2010-02-25 19:52:17

by Ian Campbell

[permalink] [raw]
Subject: VMI and kmap_atomic_pte paravirt_op

Hi Alok,

How critical is the vmi_kmap_atomic_pte pv_mmu_op hook to VMI? Is it
required for correct functionality or is it simply an optimisation?

We think we may be able to do away with the Xen version of the hook and
if VMI (which is deprecated) can live without it and you agree then I'll
cook up a patch to remove it altogether.

Thanks,
Ian.


2010-02-25 20:15:57

by Jeremy Fitzhardinge

[permalink] [raw]
Subject: Re: VMI and kmap_atomic_pte paravirt_op

On 02/25/2010 11:52 AM, Ian Campbell wrote:
> How critical is the vmi_kmap_atomic_pte pv_mmu_op hook to VMI? Is it
> required for correct functionality or is it simply an optimisation?
>

Or disable high ptes for VMI too.



J

2010-02-25 20:32:13

by Alok Kataria

[permalink] [raw]
Subject: Re: VMI and kmap_atomic_pte paravirt_op

Hi Ian,

On Thu, 2010-02-25 at 11:52 -0800, Ian Campbell wrote:
> Hi Alok,
>
> How critical is the vmi_kmap_atomic_pte pv_mmu_op hook to VMI? Is it
> required for correct functionality or is it simply an optimisation?
>
> We think we may be able to do away with the Xen version of the hook and
> if VMI (which is deprecated) can live without it and you agree then I'll
> cook up a patch to remove it altogether.

As Jeremy suggests we will be fine with disabling high ptes for VMI.
Thanks for doing this.

Alok
>
> Thanks,
> Ian.
>
>

2010-02-26 09:04:39

by Ian Campbell

[permalink] [raw]
Subject: Re: VMI and kmap_atomic_pte paravirt_op

On Thu, 2010-02-25 at 20:32 +0000, Alok Kataria wrote:
> Hi Ian,
>
> On Thu, 2010-02-25 at 11:52 -0800, Ian Campbell wrote:
> > Hi Alok,
> >
> > How critical is the vmi_kmap_atomic_pte pv_mmu_op hook to VMI? Is it
> > required for correct functionality or is it simply an optimisation?
> >
> > We think we may be able to do away with the Xen version of the hook and
> > if VMI (which is deprecated) can live without it and you agree then I'll
> > cook up a patch to remove it altogether.
>
> As Jeremy suggests we will be fine with disabling high ptes for VMI.
> Thanks for doing this.

OK, I'll do that.

Thanks,
Ian.