Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753501Ab3FYSkk (ORCPT ); Tue, 25 Jun 2013 14:40:40 -0400 Received: from relay3.sgi.com ([192.48.152.1]:52582 "EHLO relay.sgi.com" rhost-flags-OK-OK-OK-FAIL) by vger.kernel.org with ESMTP id S1753291Ab3FYSki (ORCPT ); Tue, 25 Jun 2013 14:40:38 -0400 Message-ID: <51C9E422.6040304@sgi.com> Date: Tue, 25 Jun 2013 11:40:34 -0700 From: Mike Travis User-Agent: Mozilla/5.0 (Windows NT 6.1; WOW64; rv:17.0) Gecko/20130509 Thunderbird/17.0.6 MIME-Version: 1.0 To: "H. Peter Anvin" CC: Yinghai Lu , Greg KH , Nathan Zimmer , Robin Holt , Rob Landley , Thomas Gleixner , Ingo Molnar , Andrew Morton , the arch/x86 maintainers , linux-doc@vger.kernel.org, Linux Kernel Mailing List Subject: Re: [RFC 0/2] Delay initializing of large sections of memory References: <1371831934-156971-1-git-send-email-nzimmer@sgi.com> <20130621165142.GA32125@kroah.com> <51C48745.9030304@zytor.com> <20130621185056.GA23473@kroah.com> <51C4C5F3.3050800@sgi.com> <51C9D4ED.5070805@sgi.com> <51C9DEC1.6030602@zytor.com> In-Reply-To: <51C9DEC1.6030602@zytor.com> Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 2145 Lines: 53 On 6/25/2013 11:17 AM, H. Peter Anvin wrote: > On 06/25/2013 10:35 AM, Mike Travis wrote: >> >> The two params that I couldn't figure out how to provide except via kernel >> param option was the memory block size (128M or 2G) and the physical >> address space per node. The other 3 params can be automatically >> setup by a script when the total system size is known. As soon as we >> verify on the 32TB system and surmise what will be needed for 64TB, >> then those 3 params can probably disappear. >> > > "Setup by script" is a no-go. You *have* the total system size already, > it is in the e820 tables (anything which isn't in e820 is hotplug, that > automagically gets deferred.) Okay, I'll figure something out. If Yinghai's SRAT patch can help with the node address space, then I might be able to determine if the system is a UV which is the only system I see that uses 2G memory blocks. (Or make get_memory_block_size() a global.) Then a simple param to start the insertion early or defer it until the system is fully up is still useful, and that's easy to understand. [I think we still want to keep the actual process of moving memory to the absent list an option, yes? If for no other reason except to rule out this code when a problem crops up. Or at least have a way to disable the process if it's CONFIG'd in.] > > However, please consider Ingo's counterproposal of doing this via the > buddy allocator, i.e. hugepages being broken on demand. That is a > *very* powerful model, although would require more infrastructure. We will certainly continue to make improvements as larger system sizes become more commonplace (and customers continue to complain :). But we are cutting it close to including this into the nextgen distro releases, so it would have to be a follow on project. (I've been working on this patch since last November.) Thanks, Mike > > -hpa > > -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/