Subject: Re: Panic when bringing up network bridge
To: None <port-cobalt@NetBSD.org>
From: Izumi Tsutsui <tsutsui@ceres.dti.ne.jp>
List: port-cobalt
Date: 04/26/2006 18:51:07
In article <444EC918.4020204@geekrants.org>
netbsd@geekrants.org wrote:

> I just got 200604250000Z --> 3.99.18 (GENERIC) #0: Tue Apr 25 10:44:05
> 
> This one does not crash, however things are not perfect :
> 
> pickle ~ # brconfig bridge0 up
> pickle ~ # tlp0: transmit timeout
> tlp1: transmit timeout
> Apr 25 20:33:14 pickle /netbsd tlp0: link state changed to DOWN
> Apr 25 20:33:14 pickle /netbsd tlp1: link state changed to DOWN

Okay, "transmit timeout" is caused by lost TX interrupts,
and "link state changed" is caused by reset of tlp after timeout.
Maybe some stray interrupts block tlp interrupts,
or inappropriate interrupt level is set in some place?

Could you check interrupt/network statistics before and after brconfig,
i.e. could you post result of commands:
# netstat -i; vmstat -c 3 -i 1; brconfig bridge0 up; vmstat -c 3 -i 1
or something?
---
Izumi Tsutsui