Mailing List Archive

Andi Kleen wrote:
> Do you have some lmbench numbers before/after this change?
> iirc at least fork and exit do a lot of pte accesses in various forms.
> If it's measurable it might be needed to patch those for the native case.
>

I don't. I think Rusty ran some numbers and found the pte accessors
became the most commonly used paravirt ops.

I'm working on revamping the pv_ops patching infrastructure at the
moment, so we should be able to deal with these fairly efficiently.

J
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

On Wed, 2007-02-21 at 23:15 +0100, Andi Kleen wrote:
> On Wednesday 21 February 2007 21:52, Jeremy Fitzhardinge wrote:
> > Add a set of accessors to pack, unpack and modify page table entries
> > (at all levels). This allows a paravirt implementation to control the
> > contents of pgd/pmd/pte entries. For example, Xen uses this to
> > convert the (pseudo-)physical address into a machine address when
> > populating a pagetable entry, and converting back to pphys address
> > when an entry is read.
>
> Do you have some lmbench numbers before/after this change?
> iirc at least fork and exit do a lot of pte accesses in various forms.
> If it's measurable it might be needed to patch those for the native case.

Yes, __mkpte must be patched to avoid performance embarrassment.

Jeremy, did you want me to do this, or are you happy to?
Rusty.
PS. I really must revise my "paravirt-ops counter" patch which tallies
how much each op is getting called.


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

Rusty Russell wrote:
> On Wed, 2007-02-21 at 23:15 +0100, Andi Kleen wrote:
>
>> On Wednesday 21 February 2007 21:52, Jeremy Fitzhardinge wrote:
>>
>>> Add a set of accessors to pack, unpack and modify page table entries
>>> (at all levels). This allows a paravirt implementation to control the
>>> contents of pgd/pmd/pte entries. For example, Xen uses this to
>>> convert the (pseudo-)physical address into a machine address when
>>> populating a pagetable entry, and converting back to pphys address
>>> when an entry is read.
>>>
>> Do you have some lmbench numbers before/after this change?
>> iirc at least fork and exit do a lot of pte accesses in various forms.
>> If it's measurable it might be needed to patch those for the native case.
>>
>
> Yes, __mkpte must be patched to avoid performance embarrassment.
>
> Jeremy, did you want me to do this, or are you happy to?
>
I'm working on it now. I'll mail out a proposal in a little bit.

J
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/