Subject: Re: kernel panics since 3/1: m_copym errors
To: Kentaro A. Kurahone <kurahone@sigusr1.org>
From: Hisashi T Fujinaka <htodd@twofifty.com>
List: tech-net
Date: 03/06/2005 22:31:58
On Mon, 7 Mar 2005, Kentaro A. Kurahone wrote:

> On Fri, Mar 04, 2005 at 10:07:26AM -0800, Hisashi T Fujinaka wrote:
>> So I have a couple of simple questions, has anything major changed in
>> the past week? Is m_copym or m_copydata used directly from userland
>> at all? Is the mbuf copying code used in anything besides the network
>> stack?
>
> Yes.  TCP/SACK got folded in.  I'm looking over the SACK code now to
> see if there's anything that can cause this, but in the meanwhile,
> could you disable sack via sysctl and tell me if the problem goes away?
> (net.inet.tcp.sack=0)

I have a hard time reliably reproducing the error, though it occurs more
than once a day. After a panic the system isn't completely locked up so
I can get a stack trace, but I haven't figured out how to save it. The
only thing I remember from the stack trace is soo_close.

I'm turning off tcp sack for now.

Thanks.

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