Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752524AbdHBIsy (ORCPT ); Wed, 2 Aug 2017 04:48:54 -0400 Received: from szxga02-in.huawei.com ([45.249.212.188]:10821 "EHLO szxga02-in.huawei.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752214AbdHBIsx (ORCPT ); Wed, 2 Aug 2017 04:48:53 -0400 Subject: Re: [PATCH] f2fs: update cur_valid_map_mir together with cur_valid_map To: Yunlong Song , Yunlong Song , Chao Yu CC: jaegeuk , miaoxie , "bintian.wang" , linux-fsdevel , linux-f2fs-devel , linux-kernel References: <1501574206-150946-1-git-send-email-yunlong.song@huawei.com> <8be82ea4-db80-8f87-9029-e63d2926f7d1@huawei.com> <7e8346ed-8484-2ab9-85a1-03d232b16865@huawei.com> From: Chao Yu Message-ID: <3cbec840-a4ee-7d93-6b02-8f8f1483b7f6@huawei.com> Date: Wed, 2 Aug 2017 16:47:33 +0800 User-Agent: Mozilla/5.0 (Windows NT 6.1; WOW64; rv:52.0) Gecko/20100101 Thunderbird/52.2.1 MIME-Version: 1.0 In-Reply-To: <7e8346ed-8484-2ab9-85a1-03d232b16865@huawei.com> Content-Type: text/plain; charset="utf-8" Content-Language: en-US Content-Transfer-Encoding: 7bit X-Originating-IP: [10.134.22.195] X-CFilter-Loop: Reflected X-Mirapoint-Virus-RAPID-Raw: score=unknown(0), refid=str=0001.0A020206.598191ED.008F,ss=1,re=0.000,recu=0.000,reip=0.000,cl=1,cld=1,fgs=0, ip=0.0.0.0, so=2014-11-16 11:51:01, dmn=2013-03-21 17:37:32 X-Mirapoint-Loop-Id: 51d3088cdb600446f39254f2f11cef9c Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 7128 Lines: 177 Hi Yunlong, On 2017/8/2 16:02, Yunlong Song wrote: > Hi Chao, > For the memory overflow or cache bit-transition case, when CHECK_FS > is on, should f2fs > enter panic or just provide WARNing information? I prefer WARNing, > because it is not f2fs's > fault, f2fs does not need to pay for it, why not use WARN_ON instead? Actually, if that happens, I think both original and mirror bitmap datas are not trustable, it's harmful to continue to trigger write IOs based on those bitmap datas. So IMO, it would be better to trigger bug_on. At least, if only warning is showed, we'd better to turn f2fs to readonly and fail current write IO. Thanks, > > On 2017/8/2 9:34, Chao Yu wrote: >> Hi Yunlong, >> >> On 2017/8/2 0:59, Yunlong Song wrote: >>> Hi Chao, >>> I think there is no need to test mirror bitmap when original bitmap's check >>> get passed, it is an instruction waste for "test". By the way, previous patch >>> uses WARN to skip trigger panic when the memory is flipped, I think it is proper >>> to not trigger panic in this situation, because it is not code bug at all in >>> such case. >> Original idea is trying to use bitmap mirror to detect bitmap corruption caused >> by memory overflow or bit-transition of cache. >> >> So if we encounter inconsistency in between original bitmap and mirror bitmap, >> there may be memory overflow or cache bit-transition, we need to detect that; >> another case is that both bitmap is consistent, but we are trying to reuse valid >> block address or free invalid block address, indeed that is a bug. >> >> Below modification can cover above two cases, which make original idea of >> introducing mirror bitmap check working correctly. >> >> Thanks, >> >>> On 08/01/2017 23:44, Chao Yu wrote: >>> >>> Hi Yunlong, >>> >>> How about checking consistence in between original and mirror bitmap all >>> the time as below? >>> >>> --- >>> fs/f2fs/segment.c | 41 +++++++++++++++++++++++++++-------------- >>> 1 file changed, 27 insertions(+), 14 deletions(-) >>> >>> diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c >>> index a26c24dae70c..f32a19cf486a 100644 >>> --- a/fs/f2fs/segment.c >>> +++ b/fs/f2fs/segment.c >>> @@ -1510,6 +1510,7 @@ static void update_sit_entry(struct f2fs_sb_info *sbi, >>> block_t blkaddr, int del) >>> struct seg_entry *se; >>> unsigned int segno, offset; >>> long int new_vblocks; >>> + bool exist, mir_exist; >>> >>> segno = GET_SEGNO(sbi, blkaddr); >>> >>> @@ -1526,17 +1527,23 @@ static void update_sit_entry(struct f2fs_sb_info >>> *sbi, block_t blkaddr, int del) >>> >>> /* Update valid block bitmap */ >>> if (del > 0) { >>> - if (f2fs_test_and_set_bit(offset, se->cur_valid_map)) { >>> + exist = f2fs_test_and_set_bit(offset, se->cur_valid_map); >>> #ifdef CONFIG_F2FS_CHECK_FS >>> - if (f2fs_test_and_set_bit(offset, >>> - se->cur_valid_map_mir)) >>> - f2fs_bug_on(sbi, 1); >>> - else >>> - WARN_ON(1); >>> -#else >>> + mir_exist = f2fs_test_and_set_bit(offset, >>> + se->cur_valid_map_mir); >>> + if (exist != mir_exist) { >>> + f2fs_msg(sbi->sb, KERN_ERR, "Inconsistent error " >>> + "when setting bitmap, blk:%u, old bit:%d", >>> + blkaddr, exist); >>> f2fs_bug_on(sbi, 1); >>> + } >>> #endif >>> + if (exist) { >>> + f2fs_msg(sbi->sb, KERN_ERR, >>> + "Bitmap was set, blk:%u", blkaddr); >>> + f2fs_bug_on(sbi, 1); >>> } >>> + >>> if (f2fs_discard_en(sbi) && >>> !f2fs_test_and_set_bit(offset, se->discard_map)) >>> sbi->discard_blks--; >>> @@ -1547,17 +1554,23 @@ static void update_sit_entry(struct f2fs_sb_info >>> *sbi, block_t blkaddr, int del) >>> se->ckpt_valid_blocks++; >>> } >>> } else { >>> - if (!f2fs_test_and_clear_bit(offset, se->cur_valid_map)) { >>> + exist = f2fs_test_and_clear_bit(offset, se->cur_valid_map); >>> #ifdef CONFIG_F2FS_CHECK_FS >>> - if (!f2fs_test_and_clear_bit(offset, >>> - se->cur_valid_map_mir)) >>> - f2fs_bug_on(sbi, 1); >>> - else >>> - WARN_ON(1); >>> -#else >>> + mir_exist = f2fs_test_and_clear_bit(offset, >>> + se->cur_valid_map_mir); >>> + if (exist != mir_exist) { >>> + f2fs_msg(sbi->sb, KERN_ERR, "Inconsistent error " >>> + "when clearing bitmap, blk:%u, old bit:%d", >>> + blkaddr, exist); >>> f2fs_bug_on(sbi, 1); >>> + } >>> #endif >>> + if (!exist) { >>> + f2fs_msg(sbi->sb, KERN_ERR, >>> + "Bitmap was cleared, blk:%u", blkaddr); >>> + f2fs_bug_on(sbi, 1); >>> } >>> + >>> if (f2fs_discard_en(sbi) && >>> f2fs_test_and_clear_bit(offset, se->discard_map)) >>> sbi->discard_blks++; >>> -- >>> 2.13.0.90.g1eb437020 >>> >>> On 2017/8/1 15:56, Yunlong Song wrote: >>> > When cur_valid_map passes the f2fs_test_and_set(,clear)_bit test, >>> > cur_valid_map_mir update is skipped unlikely, so fix it. >>> > >>> > Signed-off-by: Yunlong Song >>> > --- >>> > fs/f2fs/segment.c | 8 ++++++++ >>> > 1 file changed, 8 insertions(+) >>> > >>> > diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c >>> > index 151968e..6f7731a 100644 >>> > --- a/fs/f2fs/segment.c >>> > +++ b/fs/f2fs/segment.c >>> > @@ -1535,6 +1535,10 @@ static void update_sit_entry(struct f2fs_sb_info >>> *sbi, block_t blkaddr, int del) >>> > f2fs_bug_on(sbi, 1); >>> > #endif >>> > } >>> > +#ifdef CONFIG_F2FS_CHECK_FS >>> > + else >>> > + f2fs_set_bit(offset, se->cur_valid_map_mir); >>> > +#endif >>> > if (f2fs_discard_en(sbi) && >>> > !f2fs_test_and_set_bit(offset, se->discard_map)) >>> > sbi->discard_blks--; >>> > @@ -1556,6 +1560,10 @@ static void update_sit_entry(struct f2fs_sb_info >>> *sbi, block_t blkaddr, int del) >>> > f2fs_bug_on(sbi, 1); >>> > #endif >>> > } >>> > +#ifdef CONFIG_F2FS_CHECK_FS >>> > + else >>> > + f2fs_clear_bit(offset, se->cur_valid_map_mir); >>> > +#endif >>> > if (f2fs_discard_en(sbi) && >>> > f2fs_test_and_clear_bit(offset, se->discard_map)) >>> > sbi->discard_blks++; >>> > >>> >> >> . >> >