From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail172.messagelabs.com (mail172.messagelabs.com [216.82.254.3]) by kanga.kvack.org (Postfix) with ESMTP id 9BD866B007D for ; Tue, 26 Jan 2010 15:04:02 -0500 (EST) Date: Tue, 26 Jan 2010 20:03:47 +0000 From: Mel Gorman Subject: Re: [PATCH 21 of 31] split_huge_page_mm/vma Message-ID: <20100126200347.GW16468@csn.ul.ie> References: <9cb2a8a61d32163dced8.1264513936@v2.random> <20100126173450.GE16468@csn.ul.ie> <20100126194947.GV30452@random.random> MIME-Version: 1.0 Content-Type: text/plain; charset=iso-8859-15 Content-Disposition: inline In-Reply-To: <20100126194947.GV30452@random.random> Sender: owner-linux-mm@kvack.org To: Andrea Arcangeli Cc: linux-mm@kvack.org, Marcelo Tosatti , Adam Litke , Avi Kivity , Izik Eidus , Hugh Dickins , Nick Piggin , Rik van Riel , Andi Kleen , Dave Hansen , Benjamin Herrenschmidt , Ingo Molnar , Mike Travis , KAMEZAWA Hiroyuki , Christoph Lameter , Chris Wright , Andrew Morton , bpicco@redhat.com, Christoph Hellwig , KOSAKI Motohiro List-ID: On Tue, Jan 26, 2010 at 08:49:47PM +0100, Andrea Arcangeli wrote: > On Tue, Jan 26, 2010 at 05:34:51PM +0000, Mel Gorman wrote: > > I guess this is the part that breaks huge pages when smaps is read. That > > is a bit of a snag as a normal user could cause a lot of churn by > > reading those files a lot. > > yep, this is the highest priority split_huge_page to remove, along > with fixing lru stats. mprotect and mremap are much lower prio. I > think I mentioned the removal of split_huge_page from smaps in earlier > emails too. > Probably. It rings a bell for sure. > > In the event that gets fixed up, it's worth considering what KernelPageSize: > > and MMUPageSize: should be printing in smaps for regions of memory backed > > by a mix of base and huge pages. > > Suggestions and patches welcome. I just deferred it for later and no > app on my system seems to cat those files but it's definitely > something to fix or it's just spurious overhead given to khugepaged... > I'd be very surprised if there are apps that cat those files because they are expensive to read from a kernel perspective and not the cheapest to parse either. It's intended for monitoring tools that report on hugepage usage on a per-application basis. Not a high priority and I'll give it some thought. Just mentioning it in case I forgot to bring it up later. Otherwise; Acked-by: Mel Gorman -- Mel Gorman Part-time Phd Student Linux Technology Center University of Limerick IBM Dublin Software Lab -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: email@kvack.org