From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from e39.co.us.ibm.com (e39.co.us.ibm.com [32.97.110.160]) (using TLSv1 with cipher DHE-RSA-AES256-SHA (256/256 bits)) (No client certificate requested) by lists.ozlabs.org (Postfix) with ESMTPS id 9846D1A0590 for ; Wed, 28 May 2014 09:44:31 +1000 (EST) Received: from /spool/local by e39.co.us.ibm.com with IBM ESMTP SMTP Gateway: Authorized Use Only! Violators will be prosecuted for from ; Tue, 27 May 2014 17:44:27 -0600 Received: from b01cxnp22035.gho.pok.ibm.com (b01cxnp22035.gho.pok.ibm.com [9.57.198.25]) by d01dlp01.pok.ibm.com (Postfix) with ESMTP id C9D4538C8027 for ; Tue, 27 May 2014 19:44:24 -0400 (EDT) Received: from d01av01.pok.ibm.com (d01av01.pok.ibm.com [9.56.224.215]) by b01cxnp22035.gho.pok.ibm.com (8.13.8/8.13.8/NCO v10.0) with ESMTP id s4RNiO477012686 for ; Tue, 27 May 2014 23:44:24 GMT Received: from d01av01.pok.ibm.com (localhost [127.0.0.1]) by d01av01.pok.ibm.com (8.14.4/8.14.4/NCO v10.0 AVout) with ESMTP id s4RNiNHX015683 for ; Tue, 27 May 2014 19:44:24 -0400 Date: Tue, 27 May 2014 16:44:20 -0700 From: Nishanth Aravamudan To: Andrew Morton Subject: Re: [RFC PATCH v2 1/2] powerpc: numa: enable USE_PERCPU_NUMA_NODE_ID Message-ID: <20140527234420.GE4104@linux.vnet.ibm.com> References: <20140516233945.GI8941@linux.vnet.ibm.com> <20140519181423.GL8941@linux.vnet.ibm.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii In-Reply-To: <20140519181423.GL8941@linux.vnet.ibm.com> Cc: Lee Schermerhorn , Christoph Lameter , Mel Gorman , linux-mm@kvack.org, Anton Blanchard , David Rientjes , Joonsoo Kim , linuxppc-dev@lists.ozlabs.org List-Id: Linux on PowerPC Developers Mail List List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , On 19.05.2014 [11:14:23 -0700], Nishanth Aravamudan wrote: > Hi Andrew, > > I found one issue with my patch, fixed below... > > On 16.05.2014 [16:39:45 -0700], Nishanth Aravamudan wrote: > > Based off 3bccd996 for ia64, convert powerpc to use the generic per-CPU > > topology tracking, specifically: > > > > initialize per cpu numa_node entry in start_secondary > > remove the powerpc cpu_to_node() > > define CONFIG_USE_PERCPU_NUMA_NODE_ID if NUMA > > > > Signed-off-by: Nishanth Aravamudan > > > > > diff --git a/arch/powerpc/kernel/smp.c b/arch/powerpc/kernel/smp.c > > index e2a4232..b95be24 100644 > > --- a/arch/powerpc/kernel/smp.c > > +++ b/arch/powerpc/kernel/smp.c > > @@ -750,6 +750,11 @@ void start_secondary(void *unused) > > } > > traverse_core_siblings(cpu, true); > > > > + /* > > + * numa_node_id() works after this. > > + */ > > + set_numa_node(numa_cpu_lookup_table[cpu]); > > + > > Similar change is needed for the boot CPU. Update patch: > > > powerpc: numa: enable USE_PERCPU_NUMA_NODE_ID > > Based off 3bccd996 for ia64, convert powerpc to use the generic per-CPU > topology tracking, specifically: > > initialize per cpu numa_node entry in start_secondary > remove the powerpc cpu_to_node() > define CONFIG_USE_PERCPU_NUMA_NODE_ID if NUMA > > Signed-off-by: Nishanth Aravamudan Ping on this and patch 2/2. Ben, would you be willing to pull these into your -next branch so they'd get some testing? http://patchwork.ozlabs.org/patch/350368/ http://patchwork.ozlabs.org/patch/349838/ Without any further changes, these two help quite a bit with the slab consumption on CONFIG_SLUB kernels when memoryless nodes are present. Thanks, Nish