On Mon, 18 Apr 2005, Hisashi T Fujinaka wrote:

> On a i386-current system from about 0000GMT on 4/18, the m_copy_m hang
> showed up again. I have TCP_SACK turned off. The previous kernel, from
> 4/2 sources, ran perfectly until I replaced it with this one.
>
> Unfortunately, I have few other details, so this is probably a useless
> data point.

I'm going to have to set up serial console, since no one can tell me how
to get the backtrace output from the debugger to save to disk.

I had another hang, with -current sources from 4/19.

The only line I saved is:
panic: m_copydata: m == 0, len 1

and I remember a function similar to tcp_sack_buildpacket. Or
tcp_sack_ack_buildpacket. I've turned of tcp_sack in sysctl
net.inet.tcp.sack = 0.

Did things change since April 2?

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