Re: [PATCH 4/7] ide: use the PCI memory access interface

[Avi Kivity <avi@redhat.com>]

  On 09/02/2010 06:01 PM, Eduard - Gabriel Munteanu wrote:
> On Thu, Sep 02, 2010 at 12:58:13PM +0300, Michael S. Tsirkin wrote:
>> On Thu, Sep 02, 2010 at 12:12:00PM +0300, Eduard - Gabriel Munteanu wrote:
>>> On Thu, Sep 02, 2010 at 08:19:11AM +0300, Michael S. Tsirkin wrote:
>>>> On Sat, Aug 28, 2010 at 05:54:55PM +0300, Eduard - Gabriel Munteanu wrote:
>> I don't insist on this solution, but what other way do you propose to
>> avoid the overhead for everyone not using an iommu?
>> I'm all for a solution that would help iommu as well,
>> but one wasn't yet proposed.
>>
> Hm, we could get even better performance by simply making the IOMMU a
> compile-time option. It also avoids problems in case some device hasn't
> been converted yet, and involves little to no tradeoffs. What do you
> think?
>
> AFAICT, there are few uses for the IOMMU besides development and
> avantgarde stuff, as you note. So distributions can continue supplying
> prebuilt QEMU/KVM packages compiled with the IOMMU turned off for the
> time being. The only practical (commercial) use right now would be in
> the case of private virtual servers, which could be divided further into
> nested guests (though real IOMMU hardware isn't widespread yet).
>
> Blue Swirl, in the light of this, do you agree on making it a
> compile-time option?

That's not a practical long term solution.  Eventually everything gets 
turned on.

I don't really see a problem with the additional indirection.  By the 
time we reach actual hardware to satisfy the request, we'll have gone 
through many such indirections; modern processors deal very well with them.

-- 
error compiling committee.c: too many arguments to function