Re: [PATCH v3 00/14] Adding GAUDI NIC code to habanalabs driver

[Jakub Kicinski]

On Mon, 21 Sep 2020 08:52:39 -0300 Jason Gunthorpe wrote:
> I don't like this idea of backdooring a bunch of proprietary closed
> source RDMA userspace through drivers/misc, and if you don't have a
> clear idea how to get something equal for drivers/misc you should not
> accept the H_IOCTL_NIC.
> 
> Plus RoCE is complicated, there is a bunch of interaction with netdev
> and rules related to that that really needs to be respected.

+1

To me this code quite clearly fits the description of vendor SDK which
runs proprietary stuff on top. It's such an vendor SDK thing to do to
pick the parts of our infrastructure they like and "simplify the rest"
with its own re-implementation.

I'd wager the only reason you expose the netdevs at all is for link
settings, stats, packet capture and debug. You'd never run TCP traffic
over those links. And you're fighting against using Linux APIs for the
only real traffic that runs on those links - RDMA(ish) traffic.

Greg - I'm probably the least experience of the folks involved in this
conversation - could you ELI5 what's the benefit to the community from
merging this code?