On Tue, Mar 01, 2016 at 09:07:42PM +0200, Eli Cohen wrote: > On Tue, Mar 01, 2016 at 11:32:56AM -0700, Jason Gunthorpe wrote: > > > > How does this interact with the existing SRIOV stuff? It is already > > very annoying that mlx5 is incompatible with the old scheme. Is the > > proposal to shift all IB to this new scheme, or still keep mlx4/mlx5 with > > different approaches? > > Now we have proper interfaces so we can use these interfaces to > implement the same functionality in mlx4 which is there but accessible > in a different manner. > > We can't just have the kernel become incompatible with existing > > SMs. Eg opensm only supports the old scheme last I looked. > > You need a SM which support virtualizaion to have virtualization > supported but if you don't you can still work with physical functions > in the same way you did before so we don't break anything, we just > adding new functionality. I mean opensm supports the GUID Alias scheme for virtualization, this new virtualization scheme is not compatible, and we shouldn't have the kernel drop support for existing working SMs, by, eg, replacing the mlx4 guid alias scheme with this new scheme. I'm guessing a user controlled switch is going to be necessary here to pick GUID alias or port port virtualization. Jason -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
