From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail144.messagelabs.com (mail144.messagelabs.com [216.82.254.51]) by kanga.kvack.org (Postfix) with SMTP id 0C9F16B008A for ; Mon, 28 Sep 2009 11:50:08 -0400 (EDT) Date: Mon, 28 Sep 2009 16:44:01 +0800 From: Wu Fengguang Subject: Re: [RFC][PATCH] HWPOISON: remove the unsafe __set_page_locked() Message-ID: <20090928084401.GA22131@localhost> References: <20090926031537.GA10176@localhost> <20090926034936.GK30185@one.firstfloor.org> <20090926105259.GA5496@localhost> <20090926113156.GA12240@localhost> <20090927104739.GA1666@localhost> <20090927192025.GA6327@wotan.suse.de> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20090927192025.GA6327@wotan.suse.de> Sender: owner-linux-mm@kvack.org To: Nick Piggin Cc: Andi Kleen , Andrew Morton , Hugh Dickins , "linux-mm@kvack.org" , LKML List-ID: On Mon, Sep 28, 2009 at 03:20:25AM +0800, Nick Piggin wrote: > On Sun, Sep 27, 2009 at 06:47:39PM +0800, Wu Fengguang wrote: > > > > > > And standard deviation is 0.04%, much larger than the difference 0.008% .. > > > > Sorry that's not correct. I improved the accounting by treating > > function0+function1 from two CPUs as an integral entity: > > > > total time add_to_page_cache_lru percent stddev > > before 3880166848.722 9683329.610 0.250% 0.014% > > after 3828516894.376 9778088.870 0.256% 0.012% > > delta 0.006% > > I don't understand why you're doing this NFS workload to measure? Because it is the first convenient workload hit my mind, and avoids real disk IO :) > I see significant nfs, networking protocol and device overheads in > your profiles, also you're hitting some locks or something which > is causing massive context switching. So I don't think this is a > good test. Yes there are overheads. However it is a real and common workload. > But anyway as Hugh points out, you need to compare with a > *completely* fixed kernel, which includes auditing all users of page > flags non-atomically (slab, notably, but possibly also other > places). That's good point. We can do more benchmarks when more fixes are available. However I suspect their design goal will be "fix them without introducing noticeable overheads" :) > One other thing to keep in mind that I will mention is that I am > going to push in a patch to the page allocator to allow callers > to avoid the refcounting (atomic_dec_and_test) in page lifetime, > which is especially important for SLUB and takes more cycles off > the page allocator... > > I don't know exactly what you're going to do after that to get a > stable reference to slab pages. I guess you can read the page > flags and speculatively take some slab locks and recheck etc... For reliably we could skip page lock on zero refcounted pages. We may lose the PG_hwpoison bit on races with __SetPageSlub*, however it should be an acceptable imperfection. Thanks, Fengguang -- 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