Re: tcp template removal / scalability patch

2001-06-22 Thread Bosko Milekic
> - bzero(ip, sizeof(struct ip)); /* XXX overkill? */ > + bzero(ip, sizeof(struct ip)); /* XXX overkill? */ I pointed this out the first time; the change looks unnecessary. You changed what were tabs to spaces. Unless there's a good reason, it looks unnecessar

Re: tcp template removal / scalability patch

2001-06-21 Thread Mike Silbersack
Ok, here's the updated tcp template removal patch. It's functionally the same, which is why I'm not including an updated -stable patch. Also, I did test it with IPv6, and it seems good. No IPSEC testing, but I see no reason why it would break if v4/v6 didn't. What did change: - Formatting al

Re: tcp template removal / scalability patch

2001-06-19 Thread Jesper Skriver
On Tue, Jun 19, 2001 at 09:24:00PM -0500, Mike Silbersack wrote: > > On Wed, 20 Jun 2001, Jesper Skriver wrote: > > > I think we should leave TCP_COMPAT_42 in RELENG_4, so 4.x users > > won't be surprised if it's suddenly gone ... > > > > /Jesper > > Actually, TCP_COMPAT_42 was a virtual no-op

Re: tcp template removal / scalability patch

2001-06-19 Thread Mike Silbersack
On Tue, 19 Jun 2001, Mike Silbersack wrote: > Looking back, I should change the keepalive case so that it never needs > the tcp template; this will require simple mods to tcp_respond. I'll > change this and make a new patch soon. Blech. tcp_respond doesn't look friendly, and the case where a

Re: tcp template removal / scalability patch

2001-06-19 Thread Mike Silbersack
On Wed, 20 Jun 2001, Jesper Skriver wrote: > I think we should leave TCP_COMPAT_42 in RELENG_4, so 4.x users > won't be surprised if it's suddenly gone ... > > /Jesper Actually, TCP_COMPAT_42 was a virtual no-op until Kris imported the OpenBSD sequence number generation system. The only effect

Re: tcp template removal / scalability patch

2001-06-19 Thread Jesper Skriver
On Tue, Jun 19, 2001 at 05:48:14PM -0500, Mike Silbersack wrote: > > On Tue, 19 Jun 2001, Bosko Milekic wrote: > > > > I've attached two patches; one for current, and one for stable. Please > > > review / test, _especially_ if you're using IPv6 or IPSec - while those > > > cases look correct, I

Re: tcp template removal / scalability patch

2001-06-19 Thread Mike Silbersack
On Tue, 19 Jun 2001, Bosko Milekic wrote: > > I've attached two patches; one for current, and one for stable. Please > > review / test, _especially_ if you're using IPv6 or IPSec - while those > > cases look correct, I'm not running either and haven't tested them. > > I've spotted some pa

tcp template removal / scalability patch

2001-06-19 Thread Mike Silbersack
As suggested by Terry, I've cooked up a patch which halts the use of mbufs for storing tcp template structures. The structure was only used in two places; tcp_output.c when sending packets, and tcp_timer.c when sending keepalives. tcp_output now pulls the info directly from the tcpcb, while tcp_