Re: [pooka%netbsd.org@localhost: CVS commit: src/sys]

[Joerg Sonnenberger <joerg%britannica.bec.de@localhost>]

On Thu, Jan 21, 2010 at 03:29:59PM +0200, Antti Kantee wrote:
> If you are concerned with dhcp machine performance, may I suggest
> you start by optimizing the thousands of instructions that bpf
> executes for every packet while trying to decide that it's not
> handling dhcp input.

The call is also done for *output*, which does not run the filter it
all in the DHCP case.

> Also, you'd need to fix bus_space on most
> architectures to keep the damn nic drivers from doing *multiple*
> indirect calls (start with sun2, because this project will benefit
> the most from you working on that).  Finally, follow up with
> binary-patching away that pesky if_input method.

So just to make sure I understand you correctly. You are introducing
overhead for a lot of use casei to make it possible to conditionally
load 20KB after some non-trivial issues are solved. You are justifying
that because some other places either actually do use the abstraction
or are on some architectures slow?

Joerg