When we take full hotplug to recover from EEH errors, PCI buses
could be involved. For the case, the child devices of involved
PCI buses can't be attached to IOMMU group properly, which is
caused by commit 3f28c5a ("powerpc/powernv: Reduce multi-hit of
iommu_add_device()").
When adding the PCI dev
I didn't manage to test this one PHB3, since some network issue, I can't
access the machine in Austin.
Will reply after I test this on PHB3.
On Tue, Aug 05, 2014 at 06:27:38PM +1000, Gavin Shan wrote:
>When we take full hotplug to recover from EEH errors, PCI buses
>could be involved. For the cas
I just give a shot with patch on PHB3 with hotplug case. From the functional
point of view, it works. The iommu group is created and attached with device.
While it still has the problem we had which commit 3f28c5a tried to solve.
In the test case, the pci bus is not removed. This means in
pcibios_
On Tue, Aug 05, 2014 at 06:00:36PM +0800, Wei Yang wrote:
>I didn't manage to test this one PHB3, since some network issue, I can't
>access the machine in Austin.
>
>Will reply after I test this on PHB3.
>
Yep, thanks. By the way, I tested it on P7 box and looks good
for partial/full hotplug cases
On Wed, Aug 06, 2014 at 12:03:23AM +0800, Wei Yang wrote:
>I just give a shot with patch on PHB3 with hotplug case. From the functional
>point of view, it works. The iommu group is created and attached with device.
>While it still has the problem we had which commit 3f28c5a tried to solve.
>
>In th
On Wed, Aug 06, 2014 at 12:06:04PM +1000, Gavin Shan wrote:
>On Wed, Aug 06, 2014 at 12:03:23AM +0800, Wei Yang wrote:
>>I just give a shot with patch on PHB3 with hotplug case. From the functional
>>point of view, it works. The iommu group is created and attached with device.
>>While it still has