From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1760851AbZE0XR1 (ORCPT ); Wed, 27 May 2009 19:17:27 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1756391AbZE0XRT (ORCPT ); Wed, 27 May 2009 19:17:19 -0400 Received: from smtp1.linux-foundation.org ([140.211.169.13]:51118 "EHLO smtp1.linux-foundation.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753746AbZE0XRT (ORCPT ); Wed, 27 May 2009 19:17:19 -0400 Date: Wed, 27 May 2009 16:15:35 -0700 From: Andrew Morton To: Johannes Weiner Cc: linux-mm@kvack.org, linux-kernel@vger.kernel.org, kamezawa.hiroyu@jp.fujitsu.com, riel@redhat.com, hugh.dickins@tiscali.co.uk Subject: Re: [patch] mm: release swap slots for actively used pages Message-Id: <20090527161535.ac2dd1ba.akpm@linux-foundation.org> In-Reply-To: <1243388859-9760-1-git-send-email-hannes@cmpxchg.org> References: <1243388859-9760-1-git-send-email-hannes@cmpxchg.org> X-Mailer: Sylpheed version 2.2.4 (GTK+ 2.8.20; i486-pc-linux-gnu) Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Wed, 27 May 2009 03:47:39 +0200 Johannes Weiner wrote: > For anonymous pages activated by the reclaim scan or faulted from an > evicted page table entry we should always try to free up swap space. > > Both events indicate that the page is in active use and a possible > change in the working set. Thus removing the slot association from > the page increases the chance of the page being placed near its new > LRU buddies on the next eviction and helps keeping the amount of stale > swap cache entries low. > > try_to_free_swap() inherently only succeeds when the last user of the > swap slot vanishes so it is safe to use from places where that single > mapping just brought the page back to life. > Seems that this has a risk of worsening swap fragmentation for some situations. Or not, I have no way of knowing, really. > diff --git a/mm/memory.c b/mm/memory.c > index 8b4e40e..407ebf7 100644 > --- a/mm/memory.c > +++ b/mm/memory.c > @@ -2671,8 +2671,7 @@ static int do_swap_page(struct mm_struct *mm, struct vm_area_struct *vma, > mem_cgroup_commit_charge_swapin(page, ptr); > > swap_free(entry); > - if (vm_swap_full() || (vma->vm_flags & VM_LOCKED) || PageMlocked(page)) > - try_to_free_swap(page); > + try_to_free_swap(page); > unlock_page(page); > > if (write_access) { > diff --git a/mm/vmscan.c b/mm/vmscan.c > index 621708f..2f0549d 100644 > --- a/mm/vmscan.c > +++ b/mm/vmscan.c > @@ -788,7 +788,7 @@ cull_mlocked: > > activate_locked: > /* Not a candidate for swapping, so reclaim swap space. */ > - if (PageSwapCache(page) && vm_swap_full()) > + if (PageSwapCache(page)) > try_to_free_swap(page); > VM_BUG_ON(PageActive(page)); > SetPageActive(page); How are we to know that this is a desirable patch for Linux??