Re: [PATCH] VT-d: add iommu=igfx_off option to workaround graphics issues

From: "Tian, Kevin" <kevin.tian@intel.com>
To: Andrew Cooper <andrew.cooper3@citrix.com>,
	Ting-Wei Lan <lantw44@gmail.com>,
	"xen-devel@lists.xen.org" <xen-devel@lists.xen.org>
Cc: "Zhang, Yang Z" <yang.z.zhang@intel.com>,
	Jan Beulich <jbeulich@suse.com>
Subject: Re: [PATCH] VT-d: add iommu=igfx_off option to workaround graphics issues
Date: Tue, 21 Jul 2015 00:57:16 +0000	[thread overview]
Message-ID: <AADFC41AFE54684AB9EE6CBC0274A5D1262CDE56@SHSMSX101.ccr.corp.intel.com> (raw)
In-Reply-To: <55ACAF80.4000608@citrix.com>

> From: Andrew Cooper [mailto:amc96@hermes.cam.ac.uk] On Behalf Of Andrew Cooper
> Sent: Monday, July 20, 2015 4:21 PM
> 
> On 20/07/2015 02:28, Tian, Kevin wrote:
> >> From: Ting-Wei Lan [mailto:lantw44@gmail.com]
> >> Sent: Saturday, July 18, 2015 3:06 AM
> >>
> >> When using Linux >= 3.19 (commit 47591df) as dom0 on some Intel Ironlake
> >> devices, It is possible to encounter graphics issues that make screen
> >> unreadable or crash the system. It was reported in freedesktop bugzilla:
> >>
> >> https://bugs.freedesktop.org/show_bug.cgi?id=90037
> >>
> >> As we still cannot find a proper fix for this problem, this patch adds
> >> iommu=igfx_off option that is similar to Linux intel_iommu=igfx_off for
> >> users to manually workaround the problem.
> >>
> >> Signed-off-by: Ting-Wei Lan <lantw44@gmail.com>
> > Since igfx works before, I'd think a more proper fix should be on the
> > bisected Linux commit or i915 to have two working correctly together.
> > Otherwise this patch is just hiding problem.
> 
> The linux commit is the one which actually fixes PAT support for Linux
> under Xen.
> 
> It will cause the i915 driver to actually get WC mappings when it asks
> for them.

This is the part which I don't quite understand. WC is essentially an UC
attribute with write buffer to accelerate the write efficiency. There 
should be no correctness problem to use either WC or UC if i915 driver
wants WC.

> 
> > There is one possible usage to do selective IOMMU disable other than
> > global "iommu=off" switch. Then making this option general would
> > be better than igfx_off, e.g. based on BDF. But I'm not sure how it
> > is useful in reality.
> 
> It is curious that just disabling the IOMMU appears to fix the problem.
> Are there any errata you are aware of on this class of system?
> 

No such errata in my mind. It's not caused by IOMMU HW, since 
dom0-passthrough can't solve the problem neither. It's possible our 
pvMMU code may explicitly check presence of IOMMU to behave 
differently on PAT setting, but I'm not familiar with that part.

Thanks
Kevin