[PATCH 0/3, v2] x86: xor-block handling adjustments

[PATCH 0/3, v2] x86: xor-block handling adjustments

[Jan Beulich]

v2 of this series is merely updated on top of the changes between
3.6 and 3.7-rc (which includes the dropping of what previously was
the second patch in a four patch series).

1: unify SSE-base xor-block routines
2: add alternative SSE implementation only prefetching once per 64-byte line
3: make virtualization friendly

Signed-off-by: Jan Beulich <jbeulich@suse.com>

Re: [PATCH 0/3, v2] x86: xor-block handling adjustments

[Ingo Molnar]

* Jan Beulich <JBeulich@suse.com> wrote:

> v2 of this series is merely updated on top of the changes between
> 3.6 and 3.7-rc (which includes the dropping of what previously was
> the second patch in a four patch series).
> 
> 1: unify SSE-base xor-block routines
> 2: add alternative SSE implementation only prefetching once per 64-byte line
> 3: make virtualization friendly
> 
> Signed-off-by: Jan Beulich <jbeulich@suse.com>

Looks useful. Wondering what the status is: hpa, was your 
concern resolved, do you think we can apply these?

Thanks,

	Ingo

Re: [PATCH 0/3, v2] x86: xor-block handling adjustments

[H. Peter Anvin]

On 01/24/2013 11:15 AM, Ingo Molnar wrote:
> 
> * Jan Beulich <JBeulich@suse.com> wrote:
> 
>> v2 of this series is merely updated on top of the changes between
>> 3.6 and 3.7-rc (which includes the dropping of what previously was
>> the second patch in a four patch series).
>>
>> 1: unify SSE-base xor-block routines
>> 2: add alternative SSE implementation only prefetching once per 64-byte line
>> 3: make virtualization friendly
>>
>> Signed-off-by: Jan Beulich <jbeulich@suse.com>
> 
> Looks useful. Wondering what the status is: hpa, was your 
> concern resolved, do you think we can apply these?
> 

I don't see anything wrong, except that I can't *find* patch 3/3 either
in my inbox nor on LKML...

	-hpa

Re: [PATCH 0/3, v2] x86: xor-block handling adjustments

[Jan Beulich]

>>> On 24.01.13 at 18:32, "H. Peter Anvin" <hpa@zytor.com> wrote:
> On 01/24/2013 11:15 AM, Ingo Molnar wrote:
>> 
>> * Jan Beulich <JBeulich@suse.com> wrote:
>> 
>>> v2 of this series is merely updated on top of the changes between
>>> 3.6 and 3.7-rc (which includes the dropping of what previously was
>>> the second patch in a four patch series).
>>>
>>> 1: unify SSE-base xor-block routines
>>> 2: add alternative SSE implementation only prefetching once per 64-byte line
>>> 3: make virtualization friendly
>>>
>>> Signed-off-by: Jan Beulich <jbeulich@suse.com>
>> 
>> Looks useful. Wondering what the status is: hpa, was your 
>> concern resolved, do you think we can apply these?
>> 
> 
> I don't see anything wrong, except that I can't *find* patch 3/3 either
> in my inbox nor on LKML...

I can certainly resend the whole set, but our email system even
got a delivery confirmation from yours on patch 3 (other than
e.g. from Ingo's, and all of this identical to 0...2).

Or see https://patchwork.kernel.org/patch/1688991/, according
to which you even responded to it (which is the concern Ingo was
referring to).

Jan

Re: [PATCH 0/3, v2] x86: xor-block handling adjustments

[Ingo Molnar]

* Jan Beulich <JBeulich@suse.com> wrote:

> >>> On 24.01.13 at 18:32, "H. Peter Anvin" <hpa@zytor.com> wrote:
> > On 01/24/2013 11:15 AM, Ingo Molnar wrote:
> >> 
> >> * Jan Beulich <JBeulich@suse.com> wrote:
> >> 
> >>> v2 of this series is merely updated on top of the changes between
> >>> 3.6 and 3.7-rc (which includes the dropping of what previously was
> >>> the second patch in a four patch series).
> >>>
> >>> 1: unify SSE-base xor-block routines
> >>> 2: add alternative SSE implementation only prefetching once per 64-byte line
> >>> 3: make virtualization friendly
> >>>
> >>> Signed-off-by: Jan Beulich <jbeulich@suse.com>
> >> 
> >> Looks useful. Wondering what the status is: hpa, was your 
> >> concern resolved, do you think we can apply these?
> >> 
> > 
> > I don't see anything wrong, except that I can't *find* patch 
> > 3/3 either in my inbox nor on LKML...

I think it's just that this is an older thread I found, possibly 
out of your folder already.

They are looking good to me and I've applied them to x86/asm to 
give it some testing as well. Please holler if you see something 
funny.

Thanks,

	Ingo

Re: [PATCH 0/3, v2] x86: xor-block handling adjustments

[H. Peter Anvin]

On 01/25/2013 12:21 AM, Jan Beulich wrote:
>>
>> I don't see anything wrong, except that I can't *find* patch 3/3 either
>> in my inbox nor on LKML...
> 
> I can certainly resend the whole set, but our email system even
> got a delivery confirmation from yours on patch 3 (other than
> e.g. from Ingo's, and all of this identical to 0...2).
> 
> Or see https://patchwork.kernel.org/patch/1688991/, according
> to which you even responded to it (which is the concern Ingo was
> referring to).
> 

Ah yes, now I have the context.

And with that, I have to NAK the patch.  I did talk to the KVM people
and they confirmed that this is NOT something they want.  They do not
program VT-x to VMEXIT on CR0 changes.

So this is Xen-specific, and penalizing other hypervisors is wrong.

	-hpa

Re: [PATCH 0/3, v2] x86: xor-block handling adjustments

[Ingo Molnar]

* H. Peter Anvin <hpa@zytor.com> wrote:

> On 01/25/2013 12:21 AM, Jan Beulich wrote:
> >>
> >> I don't see anything wrong, except that I can't *find* patch 3/3 either
> >> in my inbox nor on LKML...
> > 
> > I can certainly resend the whole set, but our email system even
> > got a delivery confirmation from yours on patch 3 (other than
> > e.g. from Ingo's, and all of this identical to 0...2).
> > 
> > Or see https://patchwork.kernel.org/patch/1688991/, according
> > to which you even responded to it (which is the concern Ingo was
> > referring to).
> > 
> 
> Ah yes, now I have the context.
> 
> And with that, I have to NAK the patch.  I did talk to the KVM 
> people and they confirmed that this is NOT something they 
> want.  They do not program VT-x to VMEXIT on CR0 changes.
> 
> So this is Xen-specific, and penalizing other hypervisors is 
> wrong.

Are patches #1 and #2 still OK?

Thanks,

	Ingo

Re: [PATCH 0/3, v2] x86: xor-block handling adjustments

[H. Peter Anvin]

Yes.

Ingo Molnar <mingo@kernel.org> wrote:

>
>* H. Peter Anvin <hpa@zytor.com> wrote:
>
>> On 01/25/2013 12:21 AM, Jan Beulich wrote:
>> >>
>> >> I don't see anything wrong, except that I can't *find* patch 3/3
>either
>> >> in my inbox nor on LKML...
>> > 
>> > I can certainly resend the whole set, but our email system even
>> > got a delivery confirmation from yours on patch 3 (other than
>> > e.g. from Ingo's, and all of this identical to 0...2).
>> > 
>> > Or see https://patchwork.kernel.org/patch/1688991/, according
>> > to which you even responded to it (which is the concern Ingo was
>> > referring to).
>> > 
>> 
>> Ah yes, now I have the context.
>> 
>> And with that, I have to NAK the patch.  I did talk to the KVM 
>> people and they confirmed that this is NOT something they 
>> want.  They do not program VT-x to VMEXIT on CR0 changes.
>> 
>> So this is Xen-specific, and penalizing other hypervisors is 
>> wrong.
>
>Are patches #1 and #2 still OK?
>
>Thanks,
>
>	Ingo

-- 
Sent from my mobile phone. Please excuse brevity and lack of formatting.