On Wed, Mar 14, 2018 at 12:03 PM, Logan Gunthorpe [off-list ref] wrote:

On 14/03/18 12:51 PM, Bjorn Helgaas wrote:

quoted

You are focused on PCIe systems, and in those systems, most topologies
do have an upstream switch, which means two upstream bridges.  I'm
trying to remove that assumption because I don't think there's a
requirement for it in the spec.  Enforcing this assumption complicates
the code and makes it harder to understand because the reader says
"huh, I know peer-to-peer DMA should work inside any PCI hierarchy*,
so why do we need these two bridges?"

Yes, as I've said, we focused on being behind a single PCIe Switch
because it's easier and vaguely safer (we *know* switches will work but
other types of topology we have to assume will work based on the spec).
Also, I have my doubts that anyone will ever have a use for this with
non-PCIe devices.

P2P over PCI/PCI-X is quite common in devices like raid controllers.
It would be useful if those configurations were not left behind so
that Linux could feasibly deploy offload code to a controller in the
PCI domain.