Libguestfs - Jan 2014 - [PATCH 1/3] lib: Add internal function to calculate strlen for strings encoded in Latin1 or UTF-16LE

On Windows, there exist at least two APIs for dealing with the
Registry: The Win32 API (RegCreateKeyA, RegCreateKeyW, etc.) works
with null-terminated ASCII or UTF-16 strings. The native API
(ZwCreateKey, etc.), on the other hand works with UTF-16 strings that
are stored as buffers+length and may contain null characters. Malware
authors have been relying on the Win32 API's inability to properly
work with such names for several years.

These changes make such names accessible from hivex.

---
 lib/hivex-internal.h |  1 +
 lib/utf16.c          | 10 ++++++++++
 2 files changed, 11 insertions(+)

diff --git a/lib/hivex-internal.h b/lib/hivex-internal.h
index 6bc8638..7f4cc3c 100644
--- a/lib/hivex-internal.h
+++ b/lib/hivex-internal.h
@@ -277,6 +277,7 @@ extern char * _hivex_recode (const char *input_encoding,
   _hivex_recode ("LATIN1", _input, _len, "UTF-8", NULL)
 extern char* _hivex_encode_string(const char *str, size_t *size, int *utf16);
 extern size_t _hivex_utf16_string_len_in_bytes_max (const char *str, size_t
len);
+extern size_t _hivex_utf8_strlen (const char* str, size_t len, int utf16);
 
 /* util.c */
 extern void _hivex_free_strings (char **argv);
diff --git a/lib/utf16.c b/lib/utf16.c
index 3641580..7dde9e5 100644
--- a/lib/utf16.c
+++ b/lib/utf16.c
@@ -123,3 +123,13 @@ _hivex_utf16_string_len_in_bytes_max (const char *str,
size_t len)
 
   return ret;
 }
+
+size_t
+_hivex_utf8_strlen (const char* str, size_t len, int utf16)
+{
+  char* encoding = utf16 ? "UTF-16LE" : "LATIN1";
+  size_t ret;
+  char* buf = _hivex_recode(encoding, str, len, "UTF-8", &ret);
+  free(buf);
+  return ret;
+}
-- 
1.8.5.2

Since iconv() does not treat null characters as special, this makes it
possible to read node names and value keys with embedded null
characters. which are not displayed at all by Windows Regedit.
---
 generator/generator.ml | 19 ++++++++++++++-----
 lib/node.c             | 24 ++++++++++++++++++++++++
 lib/value.c            |  9 +++++----
 3 files changed, 43 insertions(+), 9 deletions(-)

diff --git a/generator/generator.ml b/generator/generator.ml
index 54e5f0f..1c62129 100755
--- a/generator/generator.ml
+++ b/generator/generator.ml
@@ -179,7 +179,14 @@ C<$$$PROTO.HIV> (other names are possible: it seems
to depend on the
 tool or program that created the hive in the first place).  You can
 only know the \"real\" name of the root node by knowing which
registry
 file this hive originally comes from, which is knowledge that is
-outside the scope of this library.";
+outside the scope of this library.
+
+The name is recoded to UTF-8 and may contain embedded NUL characters.";
+
+  "node_name_len", (RSize, [AHive; ANode "node"]),
+    "return the length of a node's name",
+    "\
+Return the length of the node name as produced by
C<hivex_node_name>.";
 
   "node_timestamp", (RInt64, [AHive; ANode "node"]),
     "return the modification time of the node",
@@ -233,13 +240,15 @@ default key.";
   "value_key_len", (RSize, [AHive; AValue "val"]),
     "return the length of a value's key",
     "\
-Return the length of the key (name) of a (key, value) pair.  The
-length can legitimately be 0, so errno is the necesary mechanism
-to check for errors.
+Return the length of the key (name) of a (key, value) pair as produced
+by C<hivex_value_key>. The length can legitimately be 0, so errno is 
+the necesary mechanism to check for errors.
 
 In the context of Windows Registries, a zero-length name means
 that this value is the default key for this node in the tree.
-This is usually written as C<\"@\">.";
+This is usually written as C<\"@\">.
+
+The key is recoded to UTF-8 and may contain embedded NUL characters.";
 
   "value_key", (RString, [AHive; AValue "val"]),
     "return the key of a (key, value) pair",
diff --git a/lib/node.c b/lib/node.c
index fcd7442..22d1861 100644
--- a/lib/node.c
+++ b/lib/node.c
@@ -96,6 +96,30 @@ hivex_node_name (hive_h *h, hive_node_h node)
   }
 }
 
+size_t
+hivex_node_name_len (hive_h *h, hive_node_h node)
+{
+  if (!IS_VALID_BLOCK (h, node) || !BLOCK_ID_EQ (h, node, "nk")) {
+    SET_ERRNO (EINVAL, "invalid block or not an 'nk' block");
+    return 0;
+  }
+  struct ntreg_nk_record *nk +    (struct ntreg_nk_record *) ((char *)
h->addr + node);
+
+  /* nk->name_len is unsigned, 16 bit, so this is safe ...  However
+   * we have to make sure the length doesn't exceed the block length.
+   */
+  size_t len = le16toh (nk->name_len);
+  size_t seg_len = block_len (h, node, NULL);
+  if (sizeof (struct ntreg_nk_record) + len - 1 > seg_len) {
+    SET_ERRNO (EFAULT, "node name is too long (%zu, %zu)", len,
seg_len);
+    return 0;
+  }
+
+  return _hivex_utf8_strlen (nk->name, len, ! (le16toh (nk->flags) &
0x20));
+}
+
+
 static int64_t
 timestamp_check (hive_h *h, hive_node_h node, int64_t timestamp)
 {
diff --git a/lib/value.c b/lib/value.c
index 3460a8c..65404d7 100644
--- a/lib/value.c
+++ b/lib/value.c
@@ -186,13 +186,14 @@ hivex_value_key_len (hive_h *h, hive_value_h value)
   /* vk->name_len is unsigned, 16 bit, so this is safe ...  However
    * we have to make sure the length doesn't exceed the block length.
    */
-  size_t ret = le16toh (vk->name_len);
+  size_t len = le16toh (vk->name_len);
+
   size_t seg_len = block_len (h, value, NULL);
-  if (sizeof (struct ntreg_vk_record) + ret - 1 > seg_len) {
-    SET_ERRNO (EFAULT, "key length is too long (%zu, %zu)", ret,
seg_len);
+  if (sizeof (struct ntreg_vk_record) + len - 1 > seg_len) {
+    SET_ERRNO (EFAULT, "key length is too long (%zu, %zu)", len,
seg_len);
     return 0;
   }
-  return ret;
+  return _hivex_utf8_strlen (vk->name, len, ! (le16toh (vk->flags) &
0x01));
 }
 
 char *
-- 
1.8.5.2

---
 generator/generator.ml | 42 ++++++++++++++++++++++++++++++++++--------
 1 file changed, 34 insertions(+), 8 deletions(-)

diff --git a/generator/generator.ml b/generator/generator.ml
index 1c62129..6b1eef7 100755
--- a/generator/generator.ml
+++ b/generator/generator.ml
@@ -380,6 +380,11 @@ new key is added.  Key matching is case insensitive.
 C<node> is the node to modify.";
 ]
 
+let f_len_exists n +  List.exists
+    (function (cand, _, _, _) -> cand = (String.concat "" [n;
"_len"]))
+    functions
+
 (* Useful functions.
  * Note we don't want to use any external OCaml libraries which
  * makes this a bit harder than it should be.
@@ -1839,9 +1844,7 @@ static void raise_closed (const char *) Noreturn;
        * call, so don't do it.  XXX
        *)
       (*pr "  caml_enter_blocking_section ();\n";*)
-      pr "  r = hivex_%s (%s" name (List.hd c_params);
-      List.iter (pr ", %s") (List.tl c_params);
-      pr ");\n";
+      pr "  r = hivex_%s (%s);\n" name (String.concat ", "
c_params);
       (*pr "  caml_leave_blocking_section ();\n";*)
       pr "\n";
 
@@ -1890,7 +1893,13 @@ static void raise_closed (const char *) Noreturn;
            pr "  rv = copy_int_array (r);\n";
            pr "  free (r);\n"
        | RString ->
-           pr "  rv = caml_copy_string (r);\n";
+           if f_len_exists name then (
+             pr "  size_t sz;\n  sz = hivex_%s_len (%s);\n"
+               name (String.concat ", " c_params);
+             pr "  rv = caml_alloc_string (sz);\n";
+             pr "  memcpy (String_val (rv), r, sz);\n"
+           ) else
+             pr "  rv = caml_copy_string (r);\n";
            pr "  free (r);\n"
        | RStringList ->
            pr "  rv = caml_copy_string_array ((const char **) r);\n";
@@ -2638,7 +2647,11 @@ DESTROY (h)
              pr "      if (r == NULL)\n";
              pr "        croak (\"%%s: %%s\", \"%s\",
strerror (errno));\n"
 	       name;
-             pr "      RETVAL = newSVpv (r, 0);\n";
+             if f_len_exists name then
+               pr "      RETVAL = newSVpvn (r, hivex_%s_len (%s));\n"
+                 name (String.concat ", " c_params)
+             else
+               pr "      RETVAL = newSVpv (r, 0);\n";
 	     pr "      free (r);\n";
              pr " OUTPUT:\n";
              pr "      RETVAL\n"
@@ -3178,10 +3191,19 @@ put_val_type (char *val, size_t len, hive_type t)
        | RValue ->
            pr "  py_r = PyLong_FromLongLong (r);\n"
        | RString ->
+           if f_len_exists name then
+             pr "  size_t sz = hivex_%s_len (%s);\n"
+               name (String.concat ", " c_params);
            pr "#ifdef HAVE_PYSTRING_ASSTRING\n";
-           pr "  py_r = PyString_FromString (r);\n";
+           if f_len_exists name then
+             pr "  py_r = PyString_FromStringAndSize (r, sz);\n"
+           else
+             pr "  py_r = PyString_FromString (r);\n";
            pr "#else\n";
-           pr "  py_r = PyUnicode_FromString (r);\n";
+           if f_len_exists name then
+             pr "  py_r = PyUnicode_FromStringAndSize (r, sz);\n"
+           else
+             pr "  py_r = PyUnicode_FromString (r);\n";
            pr "#endif\n";
            pr "  free (r);"
        | RStringList ->
@@ -3633,7 +3655,11 @@ get_values (VALUE valuesv, size_t *nr_values)
         pr "  free (r);\n";
         pr "  return rv;\n"
       | RString ->
-        pr "  VALUE rv = rb_str_new2 (r);\n";
+        if f_len_exists name then (
+          pr "  size_t sz = hivex_%s_len (%s);\n" name (String.concat
", " c_params);
+          pr "  VALUE rv = rb_str_new (r, sz);\n"
+        ) else
+          pr "  VALUE rv = rb_str_new2 (r);\n";
         pr "  free (r);\n";
         pr "  return rv;\n"
       | RStringList ->
-- 
1.8.5.2

On Wed, Jan 08, 2014 at 01:26:23AM +0100, Hilko Bengen
wrote:
> On Windows, there exist at least two APIs for dealing with the
> Registry: The Win32 API (RegCreateKeyA, RegCreateKeyW, etc.) works
> with null-terminated ASCII or UTF-16 strings. The native API
> (ZwCreateKey, etc.), on the other hand works with UTF-16 strings that
> are stored as buffers+length and may contain null characters. Malware
> authors have been relying on the Win32 API's inability to properly
> work with such names for several years.
> 
> These changes make such names accessible from hivex.
ACK to all 3 patches.

It be nice to have some sort of test coverage of these.

Thanks,

Rich.

-- 
Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones
virt-df lists disk usage of guests without needing to install any
software inside the virtual machine.  Supports Linux and Windows.
http://people.redhat.com/~rjones/virt-df/

---
 images/README          |  15 +++++++++++++++
 images/mkzero/Makefile |   7 +++++++
 images/mkzero/mkzero.c |  48 ++++++++++++++++++++++++++++++++++++++++++++++++
 images/special         | Bin 0 -> 8192 bytes
 4 files changed, 70 insertions(+)
 create mode 100644 images/mkzero/Makefile
 create mode 100644 images/mkzero/mkzero.c
 create mode 100644 images/special

diff --git a/images/README b/images/README
index 2131885..34c65f3 100644
--- a/images/README
+++ b/images/README
@@ -11,3 +11,18 @@ hand-crafted binary blob.
 tests.
 
 - Richard W.M. Jones 2010-02-24.
+
+'special' was created by importing 'minimal' into a VM running
Windows
+XP and loading it into HKEY_LOCAL_MACHINE\minimal using regedit.exe
+(File/Load Hive...)
+
+- A subkey 'asdf_äöüß' was created in the root node
+  - An empty REG_STRING value 'asdf_äöüß' was created within that node.
+- A subkey 'weird™' was created in the root node.
+  - An empty REG_STRING value 'symbols $£₤₧€' (SMALL DOLLAR SIGN,
+    FULLWIDTH POUND SIGN, PESETA SIGN, EURO SIGN) was created within
+    that node.
+- A subkey 'zero\0key' with an REG_DWORD value 'zero\0val'
+  was created using the 'mkzero/mkzero.c'. (\0 = zero character) 
+
+- Hilko Bengen 2014-01-10.
diff --git a/images/mkzero/Makefile b/images/mkzero/Makefile
new file mode 100644
index 0000000..4d33ade
--- /dev/null
+++ b/images/mkzero/Makefile
@@ -0,0 +1,7 @@
+CROSS=i686-w64-mingw32-
+
+all: mkzero.exe
+mkzero.exe: mkzero.o
+	$(CROSS)gcc -o $@ $< -lntdll
+%.o: %.c
+	$(CROSS)gcc -o $@ -c $<
diff --git a/images/mkzero/mkzero.c b/images/mkzero/mkzero.c
new file mode 100644
index 0000000..1b50b22
--- /dev/null
+++ b/images/mkzero/mkzero.c
@@ -0,0 +1,48 @@
+/* use the NT native API to create registry key and value that contain
+   a zero character */
+
+#include <ntdef.h>
+#include <stdio.h>
+#include <ddk/wdm.h>
+#include <windef.h>
+
+int main (int argc, char **argv)
+{
+  NTSTATUS rc;
+
+  UNICODE_STRING root_key_name;
+  RtlInitUnicodeString(&root_key_name,
L"\\Registry\\Machine\\minimal");
+  OBJECT_ATTRIBUTES root_key_obj;
+  InitializeObjectAttributes (&root_key_obj, &root_key_name,
+                              OBJ_OPENIF | OBJ_CASE_INSENSITIVE,
+                              NULL, NULL);
+  HANDLE minimal_key_handle;
+  rc = ZwCreateKey (&minimal_key_handle, KEY_ALL_ACCESS, &root_key_obj,
+                    0, NULL, REG_OPTION_NON_VOLATILE, NULL);
+  if (!NT_SUCCESS (rc)) {
+    printf("error: CreateKey <HKLM\\minimal>: 0x%08x\n", rc);
+    exit(1);
+  }
+
+  UNICODE_STRING key_name = {16, 16, L"zero\0key"};
+  OBJECT_ATTRIBUTES key_obj;
+  InitializeObjectAttributes (&key_obj, &key_name, 
+                              OBJ_OPENIF | OBJ_CASE_INSENSITIVE,
+                              minimal_key_handle, NULL);
+  HANDLE key_handle;
+  rc = ZwCreateKey (&key_handle, KEY_ALL_ACCESS, &key_obj,
+                    0, NULL, REG_OPTION_NON_VOLATILE, NULL);
+  if (!NT_SUCCESS (rc)) {
+    printf("error: CreateKey: 0x%08x\n", rc);
+    exit(1);
+  }
+
+  UNICODE_STRING value_name = {16, 16, L"zero\0val"};
+  DWORD value = 0;
+  rc = ZwSetValueKey (key_handle, &value_name, 0,
+                      REG_DWORD, &value, sizeof(value));
+  if (!NT_SUCCESS (rc)) {
+    printf("error: SetValueKey: 0x%08x\n", rc);
+    exit(1);
+  }
+}
diff --git a/images/special b/images/special
new file mode 100644
index
0000000000000000000000000000000000000000..8aa4f2254af7b52f0a79061a7288c2128eee7b63
GIT binary patch
literal 8192
zcmeHLziU%b6h28COhb#$Ul4=fYx@MLKL!^^6)M(2Dk4tBk~GnVmnM=(q~Z`n)IUHs
zhkyk;I61kvcM)}{1Q$C94muRkP^1{Y@7_GxmnJ$170!YC?z!ijckape-TNS?oV$K)
zEF!{jqOm&ic<4<8s>CoT)Lkl$)hjY3XCy0gl9wVzMQ%!2(s-6IPszCCVDruT2G&8`
z9AXO0qq1!fXx39_w@ooxh$5f}C<2OrBA^H;0*Zhlpa>`eioicbU^bI4nho&(K9Haf
zz;(UVY8CyX^5XN4gX|5^c6x_M9~H=AzX(EF)H&DM^IjS{WDsl5^DbVRygYe);`|l5
zmwvTc6<;J7wSyZRz#}64kJx;0%#I?=7nQG);7gK}{a8it4kYH{F-$#?#6xOZ>|NqV
zd&C<%;!nxWP=0qK-r1br*ioH{x3$K0vx<a4z%k7f)M#uy^M+y#w4Hl|^N6ef>#83Y
zz+dBbm`$(dcCK8KYPzrn|4Y=<FmVGw%f_)pj&x=wJN>D-{t;aqV!DdB6*xGaxzF8t
zo0l4V94u-aKJapH5c6RSisfa^^5XR1CDntMF7XEcF7d{W_^!Oz_^rG+U3f`_u(rH7
zp3Q5UUa@`nU5h8*=k9#*hDN7n{O|m&;rA>@e~6Ft@U?kmJ59(>9P_z7$1d1=Ci3&h
zu4c7;&s_Kd0X;?x>>!)>g@(ZILg5jm@Sfd5A1;rMTvkSgX(I!L*{Io*`#r=jSYEha
zBJU3JaZr}9uKKCad7-@`^XTPe(B&&Ycg2&Zvf%Qh&>MUwTv>ne?$F)8)E`Hoo%OT6
zp#A~q8|RbR?Kv*Y_Q?e5yt@40bx{9XtY3ZeKG-PmhWd-X8~R?kffF;_Mu=krkIp*C
oOB)Kqjk!;LZoXkZJexA9st70oihv@Z2q*%IfFhs>{C^1i23(Tu`2YX_

literal 0
HcmV?d00001

-- 
1.8.5.2