Re: [Qemu-devel] [Qemu-ppc] [PATCH v2 04/22] ppc/xics: add an InterruptStatsProvider interface to ICS and ICP objects

[Cédric Le Goater]

On 02/27/2017 12:43 AM, David Gibson wrote:
> On Fri, Feb 24, 2017 at 11:52:01AM +0100, Cédric Le Goater wrote:
>> On 02/23/2017 03:15 AM, David Gibson wrote:
>>> On Thu, Feb 16, 2017 at 02:47:27PM +0100, Cédric Le Goater wrote:
>>>> This is, again, to reduce the use of the list of ICS objects. Let's
>>>> make each individual ICS and ICP object an InterruptStatsProvider and
>>>> remove this same interface from XICSState.
>>>>
>>>> Signed-off-by: Cédric Le Goater <address@hidden>
>>>
>>> I'm a little hesitant about this, because it means that getting the
>>> interrupt stats information is now spread out over the qom tree,
>>> whereas previously there was a single location to get a good summary
>>> of the systems overall interrupt status.  The previous behaviour seems
>>> like it would be more convenient for debugging.
>>>
>>> That said, I see the structural advantages of this split.  Hmm.. still
>>> thinking..
>>
>> This is true. Another argument in favour of what you are saying 
>> is the order in which these are printed. See below.  
>>
>> What we could do after the cleanup is to make the machine an 
>> InterruptStatsProvider to clarify things.
> 
> Right.  So "info pic" does at least iterate through all the providers,
> but the semi-random order is pretty icky.  I think putting the stats
> provider on the machine would be a better idea - I guess it should be
> easy enough if the xics code provides a helper.

OK. This is on my TODO list for the next version of the patchset but as 
a followup patch. I don't want to change too much the initial cleanups
as he took me a while to find a working path to redo XICS. 

Thanks,

C.