On Fri, Dec 12, 2014 at 12:57:08PM -0500, Ted Unangst wrote:
> On Wed, Dec 10, 2014 at 21:51, Max Fillinger wrote:
> > Here's a version without malloc'ing the key.
>
> Looks like it still does?
>
> > + if (sip_key == NULL) {
> > + sip_key = malloc(SIPHASH_KEY_LENGTH);
> > + arc4random_buf(sip_key,
> > SIPHASH_KEY_LENGTH);
>
> > @@ -515,8 +516,8 @@ struct rsession {
> > struct ctl_relay_event se_in;
> > struct ctl_relay_event se_out;
> > void *se_priv;
> > - u_int32_t se_hashkey;
> > - int se_hashkeyset;
> > + SIPHASH_CTX se_hashctx;
> > + int se_hashctxset;
>
> This isn't really how the hash ctx is supposed to be used. relayd is
> doing something a little different here. I'll look at it eventually,
> but this isn't as straightforward as a mechanical conversion.
>
I saw the diff and I will take care of it. It needs some work but
that comes along with other changes that I'm doing in the hash code as
well.
Reyk
