Return-Path: linux-nfs-owner@vger.kernel.org Received: from oz.csail.mit.edu ([128.30.30.239]:36780 "EHLO mail.mgebm.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750944Ab2GCAKx (ORCPT ); Mon, 2 Jul 2012 20:10:53 -0400 Date: Mon, 2 Jul 2012 20:10:51 -0400 From: Eric B Munson To: Mel Gorman Cc: Andrew Morton , Linux-MM , Linux-Netdev , Linux-NFS , LKML , David Miller , Trond Myklebust , Neil Brown , Christoph Hellwig , Peter Zijlstra , Mike Christie , Sebastian Andrzej Siewior Subject: Re: [PATCH 00/12] Swap-over-NFS without deadlocking V8 Message-ID: <20120703001051.GA5508@mgebm.net> References: <1340976805-5799-1-git-send-email-mgorman@suse.de> <20120701172254.GB2470@mgebm.net> <20120702143556.GT14154@suse.de> MIME-Version: 1.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="ReaqsoxgOBHFXBhH" In-Reply-To: <20120702143556.GT14154@suse.de> Sender: linux-nfs-owner@vger.kernel.org List-ID: --ReaqsoxgOBHFXBhH Content-Type: text/plain; charset=us-ascii Content-Disposition: inline Content-Transfer-Encoding: quoted-printable On Mon, 02 Jul 2012, Mel Gorman wrote: > On Sun, Jul 01, 2012 at 01:22:54PM -0400, Eric B Munson wrote: > > On Fri, 29 Jun 2012, Mel Gorman wrote: > >=20 > > > Changelog since V7 > > > o Rebase to linux-next 20120629 > > > o bi->page_dma instead of bi->page in intel driver > > > o Build fix for !CONFIG_NET (sebastian) > > > o Restore PF_MEMALLOC flags correctly in all cases (jlayton) > > >=20 > > > Changelog since V6 > > > o Rebase to linux-next 20120622 > > >=20 > > > Changelog since V5 > > > o Rebase to v3.5-rc3 > > >=20 > > > Changelog since V4 > > > o Catch if SOCK_MEMALLOC flag is cleared with rmem tokens (davem) > > >=20 > > > Changelog since V3 > > > o Rebase to 3.4-rc5 > > > o kmap pages for writing to swap (akpm) > > > o Move forward declaration to reduce chance of duplication (akpm) > > >=20 > > > Changelog since V2 > > > o Nothing significant, just rebases. A radix tree lookup is replace= d with > > > a linear search would be the biggest rebase artifact > > >=20 > > > This patch series is based on top of "Swap-over-NBD without deadlocki= ng v14" > > > as it depends on the same reservation of PF_MEMALLOC reserves logic. > > >=20 > > > When a user or administrator requires swap for their application, they > > > create a swap partition and file, format it with mkswap and activate = it with > > > swapon. In diskless systems this is not an option so if swap if requi= red > > > then swapping over the network is considered. The two likely scenari= os > > > are when blade servers are used as part of a cluster where the form f= actor > > > or maintenance costs do not allow the use of disks and thin clients. > > >=20 > > > The Linux Terminal Server Project recommends the use of the Network > > > Block Device (NBD) for swap but this is not always an option. There = is > > > no guarantee that the network attached storage (NAS) device is running > > > Linux or supports NBD. However, it is likely that it supports NFS so = there > > > are users that want support for swapping over NFS despite any perform= ance > > > concern. Some distributions currently carry patches that support swap= ping > > > over NFS but it would be preferable to support it in the mainline ker= nel. > > >=20 > > > Patch 1 avoids a stream-specific deadlock that potentially affects TC= P. > > >=20 > > > Patch 2 is a small modification to SELinux to avoid using PFMEMALLOC > > > reserves. > > >=20 > > > Patch 3 adds three helpers for filesystems to handle swap cache pages. > > > For example, page_file_mapping() returns page->mapping for > > > file-backed pages and the address_space of the underlying > > > swap file for swap cache pages. > > >=20 > > > Patch 4 adds two address_space_operations to allow a filesystem > > > to pin all metadata relevant to a swapfile in memory. Upon > > > successful activation, the swapfile is marked SWP_FILE and > > > the address space operation ->direct_IO is used for writing > > > and ->readpage for reading in swap pages. > > >=20 > > > Patch 5 notes that patch 3 is bolting > > > filesystem-specific-swapfile-support onto the side and that > > > the default handlers have different information to what > > > is available to the filesystem. This patch refactors the > > > code so that there are generic handlers for each of the new > > > address_space operations. > > >=20 > > > Patch 6 adds an API to allow a vector of kernel addresses to be > > > translated to struct pages and pinned for IO. > > >=20 > > > Patch 7 adds support for using highmem pages for swap by kmapping > > > the pages before calling the direct_IO handler. > > >=20 > > > Patch 8 updates NFS to use the helpers from patch 3 where necessary. > > >=20 > > > Patch 9 avoids setting PF_private on PG_swapcache pages within NFS. > > >=20 > > > Patch 10 implements the new swapfile-related address_space operations > > > for NFS and teaches the direct IO handler how to manage > > > kernel addresses. > > >=20 > > > Patch 11 prevents page allocator recursions in NFS by using GFP_NOIO > > > where appropriate. > > >=20 > > > Patch 12 fixes a NULL pointer dereference that occurs when using > > > swap-over-NFS. > > >=20 > > > With the patches applied, it is possible to mount a swapfile that is = on an > > > NFS filesystem. Swap performance is not great with a swap stress test= taking > > > roughly twice as long to complete than if the swap device was backed = by NBD. > >=20 > > To test this set I am using memory cgroups to force swap usage. I am s= eeing > > the cgroup controller killing my processes instead of using the nfs swa= pfile. > >=20 >=20 > How sure are you that this is not a cgroup bug? For dirty file data on so= me > kernels, cgroups can prematurely kill processes if pages are not being > cleaned fast enough. I would not expect the same problem for anonymous > pages but it's worth considering. Please also test with a normal swapfile. >=20 > If OOM is disabled and the process hangs, try capturing a sysrq+t and > see where the process is stuck. >=20 It looks like the problem is with cgroups, when I run without cgroups and l= imit memory on the boot command line everything works fine. To test I limited t= he machine to 1G of ram then ran several memory benchmarks with work set sizes= of 1.5G, all completed successfully with my swap file located on an NFS share. Tested-by: Eric B Munson --ReaqsoxgOBHFXBhH Content-Type: application/pgp-signature; name="signature.asc" Content-Description: Digital signature -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.11 (GNU/Linux) iQEcBAEBAgAGBQJP8jiLAAoJEH65iIruGRnNH5MIAIUzripnqPwPKbTI/GQLicbh EgKVK+9r8lSKmrYWFfvVP+mqBV1udMV69SnsFCAmOXzCmP94kxT8oWdaRg/Rto0r qVNtR6K/pTIIlSqOBRGxG5k+6iUBDxRuBIqPEm27hH0s18oCc7hauokpceTo5bFh YDGSR3cTZK44j6QapunkDmk6+zlwr6iIc/OuAQDEo06Oij0c9fZcyyvwFHo0WbBR EqNSmp/HqLqbc3PLyX3yCq6GIS8zkSlgPGpMGgmMqjqNPcAdGVzrSHVly6EhSh2v Xvl0BqsC2MVgsfxLvyu2PWs7g14rAHGPq0r3zCf78H4wiZIhwoGNyngG+Puzd0Y= =6W1z -----END PGP SIGNATURE----- --ReaqsoxgOBHFXBhH--