1 /* hivex - Windows Registry "hive" extraction library.
2 * Copyright (C) 2009-2010 Red Hat Inc.
3 * Derived from code by Petter Nordahl-Hagen under a compatible license:
4 * Copyright (c) 1997-2007 Petter Nordahl-Hagen.
5 * Derived from code by Markus Stephany under a compatible license:
6 * Copyright (c) 2000-2004, Markus Stephany.
8 * This library is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation;
11 * version 2.1 of the License.
13 * This library is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
18 * See file LICENSE for the full license.
37 #define STREQ(a,b) (strcmp((a),(b)) == 0)
38 #define STRCASEEQ(a,b) (strcasecmp((a),(b)) == 0)
39 //#define STRNEQ(a,b) (strcmp((a),(b)) != 0)
40 //#define STRCASENEQ(a,b) (strcasecmp((a),(b)) != 0)
41 #define STREQLEN(a,b,n) (strncmp((a),(b),(n)) == 0)
42 //#define STRCASEEQLEN(a,b,n) (strncasecmp((a),(b),(n)) == 0)
43 //#define STRNEQLEN(a,b,n) (strncmp((a),(b),(n)) != 0)
44 //#define STRCASENEQLEN(a,b,n) (strncasecmp((a),(b),(n)) != 0)
45 //#define STRPREFIX(a,b) (strncmp((a),(b),strlen((b))) == 0)
48 #include "byte_conversions.h"
50 static char *windows_utf16_to_utf8 (/* const */ char *input, size_t len);
58 /* Memory-mapped (readonly) registry file. */
61 struct ntreg_header *hdr;
64 /* Use a bitmap to store which file offsets are valid (point to a
65 * used block). We only need to store 1 bit per 32 bits of the file
66 * (because blocks are 4-byte aligned). We found that the average
67 * block size in a registry file is ~50 bytes. So roughly 1 in 12
68 * bits in the bitmap will be set, making it likely a more efficient
69 * structure than a hash table.
72 #define BITMAP_SET(bitmap,off) (bitmap[(off)>>5] |= 1 << (((off)>>2)&7))
73 #define BITMAP_CLR(bitmap,off) (bitmap[(off)>>5] &= ~ (1 << (((off)>>2)&7)))
74 #define BITMAP_TST(bitmap,off) (bitmap[(off)>>5] & (1 << (((off)>>2)&7)))
75 #define IS_VALID_BLOCK(h,off) \
76 (((off) & 3) == 0 && \
78 (off) < (h)->size && \
79 BITMAP_TST((h)->bitmap,(off)))
81 /* Fields from the header, extracted from little-endianness hell. */
82 size_t rootoffs; /* Root key offset (always an nk-block). */
83 size_t endpages; /* Offset of end of pages. */
86 /* NB. All fields are little endian. */
88 char magic[4]; /* "regf" */
91 char last_modified[8];
92 uint32_t major_ver; /* 1 */
93 uint32_t minor_ver; /* 3 */
94 uint32_t unknown5; /* 0 */
95 uint32_t unknown6; /* 1 */
96 uint32_t offset; /* offset of root key record - 4KB */
97 uint32_t blocks; /* pointer AFTER last hbin in file - 4KB */
98 uint32_t unknown7; /* 1 */
100 char name[64]; /* original file name of hive */
101 char unknown_guid1[16];
102 char unknown_guid2[16];
105 char unknown_guid3[16];
110 uint32_t csum; /* checksum: xor of dwords 0-0x1fb. */
112 char unknown11[3528];
114 char unknown_guid4[16];
115 char unknown_guid5[16];
116 char unknown_guid6[16];
120 } __attribute__((__packed__));
122 struct ntreg_hbin_page {
123 char magic[4]; /* "hbin" */
124 uint32_t offset_first; /* offset from 1st block */
125 uint32_t page_size; /* size of this page (multiple of 4KB) */
127 /* Linked list of blocks follows here. */
128 } __attribute__((__packed__));
130 struct ntreg_hbin_block {
131 int32_t seg_len; /* length of this block (-ve for used block) */
132 char id[2]; /* the block type (eg. "nk" for nk record) */
133 /* Block data follows here. */
134 } __attribute__((__packed__));
136 #define BLOCK_ID_EQ(h,offs,eqid) \
137 (STREQLEN (((struct ntreg_hbin_block *)((h)->addr + (offs)))->id, (eqid), 2))
140 block_len (hive_h *h, size_t blkoff, int *used)
142 struct ntreg_hbin_block *block;
143 block = (struct ntreg_hbin_block *) (h->addr + blkoff);
145 int32_t len = le32toh (block->seg_len);
156 struct ntreg_nk_record {
157 int32_t seg_len; /* length (always -ve because used) */
158 char id[2]; /* "nk" */
162 uint32_t parent; /* offset of owner/parent */
163 uint32_t nr_subkeys; /* number of subkeys */
164 uint32_t nr_subkeys_volatile;
165 uint32_t subkey_lf; /* lf record containing list of subkeys */
166 uint32_t subkey_lf_volatile;
167 uint32_t nr_values; /* number of values */
168 uint32_t vallist; /* value-list record */
169 uint32_t sk; /* offset of sk-record */
170 uint32_t classname; /* offset of classname record */
171 uint16_t max_subkey_name_len; /* maximum length of a subkey name in bytes
172 if the subkey was reencoded as UTF-16LE */
175 uint32_t max_vk_name_len; /* maximum length of any vk name in bytes
176 if the name was reencoded as UTF-16LE */
177 uint32_t max_vk_data_len; /* maximum length of any vk data in bytes */
179 uint16_t name_len; /* length of name */
180 uint16_t classname_len; /* length of classname */
181 char name[1]; /* name follows here */
182 } __attribute__((__packed__));
184 struct ntreg_lf_record {
186 char id[2]; /* "lf" */
187 uint16_t nr_keys; /* number of keys in this record */
189 uint32_t offset; /* offset of nk-record for this subkey */
190 char hash[4]; /* hash of subkey name */
192 } __attribute__((__packed__));
194 struct ntreg_ri_record {
196 char id[2]; /* "ri" */
197 uint16_t nr_offsets; /* number of pointers to lh records */
198 uint32_t offset[1]; /* list of pointers to lh records */
199 } __attribute__((__packed__));
201 /* This has no ID header. */
202 struct ntreg_value_list {
204 uint32_t offset[1]; /* list of pointers to vk records */
205 } __attribute__((__packed__));
207 struct ntreg_vk_record {
208 int32_t seg_len; /* length (always -ve because used) */
209 char id[2]; /* "vk" */
210 uint16_t name_len; /* length of name */
211 /* length of the data:
212 * If data_len is <= 4, then it's stored inline.
213 * If data_len is 0x80000000, then it's an inline dword.
214 * Top bit may be set or not set at random.
217 uint32_t data_offset; /* pointer to the data (or data if inline) */
218 uint32_t data_type; /* type of the data */
219 uint16_t flags; /* bit 0 set => key name ASCII,
220 bit 0 clr => key name UTF-16.
221 Only seen ASCII here in the wild. */
223 char name[1]; /* key name follows here */
224 } __attribute__((__packed__));
227 header_checksum (const hive_h *h)
229 uint32_t *daddr = (uint32_t *) h->addr;
233 for (i = 0; i < 0x1fc / 4; ++i) {
234 sum ^= le32toh (*daddr);
242 hivex_open (const char *filename, int flags)
246 assert (sizeof (struct ntreg_header) == 0x1000);
247 assert (offsetof (struct ntreg_header, csum) == 0x1fc);
249 h = calloc (1, sizeof *h);
253 h->msglvl = flags & HIVEX_OPEN_MSGLVL_MASK;
255 const char *debug = getenv ("HIVEX_DEBUG");
256 if (debug && STREQ (debug, "1"))
260 fprintf (stderr, "hivex_open: created handle %p\n", h);
262 h->filename = strdup (filename);
263 if (h->filename == NULL)
266 h->fd = open (filename, O_RDONLY);
271 if (fstat (h->fd, &statbuf) == -1)
274 h->size = statbuf.st_size;
276 h->addr = mmap (NULL, h->size, PROT_READ, MAP_SHARED, h->fd, 0);
277 if (h->addr == MAP_FAILED)
281 fprintf (stderr, "hivex_open: mapped file at %p\n", h->addr);
284 if (h->hdr->magic[0] != 'r' ||
285 h->hdr->magic[1] != 'e' ||
286 h->hdr->magic[2] != 'g' ||
287 h->hdr->magic[3] != 'f') {
288 fprintf (stderr, "hivex: %s: not a Windows NT Registry hive file\n",
294 /* Check major version. */
295 uint32_t major_ver = le32toh (h->hdr->major_ver);
296 if (major_ver != 1) {
298 "hivex: %s: hive file major version %" PRIu32 " (expected 1)\n",
299 filename, major_ver);
304 h->bitmap = calloc (1 + h->size / 32, 1);
305 if (h->bitmap == NULL)
308 /* Header checksum. */
309 uint32_t sum = header_checksum (h);
310 if (sum != le32toh (h->hdr->csum)) {
311 fprintf (stderr, "hivex: %s: bad checksum in hive header\n", filename);
316 if (h->msglvl >= 2) {
317 char *name = windows_utf16_to_utf8 (h->hdr->name, 64);
320 "hivex_open: header fields:\n"
321 " file version %" PRIu32 ".%" PRIu32 "\n"
322 " sequence nos %" PRIu32 " %" PRIu32 "\n"
323 " (sequences nos should match if hive was synched at shutdown)\n"
324 " original file name %s\n"
325 " (only 32 chars are stored, name is probably truncated)\n"
326 " root offset 0x%x + 0x1000\n"
327 " end of last page 0x%x + 0x1000 (total file size 0x%zx)\n"
328 " checksum 0x%x (calculated 0x%x)\n",
329 major_ver, le32toh (h->hdr->minor_ver),
330 le32toh (h->hdr->sequence1), le32toh (h->hdr->sequence2),
331 name ? name : "(conversion failed)",
332 le32toh (h->hdr->offset),
333 le32toh (h->hdr->blocks), h->size,
334 le32toh (h->hdr->csum), sum);
338 h->rootoffs = le32toh (h->hdr->offset) + 0x1000;
339 h->endpages = le32toh (h->hdr->blocks) + 0x1000;
342 fprintf (stderr, "hivex_open: root offset = 0x%zx\n", h->rootoffs);
344 /* We'll set this flag when we see a block with the root offset (ie.
347 int seen_root_block = 0, bad_root_block = 0;
349 /* Collect some stats. */
350 size_t pages = 0; /* Number of hbin pages read. */
351 size_t smallest_page = SIZE_MAX, largest_page = 0;
352 size_t blocks = 0; /* Total number of blocks found. */
353 size_t smallest_block = SIZE_MAX, largest_block = 0, blocks_bytes = 0;
354 size_t used_blocks = 0; /* Total number of used blocks found. */
355 size_t used_size = 0; /* Total size (bytes) of used blocks. */
357 /* Read the pages and blocks. The aim here is to be robust against
358 * corrupt or malicious registries. So we make sure the loops
359 * always make forward progress. We add the address of each block
360 * we read to a hash table so pointers will only reference the start
364 struct ntreg_hbin_page *page;
365 for (off = 0x1000; off < h->size; off += le32toh (page->page_size)) {
366 if (off >= h->endpages)
369 page = (struct ntreg_hbin_page *) (h->addr + off);
370 if (page->magic[0] != 'h' ||
371 page->magic[1] != 'b' ||
372 page->magic[2] != 'i' ||
373 page->magic[3] != 'n') {
374 fprintf (stderr, "hivex: %s: trailing garbage at end of file (at 0x%zx, after %zu pages)\n",
375 filename, off, pages);
380 size_t page_size = le32toh (page->page_size);
382 fprintf (stderr, "hivex_open: page at 0x%zx, size %zu\n", off, page_size);
384 if (page_size < smallest_page) smallest_page = page_size;
385 if (page_size > largest_page) largest_page = page_size;
387 if (page_size <= sizeof (struct ntreg_hbin_page) ||
388 (page_size & 0x0fff) != 0) {
389 fprintf (stderr, "hivex: %s: page size %zu at 0x%zx, bad registry\n",
390 filename, page_size, off);
395 /* Read the blocks in this page. */
397 struct ntreg_hbin_block *block;
399 for (blkoff = off + 0x20;
400 blkoff < off + page_size;
404 int is_root = blkoff == h->rootoffs;
408 block = (struct ntreg_hbin_block *) (h->addr + blkoff);
410 seg_len = block_len (h, blkoff, &used);
411 if (seg_len <= 4 || (seg_len & 3) != 0) {
412 fprintf (stderr, "hivex: %s: block size %" PRIu32 " at 0x%zx, bad registry\n",
413 filename, le32toh (block->seg_len), blkoff);
419 fprintf (stderr, "hivex_open: %s block id %d,%d at 0x%zx size %zu%s\n",
420 used ? "used" : "free", block->id[0], block->id[1], blkoff,
421 seg_len, is_root ? " (root)" : "");
423 blocks_bytes += seg_len;
424 if (seg_len < smallest_block) smallest_block = seg_len;
425 if (seg_len > largest_block) largest_block = seg_len;
427 if (is_root && !used)
432 used_size += seg_len;
434 /* Root block must be an nk-block. */
435 if (is_root && (block->id[0] != 'n' || block->id[1] != 'k'))
438 /* Note this blkoff is a valid address. */
439 BITMAP_SET (h->bitmap, blkoff);
444 if (!seen_root_block) {
445 fprintf (stderr, "hivex: %s: no root block found\n", filename);
450 if (bad_root_block) {
451 fprintf (stderr, "hivex: %s: bad root block (free or not nk)\n", filename);
458 "hivex_open: successfully read Windows Registry hive file:\n"
459 " pages: %zu [sml: %zu, lge: %zu]\n"
460 " blocks: %zu [sml: %zu, avg: %zu, lge: %zu]\n"
461 " blocks used: %zu\n"
462 " bytes used: %zu\n",
463 pages, smallest_page, largest_page,
464 blocks, smallest_block, blocks_bytes / blocks, largest_block,
465 used_blocks, used_size);
473 if (h->addr && h->size && h->addr != MAP_FAILED)
474 munmap (h->addr, h->size);
485 hivex_close (hive_h *h)
490 munmap (h->addr, h->size);
499 hivex_root (hive_h *h)
501 hive_node_h ret = h->rootoffs;
502 if (!IS_VALID_BLOCK (h, ret)) {
510 hivex_node_name (hive_h *h, hive_node_h node)
512 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
517 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
519 /* AFAIK the node name is always plain ASCII, so no conversion
520 * to UTF-8 is necessary. However we do need to nul-terminate
524 /* nk->name_len is unsigned, 16 bit, so this is safe ... However
525 * we have to make sure the length doesn't exceed the block length.
527 size_t len = le16toh (nk->name_len);
528 size_t seg_len = block_len (h, node, NULL);
529 if (sizeof (struct ntreg_nk_record) + len - 1 > seg_len) {
531 fprintf (stderr, "hivex_node_name: returning EFAULT because node name is too long (%zu, %zu)\n",
537 char *ret = malloc (len + 1);
540 memcpy (ret, nk->name, len);
546 /* I think the documentation for the sk and classname fields in the nk
547 * record is wrong, or else the offset field is in the wrong place.
548 * Otherwise this makes no sense. Disabled this for now -- it's not
549 * useful for reading the registry anyway.
553 hivex_node_security (hive_h *h, hive_node_h node)
555 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
560 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
562 hive_node_h ret = le32toh (nk->sk);
564 if (!IS_VALID_BLOCK (h, ret)) {
572 hivex_node_classname (hive_h *h, hive_node_h node)
574 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
579 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
581 hive_node_h ret = le32toh (nk->classname);
583 if (!IS_VALID_BLOCK (h, ret)) {
591 /* Structure for returning 0-terminated lists of offsets (nodes,
601 init_offset_list (struct offset_list *list)
605 list->offsets = NULL;
608 #define INIT_OFFSET_LIST(name) \
609 struct offset_list name; \
610 init_offset_list (&name)
612 /* Preallocates the offset_list, but doesn't make the contents longer. */
614 grow_offset_list (struct offset_list *list, size_t alloc)
616 assert (alloc >= list->len);
617 size_t *p = realloc (list->offsets, alloc * sizeof (size_t));
626 add_to_offset_list (struct offset_list *list, size_t offset)
628 if (list->len >= list->alloc) {
629 if (grow_offset_list (list, list->alloc ? list->alloc * 2 : 4) == -1)
632 list->offsets[list->len] = offset;
638 free_offset_list (struct offset_list *list)
640 free (list->offsets);
644 return_offset_list (struct offset_list *list)
646 if (add_to_offset_list (list, 0) == -1)
648 return list->offsets; /* caller frees */
651 /* Iterate over children, returning child nodes and intermediate blocks. */
653 get_children (hive_h *h, hive_node_h node,
654 hive_node_h **children_ret, size_t **blocks_ret)
656 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
661 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
663 size_t nr_subkeys_in_nk = le32toh (nk->nr_subkeys);
665 INIT_OFFSET_LIST (children);
666 INIT_OFFSET_LIST (blocks);
668 /* Deal with the common "no subkeys" case quickly. */
669 if (nr_subkeys_in_nk == 0)
672 /* Arbitrarily limit the number of subkeys we will ever deal with. */
673 if (nr_subkeys_in_nk > 1000000) {
678 /* Preallocate space for the children. */
679 if (grow_offset_list (&children, nr_subkeys_in_nk) == -1)
682 /* The subkey_lf field can point either to an lf-record, which is
683 * the common case, or if there are lots of subkeys, to an
686 size_t subkey_lf = le32toh (nk->subkey_lf);
688 if (!IS_VALID_BLOCK (h, subkey_lf)) {
690 fprintf (stderr, "hivex_node_children: returning EFAULT because subkey_lf is not a valid block (%zu)\n",
696 if (add_to_offset_list (&blocks, subkey_lf) == -1)
699 struct ntreg_hbin_block *block =
700 (struct ntreg_hbin_block *) (h->addr + subkey_lf);
702 /* Points to lf-record? (Note, also "lh" but that is basically the
703 * same as "lf" as far as we are concerned here).
705 if (block->id[0] == 'l' && (block->id[1] == 'f' || block->id[1] == 'h')) {
706 struct ntreg_lf_record *lf = (struct ntreg_lf_record *) block;
708 /* Check number of subkeys in the nk-record matches number of subkeys
711 size_t nr_subkeys_in_lf = le16toh (lf->nr_keys);
714 fprintf (stderr, "hivex_node_children: nr_subkeys_in_nk = %zu, nr_subkeys_in_lf = %zu\n",
715 nr_subkeys_in_nk, nr_subkeys_in_lf);
717 if (nr_subkeys_in_nk != nr_subkeys_in_lf) {
722 size_t len = block_len (h, subkey_lf, NULL);
723 if (8 + nr_subkeys_in_lf * 8 > len) {
725 fprintf (stderr, "hivex_node_children: returning EFAULT because too many subkeys (%zu, %zu)\n",
726 nr_subkeys_in_lf, len);
732 for (i = 0; i < nr_subkeys_in_lf; ++i) {
733 hive_node_h subkey = le32toh (lf->keys[i].offset);
735 if (!IS_VALID_BLOCK (h, subkey)) {
737 fprintf (stderr, "hivex_node_children: returning EFAULT because subkey is not a valid block (0x%zx)\n",
742 if (add_to_offset_list (&children, subkey) == -1)
747 /* Points to ri-record? */
748 else if (block->id[0] == 'r' && block->id[1] == 'i') {
749 struct ntreg_ri_record *ri = (struct ntreg_ri_record *) block;
751 size_t nr_offsets = le16toh (ri->nr_offsets);
753 /* Count total number of children. */
755 for (i = 0; i < nr_offsets; ++i) {
756 hive_node_h offset = ri->offset[i];
758 if (!IS_VALID_BLOCK (h, offset)) {
760 fprintf (stderr, "hivex_node_children: returning EFAULT because ri-offset is not a valid block (0x%zx)\n",
765 if (!BLOCK_ID_EQ (h, offset, "lf") && !BLOCK_ID_EQ (h, offset, "lh")) {
770 if (add_to_offset_list (&blocks, offset) == -1)
773 struct ntreg_lf_record *lf =
774 (struct ntreg_lf_record *) (h->addr + offset);
776 count += le16toh (lf->nr_keys);
780 fprintf (stderr, "hivex_node_children: nr_subkeys_in_nk = %zu, counted = %zu\n",
781 nr_subkeys_in_nk, count);
783 if (nr_subkeys_in_nk != count) {
788 /* Copy list of children. Note nr_subkeys_in_nk is limited to
789 * something reasonable above.
791 for (i = 0; i < nr_offsets; ++i) {
792 hive_node_h offset = ri->offset[i];
794 if (!IS_VALID_BLOCK (h, offset)) {
796 fprintf (stderr, "hivex_node_children: returning EFAULT because ri-offset is not a valid block (0x%zx)\n",
801 if (!BLOCK_ID_EQ (h, offset, "lf") && !BLOCK_ID_EQ (h, offset, "lh")) {
806 struct ntreg_lf_record *lf =
807 (struct ntreg_lf_record *) (h->addr + offset);
810 for (j = 0; j < le16toh (lf->nr_keys); ++j) {
811 hive_node_h subkey = le32toh (lf->keys[j].offset);
813 if (!IS_VALID_BLOCK (h, subkey)) {
815 fprintf (stderr, "hivex_node_children: returning EFAULT because indirect subkey is not a valid block (0x%zx)\n",
820 if (add_to_offset_list (&children, subkey) == -1)
826 /* else not supported, set errno and fall through */
829 free_offset_list (&children);
830 free_offset_list (&blocks);
834 *children_ret = return_offset_list (&children);
835 *blocks_ret = return_offset_list (&blocks);
836 if (!*children_ret || !*blocks_ret)
842 hivex_node_children (hive_h *h, hive_node_h node)
844 hive_node_h *children;
847 if (get_children (h, node, &children, &blocks) == -1)
854 /* Very inefficient, but at least having a separate API call
855 * allows us to make it more efficient in future.
858 hivex_node_get_child (hive_h *h, hive_node_h node, const char *nname)
860 hive_node_h *children = NULL;
864 children = hivex_node_children (h, node);
865 if (!children) goto error;
868 for (i = 0; children[i] != 0; ++i) {
869 name = hivex_node_name (h, children[i]);
870 if (!name) goto error;
871 if (STRCASEEQ (name, nname)) {
875 free (name); name = NULL;
885 hivex_node_parent (hive_h *h, hive_node_h node)
887 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
892 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
894 hive_node_h ret = le32toh (nk->parent);
896 if (!IS_VALID_BLOCK (h, ret)) {
898 fprintf (stderr, "hivex_node_parent: returning EFAULT because parent is not a valid block (0x%zx)\n",
907 get_values (hive_h *h, hive_node_h node,
908 hive_value_h **values_ret, size_t **blocks_ret)
910 if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
915 struct ntreg_nk_record *nk = (struct ntreg_nk_record *) (h->addr + node);
917 size_t nr_values = le32toh (nk->nr_values);
920 fprintf (stderr, "hivex_node_values: nr_values = %zu\n", nr_values);
922 INIT_OFFSET_LIST (values);
923 INIT_OFFSET_LIST (blocks);
925 /* Deal with the common "no values" case quickly. */
929 /* Arbitrarily limit the number of values we will ever deal with. */
930 if (nr_values > 100000) {
935 /* Preallocate space for the values. */
936 if (grow_offset_list (&values, nr_values) == -1)
939 /* Get the value list and check it looks reasonable. */
940 size_t vlist_offset = le32toh (nk->vallist);
941 vlist_offset += 0x1000;
942 if (!IS_VALID_BLOCK (h, vlist_offset)) {
944 fprintf (stderr, "hivex_node_values: returning EFAULT because value list is not a valid block (0x%zx)\n",
950 if (add_to_offset_list (&blocks, vlist_offset) == -1)
953 struct ntreg_value_list *vlist =
954 (struct ntreg_value_list *) (h->addr + vlist_offset);
956 size_t len = block_len (h, vlist_offset, NULL);
957 if (4 + nr_values * 4 > len) {
959 fprintf (stderr, "hivex_node_values: returning EFAULT because value list is too long (%zu, %zu)\n",
966 for (i = 0; i < nr_values; ++i) {
967 hive_node_h value = vlist->offset[i];
969 if (!IS_VALID_BLOCK (h, value)) {
971 fprintf (stderr, "hivex_node_values: returning EFAULT because value is not a valid block (0x%zx)\n",
976 if (add_to_offset_list (&values, value) == -1)
981 *values_ret = return_offset_list (&values);
982 *blocks_ret = return_offset_list (&blocks);
983 if (!*values_ret || !*blocks_ret)
988 free_offset_list (&values);
989 free_offset_list (&blocks);
994 hivex_node_values (hive_h *h, hive_node_h node)
996 hive_value_h *values;
999 if (get_values (h, node, &values, &blocks) == -1)
1006 /* Very inefficient, but at least having a separate API call
1007 * allows us to make it more efficient in future.
1010 hivex_node_get_value (hive_h *h, hive_node_h node, const char *key)
1012 hive_value_h *values = NULL;
1014 hive_value_h ret = 0;
1016 values = hivex_node_values (h, node);
1017 if (!values) goto error;
1020 for (i = 0; values[i] != 0; ++i) {
1021 name = hivex_value_key (h, values[i]);
1022 if (!name) goto error;
1023 if (STRCASEEQ (name, key)) {
1027 free (name); name = NULL;
1037 hivex_value_key (hive_h *h, hive_value_h value)
1039 if (!IS_VALID_BLOCK (h, value) || !BLOCK_ID_EQ (h, value, "vk")) {
1044 struct ntreg_vk_record *vk = (struct ntreg_vk_record *) (h->addr + value);
1046 /* AFAIK the key is always plain ASCII, so no conversion to UTF-8 is
1047 * necessary. However we do need to nul-terminate the string.
1050 /* vk->name_len is unsigned, 16 bit, so this is safe ... However
1051 * we have to make sure the length doesn't exceed the block length.
1053 size_t len = le16toh (vk->name_len);
1054 size_t seg_len = block_len (h, value, NULL);
1055 if (sizeof (struct ntreg_vk_record) + len - 1 > seg_len) {
1057 fprintf (stderr, "hivex_value_key: returning EFAULT because key length is too long (%zu, %zu)\n",
1063 char *ret = malloc (len + 1);
1066 memcpy (ret, vk->name, len);
1072 hivex_value_type (hive_h *h, hive_value_h value, hive_type *t, size_t *len)
1074 if (!IS_VALID_BLOCK (h, value) || !BLOCK_ID_EQ (h, value, "vk")) {
1079 struct ntreg_vk_record *vk = (struct ntreg_vk_record *) (h->addr + value);
1082 *t = le32toh (vk->data_type);
1085 *len = le32toh (vk->data_len);
1086 if (*len == 0x80000000) { /* special case */
1088 if (t) *t = hive_t_dword;
1097 hivex_value_value (hive_h *h, hive_value_h value,
1098 hive_type *t_rtn, size_t *len_rtn)
1100 if (!IS_VALID_BLOCK (h, value) || !BLOCK_ID_EQ (h, value, "vk")) {
1105 struct ntreg_vk_record *vk = (struct ntreg_vk_record *) (h->addr + value);
1110 t = le32toh (vk->data_type);
1112 len = le32toh (vk->data_len);
1113 if (len == 0x80000000) { /* special case */
1120 fprintf (stderr, "hivex_value_value: value=0x%zx, t=%d, len=%zu\n",
1128 /* Arbitrarily limit the length that we will read. */
1129 if (len > 1000000) {
1134 char *ret = malloc (len);
1138 /* If length is <= 4 it's always stored inline. */
1140 memcpy (ret, (char *) &vk->data_offset, len);
1144 size_t data_offset = le32toh (vk->data_offset);
1145 data_offset += 0x1000;
1146 if (!IS_VALID_BLOCK (h, data_offset)) {
1148 fprintf (stderr, "hivex_value_value: returning EFAULT because data offset is not a valid block (0x%zx)\n",
1155 /* Check that the declared size isn't larger than the block its in. */
1156 size_t blen = block_len (h, data_offset, NULL);
1157 if (len > blen - 4 /* subtract 4 for block header */) {
1159 fprintf (stderr, "hivex_value_value: returning EFAULT because data is longer than its block (data 0x%zx, data len %zu, block len %zu)\n",
1160 data_offset, len, blen);
1166 char *data = h->addr + data_offset + 4;
1167 memcpy (ret, data, len);
1172 windows_utf16_to_utf8 (/* const */ char *input, size_t len)
1174 iconv_t ic = iconv_open ("UTF-8", "UTF-16");
1175 if (ic == (iconv_t) -1)
1178 /* iconv(3) has an insane interface ... */
1180 /* Mostly UTF-8 will be smaller, so this is a good initial guess. */
1181 size_t outalloc = len;
1185 size_t outlen = outalloc;
1186 char *out = malloc (outlen + 1);
1196 size_t r = iconv (ic, &inp, &inlen, &outp, &outlen);
1197 if (r == (size_t) -1) {
1198 if (errno == E2BIG) {
1199 size_t prev = outalloc;
1200 /* Try again with a larger output buffer. */
1203 if (outalloc < prev)
1208 /* Else some conversion failure, eg. EILSEQ, EINVAL. */
1224 hivex_value_string (hive_h *h, hive_value_h value)
1228 char *data = hivex_value_value (h, value, &t, &len);
1233 if (t != hive_t_string && t != hive_t_expand_string && t != hive_t_link) {
1239 char *ret = windows_utf16_to_utf8 (data, len);
1248 free_strings (char **argv)
1253 for (i = 0; argv[i] != NULL; ++i)
1259 /* Get the length of a UTF-16 format string. Handle the string as
1260 * pairs of bytes, looking for the first \0\0 pair.
1263 utf16_string_len_in_bytes (const char *str)
1267 while (str[0] || str[1]) {
1275 /* http://blogs.msdn.com/oldnewthing/archive/2009/10/08/9904646.aspx */
1277 hivex_value_multiple_strings (hive_h *h, hive_value_h value)
1281 char *data = hivex_value_value (h, value, &t, &len);
1286 if (t != hive_t_multiple_strings) {
1292 size_t nr_strings = 0;
1293 char **ret = malloc ((1 + nr_strings) * sizeof (char *));
1303 while (p < data + len && (plen = utf16_string_len_in_bytes (p)) > 0) {
1305 char **ret2 = realloc (ret, (1 + nr_strings) * sizeof (char *));
1313 ret[nr_strings-1] = windows_utf16_to_utf8 (p, plen);
1314 ret[nr_strings] = NULL;
1315 if (ret[nr_strings-1] == NULL) {
1321 p += plen + 2 /* skip over UTF-16 \0\0 at the end of this string */;
1329 hivex_value_dword (hive_h *h, hive_value_h value)
1333 char *data = hivex_value_value (h, value, &t, &len);
1338 if ((t != hive_t_dword && t != hive_t_dword_be) || len != 4) {
1344 int32_t ret = *(int32_t*)data;
1346 if (t == hive_t_dword) /* little endian */
1347 ret = le32toh (ret);
1349 ret = be32toh (ret);
1355 hivex_value_qword (hive_h *h, hive_value_h value)
1359 char *data = hivex_value_value (h, value, &t, &len);
1364 if (t != hive_t_qword || len != 8) {
1370 int64_t ret = *(int64_t*)data;
1372 ret = le64toh (ret); /* always little endian */
1378 hivex_visit (hive_h *h, const struct hivex_visitor *visitor, size_t len,
1379 void *opaque, int flags)
1381 return hivex_visit_node (h, hivex_root (h), visitor, len, opaque, flags);
1384 static int hivex__visit_node (hive_h *h, hive_node_h node, const struct hivex_visitor *vtor, char *unvisited, void *opaque, int flags);
1387 hivex_visit_node (hive_h *h, hive_node_h node,
1388 const struct hivex_visitor *visitor, size_t len, void *opaque,
1391 struct hivex_visitor vtor;
1392 memset (&vtor, 0, sizeof vtor);
1394 /* Note that len might be larger *or smaller* than the expected size. */
1395 size_t copysize = len <= sizeof vtor ? len : sizeof vtor;
1396 memcpy (&vtor, visitor, copysize);
1398 /* This bitmap records unvisited nodes, so we don't loop if the
1399 * registry contains cycles.
1401 char *unvisited = malloc (1 + h->size / 32);
1402 if (unvisited == NULL)
1404 memcpy (unvisited, h->bitmap, 1 + h->size / 32);
1406 int r = hivex__visit_node (h, node, &vtor, unvisited, opaque, flags);
1412 hivex__visit_node (hive_h *h, hive_node_h node,
1413 const struct hivex_visitor *vtor, char *unvisited,
1414 void *opaque, int flags)
1416 int skip_bad = flags & HIVEX_VISIT_SKIP_BAD;
1418 hive_value_h *values = NULL;
1419 hive_node_h *children = NULL;
1425 /* Return -1 on all callback errors. However on internal errors,
1426 * check if skip_bad is set and suppress those errors if so.
1430 if (!BITMAP_TST (unvisited, node)) {
1432 fprintf (stderr, "hivex__visit_node: contains cycle: visited node 0x%zx already\n",
1436 return skip_bad ? 0 : -1;
1438 BITMAP_CLR (unvisited, node);
1440 name = hivex_node_name (h, node);
1441 if (!name) return skip_bad ? 0 : -1;
1442 if (vtor->node_start && vtor->node_start (h, opaque, node, name) == -1)
1445 values = hivex_node_values (h, node);
1447 ret = skip_bad ? 0 : -1;
1451 for (i = 0; values[i] != 0; ++i) {
1455 if (hivex_value_type (h, values[i], &t, &len) == -1) {
1456 ret = skip_bad ? 0 : -1;
1460 key = hivex_value_key (h, values[i]);
1462 ret = skip_bad ? 0 : -1;
1466 if (vtor->value_any) {
1467 str = hivex_value_value (h, values[i], &t, &len);
1469 ret = skip_bad ? 0 : -1;
1472 if (vtor->value_any (h, opaque, node, values[i], t, len, key, str) == -1)
1474 free (str); str = NULL;
1479 str = hivex_value_value (h, values[i], &t, &len);
1481 ret = skip_bad ? 0 : -1;
1484 if (t != hive_t_none) {
1485 ret = skip_bad ? 0 : -1;
1488 if (vtor->value_none &&
1489 vtor->value_none (h, opaque, node, values[i], t, len, key, str) == -1)
1491 free (str); str = NULL;
1495 case hive_t_expand_string:
1497 str = hivex_value_string (h, values[i]);
1499 if (errno != EILSEQ && errno != EINVAL) {
1500 ret = skip_bad ? 0 : -1;
1503 if (vtor->value_string_invalid_utf16) {
1504 str = hivex_value_value (h, values[i], &t, &len);
1505 if (vtor->value_string_invalid_utf16 (h, opaque, node, values[i], t, len, key, str) == -1)
1507 free (str); str = NULL;
1511 if (vtor->value_string &&
1512 vtor->value_string (h, opaque, node, values[i], t, len, key, str) == -1)
1514 free (str); str = NULL;
1518 case hive_t_dword_be: {
1519 int32_t i32 = hivex_value_dword (h, values[i]);
1520 if (vtor->value_dword &&
1521 vtor->value_dword (h, opaque, node, values[i], t, len, key, i32) == -1)
1526 case hive_t_qword: {
1527 int64_t i64 = hivex_value_qword (h, values[i]);
1528 if (vtor->value_qword &&
1529 vtor->value_qword (h, opaque, node, values[i], t, len, key, i64) == -1)
1535 str = hivex_value_value (h, values[i], &t, &len);
1537 ret = skip_bad ? 0 : -1;
1540 if (t != hive_t_binary) {
1541 ret = skip_bad ? 0 : -1;
1544 if (vtor->value_binary &&
1545 vtor->value_binary (h, opaque, node, values[i], t, len, key, str) == -1)
1547 free (str); str = NULL;
1550 case hive_t_multiple_strings:
1551 strs = hivex_value_multiple_strings (h, values[i]);
1553 if (errno != EILSEQ && errno != EINVAL) {
1554 ret = skip_bad ? 0 : -1;
1557 if (vtor->value_string_invalid_utf16) {
1558 str = hivex_value_value (h, values[i], &t, &len);
1559 if (vtor->value_string_invalid_utf16 (h, opaque, node, values[i], t, len, key, str) == -1)
1561 free (str); str = NULL;
1565 if (vtor->value_multiple_strings &&
1566 vtor->value_multiple_strings (h, opaque, node, values[i], t, len, key, strs) == -1)
1568 free_strings (strs); strs = NULL;
1571 case hive_t_resource_list:
1572 case hive_t_full_resource_description:
1573 case hive_t_resource_requirements_list:
1575 str = hivex_value_value (h, values[i], &t, &len);
1577 ret = skip_bad ? 0 : -1;
1580 if (vtor->value_other &&
1581 vtor->value_other (h, opaque, node, values[i], t, len, key, str) == -1)
1583 free (str); str = NULL;
1588 free (key); key = NULL;
1591 children = hivex_node_children (h, node);
1592 if (children == NULL) {
1593 ret = skip_bad ? 0 : -1;
1597 for (i = 0; children[i] != 0; ++i) {
1599 fprintf (stderr, "hivex__visit_node: %s: visiting subkey %d (0x%zx)\n",
1600 name, i, children[i]);
1602 if (hivex__visit_node (h, children[i], vtor, unvisited, opaque, flags) == -1)
1606 if (vtor->node_end && vtor->node_end (h, opaque, node, name) == -1)
1617 free_strings (strs);