From: Chao Yu <chao@kernel.org>
To: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: linux-kernel@vger.kernel.org, linux-f2fs-devel@lists.sourceforge.net
Subject: Re: [f2fs-dev] [PATCH v2] f2fs: fix to avoid racing in between read and OPU dio write
Date: Wed, 15 May 2024 14:38:39 +0800 [thread overview]
Message-ID: <efae597c-d334-498b-9050-1a21bf40e21d@kernel.org> (raw)
In-Reply-To: <ZkQ9Uo5713Xpr2n7@google.com>
On 2024/5/15 12:42, Jaegeuk Kim wrote:
> On 05/15, Chao Yu wrote:
>> On 2024/5/15 0:09, Jaegeuk Kim wrote:
>>> On 05/10, Chao Yu wrote:
>>>> If lfs mode is on, buffered read may race w/ OPU dio write as below,
>>>> it may cause buffered read hits unwritten data unexpectly, and for
>>>> dio read, the race condition exists as well.
>>>>
>>>> Thread A Thread B
>>>> - f2fs_file_write_iter
>>>> - f2fs_dio_write_iter
>>>> - __iomap_dio_rw
>>>> - f2fs_iomap_begin
>>>> - f2fs_map_blocks
>>>> - __allocate_data_block
>>>> - allocated blkaddr #x
>>>> - iomap_dio_submit_bio
>>>> - f2fs_file_read_iter
>>>> - filemap_read
>>>> - f2fs_read_data_folio
>>>> - f2fs_mpage_readpages
>>>> - f2fs_map_blocks
>>>> : get blkaddr #x
>>>> - f2fs_submit_read_bio
>>>> IRQ
>>>> - f2fs_read_end_io
>>>> : read IO on blkaddr #x complete
>>>> IRQ
>>>> - iomap_dio_bio_end_io
>>>> : direct write IO on blkaddr #x complete
>>>>
>>>> This patch introduces a new per-inode i_opu_rwsem lock to avoid
>>>> such race condition.
>>>
>>> Wasn't this supposed to be managed by user-land?
>>
>> Actually, the test case is:
>>
>> 1. mount w/ lfs mode
>> 2. touch file;
>> 3. initialize file w/ 4k zeroed data; fsync;
>> 4. continue triggering dio write 4k zeroed data to file;
>> 5. and meanwhile, continue triggering buf/dio 4k read in file,
>> use md5sum to verify the 4k data;
>>
>> It expects data is all zero, however it turned out it's not.
>
> Can we check outstanding write bios instead of abusing locks?
I didn't figure out a way to solve this w/o lock, due to:
- write bios can be issued after outstanding write bios check condition,
result in the race.
- once read() detects that there are outstanding write bios, we need to
delay read flow rather than fail it, right? It looks using a lock is more
proper here?
Any suggestion?
Thanks,
>
>>
>> Thanks,
>>
>>>
>>>>
>>>> Fixes: f847c699cff3 ("f2fs: allow out-place-update for direct IO in LFS mode")
>>>> Signed-off-by: Chao Yu <chao@kernel.org>
>>>> ---
>>>> v2:
>>>> - fix to cover dio read path w/ i_opu_rwsem as well.
>>>> fs/f2fs/f2fs.h | 1 +
>>>> fs/f2fs/file.c | 28 ++++++++++++++++++++++++++--
>>>> fs/f2fs/super.c | 1 +
>>>> 3 files changed, 28 insertions(+), 2 deletions(-)
>>>>
>>>> diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
>>>> index 30058e16a5d0..91cf4b3d6bc6 100644
>>>> --- a/fs/f2fs/f2fs.h
>>>> +++ b/fs/f2fs/f2fs.h
>>>> @@ -847,6 +847,7 @@ struct f2fs_inode_info {
>>>> /* avoid racing between foreground op and gc */
>>>> struct f2fs_rwsem i_gc_rwsem[2];
>>>> struct f2fs_rwsem i_xattr_sem; /* avoid racing between reading and changing EAs */
>>>> + struct f2fs_rwsem i_opu_rwsem; /* avoid racing between buf read and opu dio write */
>>>>
>>>> int i_extra_isize; /* size of extra space located in i_addr */
>>>> kprojid_t i_projid; /* id for project quota */
>>>> diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
>>>> index 72ce1a522fb2..4ec260af321f 100644
>>>> --- a/fs/f2fs/file.c
>>>> +++ b/fs/f2fs/file.c
>>>> @@ -4445,6 +4445,7 @@ static ssize_t f2fs_dio_read_iter(struct kiocb *iocb, struct iov_iter *to)
>>>> const loff_t pos = iocb->ki_pos;
>>>> const size_t count = iov_iter_count(to);
>>>> struct iomap_dio *dio;
>>>> + bool do_opu = f2fs_lfs_mode(sbi);
>>>> ssize_t ret;
>>>>
>>>> if (count == 0)
>>>> @@ -4457,8 +4458,14 @@ static ssize_t f2fs_dio_read_iter(struct kiocb *iocb, struct iov_iter *to)
>>>> ret = -EAGAIN;
>>>> goto out;
>>>> }
>>>> + if (do_opu && !f2fs_down_read_trylock(&fi->i_opu_rwsem)) {
>>>> + f2fs_up_read(&fi->i_gc_rwsem[READ]);
>>>> + ret = -EAGAIN;
>>>> + goto out;
>>>> + }
>>>> } else {
>>>> f2fs_down_read(&fi->i_gc_rwsem[READ]);
>>>> + f2fs_down_read(&fi->i_opu_rwsem);
>>>> }
>>>>
>>>> /*
>>>> @@ -4477,6 +4484,7 @@ static ssize_t f2fs_dio_read_iter(struct kiocb *iocb, struct iov_iter *to)
>>>> ret = iomap_dio_complete(dio);
>>>> }
>>>>
>>>> + f2fs_up_read(&fi->i_opu_rwsem);
>>>> f2fs_up_read(&fi->i_gc_rwsem[READ]);
>>>>
>>>> file_accessed(file);
>>>> @@ -4523,7 +4531,13 @@ static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *to)
>>>> if (f2fs_should_use_dio(inode, iocb, to)) {
>>>> ret = f2fs_dio_read_iter(iocb, to);
>>>> } else {
>>>> + bool do_opu = f2fs_lfs_mode(F2FS_I_SB(inode));
>>>> +
>>>> + if (do_opu)
>>>> + f2fs_down_read(&F2FS_I(inode)->i_opu_rwsem);
>>>> ret = filemap_read(iocb, to, 0);
>>>> + if (do_opu)
>>>> + f2fs_up_read(&F2FS_I(inode)->i_opu_rwsem);
>>>> if (ret > 0)
>>>> f2fs_update_iostat(F2FS_I_SB(inode), inode,
>>>> APP_BUFFERED_READ_IO, ret);
>>>> @@ -4748,14 +4762,22 @@ static ssize_t f2fs_dio_write_iter(struct kiocb *iocb, struct iov_iter *from,
>>>> ret = -EAGAIN;
>>>> goto out;
>>>> }
>>>> + if (do_opu && !f2fs_down_write_trylock(&fi->i_opu_rwsem)) {
>>>> + f2fs_up_read(&fi->i_gc_rwsem[READ]);
>>>> + f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
>>>> + ret = -EAGAIN;
>>>> + goto out;
>>>> + }
>>>> } else {
>>>> ret = f2fs_convert_inline_inode(inode);
>>>> if (ret)
>>>> goto out;
>>>>
>>>> f2fs_down_read(&fi->i_gc_rwsem[WRITE]);
>>>> - if (do_opu)
>>>> + if (do_opu) {
>>>> f2fs_down_read(&fi->i_gc_rwsem[READ]);
>>>> + f2fs_down_write(&fi->i_opu_rwsem);
>>>> + }
>>>> }
>>>>
>>>> /*
>>>> @@ -4779,8 +4801,10 @@ static ssize_t f2fs_dio_write_iter(struct kiocb *iocb, struct iov_iter *from,
>>>> ret = iomap_dio_complete(dio);
>>>> }
>>>>
>>>> - if (do_opu)
>>>> + if (do_opu) {
>>>> + f2fs_up_write(&fi->i_opu_rwsem);
>>>> f2fs_up_read(&fi->i_gc_rwsem[READ]);
>>>> + }
>>>> f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
>>>>
>>>> if (ret < 0)
>>>> diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
>>>> index daf2c4dbe150..b4ed3b094366 100644
>>>> --- a/fs/f2fs/super.c
>>>> +++ b/fs/f2fs/super.c
>>>> @@ -1428,6 +1428,7 @@ static struct inode *f2fs_alloc_inode(struct super_block *sb)
>>>> init_f2fs_rwsem(&fi->i_gc_rwsem[READ]);
>>>> init_f2fs_rwsem(&fi->i_gc_rwsem[WRITE]);
>>>> init_f2fs_rwsem(&fi->i_xattr_sem);
>>>> + init_f2fs_rwsem(&fi->i_opu_rwsem);
>>>>
>>>> /* Will be used by directory only */
>>>> fi->i_dir_level = F2FS_SB(sb)->dir_level;
>>>> --
>>>> 2.40.1
_______________________________________________
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:[~2024-05-15 6:38 UTC|newest]
Thread overview: 8+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-05-10 2:39 [f2fs-dev] [PATCH v2] f2fs: fix to avoid racing in between read and OPU dio write Chao Yu
2024-05-14 16:09 ` Jaegeuk Kim
2024-05-15 1:42 ` Chao Yu
2024-05-15 4:42 ` Jaegeuk Kim
2024-05-15 6:38 ` Chao Yu [this message]
2024-05-15 8:32 ` Wu Bo via Linux-f2fs-devel
2024-05-15 8:40 ` Markus Elfring via Linux-f2fs-devel
2024-05-17 8:15 ` kernel test robot
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=efae597c-d334-498b-9050-1a21bf40e21d@kernel.org \
--to=chao@kernel.org \
--cc=jaegeuk@kernel.org \
--cc=linux-f2fs-devel@lists.sourceforge.net \
--cc=linux-kernel@vger.kernel.org \
/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: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).