On Tue, 2007-09-04 at 13:02 +0800, Cathy Zhou wrote:
> >>>> - Why we need the notification quiesce twice? One on Line 1426-1429,
> >>>> another
> >>>> on Line 1439-1448?
> >>> The first one isn't strictly needed, and I could remove it with no real
> >>> harm, but it does allow for any pending notifications to finish delivery
> >>> before shutting down the thread. This isn't on a critical hot path, so
> >>> I figured it was more polite to allow them to complete than to nuke 'em
> >>> from orbit.
> >> Can this be achieved by changing i_mac_notify_thread() to first process
> >> all
> >> pending notifications before it exits the thread?
> >
> > Well, there is also the problem that I'd really prefer the notifications
> > to finish *before* commiting to destroy the thread, but I'd prefer not
> > to commit to destroy the thread until after the dls_destroy() is called.
> > This helps with a more graceful recovery if dls_destroy() fails for some
> > reason.
> >
> I meant to change i_mac_notify_thread, and delete line 1424-1429. So that
> you donot need to recovery if dls_destroy() fails.
Right, but in that case I'm sending notifications *after* dls_destroy().
Is that safe? I'm not sure about that. Admittedly the relationship
between dld, dls, and mac sort of "mystifies" me a bit...
>
> > Right now it Just Works, so unless you really think it should be
> > altered, I'd prefer to leave it alone.
> >
> Okay. I don't feel strongly about it.
Cool. I'll leave it as is.
-- Garrett
>
> Thanks
> - Cathy
>
>
> _______________________________________________
> networking-discuss mailing list
> [email protected]
_______________________________________________
networking-discuss mailing list
[email protected]
