[PATCH] arm64: KVM: fix unmapping with 48-bit VAs

[Christoffer Dall]

On Thu, Oct 30, 2014 at 10:23 AM, Marc Zyngier <marc.zyngier at arm.com> wrote:
> On 30/10/14 09:15, Christoffer Dall wrote:
>> On Thu, Oct 30, 2014 at 10:08 AM, Marc Zyngier <marc.zyngier at arm.com> wrote:
>>> On 30/10/14 09:00, Christoffer Dall wrote:
>>>> On Wed, Oct 29, 2014 at 10:39:45AM +0000, Marc Zyngier wrote:
>>>>> Hi Mark,
>>>>>
>>>>> On 28/10/14 19:36, Mark Rutland wrote:
>>>>>> Hi All,
>>>>>>
>>>>>> I'm not at all familiar with this code, so my analysis below might be off. Does
>>>>>> this look right or am I just masking a more fundamental issue?
>>>>>>
>>>>>> I've given this a go with 4KiB pages on Juno with {39,48}-bit VAs and the
>>>>>> system seems to boot fine.
>>>>>
>>>>> Thanks for looking into this.
>>>>>
>>>>
>>>> yeah, sorry for missing this!
>>>>
>>>> [...]
>>>>
>>>>>
>>>>> Looks good to me!
>>>>>
>>>>> Acked-by: Marc Zyngier <marc.zyngier at arm.com>
>>>>>
>>>>> Christoffer, if you're OK with that one, I'll queue it up for the next -rc.
>>>>>
>>>> Yup (for the record we're not breaking anything because 48 bit VA still
>>>> depends on broken, right?):
>>>
>>> Since 04f905a (arm64: Allow 48-bits VA space without ARM_SMMU), we can
>>> select 48bit VA if the SMMU is not selected.
>>>
>> Did Catlin merge this separately or did I accidentally do that?  I
>> thought we were holding off on that until we had done more testing...
>
> Catalin felt very brave, did some testing (LTP and stuff), fixed another
> nit (3dec0fe), sent a pull request to Linus and went on holiday!
>
> So far, it looks good. :-)
>
great, hellelujah for progress.