<br><br>
<div class="gmail_quote">On Fri, Jul 31, 2009 at 2:51 PM, Yossi Etigin <span dir="ltr"><<a href="mailto:yosefe@voltaire.com">yosefe@voltaire.com</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="PADDING-LEFT: 1ex; MARGIN: 0px 0px 0px 0.8ex; BORDER-LEFT: #ccc 1px solid">
<div class="im">On 31/07/09 20:50, Hal Rosenstock wrote:<br>><br>><br>> On Fri, Jul 31, 2009 at 12:13 PM, Yossi Etigin <<a href="mailto:yosefe@voltaire.com">yosefe@voltaire.com</a><br></div>
<div class="im">> <mailto:<a href="mailto:yosefe@voltaire.com">yosefe@voltaire.com</a>>> wrote:<br>><br>> On 29/07/09 19:35, Hal Rosenstock wrote:<br>> ><br>> ><br>> > On Wed, Jul 29, 2009 at 10:22 AM, Moni Shoua <<a href="mailto:monis@voltaire.com">monis@voltaire.com</a><br>
> <mailto:<a href="mailto:monis@voltaire.com">monis@voltaire.com</a>><br></div>
<div>
<div></div>
<div class="h5">> > <mailto:<a href="mailto:monis@voltaire.com">monis@voltaire.com</a> <mailto:<a href="mailto:monis@voltaire.com">monis@voltaire.com</a>>>> wrote:<br>> ><br>> > Hal Rosenstock wrote:<br>
> > ><br>> > ><br>> > > On Wed, Jul 29, 2009 at 8:16 AM, Moni Shoua<br>> <<a href="mailto:monis@voltaire.com">monis@voltaire.com</a> <mailto:<a href="mailto:monis@voltaire.com">monis@voltaire.com</a>><br>
> > <mailto:<a href="mailto:monis@voltaire.com">monis@voltaire.com</a> <mailto:<a href="mailto:monis@voltaire.com">monis@voltaire.com</a>>><br>> > > <mailto:<a href="mailto:monis@voltaire.com">monis@voltaire.com</a> <mailto:<a href="mailto:monis@voltaire.com">monis@voltaire.com</a>><br>
> <mailto:<a href="mailto:monis@voltaire.com">monis@voltaire.com</a> <mailto:<a href="mailto:monis@voltaire.com">monis@voltaire.com</a>>>>> wrote:<br>> > ><br>> > ><br>
> > > > > Is the LMC in the above from the local port ?<br>> > Unfortunately, LMC<br>> > > > is not<br>> > > > > required to be uniform across the subnet so the<br>
> remote<br>> > > port's LMC may<br>> > > > > not be the same as that on the local port.<br>> > > ><br>> > > > opensm configures the same LMC to all endports<br>
> (CA) in the<br>> > > fabric so<br>> > > > in which case do you suspect that<br>> > > > it will be different?<br>> > > ><br>
> > > ><br>> > > > I wasn't talking about the simplification that OpenSM uses<br>> > but rather<br>> > > > what IBA allows. There may be other SMs which do this<br>
> or OpenSM<br>> > > could be<br>> > > > extended (with additional configuration for this).<br>> > > ><br>> > > > -- Hal<br>
> > ><br>> > > So I guess a possible solutions for that is to query each<br>> > suspected<br>> > > node before making a decision to flush it.<br>> > > When getting the node info response the true LMC can be<br>
> put in the<br>> > > LID check function.<br>> > ><br>> > ><br>> > > I'm not following how you are saying to determine the true<br>> LMC (of the<br>
> > > remote port).<br>> > ><br>> > Send PortInfo query for that port<br>> ><br>> ><br>> > from the IPoIB client ? I think this ends up needing to contact the SA<br>
> > to get this info rather than the port directly. Isn't that what<br>> you were<br>> > trying to avoid originally ? If that's the case, one way would've been<br>> > if ARPs carried the LMC as well as the LID.<br>
> ><br>><br>> What if we query the remote port LMC once, when the path is<br>> resolved, and then<br>> use it to mask the LID until the path is refreshed again?<br>><br>><br>> Just like the LID, remote LMC could change.<br>
><br><br></div></div>Yes, but AFAIK the only "bad" case is if the LID stays the same but LMC changes to a lower<br>value. In this case the path refresh will not happen when it is supposed to.</blockquote>
<div> </div>
<div>What's the impact of that ?</div>
<div> </div>
<div>Also the LID can change at the same time as the LMC.</div>
<div> </div>
<div>I can't tell if all the possible cases are handled properly. Are they ?</div>
<div> </div>
<blockquote class="gmail_quote" style="PADDING-LEFT: 1ex; MARGIN: 0px 0px 0px 0.8ex; BORDER-LEFT: #ccc 1px solid"><span id=""></span>If LMC changes to a higher value it might trigger an unneeded path refresh, which is not bad because it will refresh the LMC as well.</blockquote>
<div> </div></div>