From: Chao Yu <chao@kernel.org> To: Eric Biggers <ebiggers@kernel.org>, Jaegeuk Kim <jaegeuk@kernel.org> Cc: linux-f2fs-devel@lists.sourceforge.net, linux-fsdevel@vger.kernel.org, linux-xfs@vger.kernel.org, Satya Tangirala <satyaprateek2357@gmail.com>, Changheun Lee <nanich.lee@samsung.com>, Matthew Bobrowski <mbobrowski@mbobrowski.org> Subject: Re: [PATCH 3/9] f2fs: rework write preallocations Date: Tue, 27 Jul 2021 11:23:03 +0800 [thread overview] Message-ID: <70f16fec-02f6-cb19-c407-856101cacc23@kernel.org> (raw) In-Reply-To: <YP9oou9sx4oJF1sc@google.com> On 2021/7/27 10:00, Jaegeuk Kim wrote: > On 07/25, Eric Biggers wrote: >> On Sun, Jul 25, 2021 at 06:50:51PM +0800, Chao Yu wrote: >>> On 2021/7/16 22:39, Eric Biggers wrote: >>>> From: Eric Biggers <ebiggers@google.com> >>>> >>>> f2fs_write_begin() assumes that all blocks were preallocated by >>>> default unless FI_NO_PREALLOC is explicitly set. This invites data >>>> corruption, as there are cases in which not all blocks are preallocated. >>>> Commit 47501f87c61a ("f2fs: preallocate DIO blocks when forcing >>>> buffered_io") fixed one case, but there are others remaining. >>> >>> Could you please explain which cases we missed to handle previously? >>> then I can check those related logic before and after the rework. >> >> Any case where a buffered write happens while not all blocks were preallocated >> but FI_NO_PREALLOC wasn't set. For example when ENOSPC was hit in the middle of >> the preallocations for a direct write that will fall back to a buffered write, >> e.g. due to f2fs_force_buffered_io() or page cache invalidation failure. Indeed, IIUC, the buggy code is as below, if any preallocation failed, we need to set FI_NO_PREALLOC flag. map_blocks: err = f2fs_map_blocks(inode, &map, 1, flag); if (map.m_len > 0 && err == -ENOSPC) { if (!direct_io) <---- set_inode_flag(inode, FI_NO_PREALLOC); err = 0; } BTW, it will be better to include above issue details you explained into commit message? >> >>> >>>> - /* >>>> - * If force_buffere_io() is true, we have to allocate >>>> - * blocks all the time, since f2fs_direct_IO will fall >>>> - * back to buffered IO. >>>> - */ >>>> - if (!f2fs_force_buffered_io(inode, iocb, from) && >>>> - f2fs_lfs_mode(F2FS_I_SB(inode))) >>>> - goto write; >>> >>> We should keep this OPU DIO logic, otherwise, in lfs mode, write dio >>> will always allocate two block addresses for each 4k append IO. >>> >>> I jsut test based on codes of last f2fs dev-test branch. >> >> Yes, I had misread that due to the weird goto and misleading comment and >> translated it into: >> >> /* If it will be an in-place direct write, don't bother. */ >> if (dio && !f2fs_lfs_mode(sbi)) >> return 0; >> >> It should be: >> >> if (dio && f2fs_lfs_mode(sbi)) >> return 0; > > Hmm, this addresses my 250 failure. And, I think the below commit can explain > the case. > > commit 47501f87c61ad2aa234add63e1ae231521dbc3f5 > Author: Jaegeuk Kim <jaegeuk@kernel.org> > Date: Tue Nov 26 15:01:42 2019 -0800 > > f2fs: preallocate DIO blocks when forcing buffered_io > > The previous preallocation and DIO decision like below. > > allow_outplace_dio !allow_outplace_dio > f2fs_force_buffered_io (*) No_Prealloc / Buffered_IO Prealloc / Buffered_IO > !f2fs_force_buffered_io No_Prealloc / DIO Prealloc / DIO > > But, Javier reported Case (*) where zoned device bypassed preallocation but > fell back to buffered writes in f2fs_direct_IO(), resulting in stale data > being read. > > In order to fix the issue, actually we need to preallocate blocks whenever > we fall back to buffered IO like this. No change is made in the other cases. > > allow_outplace_dio !allow_outplace_dio > f2fs_force_buffered_io (*) Prealloc / Buffered_IO Prealloc / Buffered_IO > !f2fs_force_buffered_io No_Prealloc / DIO Prealloc / DIO > > Reported-and-tested-by: Javier Gonzalez <javier@javigon.com> > Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> > Tested-by: Shin'ichiro Kawasaki <shinichiro.kawasaki@wdc.com> > Reviewed-by: Chao Yu <yuchao0@huawei.com> > Reviewed-by: Javier González <javier@javigon.com> > Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org> > Thanks for the explain. > >> >> Do you have a proper explanation for why preallocations shouldn't be done in See commit f847c699cff3 ("f2fs: allow out-place-update for direct IO in LFS mode"), f2fs_map_blocks() logic was changed to force allocating a new block address no matter previous block address was existed if it is called from write path of DIO. So, in such condition, if we preallocate new block address in f2fs_file_write_iter(), we will suffer the problem which my trace indicates. >> this case? Note that preallocations are still done for buffered writes, which >> may be out-of-place as well; how are those different? Got your concern. For buffered IO, we use F2FS_GET_BLOCK_PRE_AIO, in this mode, we just preserve filesystem block count and tag NEW_ADDR in dnode block, so, it's fine, double new block address allocation won't happen during data page writeback. For direct IO, we use F2FS_GET_BLOCK_PRE_DIO, in this mode, we will allocate physical block address, after preallocation, if we fallback to buffered IO, we may suffer double new block address allocation issue... IIUC. Well, can we relocate preallocation into f2fs_direct_IO() after all cases which may cause fallbacking DIO to buffered IO? Thanks, >> >> - Eric
WARNING: multiple messages have this Message-ID (diff)
From: Chao Yu <chao@kernel.org> To: Eric Biggers <ebiggers@kernel.org>, Jaegeuk Kim <jaegeuk@kernel.org> Cc: Satya Tangirala <satyaprateek2357@gmail.com>, linux-f2fs-devel@lists.sourceforge.net, linux-xfs@vger.kernel.org, Matthew Bobrowski <mbobrowski@mbobrowski.org>, Changheun Lee <nanich.lee@samsung.com>, linux-fsdevel@vger.kernel.org Subject: Re: [f2fs-dev] [PATCH 3/9] f2fs: rework write preallocations Date: Tue, 27 Jul 2021 11:23:03 +0800 [thread overview] Message-ID: <70f16fec-02f6-cb19-c407-856101cacc23@kernel.org> (raw) In-Reply-To: <YP9oou9sx4oJF1sc@google.com> On 2021/7/27 10:00, Jaegeuk Kim wrote: > On 07/25, Eric Biggers wrote: >> On Sun, Jul 25, 2021 at 06:50:51PM +0800, Chao Yu wrote: >>> On 2021/7/16 22:39, Eric Biggers wrote: >>>> From: Eric Biggers <ebiggers@google.com> >>>> >>>> f2fs_write_begin() assumes that all blocks were preallocated by >>>> default unless FI_NO_PREALLOC is explicitly set. This invites data >>>> corruption, as there are cases in which not all blocks are preallocated. >>>> Commit 47501f87c61a ("f2fs: preallocate DIO blocks when forcing >>>> buffered_io") fixed one case, but there are others remaining. >>> >>> Could you please explain which cases we missed to handle previously? >>> then I can check those related logic before and after the rework. >> >> Any case where a buffered write happens while not all blocks were preallocated >> but FI_NO_PREALLOC wasn't set. For example when ENOSPC was hit in the middle of >> the preallocations for a direct write that will fall back to a buffered write, >> e.g. due to f2fs_force_buffered_io() or page cache invalidation failure. Indeed, IIUC, the buggy code is as below, if any preallocation failed, we need to set FI_NO_PREALLOC flag. map_blocks: err = f2fs_map_blocks(inode, &map, 1, flag); if (map.m_len > 0 && err == -ENOSPC) { if (!direct_io) <---- set_inode_flag(inode, FI_NO_PREALLOC); err = 0; } BTW, it will be better to include above issue details you explained into commit message? >> >>> >>>> - /* >>>> - * If force_buffere_io() is true, we have to allocate >>>> - * blocks all the time, since f2fs_direct_IO will fall >>>> - * back to buffered IO. >>>> - */ >>>> - if (!f2fs_force_buffered_io(inode, iocb, from) && >>>> - f2fs_lfs_mode(F2FS_I_SB(inode))) >>>> - goto write; >>> >>> We should keep this OPU DIO logic, otherwise, in lfs mode, write dio >>> will always allocate two block addresses for each 4k append IO. >>> >>> I jsut test based on codes of last f2fs dev-test branch. >> >> Yes, I had misread that due to the weird goto and misleading comment and >> translated it into: >> >> /* If it will be an in-place direct write, don't bother. */ >> if (dio && !f2fs_lfs_mode(sbi)) >> return 0; >> >> It should be: >> >> if (dio && f2fs_lfs_mode(sbi)) >> return 0; > > Hmm, this addresses my 250 failure. And, I think the below commit can explain > the case. > > commit 47501f87c61ad2aa234add63e1ae231521dbc3f5 > Author: Jaegeuk Kim <jaegeuk@kernel.org> > Date: Tue Nov 26 15:01:42 2019 -0800 > > f2fs: preallocate DIO blocks when forcing buffered_io > > The previous preallocation and DIO decision like below. > > allow_outplace_dio !allow_outplace_dio > f2fs_force_buffered_io (*) No_Prealloc / Buffered_IO Prealloc / Buffered_IO > !f2fs_force_buffered_io No_Prealloc / DIO Prealloc / DIO > > But, Javier reported Case (*) where zoned device bypassed preallocation but > fell back to buffered writes in f2fs_direct_IO(), resulting in stale data > being read. > > In order to fix the issue, actually we need to preallocate blocks whenever > we fall back to buffered IO like this. No change is made in the other cases. > > allow_outplace_dio !allow_outplace_dio > f2fs_force_buffered_io (*) Prealloc / Buffered_IO Prealloc / Buffered_IO > !f2fs_force_buffered_io No_Prealloc / DIO Prealloc / DIO > > Reported-and-tested-by: Javier Gonzalez <javier@javigon.com> > Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> > Tested-by: Shin'ichiro Kawasaki <shinichiro.kawasaki@wdc.com> > Reviewed-by: Chao Yu <yuchao0@huawei.com> > Reviewed-by: Javier González <javier@javigon.com> > Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org> > Thanks for the explain. > >> >> Do you have a proper explanation for why preallocations shouldn't be done in See commit f847c699cff3 ("f2fs: allow out-place-update for direct IO in LFS mode"), f2fs_map_blocks() logic was changed to force allocating a new block address no matter previous block address was existed if it is called from write path of DIO. So, in such condition, if we preallocate new block address in f2fs_file_write_iter(), we will suffer the problem which my trace indicates. >> this case? Note that preallocations are still done for buffered writes, which >> may be out-of-place as well; how are those different? Got your concern. For buffered IO, we use F2FS_GET_BLOCK_PRE_AIO, in this mode, we just preserve filesystem block count and tag NEW_ADDR in dnode block, so, it's fine, double new block address allocation won't happen during data page writeback. For direct IO, we use F2FS_GET_BLOCK_PRE_DIO, in this mode, we will allocate physical block address, after preallocation, if we fallback to buffered IO, we may suffer double new block address allocation issue... IIUC. Well, can we relocate preallocation into f2fs_direct_IO() after all cases which may cause fallbacking DIO to buffered IO? Thanks, >> >> - Eric _______________________________________________ Linux-f2fs-devel mailing list Linux-f2fs-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
next prev parent reply other threads:[~2021-07-27 3:23 UTC|newest] Thread overview: 66+ messages / expand[flat|nested] mbox.gz Atom feed top 2021-07-16 14:39 [PATCH 0/9] f2fs: use iomap for direct I/O Eric Biggers 2021-07-16 14:39 ` [f2fs-dev] " Eric Biggers 2021-07-16 14:39 ` [PATCH 1/9] f2fs: make f2fs_write_failed() take struct inode Eric Biggers 2021-07-16 14:39 ` [f2fs-dev] " Eric Biggers 2021-07-25 10:00 ` Chao Yu 2021-07-25 10:00 ` [f2fs-dev] " Chao Yu 2021-07-16 14:39 ` [PATCH 2/9] f2fs: remove allow_outplace_dio() Eric Biggers 2021-07-16 14:39 ` [f2fs-dev] " Eric Biggers 2021-07-19 8:41 ` Christoph Hellwig 2021-07-19 8:41 ` [f2fs-dev] " Christoph Hellwig 2021-07-16 14:39 ` [PATCH 3/9] f2fs: rework write preallocations Eric Biggers 2021-07-16 14:39 ` [f2fs-dev] " Eric Biggers 2021-07-25 10:50 ` Chao Yu 2021-07-25 10:50 ` [f2fs-dev] " Chao Yu 2021-07-25 17:57 ` Eric Biggers 2021-07-25 17:57 ` [f2fs-dev] " Eric Biggers 2021-07-27 2:00 ` Jaegeuk Kim 2021-07-27 2:00 ` [f2fs-dev] " Jaegeuk Kim 2021-07-27 3:23 ` Chao Yu [this message] 2021-07-27 3:23 ` Chao Yu 2021-07-27 7:38 ` Eric Biggers 2021-07-27 7:38 ` [f2fs-dev] " Eric Biggers 2021-07-27 8:30 ` Chao Yu 2021-07-27 8:30 ` [f2fs-dev] " Chao Yu 2021-07-27 15:33 ` Darrick J. Wong 2021-07-27 15:33 ` [f2fs-dev] " Darrick J. Wong 2021-07-29 0:26 ` Chao Yu 2021-07-29 0:26 ` [f2fs-dev] " Chao Yu 2021-07-28 2:29 ` Jaegeuk Kim 2021-07-28 2:29 ` [f2fs-dev] " Jaegeuk Kim 2021-07-25 15:35 ` Jaegeuk Kim 2021-07-25 15:35 ` [f2fs-dev] " Jaegeuk Kim 2021-07-25 15:47 ` Jaegeuk Kim 2021-07-25 15:47 ` [f2fs-dev] " Jaegeuk Kim 2021-07-25 18:01 ` Eric Biggers 2021-07-25 18:01 ` [f2fs-dev] " Eric Biggers 2021-07-26 19:04 ` Jaegeuk Kim 2021-07-26 19:04 ` [f2fs-dev] " Jaegeuk Kim 2021-07-16 14:39 ` [PATCH 4/9] f2fs: reduce indentation in f2fs_file_write_iter() Eric Biggers 2021-07-16 14:39 ` [f2fs-dev] " Eric Biggers 2021-07-16 14:39 ` [PATCH 5/9] f2fs: fix the f2fs_file_write_iter tracepoint Eric Biggers 2021-07-16 14:39 ` [f2fs-dev] " Eric Biggers 2021-07-16 14:39 ` [PATCH 6/9] f2fs: implement iomap operations Eric Biggers 2021-07-16 14:39 ` [f2fs-dev] " Eric Biggers 2021-07-19 8:59 ` Christoph Hellwig 2021-07-19 8:59 ` [f2fs-dev] " Christoph Hellwig 2021-07-22 20:47 ` Jaegeuk Kim 2021-07-22 20:47 ` [f2fs-dev] " Jaegeuk Kim 2021-07-22 20:49 ` Jaegeuk Kim 2021-07-22 20:49 ` [f2fs-dev] " Jaegeuk Kim 2021-07-22 20:54 ` Eric Biggers 2021-07-22 20:54 ` [f2fs-dev] " Eric Biggers 2021-07-22 21:57 ` Jaegeuk Kim 2021-07-22 21:57 ` [f2fs-dev] " Jaegeuk Kim 2021-07-23 1:52 ` Eric Biggers 2021-07-23 1:52 ` [f2fs-dev] " Eric Biggers 2021-07-23 5:00 ` Christoph Hellwig 2021-07-23 5:00 ` [f2fs-dev] " Christoph Hellwig 2021-07-23 8:05 ` Eric Biggers 2021-07-23 8:05 ` [f2fs-dev] " Eric Biggers 2021-07-16 14:39 ` [PATCH 7/9] f2fs: use iomap for direct I/O reads Eric Biggers 2021-07-16 14:39 ` [f2fs-dev] " Eric Biggers 2021-07-16 14:39 ` [PATCH 8/9] f2fs: use iomap for direct I/O writes Eric Biggers 2021-07-16 14:39 ` [f2fs-dev] " Eric Biggers 2021-07-16 14:39 ` [PATCH 9/9] f2fs: remove f2fs_direct_IO() Eric Biggers 2021-07-16 14:39 ` [f2fs-dev] " Eric Biggers
Reply instructions: You may reply publicly to this message via plain-text email using any one of the following methods: * Save the following mbox file, import it into your mail client, and reply-to-all from there: mbox Avoid top-posting and favor interleaved quoting: https://en.wikipedia.org/wiki/Posting_style#Interleaved_style * Reply using the --to, --cc, and --in-reply-to switches of git-send-email(1): git send-email \ --in-reply-to=70f16fec-02f6-cb19-c407-856101cacc23@kernel.org \ --to=chao@kernel.org \ --cc=ebiggers@kernel.org \ --cc=jaegeuk@kernel.org \ --cc=linux-f2fs-devel@lists.sourceforge.net \ --cc=linux-fsdevel@vger.kernel.org \ --cc=linux-xfs@vger.kernel.org \ --cc=mbobrowski@mbobrowski.org \ --cc=nanich.lee@samsung.com \ --cc=satyaprateek2357@gmail.com \ /path/to/YOUR_REPLY https://kernel.org/pub/software/scm/git/docs/git-send-email.html * If your mail client supports setting the In-Reply-To header via mailto: links, try the mailto: linkBe sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.