[ofa-general] Re: [openib-general] Fw: [PATCH] enable IPoIB only if broadcast join finish

Shirley Ma xma at us.ibm.com
Tue Mar 6 12:52:27 PST 2007






Roland Dreier <rdreier at cisco.com> wrote on 03/06/2007 12:20:55 PM:
>  > I believe even IPv6 with ethernet, the interface will be UP and
RUNNING
>  > even they have a duplicated IPv6 address so IPv4 can work. I don't
know why
>  > we do thing differently here.
>
> That's not the point.  The point is that if we bring the interface up
> before the multicast groups are joined, then IPv6 DAD has a chance of
> incorrectly not detecting duplicate addresses (think about it -- if
> IPv6 autoconf/DAD starts before all the multicast groups are ready,
> then the IPv6 stack will generate ND and router solicitation messages,
> but they will just be queued up pending multicast join completion, and
> if that doesn't happen before the timeouts occur, then the IPv6 stack
> will incorrectly conclude that a duplicate address doesn't exist, even
> if it does)
The IPv6 stack will generate ND and router solicitation messages when
sending packet. The duplicated address can be detected anytime. Am I right?
So if the multicast join completion later, the duplication address will be
detect later.

I thought none of the packets can be sent out before the interface is up
and running. The link carrier is ON, doesn't mean packets can be sent, it
only means the carrier is ready to send data. I looked at the code again,
and found that the interface is UP in the beginning of the ipoib code. We
could move the UP in the end of operation.

> The real question is what tradeoff do we want to make for broken
> fabrics where some multicast groups can never be joined.  Is it worth
> the (small) risk of breaking IPv6 autoconf on good fabrics in order to
> behave more gracefully on broken fabrics?
>
>  - R.

My point is IPoIB should behave as other networking device drivers.

Thanks
Shirley Ma
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openfabrics.org/pipermail/general/attachments/20070306/52453e07/attachment.html>


More information about the general mailing list