Periodic xl watchdog timeouts on 6.2-RELEASE

[Brian J. Conway]

On Mon, 12 Mar 2007 22:22:44 -0800
"Ted Mittelstaedt" <tedm at toybox.placo.com> wrote:

> I have found with some of the intel MBs that the latest BIOS update
> actually causes trouble.  Don't be afraid to try back-flashing to an
> older BIOS update.  Intel has all the BIOS versions up on their site for
> each board.
> 
> Ted

Tried a few things in the past couple days:

- Set the interfaces to not auto-negotiate and hard-coded them: No change.
- Tried the past 3 BIOS revisions I had been using previously: No change.

Next up, started swapping around cards.  I noticed that one of my 3 3c905C
cards hadn't been giving the watchdog timeout errors that I could
remember, even though I'm doubting 2/3 of my previously-good cards are
actually bad, but I kept that one at xl1 and tried a good 3c905B as xl0. 
This worked a little differently, now instead of watchdog timeouts, on the
previously-normal xl1 I get:

Mar 15 05:56:51 imogen kernel: xl1: transmission error: 90
Mar 15 05:56:51 imogen kernel: xl1: tx underrun, increasing tx start
threshold to 120 bytes

I know it's *technically* an informational message and not a problem, I'm
a perfectionist and would prefer it not to be there.  Perhaps I need to
start from stratch with some em* cards, they've been working well for me
everywhere else.

(Original discussion:
http://lists.freebsd.org/pipermail/freebsd-questions/2007-March/144227.html)

Brian J. Conway