Re: [RFC PATCH 3/3] tools, libxl: handle the iomem parameter with the memory_mapping hcall

[Arianna Avanzini <avanzini.arianna@gmail.com>]

On 03/03/2014 04:13 PM, Julien Grall wrote:
> 
> 
> On 03/03/14 18:32, Dario Faggioli wrote:
>> Right. So, just to make sure I understand it correctly, you're saying
>> that it is fine to have the hypercall implemented and called sort of
>> like how Arianna is doing it, modulo moving the call outside of common
>> ARM and x86 code, as Julien pointed out, is this correct? If yes, I
>> (FWIW) totally concur. :-)
> 
> I think in both solution, this hypercall should be implemented :).
> 
>> For the sake of this series, do you think there is something that
>> Arianna should do, if not to ensure 100% consistency (which is not Xen's
>> job, in this case) but, perhaps, just to put things on the safe side?
>>
>> Right. Again, and sorry for not getting it just from the above, do you
>> think this is something Arianna should take care of, when reposting the
>> series, or can/should it be done also later?
> 
> I think Arianna solution's might not work if the 1:1 mapping clash with existing
> layout. I might help Arianna on this subject because it will be useful to
> passthrough a device to the guest.
> 

Sorry for the delay and for bothering you with another question. In order to
ensure that the 1:1 mapping of a new range of I/O memory does not clash with the
existing layout, we should then check that nothing else has been mapped to that
address range in the domU's address space?
Again, sorry for the probably obvious question.



-- 
/*
 * Arianna Avanzini
 * avanzini.arianna@gmail.com
 * 73628@studenti.unimore.it
 */