On Wed, Nov 23, 2011 at 02:01:54PM +0000, David Woodhouse wrote:
On Wed, 2011-11-23 at 11:26 +0100, Daniel Vetter wrote:
On Tue, Nov 22, 2011 at 07:31:34PM -0800, Keith Packard wrote:
On Tue, 22 Nov 2011 20:15:31 +0000, Matthew Garrett mjg@redhat.com wrote:
So the user has to choose between 5W of power saving or having dmar? And we default to giving them dmar? I think that's going to come as a surprise to people.
You'd have to go into the BIOS to turn this on for most machines at least?
But, yeah, it seems like we should be turning DMAR off unless explicitly requested; I can't understand how you'd ever need this running native on the hardware. Not exactly an area I care about deeply; I've always worked hard to make sure all virtualization garbage is disabled on every machine I use.
Problem is that we need to disable dmar on the entire box, afaics. And I assume that a bunch of people abusing desktop boards as servers will call "regression" on that.
Hm, do you really have to disable it for the entire box, or just the graphics?
At least for the dmar+gfx+semaphores hang I can reproduce, just disabling dmar with intel_iommu=igfx_off is not good enough and iirc the same holds for the dmar+rc6 hangs reported.
Do we have a coherent erratum from Intel for the issues mentioned above with DMAR+gfx+RC6?
Afaik no errata applies to our dmar related troubles on snb. I've hoped that ppgtt would magically fix this, and it seems to help quite a bit for the semaphore hangs (but not everywhere). Couldn't yet look more into this. -Daniel