From: Chao Yu <chao@kernel.org>
To: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: Chao Yu <yuchao0@huawei.com>,
linux-f2fs-devel@lists.sourceforge.net,
linux-kernel@vger.kernel.org
Subject: Re: [PATCH 3/4] f2fs: fix avoid race between truncate and background GC
Date: Sun, 29 Jul 2018 14:16:29 +0800 [thread overview]
Message-ID: <17b3ee11-ad97-b313-629f-8167bc8a266a@kernel.org> (raw)
In-Reply-To: <20180729061351.GA5102@jaegeuk-macbookpro.roam.corp.google.com>
On 2018/7/29 14:13, Jaegeuk Kim wrote:
> On 07/29, Chao Yu wrote:
>> On 2018/7/29 9:58, Jaegeuk Kim wrote:
>>> On 07/27, Chao Yu wrote:
>>>> Thread A Background GC
>>>> - f2fs_setattr isize to 0
>>>> - truncate_setsize
>>>> - gc_data_segment
>>>> - f2fs_get_read_data_page page #0
>>>> - set_page_dirty
>>>> - set_cold_data
>>>> - f2fs_truncate
>>>>
>>>> - f2fs_setattr isize to 4k
>>>> - read 4k <--- hit data in cached page #0
>>>>
>>>> Above race condition can cause read out invalid data in a truncated
>>>> page, fix it by i_gc_rwsem[WRITE] lock.
>>>
>>> We can truncate pages again after f2fs_truncate()?
>>
>> I think it can fix this issue, although it looks a little strange to truncate
>> page cache twice.
>
> It'd be fine to do setsize first, and drop the cache later.
Let me rethink about this solution, since we have different case like punch_hole
which may not need setsize.
Thanks,
>
>>
>> Thanks,
>>
>>>
>>>>
>>>> Signed-off-by: Chao Yu <yuchao0@huawei.com>
>>>> ---
>>>> fs/f2fs/data.c | 4 ++++
>>>> fs/f2fs/file.c | 33 +++++++++++++++++++--------------
>>>> 2 files changed, 23 insertions(+), 14 deletions(-)
>>>>
>>>> diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
>>>> index 071224ded5f4..a29f3162b887 100644
>>>> --- a/fs/f2fs/data.c
>>>> +++ b/fs/f2fs/data.c
>>>> @@ -2214,10 +2214,14 @@ static void f2fs_write_failed(struct address_space *mapping, loff_t to)
>>>> loff_t i_size = i_size_read(inode);
>>>>
>>>> if (to > i_size) {
>>>> + down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
>>>> down_write(&F2FS_I(inode)->i_mmap_sem);
>>>> +
>>>> truncate_pagecache(inode, i_size);
>>>> f2fs_truncate_blocks(inode, i_size, true);
>>>> +
>>>> up_write(&F2FS_I(inode)->i_mmap_sem);
>>>> + up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
>>>> }
>>>> }
>>>>
>>>> diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
>>>> index 7bd2412a8c37..ed5c9b0e0d0c 100644
>>>> --- a/fs/f2fs/file.c
>>>> +++ b/fs/f2fs/file.c
>>>> @@ -796,22 +796,25 @@ int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
>>>> }
>>>>
>>>> if (attr->ia_valid & ATTR_SIZE) {
>>>> - if (attr->ia_size <= i_size_read(inode)) {
>>>> - down_write(&F2FS_I(inode)->i_mmap_sem);
>>>> - truncate_setsize(inode, attr->ia_size);
>>>> + bool to_smaller = (attr->ia_size <= i_size_read(inode));
>>>> +
>>>> + down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
>>>> + down_write(&F2FS_I(inode)->i_mmap_sem);
>>>> +
>>>> + truncate_setsize(inode, attr->ia_size);
>>>> +
>>>> + if (to_smaller)
>>>> err = f2fs_truncate(inode);
>>>> - up_write(&F2FS_I(inode)->i_mmap_sem);
>>>> - if (err)
>>>> - return err;
>>>> - } else {
>>>> - /*
>>>> - * do not trim all blocks after i_size if target size is
>>>> - * larger than i_size.
>>>> - */
>>>> - down_write(&F2FS_I(inode)->i_mmap_sem);
>>>> - truncate_setsize(inode, attr->ia_size);
>>>> - up_write(&F2FS_I(inode)->i_mmap_sem);
>>>> + /*
>>>> + * do not trim all blocks after i_size if target size is
>>>> + * larger than i_size.
>>>> + */
>>>> + up_write(&F2FS_I(inode)->i_mmap_sem);
>>>> + up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
>>>> + if (err)
>>>> + return err;
>>>>
>>>> + if (!to_smaller) {
>>>> /* should convert inline inode here */
>>>> if (!f2fs_may_inline_data(inode)) {
>>>> err = f2fs_convert_inline_inode(inode);
>>>> @@ -958,6 +961,7 @@ static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
>>>>
>>>> blk_start = (loff_t)pg_start << PAGE_SHIFT;
>>>> blk_end = (loff_t)pg_end << PAGE_SHIFT;
>>>> + down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
>>>> down_write(&F2FS_I(inode)->i_mmap_sem);
>>>> truncate_inode_pages_range(mapping, blk_start,
>>>> blk_end - 1);
>>>> @@ -966,6 +970,7 @@ static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
>>>> ret = f2fs_truncate_hole(inode, pg_start, pg_end);
>>>> f2fs_unlock_op(sbi);
>>>> up_write(&F2FS_I(inode)->i_mmap_sem);
>>>> + up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
>>>> }
>>>> }
>>>>
>>>> --
>>>> 2.18.0.rc1
next prev parent reply other threads:[~2018-07-29 6:16 UTC|newest]
Thread overview: 12+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-07-27 10:15 [PATCH 1/4] f2fs: don't keep meta pages used for block migration Chao Yu
2018-07-27 10:15 ` [PATCH 2/4] f2fs: fix to active page in lru list for read path Chao Yu
2018-07-27 10:15 ` [PATCH 3/4] f2fs: fix avoid race between truncate and background GC Chao Yu
2018-07-29 1:58 ` Jaegeuk Kim
2018-07-29 6:02 ` Chao Yu
2018-07-29 6:13 ` Jaegeuk Kim
2018-07-29 6:16 ` Chao Yu [this message]
2018-07-27 10:15 ` [PATCH 4/4] f2fs: fix to spread clear_cold_data() Chao Yu
2018-07-29 2:00 ` Jaegeuk Kim
2018-07-29 2:19 ` Chao Yu
2018-07-29 2:44 ` Jaegeuk Kim
2018-07-29 3:19 ` Chao Yu
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=17b3ee11-ad97-b313-629f-8167bc8a266a@kernel.org \
--to=chao@kernel.org \
--cc=jaegeuk@kernel.org \
--cc=linux-f2fs-devel@lists.sourceforge.net \
--cc=linux-kernel@vger.kernel.org \
--cc=yuchao0@huawei.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: 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 NNTP newsgroup(s).