[PATCH 0/4] More work towards neigh-less dsts.

[PATCH 0/4] More work towards neigh-less dsts.

[David Miller]

This continues the effort to divorce neighbour from dst_entry as
much as possible.

When we remove the routing cache, route entries will be fully generic
and not specific to any particular keyed destination.  Therefore we'll
need to lookup neigh entries dynamically, and quickly, at packet
output time.

To that end, this series:

1) Passes an explicit neighbour to packet out neigh_ops.

2) Abstracts all dst->neighbour behind helpers to ease the
   transition process.

3) Adds a new dst_ops entry, ->neigh_lookup(), to provide a transition
   scheme for dst->neighbour uses outside of the packet output path.

The next set of patches after this one will deal with all of
the dst_get_neighbour() code paths, converting them over to
dst_neigh_lookup()/neigh_release() sequences.

Then there will be a little bit of fiddling around in decnet since
it's usage of dst->neighbour is still too tight in the packet output
path.

Finally, we can kill dst->neighbour and do RCU neigh lookups in the
packet output paths for all the protocols using the neighbour cache
(ipv4, ipv6, decnet).

Re: [PATCH 0/4] More work towards neigh-less dsts.

[Steven Whitehouse]

Hi,

On Mon, 2011-07-18 at 01:24 -0700, David Miller wrote:
> This continues the effort to divorce neighbour from dst_entry as
> much as possible.
> 
> When we remove the routing cache, route entries will be fully generic
> and not specific to any particular keyed destination.  Therefore we'll
> need to lookup neigh entries dynamically, and quickly, at packet
> output time.
> 
> To that end, this series:
> 
> 1) Passes an explicit neighbour to packet out neigh_ops.
> 
> 2) Abstracts all dst->neighbour behind helpers to ease the
>    transition process.
> 
> 3) Adds a new dst_ops entry, ->neigh_lookup(), to provide a transition
>    scheme for dst->neighbour uses outside of the packet output path.
> 
> The next set of patches after this one will deal with all of
> the dst_get_neighbour() code paths, converting them over to
> dst_neigh_lookup()/neigh_release() sequences.
> 
> Then there will be a little bit of fiddling around in decnet since
> it's usage of dst->neighbour is still too tight in the packet output
> path.
> 
I've not actually tested the decnet changes, but I've read through them
all and I think it looks good. Likewise the neigh simplification patch
set.

Acked-by: Steven Whitehouse <swhiteho@redhat.com>

Steve.

> Finally, we can kill dst->neighbour and do RCU neigh lookups in the
> packet output paths for all the protocols using the neighbour cache
> (ipv4, ipv6, decnet).
> --
> To unsubscribe from this list: send the line "unsubscribe netdev" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html