On 9/25/2018 9:28 AM, Jason Gunthorpe wrote: > On Tue, Sep 25, 2018 at 08:52:48AM +0300, Leon Romanovsky wrote: >>> I was thinking on this some more, and we don't really need to wait for >>> netlink or to add uverbs stuff to netlink. >>> >>> We should add a new ioctl to return the 'driver description' from >>> uverbs. >>> >>> This changes the rdma-core flow a bit as it cannot bind the driver >>> until after opening the uverbs FD, but it sure makes more sense .. >> >> It may be helpful for rdma-core, but for Steve's work it won't. He is >> not supposed to touch uverbs file just to understand on which software >> RDMA device to operate. It means that knowledge of driver_id will be >> exposed over netlink anyway, and I think that it won't be cleanest thing >> to duplicate interfaces. > > Steve's problem is entirely inside rdma-core though.. > > Jason > Correct. I have a series that uses a new netlink messages to create rxe links, and name them. It works fine. (I'll post this as an RFC soon. But I should rebase it on Jason's device name series). The remaining unresolved issue is that the rdma core associates uverb devices with the rxe provider by requiring the rxe device name to begin with "rxe". Steve. Steve.
