[openib-general] Re: IPoIB question/problem

[Michael S. Tsirkin]

Quoting r. Roland Dreier <rolandd at cisco.com>:
> Subject: Re: IPoIB question/problem
> 
>     Michael> Does this analysis make sense? If yes, what would be the
>     Michael> best way to fix this?
> 
> I'm not sure if this could really happen or not.

This seems to be happening on some gen1 installations.

> But could we add a
> header_cache_update() method to the IPoIB struct net_device to handle
> this situation if it does occur?
> 
>  - R.
> 

It seems that we'll need to define hard_header_cache as well,
then. And having this appears to have performance implications, affecting
TCP code flow in a major way (for better or worse).
I was thinking that using the header cache may be a good replacement for
using the free space in the ha field, but this would be a major change.
What do you think?

A more modest approach that I was considering: keep a copy of the gid
as part of ipoib_neigh structure, and make sure that the gid didnt
change before posting a packet.
This seems to work for some of our gen1 clients.

-- 
MST