<html><body>
<p><tt>Jason,</tt><br>
<br>
<tt>>So, yes, currently, IPoIB is broken in that DAD for new addresses is<br>
>not synchronized to the SM join. But, DAD for startup addresses is<br>
>OK due to the trick that is played with carrier. Your patch breaks<br>
>both equally :></tt><br>
<br>
My patch doesn't break DAD that's I am tring to explain. The patch means the IPoIB link is ready to send data. That's how the net carrier defined. We shouldn't prevent the interface from UP and RUNNING because of there is a extremly small probability of duplicated link local address for IPv6. Especially when we have an IB spec for switch to support 1 to 16K MLIDs and IPv6 is a module enabled in kernel by default. For example, if the switch only supports 100 MLIDs, then we limit the fabric to 25 nodes cluster with 4 links on each.<br>
<br>
The IPoIB DAD should be fixed, like return something back to IPv6 and not set the flag to permenent. It has nothing with my patch. This patch just has a window to explore this problem.<br>
<br>
For IPv4 and IPv6, the carrier is ON has nothing to do with the IP address.<br>
<br>
Thanks<br>
Shirley Ma</body></html>