On Fri, 2015-01-09 at 10:32 +0200, Or Gerlitz wrote: > On Wed, Jan 7, 2015 at 5:04 PM, Or Gerlitz <[email protected]> wrote: > > From: Erez Shitrit <[email protected]> > > > > Following commit 016d9fb25cd9 "IPoIB: fix MCAST_FLAG_BUSY usage" > > both IPv6 traffic and for the most cases all IPv4 multicast traffic > > aren't working. > > Doug, can you ack the breakage introduced by your commit and the fix?
I haven't double checked the breakage, I'll take your word for it (at the time I did my work, I had multicast debugging on and I verified the join/leave process, but I had assumed that the process would work the same for optional multicast groups as it does for the IPoIB broadcast group and other default IPoIB groups, so I didn't specifically test additional multicast groups above and beyond the broadcast/etc groups). However, the fix is not workable. In particular, as soon as this patch is added to the kernel, you will start getting messages like this: mlx4_ib0: ipoib_mcast_leave on an in-flight join Every time you get this message, you've run into a "shouldn't ever happen" situation. If this happens, then we've lost track of the mcast flags settings or we've genuinely tried to remove a mcast group where the lower layer is still working on our join. Either way, it means we've screwed up. Further, with this patch in place, I'm seeing random acts of badness now with non-default IPoIB pkey joins. Sometimes they work, sometimes they don't. So, no, this patch doesn't work. I'll do some more investigating and report back. > Or. > -- > To unsubscribe from this list: send the line "unsubscribe linux-rdma" in > the body of a message to [email protected] > More majordomo info at http://vger.kernel.org/majordomo-info.html -- Doug Ledford <[email protected]> GPG KeyID: 0E572FDD
signature.asc
Description: This is a digitally signed message part
