From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail-qy0-f174.google.com ([209.85.216.174]:56180 "EHLO mail-qy0-f174.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755230Ab1IMQR2 convert rfc822-to-8bit (ORCPT ); Tue, 13 Sep 2011 12:17:28 -0400 Received: by qyk30 with SMTP id 30so2786615qyk.19 for ; Tue, 13 Sep 2011 09:17:28 -0700 (PDT) In-Reply-To: <1315622957-1549-2-git-send-email-bergwolf@gmail.com> References: <1315622957-1549-1-git-send-email-bergwolf@gmail.com> <1315622957-1549-2-git-send-email-bergwolf@gmail.com> From: Peng Tao Date: Wed, 14 Sep 2011 00:09:32 +0800 Message-ID: Subject: Re: [PATCH 2/2] pnfsblock: fix writeback deadlock To: Trond Myklebust Cc: Benny Halevy , Jim Rees , Peng Tao , linuxnfs Content-Type: text/plain; charset=UTF-8 Sender: linux-nfs-owner@vger.kernel.org List-ID: MIME-Version: 1.0 And this one too... Add cc nfs list. Sorry for the noise... On Sat, Sep 10, 2011 at 10:49 AM, Peng Tao wrote: > We should check if the sector is already initialized before > trying to grab the page from page cache. Otherwise when two > pages of the same block are written back by two threads each > calling from writepage_locked, it can cause deadlock like bellow. > >  [ 1080.972099] INFO: task kswapd0:25 blocked for more than 120 seconds. >  [ 1080.972377] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. >  [ 1080.972812] kswapd0         D ffff88000c4926c0     0    25      2 0x00000000 >  [ 1080.972816]  ffff88000df276b0 0000000000000046 ffff88000df27640 ffffffff81013ba7 >  [ 1080.972821]  ffff88000c492310 ffff88000df27fd8 ffff88000df27fd8 00000000001d3440 >  [ 1080.972824]  ffff88000c378000 ffff88000c492310 ffff8800175d3d40 ffff880017fc75a8 >  [ 1080.972828] Call Trace: >  [ 1080.972860]  [] ? read_tsc+0x9/0x19 >  [ 1080.972877]  [] ? lock_page+0x2b/0x2b >  [ 1080.972899]  [] io_schedule+0x63/0x7e >  [ 1080.972902]  [] sleep_on_page+0xe/0x12 >  [ 1080.972905]  [] __wait_on_bit_lock+0x46/0x8f >  [ 1080.972916]  [] ? lock_release_holdtime.part.7+0x6b/0x72 >  [ 1080.972919]  [] __lock_page+0x66/0x68 >  [ 1080.972928]  [] ? autoremove_wake_function+0x3d/0x3d >  [ 1080.972932]  [] lock_page+0x27/0x2b >  [ 1080.972934]  [] find_lock_page+0x34/0x57 >  [ 1080.972937]  [] find_or_create_page+0x34/0x8a >  [ 1080.972947]  [] bl_write_pagelist+0x205/0x6da [blocklayoutdriver] >  [ 1080.972951]  [] ? bl_free_lseg+0x38/0x38 [blocklayoutdriver] >  [ 1080.972995]  [] ? nfs_write_rpcsetup+0x118/0x123 [nfs] >  [ 1080.973033]  [] pnfs_generic_pg_writepages+0x10b/0x1f4 [nfs] >  [ 1080.973089]  [] nfs_pageio_doio+0x1a/0x43 [nfs] >  [ 1080.973098]  [] nfs_pageio_complete+0x16/0x2d [nfs] >  [ 1080.973108]  [] nfs_writepage_locked+0xa0/0xbf [nfs] >  [ 1080.973119]  [] nfs_writepage+0x16/0x2b [nfs] >  [ 1080.973122]  [] ? clear_page_dirty_for_io+0x87/0x9a >  [ 1080.973133]  [] shrink_page_list+0x39b/0x6c8 >  [ 1080.973139]  [] shrink_inactive_list+0x22c/0x39e >  [ 1080.973144]  [] ? lock_release_holdtime.part.7+0x6b/0x72 >  [ 1080.973148]  [] shrink_zone+0x445/0x588 >  [ 1080.973152]  [] balance_pgdat+0x2c2/0x56b >  [ 1080.973170]  [] ? __bitmap_weight+0x34/0x80 >  [ 1080.973175]  [] kswapd+0x2be/0x2fa >  [ 1080.973179]  [] ? __init_waitqueue_head+0x4b/0x4b >  [ 1080.973183]  [] ? balance_pgdat+0x56b/0x56b >  [ 1080.973187]  [] kthread+0xa8/0xb0 >  [ 1080.973200]  [] kernel_thread_helper+0x4/0x10 >  [ 1080.973205]  [] ? __init_kthread_worker+0x5a/0x5a >  [ 1080.973210]  [] ? gs_change+0x13/0x13 >  [ 1080.973213] no locks held by kswapd0/25. > > Signed-off-by: Peng Tao > --- >  fs/nfs/blocklayout/blocklayout.c |    8 ++++++-- >  1 files changed, 6 insertions(+), 2 deletions(-) > > diff --git a/fs/nfs/blocklayout/blocklayout.c b/fs/nfs/blocklayout/blocklayout.c > index 0920c35..cc78e62 100644 > --- a/fs/nfs/blocklayout/blocklayout.c > +++ b/fs/nfs/blocklayout/blocklayout.c > @@ -533,6 +533,11 @@ bl_write_pagelist(struct nfs_write_data *wdata, int sync) >  fill_invalid_ext: >                dprintk("%s need to zero %d pages\n", __func__, npg_zero); >                for (;npg_zero > 0; npg_zero--) { > +                       if (bl_is_sector_init(be->be_inval, isect)) { > +                               dprintk("isect %llu already init\n", > +                                       (unsigned long long)isect); > +                               goto next_page; > +                       } >                        /* page ref released in bl_end_io_write_zero */ >                        index = isect >> PAGE_CACHE_SECTOR_SHIFT; >                        dprintk("%s zero %dth page: index %lu isect %llu\n", > @@ -552,8 +557,7 @@ fill_invalid_ext: >                         * PageUptodate: It was read before >                         * sector_initialized: already written out >                         */ > -                       if (PageDirty(page) || PageWriteback(page) || > -                           bl_is_sector_init(be->be_inval, isect)) { > +                       if (PageDirty(page) || PageWriteback(page)) { >                                print_page(page); >                                unlock_page(page); >                                page_cache_release(page); > -- > 1.7.1.262.g5ef3d > > -- Thanks, -Bergwolf