On Mon, Aug 8, 2011 at 4:07 PM, Mandeep Singh Baines <msb@xxxxxxxxxxxx> wrote:
>
> There is no loss of security due to removing the memset. It would be a
> bug for the stack to leak to userspace. However, a defence-in-depth
> argument could be made for keeping the clearing of the workspace.
So I'm nervous about this just because I can see the security crazies
rising up about this.
The fact is, in our current code in drivers/char/random.c, we do have
a memset() of the workspace buffer on the stack, and any competent
compiler should actually just remove it, because it's dead memory (and
the compiler can *see* that it's dead memory). Of course, I don't know
if gcc does notice that, but it's a prime example of code that "looks"
secure, but has absolutely zero actual real security. Getting rid of
the memset() is actually better for *real* security, in that at least
it's not some kind of pointless security theater. But I can see some
people wanting to add a memory barrier or something to force the
memset() to actually take place.
So I dunno.
Arguably it's theoretically possible to find random data on the stack,
and maybe it can even be interesting (although I don't think the last
64 bytes of SHA1 state is all that exciting myself). Personally, I
consider it unlikely as hell to be relevant to anybody, and anybody
who has access to the kernel stack has *much* more direct security
holes than some random data that they can use. But the patch still
makes me worry about the brouhaha from some people.
Linus
--
To unsubscribe from this list: send the line "unsubscribe linux-crypto" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
