Re: [PATCH v3 01/11] PCI/P2PDMA: Support peer-to-peer memory

[Jason Gunthorpe <jgg-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>]

On Mon, Mar 26, 2018 at 12:11:38PM +0100, Jonathan Cameron wrote:
> On Tue, 13 Mar 2018 10:43:55 -0600
> Logan Gunthorpe <logang-OTvnGxWRz7hWk0Htik3J/w@public.gmane.org> wrote:
> 
> > On 12/03/18 09:28 PM, Sinan Kaya wrote:
> > > On 3/12/2018 3:35 PM, Logan Gunthorpe wrote:
> > > Regarding the switch business, It is amazing how much trouble you went into
> > > limit this functionality into very specific hardware.
> > > 
> > > I thought that we reached to an agreement that code would not impose
> > > any limits on what user wants.
> > > 
> > > What happened to all the emails we exchanged?  
> > 
> > It turns out that root ports that support P2P are far less common than 
> > anyone thought. So it will likely have to be a white list.
> 
> This came as a bit of a surprise to our PCIe architect.

I don't think it is a hardware problem.

I know Mellanox and Nvidia have been doing p2p on Intel root complexes
for something like 5-6 years now.. I don't have the details, but it
does seem to work.

I have heard some chips give poor performance..

Also AMD GPU SLI uses P2P these days, so this isn't exactly a niche
feature in Intel/AMD land.

I think the main issue here is that there is some BIOS involvement to
set things up properly. Eg in GPU land motherboards certify for
'crossfire' support.

> His follow up was whether it was worth raising an ECR for the PCIe spec
> to add a capability bit to allow this to be discovered.  This might
> long term avoid the need to maintain the white list for new devices.

If it is primarily a BIOS issue then it should be an ACPI thing, right?

Jason