tech-kern archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: Socket options KPI

On Sun, 27 Jan 2008, Elad Efrat wrote:

> Iain Hibbert wrote:
> > On Sun, 27 Jan 2008, Elad Efrat wrote:
> > > For the first step, I'd like to reduce the mbuf abuse.
> >
> > In that case, I think better to keep the PRCO_ and not encode the
> > direction into the sockopt structure, and using:
> >
> >   int pr_ctloutput(int req, struct socket *so, struct sockopt *opt);
> >
> > (just because I don't like the fit)
> What you're suggesting implies we will be changing the KPI that way,
> which, at the moment, is not necessarily true. I would very much like
> for the first sweep to be close to what is done in FreeBSD.

Ah, I see - I did not notice they changed it that way.

my objection really was that pr_ctloutput() seems to be designed such that
it can be extended in a per-protocol way so that protocols can be layered
in the kernel. changing it to use SOPT_ means that you can only get and
set options (though I think there are no other usages)

in fact the way that protocols are implemented is borked, you can't
actually layer them without the lower knowing about the upper and I had to
work it differently when writing the bluetooth code (for which the upper
layer is anonymous)

> Can we agree on revising this at a later date please?

thats fine


Home | Main Index | Thread Index | Old Index