RE: [Xen-devel] We still have to disable vtd for old chipset?

["Kay, Allen M" <allen.m.kay@xxxxxxxxx>]

Queued invalidation and Interrupt Remapping are for Nahalem generation
of HW.  The messages are probably too verbose.

Nested PCI bridge means you the devices behind more than 1 level of pci
bridges are found.  As we don't support this in our code yet, these
devices most likely don't work.

Try PCI-e device passthrough first.  They should work for you.

We are thinking of implementing a new "xm" tool to list the devices that
are available for passthrough to make this stuff easier.

Allen

>-----Original Message-----
>From: Neo Jia [mailto:neojia@xxxxxxxxx] 
>Sent: Tuesday, June 03, 2008 11:39 AM
>To: Ian Jackson
>Cc: Kay, Allen M; xen-devel@xxxxxxxxxxxxxxxxxxx
>Subject: Re: [Xen-devel] We still have to disable vtd for old chipset?
>
>I just tried it by removing the checking. The following is from my xm
>dmesg. Does those warning matter?
>
>
>(XEN) [VT-D]iommu.c:1722: Queued Invalidation hardware not found
>(XEN) [VT-D]iommu.c:1722: Queued Invalidation hardware not found
>(XEN) [VT-D]iommu.c:1722: Queued Invalidation hardware not found
>(XEN) [VT-D]iommu.c:1722: Queued Invalidation hardware not found
>(XEN) [VT-D]iommu.c:1730: Interrupt Remapping hardware not found
>(XEN) [VT-D]iommu.c:1730: Interrupt Remapping hardware not found
>(XEN) [VT-D]iommu.c:1730: Interrupt Remapping hardware not found
>(XEN) [VT-D]iommu.c:1730: Interrupt Remapping hardware not found
>(XEN) I/O virtualisation enabled
>(XEN) [VT-D]iommu.c:1241:d32767 context_context_mapping: nested PCI
>bridge not supported: bdf = 0:9:0 sec_bus = 3 sub_bus = 7
>(XEN) [VT-D]iommu.c:1241:d32767 context_context_mapping: nested PCI
>bridge not supported: bdf = 3:0:0 sec_bus = 4 sub_bus = 6
>(XEN) *** LOADING DOMAIN 0 ***
>
>
>Thanks,
>Neo
>
>On Tue, Jun 3, 2008 at 8:36 AM, Ian Jackson 
><Ian.Jackson@xxxxxxxxxxxxx> wrote:
>> Kay, Allen M writes ("RE: [Xen-devel] We still have to 
>disable vtd for old chipset?"):
>>> It should work assuming your BIOS is reporting VT-d 
>correctly.  At least
>>> the page table sharing should no longer be the issue.
>>
>> Are you sure ?  My understanding (admittedly half-remembered 
>from some
>> technical note that came with the motherboard replacement to 
>bring our
>> pre-C0 pre-production VT-d system up to date) was that the iommu page
>> table format is different in the pre-C0 chipsets.
>>
>> That would explain why pagetable sharing between iommu and normal mmu
>> is not possible with those chipsets.  But it also makes those pre-C0
>> chipsets completely incompatible with all of the code in Xen, because
>> Xen doesn't know anything about those pre-C0 pagetables ?  (And isn't
>> going to be taught.)
>>
>> Ian.
>>
>
>
>
>-- 
>I would remember that if researchers were not ambitious
>probably today we haven't the technology we are using!
>

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel