Hi,
bluetoothd does not check in some (all?) places that the remote name
reported by a device is valid utf8 data. e.g., extract_eir_name() in
src/dbus-hci.c.
The reception of an extended inquiry response containing a name with
invalid utf8 data can cause the dbus interface to disappear. This is
therefore a denial-of-service vulnerability (at the very least).
The following patch fixes the above problem but there are probably other
places where the check needs to be done.
--- bluez-4.51.orig/src/dbus-hci.c
+++ bluez-4.51/src/dbus-hci.c
@@ -450,6 +450,8 @@
switch (*type) {
case 0x08:
case 0x09:
+ if (!g_utf8_validate(data + 2, data[0] - 1, NULL))
+ return strdup("");
return strndup((char *) (data + 2), data[0] - 1);
}
David
--
David Vrabel, Senior Software Engineer, Drivers
CSR, Churchill House, Cambridge Business Park, Tel: +44 (0)1223 692562
Cowley Road, Cambridge, CB4 0WZ http://www.csr.com/
Member of the CSR plc group of companies. CSR plc registered in England and Wales, registered number 4187346, registered office Churchill House, Cambridge Business Park, Cowley Road, Cambridge, CB4 0WZ, United Kingdom
Hi again,
On Wed, Aug 25, 2010, Johan Hedberg wrote:
> Good catch. At least the legacy name queries are already protected
> (remote_name_information function in security.c) so I think this is the only
> place missing the UTF-8 validation. However, your patch doesn't compile cleanly
> so some fine tuning is still needed (always check compilation with
> "./bootstrap-configure && make" before sending upstream):
>
> src/dbus-hci.c: In function ‘extract_eir_name’:
> src/dbus-hci.c:466: error: pointer targets in passing argument 1 of ‘g_utf8_validate’ differ in signedness
> /usr/include/glib-2.0/glib/gunicode.h:356: note: expected ‘const gchar *’ but argument is of type ‘uint8_t *’
> make[1]: *** [src/dbus-hci.o] Error 1
Since this was the only thing blocking a new release I went ahead and
fixed the issue myself. The (fixed) patch is now upstream.
Johan
Hi David,
On Wed, Aug 25, 2010, David Vrabel wrote:
> bluetoothd does not check in some (all?) places that the remote name
> reported by a device is valid utf8 data. e.g., extract_eir_name() in
> src/dbus-hci.c.
>
> The reception of an extended inquiry response containing a name with
> invalid utf8 data can cause the dbus interface to disappear. This is
> therefore a denial-of-service vulnerability (at the very least).
>
> The following patch fixes the above problem but there are probably other
> places where the check needs to be done.
>
> --- bluez-4.51.orig/src/dbus-hci.c
> +++ bluez-4.51/src/dbus-hci.c
> @@ -450,6 +450,8 @@
> switch (*type) {
> case 0x08:
> case 0x09:
> + if (!g_utf8_validate(data + 2, data[0] - 1, NULL))
> + return strdup("");
> return strndup((char *) (data + 2), data[0] - 1);
> }
Good catch. At least the legacy name queries are already protected
(remote_name_information function in security.c) so I think this is the only
place missing the UTF-8 validation. However, your patch doesn't compile cleanly
so some fine tuning is still needed (always check compilation with
"./bootstrap-configure && make" before sending upstream):
src/dbus-hci.c: In function ‘extract_eir_name’:
src/dbus-hci.c:466: error: pointer targets in passing argument 1 of ‘g_utf8_validate’ differ in signedness
/usr/include/glib-2.0/glib/gunicode.h:356: note: expected ‘const gchar *’ but argument is of type ‘uint8_t *’
make[1]: *** [src/dbus-hci.o] Error 1
After fixing that, could you prepare the patch through git format-patch so that
I can easily apply it using git am? Thanks.
Johan