From patchwork Fri Sep 7 12:55:19 2012 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Mel Gorman X-Patchwork-Id: 182365 X-Patchwork-Delegate: davem@davemloft.net Return-Path: X-Original-To: patchwork-incoming@ozlabs.org Delivered-To: patchwork-incoming@ozlabs.org Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by ozlabs.org (Postfix) with ESMTP id D0E902C0098 for ; Fri, 7 Sep 2012 22:55:32 +1000 (EST) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752518Ab2IGMz2 (ORCPT ); Fri, 7 Sep 2012 08:55:28 -0400 Received: from cantor2.suse.de ([195.135.220.15]:41449 "EHLO mx2.suse.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751062Ab2IGMzZ (ORCPT ); Fri, 7 Sep 2012 08:55:25 -0400 Received: from relay1.suse.de (unknown [195.135.220.254]) (using TLSv1 with cipher DHE-RSA-AES256-SHA (256/256 bits)) (No client certificate requested) by mx2.suse.de (Postfix) with ESMTP id 33C8AA30B9; Fri, 7 Sep 2012 14:55:24 +0200 (CEST) Date: Fri, 7 Sep 2012 13:55:19 +0100 From: Mel Gorman To: JoonSoo Kim Cc: Andrew Morton , Linux-MM , Linux-Netdev , LKML , David Miller , Chuck Lever , Pekka Enberg , David Rientjes , Christoph Lameter Subject: Re: [PATCH 1/4] slab: do ClearSlabPfmemalloc() for all pages of slab Message-ID: <20120907125519.GB11266@suse.de> References: <1346779479-1097-1-git-send-email-mgorman@suse.de> <1346779479-1097-2-git-send-email-mgorman@suse.de> MIME-Version: 1.0 Content-Disposition: inline In-Reply-To: User-Agent: Mutt/1.5.21 (2010-09-15) Sender: netdev-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: netdev@vger.kernel.org On Fri, Sep 07, 2012 at 03:05:39AM +0900, JoonSoo Kim wrote: > Correct Pekka's mail address and resend. > Sorry. > > Add "Cc" to "Christoph Lameter" > > 2012/9/5 Mel Gorman : > > Right now, we call ClearSlabPfmemalloc() for first page of slab when we > > clear SlabPfmemalloc flag. This is fine for most swap-over-network use > > cases as it is expected that order-0 pages are in use. Unfortunately it > > is possible that that __ac_put_obj() checks SlabPfmemalloc on a tail page > > and while this is harmless, it is sloppy. This patch ensures that the head > > page is always used. > > > > This problem was originally identified by Joonsoo Kim. > > > > [js1304@gmail.com: Original implementation and problem identification] > > Signed-off-by: Mel Gorman > > --- > > mm/slab.c | 4 ++-- > > 1 file changed, 2 insertions(+), 2 deletions(-) > > > > diff --git a/mm/slab.c b/mm/slab.c > > index 811af03..d34a903 100644 > > --- a/mm/slab.c > > +++ b/mm/slab.c > > @@ -1000,7 +1000,7 @@ static void *__ac_get_obj(struct kmem_cache *cachep, struct array_cache *ac, > > l3 = cachep->nodelists[numa_mem_id()]; > > if (!list_empty(&l3->slabs_free) && force_refill) { > > struct slab *slabp = virt_to_slab(objp); > > - ClearPageSlabPfmemalloc(virt_to_page(slabp->s_mem)); > > + ClearPageSlabPfmemalloc(virt_to_head_page(slabp->s_mem)); > > clear_obj_pfmemalloc(&objp); > > recheck_pfmemalloc_active(cachep, ac); > > return objp; > > We assume that slabp->s_mem's address is always in head page, so > "virt_to_head_page" is not needed. > Fair point. I thought it would be more "obvious" later that we really always intended to use the head page but it is unnecessary. > > @@ -1032,7 +1032,7 @@ static void *__ac_put_obj(struct kmem_cache *cachep, struct array_cache *ac, > > { > > if (unlikely(pfmemalloc_active)) { > > /* Some pfmemalloc slabs exist, check if this is one */ > > - struct page *page = virt_to_page(objp); > > + struct page *page = virt_to_head_page(objp); > > if (PageSlabPfmemalloc(page)) > > set_obj_pfmemalloc(&objp); > > } > > -- > > 1.7.9.2 > > > > If we always use head page, following suggestion is more good to me. > How about you? > > diff --git a/mm/slab.c b/mm/slab.c > index f8b0d53..ce70989 100644 > --- a/mm/slab.c > +++ b/mm/slab.c > @@ -1032,7 +1032,7 @@ static void *__ac_put_obj(struct kmem_cache > *cachep, struct array_cache *ac, > { > if (unlikely(pfmemalloc_active)) { > /* Some pfmemalloc slabs exist, check if this is one */ > - struct page *page = virt_to_page(objp); > + struct page *page = virt_to_head_page(objp); > if (PageSlabPfmemalloc(page)) > set_obj_pfmemalloc(&objp); > } ok. > @@ -1921,10 +1921,9 @@ static void *kmem_getpages(struct kmem_cache > *cachep, gfp_t flags, int nodeid) > NR_SLAB_UNRECLAIMABLE, nr_pages); > for (i = 0; i < nr_pages; i++) { > __SetPageSlab(page + i); > - > - if (page->pfmemalloc) > - SetPageSlabPfmemalloc(page + i); > } > + if (page->pfmemalloc) > + SetPageSlabPfmemalloc(page); > > if (kmemcheck_enabled && !(cachep->flags & SLAB_NOTRACK)) { > kmemcheck_alloc_shadow(page, cachep->gfporder, flags, nodeid); ok. > @@ -1943,26 +1942,26 @@ static void *kmem_getpages(struct kmem_cache > *cachep, gfp_t flags, int nodeid) > */ > static void kmem_freepages(struct kmem_cache *cachep, void *addr) > { > - unsigned long i = (1 << cachep->gfporder); > + int nr_pages = (1 << cachep->gfporder); > + int i; > struct page *page = virt_to_page(addr); > - const unsigned long nr_freed = i; > > kmemcheck_free_shadow(page, cachep->gfporder); > > if (cachep->flags & SLAB_RECLAIM_ACCOUNT) > sub_zone_page_state(page_zone(page), > - NR_SLAB_RECLAIMABLE, nr_freed); > + NR_SLAB_RECLAIMABLE, nr_pages); > else > sub_zone_page_state(page_zone(page), > - NR_SLAB_UNRECLAIMABLE, nr_freed); > - while (i--) { > - BUG_ON(!PageSlab(page)); > - __ClearPageSlabPfmemalloc(page); > - __ClearPageSlab(page); > - page++; > + NR_SLAB_UNRECLAIMABLE, nr_pages); > + for (i = 0; i < nr_pages; i++) { > + BUG_ON(!PageSlab(page + i)); > + __ClearPageSlab(page + i); > } > + __ClearPageSlabPfmemalloc(page); > + > if (current->reclaim_state) > - current->reclaim_state->reclaimed_slab += nr_freed; > + current->reclaim_state->reclaimed_slab += nr_pages; > free_pages((unsigned long)addr, cachep->gfporder); > } This churns code a lot more than is necessary. How about this as a replacement patch? ---8<--- From: Joonsoo Kim Subject: [PATCH] slab: do ClearSlabPfmemalloc() for all pages of slab Right now, we call ClearSlabPfmemalloc() for first page of slab when we clear SlabPfmemalloc flag. This is fine for most swap-over-network use cases as it is expected that order-0 pages are in use. Unfortunately it is possible that that __ac_put_obj() checks SlabPfmemalloc on a tail page and while this is harmless, it is sloppy. This patch ensures that the head page is always used. [mgorman@suse.de: Easier implementation, changelog cleanup] Signed-off-by: Joonsoo Kim Signed-off-by: Mel Gorman --- mm/slab.c | 12 +++++------- 1 file changed, 5 insertions(+), 7 deletions(-) -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html diff --git a/mm/slab.c b/mm/slab.c index 811af03..590d52a 100644 --- a/mm/slab.c +++ b/mm/slab.c @@ -1032,7 +1032,7 @@ static void *__ac_put_obj(struct kmem_cache *cachep, struct array_cache *ac, { if (unlikely(pfmemalloc_active)) { /* Some pfmemalloc slabs exist, check if this is one */ - struct page *page = virt_to_page(objp); + struct page *page = virt_to_head_page(objp); if (PageSlabPfmemalloc(page)) set_obj_pfmemalloc(&objp); } @@ -1919,12 +1919,10 @@ static void *kmem_getpages(struct kmem_cache *cachep, gfp_t flags, int nodeid) else add_zone_page_state(page_zone(page), NR_SLAB_UNRECLAIMABLE, nr_pages); - for (i = 0; i < nr_pages; i++) { + for (i = 0; i < nr_pages; i++) __SetPageSlab(page + i); - - if (page->pfmemalloc) - SetPageSlabPfmemalloc(page + i); - } + if (page->pfmemalloc) + SetPageSlabPfmemalloc(page); if (kmemcheck_enabled && !(cachep->flags & SLAB_NOTRACK)) { kmemcheck_alloc_shadow(page, cachep->gfporder, flags, nodeid); @@ -1955,9 +1953,9 @@ static void kmem_freepages(struct kmem_cache *cachep, void *addr) else sub_zone_page_state(page_zone(page), NR_SLAB_UNRECLAIMABLE, nr_freed); + __ClearPageSlabPfmemalloc(page); while (i--) { BUG_ON(!PageSlab(page)); - __ClearPageSlabPfmemalloc(page); __ClearPageSlab(page); page++; }