On Wed, Dec 30, 2009 at 04:47:45PM -0600, Anthony Liguori wrote:
> On 12/30/2009 08:30 AM, Gleb Natapov wrote:
> >On Wed, Dec 30, 2009 at 04:26:11PM +0200, Gleb Natapov wrote:
> >>On Wed, Dec 30, 2009 at 08:16:30AM -0600, Anthony Liguori wrote:
> >>>On 12/30/2009 03:27 AM, Gleb Natapov wrote:
> >>>>On Tue, Dec 29, 2009 at 08:00:09PM -0600, Anthony Liguori wrote:
> >>>>>On 12/24/2009 09:02 AM, Gleb Natapov wrote:
> >>>>>>Knowing ioapic configuration is very useful for the poor soles
> >>>>>>how need to debug guest occasionally.
> >>>>>
> >>>>>Can this be implemented in terms of VMState?
> >>>>>
> >>>>What do you mean and why it whould be any better?
> >>>
> >>>We've been discussing having a command that's something like
> >>>'info-device ioapic' that would dump the device state based on
> >>>what's stored in VMState for the device.
> >>>
> >>>The advantage of a command like this is that it works for any qdev
> >>>device making it much more useful.
> >>>
> >>>The disadvantage is that at the moment, we store ioredtbl as an
> >>>array of uint64s. To get the level of info you're looking for we
> >>>would need to teach VMState how to decode this structure.
> >>>
> >>
> >>So this is completely orthogonal to my patch. There are devices already
> >>that have info command. No need to hold the patch just because sometime
> >>in the feature VMState may be extended. Other then that adding print
> >>function to VMState sounds like a good idea.
> >BTW I'd like this to be backported to 0.12 too.
>
> It's not a bug fix so that doesn't make sense.
>
It helps debug problems and it is not intrusive. Since 0.12 will be used
for a long time I want to add it there for convenience.
--
Gleb.
