On 09/28/2010 11:58 AM, Jeremy Fitzhardinge wrote:
> On 09/28/2010 11:46 AM, H. Peter Anvin wrote:
>> Well, we're specifically talking about a virtual machine which has
>> direct access to hardware, so it is concerned about the real physical
>> memory properties of real physical pages. If we can assume that
>> BIOS/Xen will always set up MTRR correctly then there shouldn't be any
>> need for the kernel to modify the MTRR itself. How true is that in
>> general? I don't know, but if we could rely on BIOS then there'd never
>> be a need to touch MTRR, would there?
>> Well, in the past MTRRs were abused for device properties mainly by the
>> X server, but other than that, no, not really. The other thing we do is
>> the MTRR cleanup (which doesn't involve /proc/mtrr) to deal with
>> brokenness in the BIOS setup, but that really belongs in the hypervisor
>> in your case since it fundamentally affects how memory is handled.
> Yeah, the hypervisor should definitely deal with that. I have no
> problem in principle with leaving MTRRs entirely to Xen, but I was just
> concerned about possible repercussions. Certainly when I first did this
> work, I was using Fedora 8 whose X server did depend on /proc/mtrr for
> good performance.
Yeah, that should all be fixed now.
Xen-devel mailing list