Subject: Re: port-xen/29999: ctrl_if_interrupt IPL problem
To: None <port-xen-maintainer@netbsd.org, gnats-admin@netbsd.org,>
From: Manuel Bouyer <bouyer@antioche.lip6.fr>
List: netbsd-bugs
Date: 04/19/2005 07:05:03
The following reply was made to PR port-xen/29999; it has been noted by GNATS.
From: Manuel Bouyer <bouyer@antioche.lip6.fr>
To: YAMAMOTO Takashi <yamt@mwd.biglobe.ne.jp>
Cc: gnats-bugs@NetBSD.org, port-xen-maintainer@NetBSD.org,
gnats-admin@NetBSD.org, netbsd-bugs@NetBSD.org
Subject: Re: port-xen/29999: ctrl_if_interrupt IPL problem
Date: Tue, 19 Apr 2005 09:04:37 +0200
On Tue, Apr 19, 2005 at 09:06:57AM +0900, YAMAMOTO Takashi wrote:
> hi,
>
> > We could move __ctrl_if_tx_tasklet() to run from the softintr. This would
> > fix the wakeup issue.
> > For softintr_schedule(), we could probably use IPL_SOFTXENEVT instead of
> > IPL_SOFTNET, and call ctrl_if_softintr_handler() from the assemby code
> > instead. This should take care of the issue.
>
> you mean (ab)using softintr() as xenevt does?
Yes
> i think it should work.
>
> (although i'm not entirely happy with how xenevt works,
> making hacks similar is a good thing :-)
>
> > IPL_CTRL needs to be very high because the console I/O is done though
> > control messages on guest domains, and there's no way to poll for keyboard
> > input. So, to get ddb working in guest domains, we need to have control
> > events enabled while in ddb.
>
> "keyboard input" meant "console input"?
Yes, sorry for the bad wordings here
>
> > I guess we could also be busy-looping waiting for the event, but it
> > would not be nice to other guests domains.
>
> if there's a way to poll events (i couldn't find any),
> i think it's better to use it because ddb needs to do busy-loop anyway.
Not here, xenconscn_getc() calls HYPERVISOR_yield() when it waits for
input, so it won't consume CPU. But HYPERVISOR_yield() won't return unless
interrupts are enabled.
--
Manuel Bouyer <bouyer@antioche.eu.org>
NetBSD: 26 ans d'experience feront toujours la difference
--