[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] AMD IOMMU disabled - No Perdev Intremap



Sunday, October 6, 2013, 4:45:18 PM, you wrote:

> On 06/10/2013 15:38, Sander Eikelenboom wrote:
>> Sunday, October 6, 2013, 3:09:54 PM, you wrote:
>>
>>> On 10/06/2013 02:58 PM, Andrew Cooper wrote:
>>>> On 06/10/2013 13:56, Ferdinand Nölscher wrote:
>>>>> Hi Sander,
>>>>>
>>>>> Thanks for your reply.
>>>>>
>>>>> On 10/06/2013 02:10 PM, Sander Eikelenboom wrote:
>>>>>> Sunday, October 6, 2013, 1:36:33 PM, you wrote:
>>>>>>
>>>>>>> Hi!
>>>>>>>> From other people posting to this list, I know that there has
>>>>>>>> been a
>>>>>>> bug related to the issue described in Xen Security Advisory 36
>>>>>>> that disables iommu for some AMD users like me.
>>>>>> What motherboard do you have ?
>>>>> I have an ASUS M5A99X EVO Rev 1.01
>>>>>
>>>>>
>>>>>> It depends if it's the north or southbridge ioapic but try using
>>>>>> the new xen boot parameter: (add it to the xen.gz line in grub)
>>>>>>
>>>>>> ivrs_ioapic[0]=00:14.0
>>>>>>
>>>>>> or
>>>>>>
>>>>>> ivrs_ioapic[0]=00:00.1
>>>>> I tried both, it still fails to enable i/o virtualisation.
>>>> Can you post a full xl dmesg, booting with iommu=debug,verbose as well?
>>>>
>>>> ~Andrew
>>>>
>>>>
>> Ah it seems i got the id's mixed up :-)
>>
>> ivrs_ioapic[9]=00:00.1 ivrs_ioapic[10]=00:14.0
>>
>> Should do it i guess .. or should the ioapic handle be in hex .. (then the 
>> last should be ivrs_ioapic[a]=00:14.0 )
>>
>> --
>> Sander
>>
>>

> parse_ivrs_ioapic uses simple_strtoul(..., 0), so decimal or 0xa is
> fine, but 'a' on its own wont be.

> Seeing as I clearly don't understand how this fix is working, do you
> mind explaining (as you seem to have worked out the problem) ?

Bios provided tables are a mess .. regarding to iommu's it seems to be a double 
mess.
And motherboard manufacturers seem to be quite incapable to provide sane tables 
for years now, and it doesn't have any kind of priority.
(ok that little rant is out :-p )

But on (current / consumer) AMD systems, the mapping always seems to be the 
same ioapic northbridge to 00:00.1 and southbridge to 00:14.0.
So if you know the ioapic handle's from the xl dmesg output, you can easily 
determine the needed override.

It's perhaps a bit unfortunate that the error without iommu=debug,verbose 
doesn't show the handle but only the id:

(XEN) IVHD Error: Invalid IO-APIC 0
(XEN) AMD-Vi: Error initialization
(XEN) I/O virtualisation disabled

And it should probably be something for the next document day to have some 
documentation for it on the iommu part of the xen-wiki.

--
Sander

> ~Andrew



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


 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.