Re: [Xen-users] Enabling AMD-Vi IOMMU panics Xen

[‍小太 <nospam@xxxxxxxx>]

On 7 July 2016 at 03:56, ‍小太 <nospam@xxxxxxxx> wrote:

On 7 July 2016 at 01:16, ‍小太 <nospam@xxxxxxxx> wrote:

So maybe the cause of the panic would be that the IVRS table is missing type 43 entries for these bridges that contain the NICs?

I'll try adding those entries to the table and see if I'm successful.

Well I gave it a try, and I was unsuccessful. I've attached the table I used.

(I also realised after creating it that I probably should have used the type 42 ("Alias select") entries instead, but it shouldn't make a difference here)

With the new IVRS table, Xen panics soon after booting dom0 with:

(XEN) ****************************************
(XEN) Panic on CPU 3:
(XEN) Xen BUG at iommu_intr.c:451
(XEN) ****************************************

 and if I boot directly into Linux, it tell me:

[    3.885914] AMD-Vi: Using IVRS reported alias 00:02.2 for device 0000:01:00.0[8086:157b], kernel reported alias 01:00.0
[    3.896731] iommu: Adding device 0000:01:00.0 to group 1
[    3.902069] AMD-Vi: Using IVRS reported alias 00:02.3 for device 0000:02:00.0[8086:157b], kernel reported alias 02:00.0
[    3.912938] iommu: Adding device 0000:02:00.0 to group 1
[    3.918271] AMD-Vi: Using IVRS reported alias 00:02.4 for device 0000:03:00.0[8086:157b], kernel reported alias 03:00.0
[    3.929096] iommu: Adding device 0000:03:00.0 to group 1

so it seems like the NICs use their own device IDs despite being behind a PCI bridge, and thus this isn't the correct solution

I've just tested PCI passthrough with KVM instead of Xen (excluding the IVRS patch), and it works fine.

This leads me to believe it's a Xen bug - should I report this to xen-devel?

_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxx
https://lists.xen.org/xen-users