Re: [PATCH v7] x86/p2m: use large pages for MMIO mappings

[George Dunlap <george.dunlap@citrix.com>]

On 10/02/16 10:06, Ian Campbell wrote:
> On Tue, 2016-02-09 at 05:35 -0700, Jan Beulich wrote:
>>> On 09.02.16 at 13:17, <george.dunlap@citrix.com> wrote:
>>> I don't think sometimes returning the number of things you did and
>>> sometimes returning zero makes any sense.  My suggestion would be
>>> either
>>> make "nr_mfns" bidirectional (as similar fields are in the other
>>> domctls) and return 0 on either full or partial success, or just return
>>> the number of mfns actually mapped either on full or partial success.
>>
>> As said - I can see your point, and I've been considering the
>> alternatives and had to decide for one. Since I've already got
>> Ian's approval for the currently implementation, and since we're
>> at v7 and I've already spent way more time on this than I had
>> expected, I hope you understand that I'm a little hesitant to
>> make more changes (perhaps even requiring re-obtaining acks,
>> which has by itself been taking long enough for this patch) than
>> absolutely necessary to get this in.
>>
>> So - Ian, do you think the alternative proposed by George
>> would make for a meaningfully better interface?
> 
> I can see his point, but for a domctl I don't think I'd be inclined to
> insist on changing it, given the reasons you explain above for not wanting
> to at this stage.
> 
> I'd most likely be inclined to ack a follow up patch (from whomsoever is
> motivated enough to produce one) which revved the API again though.

Yes, I can certainly understand just geting this off the plate.

If we fix the mmio_ro page size checks / assertion, I'm fine with the
current interface.

 -George