_be) || len != 4) {
1270 int32_t ret = *(int32_t*)data;
1272 if (t == hive_t_dword) /* little endian */
1273 ret = le32toh (ret);
1275 ret = be32toh (ret);
1281 hivex_value_qword (hive_h *h, hive_value_h value)
1285 char *data = hivex_value_value (h, value, &t, &len);
1290 if (t != hive_t_qword || len != 8) {
1296 int64_t ret = *(int64_t*)data;
1298 ret = le64toh (ret); /* always little endian */
1304 hivex_visit (hive_h *h, const struct hivex_visitor *visitor, size_t len,
1305 void *opaque, int flags)
1307 return hivex_visit_node (h, hivex_root (h), visitor, len, opaque, flags);
1310 static int hivex__visit_node (hive_h *h, hive_node_h node, const struct hivex_visitor *vtor, char *unvisited, void *opaque, int flags);
1313 hivex_visit_node (hive_h *h, hive_node_h node,
1314 const struct hivex_visitor *visitor, size_t len, void *opaque,
1317 struct hivex_visitor vtor;
1318 memset (&vtor, 0, sizeof vtor);
1320 /* Note that len might be larger *or smaller* than the expected size. */
1321 size_t copysize = len <= sizeof vtor ? len : sizeof vtor;
1322 memcpy (&vtor, visitor, copysize);
1324 /* This bitmap records unvisited nodes, so we don't loop if the
1325 * registry contains cycles.
1327 char *unvisited = malloc (1 + h->size / 32);
1328 if (unvisited == NULL)
1330 memcpy (unvisited, h->bitmap, 1 + h->size / 32);
1332 int r = hivex__visit_node (h, node, &vtor, unvisited, opaque, flags);
1338 hivex__visit_node (hive_h *h, hive_node_h node,
1339 const struct hivex_visitor *vtor, char *unvisited,
1340 void *opaque, int flags)
1342 int skip_bad = flags & HIVEX_VISIT_SKIP_BAD;
1344 hive_value_h *values = NULL;
1345 hive_node_h *children = NULL;
1351 /* Return -1 on all callback errors. However on internal errors,
1352 * check if skip_bad is set and suppress those errors if so.
1356 if (!BITMAP_TST (unvisited, node)) {
1358 fprintf (stderr, "hivex__visit_node: contains cycle: visited node 0x%zx already\n",
1362 return skip_bad ? 0 : -1;
1364 BITMAP_CLR (unvisited, node);
1366 name = hivex_node_name (h, node);
1367 if (!name) return skip_bad ? 0 : -1;
1368 if (vtor->node_start && vtor->node_start (h, opaque, node, name) == -1)
1371 values = hivex_node_values (h, node);
1373 ret = skip_bad ? 0 : -1;
1377 for (i = 0; values[i] != 0; ++i) {
1381 if (hivex_value_type (h, values[i], &t, &len) == -1) {
1382 ret = skip_bad ? 0 : -1;
1386 key = hivex_value_key (h, values[i]);
1388 ret = skip_bad ? 0 : -1;
1394 str = hivex_value_value (h, values[i], &t, &len);
1396 ret = skip_bad ? 0 : -1;
1399 if (t != hive_t_none) {
1400 ret = skip_bad ? 0 : -1;
1403 if (vtor->value_none &&
1404 vtor->value_none (h, opaque, node, values[i], t, len, key, str) == -1)
1406 free (str); str = NULL;
1410 case hive_t_expand_string:
1412 str = hivex_value_string (h, values[i]);
1414 if (errno != EILSEQ && errno != EINVAL) {
1415 ret = skip_bad ? 0 : -1;
1418 if (vtor->value_string_invalid_utf16) {
1419 str = hivex_value_value (h, values[i], &t, &len);
1420 if (vtor->value_string_invalid_utf16 (h, opaque, node, values[i], t, len, key, str) == -1)
1422 free (str); str = NULL;
1426 if (vtor->value_string &&
1427 vtor->value_string (h, opaque, node, values[i], t, len, key, str) == -1)
1429 free (str); str = NULL;
1433 case hive_t_dword_be: {
1434 int32_t i32 = hivex_value_dword (h, values[i]);
1435 if (vtor->value_dword &&
1436 vtor->value_dword (h, opaque, node, values[i], t, len, key, i32) == -1)
1441 case hive_t_qword: {
1442 int64_t i64 = hivex_value_qword (h, values[i]);
1443 if (vtor->value_qword &&
1444 vtor->value_qword (h, opaque, node, values[i], t, len, key, i64) == -1)
1450 str = hivex_value_value (h, values[i], &t, &len);
1452 ret = skip_bad ? 0 : -1;
1455 if (t != hive_t_binary) {
1456 ret = skip_bad ? 0 : -1;
1459 if (vtor->value_binary &&
1460 vtor->value_binary (h, opaque, node, values[i], t, len, key, str) == -1)
1462 free (str); str = NULL;
1465 case hive_t_multiple_strings:
1466 strs = hivex_value_multiple_strings (h, values[i]);
1468 if (errno != EILSEQ && errno != EINVAL) {
1469 ret = skip_bad ? 0 : -1;
1472 if (vtor->value_string_invalid_utf16) {
1473 str = hivex_value_value (h, values[i], &t, &len);
1474 if (vtor->value_string_invalid_utf16 (h, opaque, node, values[i], t, len, key, str) == -1)
1476 free (str); str = NULL;
1480 if (vtor->value_multiple_strings &&
1481 vtor->value_multiple_strings (h, opaque, node, values[i], t, len, key, strs) == -1)
1483 free_strings (strs); strs = NULL;
1486 case hive_t_resource_list:
1487 case hive_t_full_resource_description:
1488 case hive_t_resource_requirements_list:
1490 str = hivex_value_value (h, values[i], &t, &len);
1492 ret = skip_bad ? 0 : -1;
1495 if (vtor->value_other &&
1496 vtor->value_other (h, opaque, node, values[i], t, len, key, str) == -1)
1498 free (str); str = NULL;
1502 free (key); key = NULL;
1505 children = hivex_node_children (h, node);
1506 if (children == NULL) {
1507 ret = skip_bad ? 0 : -1;
1511 for (i = 0; children[i] != 0; ++i) {
1513 fprintf (stderr, "hivex__visit_node: %s: visiting subkey %d (0x%zx)\n",
1514 name, i, children[i]);
1516 if (hivex__visit_node (h, children[i], vtor, unvisited, opaque, flags) == -1)
1520 if (vtor->node_end && vtor->node_end (h, opaque, node, name) == -1)
1531 free_strings (strs);