On Fri, 2010-05-14 at 09:44 -0700, Yinghai Lu wrote: > On 05/14/2010 01:30 AM, David Miller wrote: > > From: Benjamin Herrenschmidt <benh@xxxxxxxxxxxxxxxxxxx> > > Date: Fri, 14 May 2010 18:19:47 +1000 > > > > > >> OK, I'll have to let Davem deal with the fine point of the sparc bits, > >> but I think basically sparc has CONFIG_ARCH_POPULATES_NODE_MAP set, but > >> the way it's NUMA affinity works, the early_node_map[] is crap, you > >> cannot rely on the ranges in there. > >> > > > > > Right, we can't use early_node_map[] on sparc, because the NUMA > > mappings are far too granular to use that kind of representation. > > > good to know. > > early_node_map[] doesn't have enough slot? I think that's the case. sparc doesn't override MAX_ACTIVE_REGIONS, which means you get the default which is 256 or 50 per node depending on MAX_NUMNODES. My understanding is that this may not be enough. Dave, what does it look like in practice tho ? Cheers, Ben. -- To unsubscribe from this list: send the line "unsubscribe linux-arch" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html