lkml.org 
[lkml]   [2019]   [Aug]   [14]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
SubjectRE: [PATCH v4,1/2] PCI: hv: Detect and fix Hyper-V PCI domain number collision
Date


> -----Original Message-----
> From: Bjorn Helgaas <helgaas@kernel.org>
> Sent: Wednesday, August 14, 2019 12:34 AM
> To: Haiyang Zhang <haiyangz@microsoft.com>
> Cc: sashal@kernel.org; lorenzo.pieralisi@arm.com; linux-
> hyperv@vger.kernel.org; linux-pci@vger.kernel.org; KY Srinivasan
> <kys@microsoft.com>; Stephen Hemminger <sthemmin@microsoft.com>;
> olaf@aepfle.de; vkuznets <vkuznets@redhat.com>; linux-
> kernel@vger.kernel.org
> Subject: Re: [PATCH v4,1/2] PCI: hv: Detect and fix Hyper-V PCI domain
> number collision
>
> Thanks for splitting these; I think that makes more sense.
>
> On Wed, Aug 14, 2019 at 12:38:54AM +0000, Haiyang Zhang wrote:
> > Currently in Azure cloud, for passthrough devices including GPU, the host
> > sets the device instance ID's bytes 8 - 15 to a value derived from the host
> > HWID, which is the same on all devices in a VM. So, the device instance
> > ID's bytes 8 and 9 provided by the host are no longer unique. This can
> > cause device passthrough to VMs to fail because the bytes 8 and 9 are used
> > as PCI domain number. Collision of domain numbers will cause the second
> > device with the same domain number fail to load.
>
> I think this patch is fine. I could be misunderstanding the commit
> log, but when you say "the ID bytes 8 and 9 are *no longer* unique",
> that suggests that they *used* to be unique but stopped being unique
> at some point, which of course raises the question of *when* they
> became non-unique.
>
> The specific information about that point would be useful to have in
> the commit log, e.g., is this related to a specific version of Azure,
> a configuration change, etc?
The host side change happened last year, rolled out to all azure hosts.
I will put "all current azure hosts" in the commit log.

> Does this problem affect GPUs more than other passthrough devices? If
> all passthrough devices are affected, why mention GPUs in particular?
> I can't tell whether that information is relevant or superfluous.

We found this issue initially on multiple passthrough GPUs, I mentioned this
just as an example. I will remove this word, because any PCI devices may
be affected.

Thanks,
- Haiyang

\
 
 \ /
  Last update: 2019-08-14 17:34    [W:0.051 / U:0.576 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site