On Tue, 19 Apr 2005, Kentaro A. Kurahone wrote:

> On Tue, Apr 19, 2005 at 02:20:36PM -0700, Hisashi T Fujinaka wrote:
>> On Tue, 19 Apr 2005, Markus W Kilbinger wrote:
>>
>>> Sounds/looks very similar similar to my
>>>
>>> kern/29787
>>>
>>> where I could catch such a panic within a serial remote console in the
>>> meantime.
>>>
>>> My presumed wm driver relation could be pure random though it's still
>>> the only machine showing such panics.
>>>
>>> Setting net.inet.tcp{,6}.sack.enable=0 seems to avoid these panics.
>>
>> They changed it from net.inet.tcp.sack? Just FYI, this error occur for
>> me on a single fxp, and also ocurred on a single Linksys (rtk? can't
>> remember).
>
> Yes, it got changed because there's more SACK related sysctls now.
>
> FWIW, the nics that I've been using to test this the most are
> rtk, iwi, and wi.  And I have yet to be able to reproduce this even
> after moderately heavy use.

Aha! Well, this probably explains why it happens again. Since the sysctl
changed, the new kernels have tcp sack turned on because my sysctl.conf
is wrong. Sounds like tcp sack is killing my system again.

I don't know what it takes to trigger the problem. The server has screen
running, a moderately busy webserver (including my sister's knitting
blog that generates an incredible amount of traffic for as simple as it
is), ssh connections to all my machines, mailman, incoming mail server
(for just a couple of people) and that's about it. Nothing special. Oh,
irssi and bitlbee, but just for myself.

And I have two other -current machines that don't have problems, but
they're just client machines.

-- 
Hisashi T Fujinaka - htodd@twofifty.com
BSEE(6/86) + BSChem(3/95) + BAEnglish(8/95) + MSCS(8/03) + $2.50 = latte