At Thu, 02 Feb 2012 17:00:00 -0800, Jeff Rizzo <riz%netbsd.org@localhost> wrote:
Subject: Re: netbsd-5 on Citrix XenClient 2.1 (on an HP EliteBook 8460p)
>
> On 2/2/12 4:24 PM, Greg A. Woods wrote:
> > re0 (and re1 when the wireless is active) is throwing "watchdog timeout"
> > warnings on the kernel which may correspond to some of the freezes.
>
> This is a known issue; most Xen implementations don't provide
> interrupts properly for the re(4) driver, but it actually works OK if
> you let the rtk(4) driver match, either by commenting out re(4) or
> using userconf to disable re, which you can do from boot.cfg. I don't
> know whether Xenserver allows you to change the emulation, but the
> wm(4) one works better. :)
Excellent hint, thanks!
Yes, preventing re(4) from being used (and thus allowing rtk(4) to be
used, solved both the networking problems, as well as the periodic
"freezing" of the system.
Unfortunately I cannot use boot.cfg to send commands to userconf(4) on
NetBSD-5 -- unless I pull up the changes needed to do so I guess.
Perhaps re(4) should be made smart enough to recognise situations where
it would cause more trouble than it would be worth and to prevent itself
from being attached in those situations.
Either that or maybe the support for the unique chips it alone can
handle should be migrated to rtk(4) and re(4) should be sent to the bit
bucket. (That's said with my only, at least even semi-recent,
experience with the both of them being now with XenClient.)
XenClient can't easily change the emulation it offers, though perhaps
behind the scenes and under the hood there's something I could tweak.
At the moment XC-2.1 doesn't even let you assign more than one CPU per
virtual machine (using the GUI management interface) because apparently
there was sometimes some instability observed when this was done.
--
Greg A. Woods
Planix, Inc.
<woods%planix.com@localhost> +1 250 762-7675 http://www.planix.com/
Attachment:
pgpN3O5zzadOP.pgp
Description: PGP signature