netdevice.h: Add RDMA private pointer to the net_device structure

netdevice.h: Add RDMA private pointer to the net_device structure

@@ -366,6 +366,7 @@ struct net_device
        void                    *ip6_ptr;       /* IPv6 specific data */
        void                    *ec_ptr;        /* Econet specific data */
        void                    *ax25_ptr;      /* AX.25 specific data */
+       void                    *rdma_ptr;      /* RDMA specific data */
 
 /*
  * Cache line mostly used on receive path (including eth_type_trans())

Who is going to use it? Is RDMA being submitted for code review?

Stephen Hemminger <[email protected]>
OSDL http://developer.osdl.org/~shemminger

IB (and ultimately RDMA) will use it.

-- Hal

Stephen> Who is going to use it? Is RDMA being submitted for code Stephen> review?

I agree that we should hold off on this until there's an in-tree user. However, just as a clarification, we're trying to move from "ib" to "rdma" nomenclature as we try to make the existing kernel InfiniBand layer a more generic layer than can support both IB and iWARP. So new code should use "rdma" names.

- R.

Hal> netdevice.h: Add RDMA private pointer to the net_device structure

I don't think there's any point in making this change until we have some code that will use the pointer.

- R.

We will have this shortly. I have been waiting for this to propose the changes to SDP et al.

-- Hal

Hal> We will have this shortly. I have been waiting for this to Hal> propose the changes to SDP et al.

OK, but I don't think it makes sense to merge this upstream until there is in-tree code that will use it.

- R.

I wanted to get this in so I could add the code to IPoIB to use this so SDP and others no longer poke at IPoIB's private data. This is a small change. Should this change be made locally (in OpenIB) first (and we'll have our own modified netdevice.h for a short time) ?

-- Hal

Hal> I wanted to get this in so I could add the code to IPoIB to Hal> use this so SDP and others no longer poke at IPoIB's private Hal> data. This is a small change. Should this change be made Hal> locally (in OpenIB) first (and we'll have our own modified Hal> netdevice.h for a short time) ?

Yes, I think that's the way to develop this sort of thing.

- R.

From: Roland Dreier <[email protected]>
Date: Mon, 03 Oct 2005 14:29:08 -0700

I definitely agree.