Hi Michał,
I will fix the style guide and spec alignment issues you pointed at elsewhere...
> -----Original Message-----
> > + if (key.phase == KEY_REFRESH_PHASE_TWO) {
> > + memcpy(out + 1, key.new_key, 16);
> > + snb_flags |= PROV_FLAG_KR;
> > + } else
> > + memcpy(out + 1, key.old_key, 16);
> > +
> > + l_put_be16(net_idx, out + 1 + 16);
> > + l_put_u8(snb_flags, out + 1 + 16 + 2);
> > + l_put_be32(iv_index, out + 1 + 16 + 2 + 1);
> > + l_put_be16(primary, out + 1 + 16 + 2 + 1 + 4);
>
> Not really related to this patchset, but a general comment about packet
> building and parsing: could we please move away from calculating byte offsets
> and use packed structs instead? Do you have any opinion on that?
You are probably right about this. Technically, this is a "Network Packet" (even though
we locally encrypt it before sending) and so we should treat it like any packet we
send over a socket (for instance, the HCI sockets we use to talk to controllers). This is
a vestige of my "old man engineering" where I couldn't always trust compilers across
all platforms to faithfully pack octets the same, but obviously if we are trusting the
packing for HCI, we should be trusting the packing for this. I will scrub the code for
opportunities for packing this (and other similar) structures.
