Re: [Xen-devel] [PATCH] amd iommu: Dump flags of IO page faults

[Sander Eikelenboom <linux@xxxxxxxxxxxxxx>]

Friday, September 7, 2012, 12:06:31 PM, you wrote:

>>>> On 07.09.12 at 12:00, Sander Eikelenboom <linux@xxxxxxxxxxxxxx> wrote:

>> Friday, September 7, 2012, 11:53:36 AM, you wrote:
>> 
>>>>>> On 07.09.12 at 09:32, Sander Eikelenboom <linux@xxxxxxxxxxxxxx> wrote:
>>>> 4.1:    Capabilities: [54] MSI: Enable- Count=1/1 Maskable- 64bit+
>>>> 4.2:    Capabilities: [54] MSI: Enable+ Count=1/1 Maskable- 64bit+
>> 
>>> No surprise you're not seeing any faults on 4.1 - there's no way
>>> they could get reported. I'm somewhat hesitant to pull the
>>> workaround patch from 4.2 into 4.1, as it's wrong for the kernel
>>> to touch the MSI setting of the IOMMU (which is under the
>>> control of Xen) in the first place, but the kernel side patch I had
>>> submitted a while ago wasn't received well. And that patch isn't
>>> really small, and it would remain to be seen what other
>>> dependencies it would have...
>> 
>> Ok so that would mean that in the 4.1 case, the IOMMU is doing nothing ?

> No, it just can't report faults (they would need to be polled for).
> Also, saying "in the 4.1 case" is wrong here - this really depends
> on whether you have an affected Dom0 kernel. Things work fine
> if the Dom0 kernel doesn't trample over Xen's setup.

Except for the IO PAGE FAULT in xl dmesg, which is reported before the kernel 
even gets loaded with xen-4.2.
I don't see that one on 4.1 either, so that wouldn't add up to it being a dom0 
kernel problem only ...


> Jan




_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel