From: Chao Yu <yuchao0@huawei.com>
To: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: kernel-team@android.com, linux-kernel@vger.kernel.org,
linux-f2fs-devel@lists.sourceforge.net
Subject: Re: [f2fs-dev] [PATCH v3] f2fs: avoid inifinite loop to wait for flushing node pages at cp_error
Date: Wed, 27 May 2020 10:35:35 +0800 [thread overview]
Message-ID: <765a1ac5-a318-14d6-666f-eab46f892d01@huawei.com> (raw)
In-Reply-To: <20200526015650.GA207949@google.com>
On 2020/5/26 9:56, Jaegeuk Kim wrote:
> On 05/26, Chao Yu wrote:
>> On 2020/5/26 9:11, Chao Yu wrote:
>>> On 2020/5/25 23:06, Jaegeuk Kim wrote:
>>>> On 05/25, Chao Yu wrote:
>>>>> On 2020/5/25 11:56, Jaegeuk Kim wrote:
>>>>>> Shutdown test is somtimes hung, since it keeps trying to flush dirty node pages
71.07% 0.01% kworker/u256:1+ [kernel.kallsyms] [k] wb_writeback
|
--71.06%--wb_writeback
|
|--68.96%--__writeback_inodes_wb
| |
| --68.95%--writeback_sb_inodes
| |
| |--65.08%--__writeback_single_inode
| | |
| | --64.35%--do_writepages
| | |
| | |--59.83%--f2fs_write_node_pages
| | | |
| | | --59.74%--f2fs_sync_node_pages
| | | |
| | | |--27.91%--pagevec_lookup_range_tag
| | | | |
| | | | --27.90%--find_get_pages_range_tag
Before umount, kworker will always hold one core, that looks not reasonable,
to avoid that, could we just allow node write, since it's out-place-update,
and cp is not allowed, we don't need to worry about its effect on data on
previous checkpoint, and it can decrease memory footprint cost by node pages.
Thanks,
>>>>>
>>>>> IMO, for umount case, we should drop dirty reference and dirty pages on meta/data
>>>>> pages like we change for node pages to avoid potential dead loop...
>>>>
>>>> I believe we're doing for them. :P
>>>
>>> Actually, I mean do we need to drop dirty meta/data pages explicitly as below:
>>>
>>> diff --git a/fs/f2fs/checkpoint.c b/fs/f2fs/checkpoint.c
>>> index 3dc3ac6fe143..4c08fd0a680a 100644
>>> --- a/fs/f2fs/checkpoint.c
>>> +++ b/fs/f2fs/checkpoint.c
>>> @@ -299,8 +299,15 @@ static int __f2fs_write_meta_page(struct page *page,
>>>
>>> trace_f2fs_writepage(page, META);
>>>
>>> - if (unlikely(f2fs_cp_error(sbi)))
>>> + if (unlikely(f2fs_cp_error(sbi))) {
>>> + if (is_sbi_flag_set(sbi, SBI_IS_CLOSE)) {
>>> + ClearPageUptodate(page);
>>> + dec_page_count(sbi, F2FS_DIRTY_META);
>>> + unlock_page(page);
>>> + return 0;
>>> + }
>>> goto redirty_out;
>>> + }
>>> if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
>>> goto redirty_out;
>>> if (wbc->for_reclaim && page->index < GET_SUM_BLOCK(sbi, 0))
>>> diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
>>> index 48a622b95b76..94b342802513 100644
>>> --- a/fs/f2fs/data.c
>>> +++ b/fs/f2fs/data.c
>>> @@ -2682,6 +2682,12 @@ int f2fs_write_single_data_page(struct page *page, int *submitted,
>>>
>>> /* we should bypass data pages to proceed the kworkder jobs */
>>> if (unlikely(f2fs_cp_error(sbi))) {
>>> + if (is_sbi_flag_set(sbi, SBI_IS_CLOSE)) {
>>> + ClearPageUptodate(page);
>>> + inode_dec_dirty_pages(inode);
>>> + unlock_page(page);
>>> + return 0;
>>> + }
>>
>> Oh, I notice previously, we will drop non-directory inode's dirty pages directly,
>> however, during umount, we'd better drop directory inode's dirty pages as well, right?
>
> Hmm, I remember I dropped them before. Need to double check.
>
>>
>>> mapping_set_error(page->mapping, -EIO);
>>> /*
>>> * don't drop any dirty dentry pages for keeping lastest
>>>
>>>>
>>>>>
>>>>> Thanks,
>>>>>
>>>>>> in an inifinite loop. Let's drop dirty pages at umount in that case.
>>>>>>
>>>>>> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
>>>>>> ---
>>>>>> v3:
>>>>>> - fix wrong unlock
>>>>>>
>>>>>> v2:
>>>>>> - fix typos
>>>>>>
>>>>>> fs/f2fs/node.c | 9 ++++++++-
>>>>>> 1 file changed, 8 insertions(+), 1 deletion(-)
>>>>>>
>>>>>> diff --git a/fs/f2fs/node.c b/fs/f2fs/node.c
>>>>>> index e632de10aedab..e0bb0f7e0506e 100644
>>>>>> --- a/fs/f2fs/node.c
>>>>>> +++ b/fs/f2fs/node.c
>>>>>> @@ -1520,8 +1520,15 @@ static int __write_node_page(struct page *page, bool atomic, bool *submitted,
>>>>>>
>>>>>> trace_f2fs_writepage(page, NODE);
>>>>>>
>>>>>> - if (unlikely(f2fs_cp_error(sbi)))
>>>>>> + if (unlikely(f2fs_cp_error(sbi))) {
>>>>>> + if (is_sbi_flag_set(sbi, SBI_IS_CLOSE)) {
>>>>>> + ClearPageUptodate(page);
>>>>>> + dec_page_count(sbi, F2FS_DIRTY_NODES);
>>>>>> + unlock_page(page);
>>>>>> + return 0;
>>>>>> + }
>>>>>> goto redirty_out;
>>>>>> + }
>>>>>>
>>>>>> if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
>>>>>> goto redirty_out;
>>>>>>
>>>> .
>>>>
>>>
>>>
>>> _______________________________________________
>>> Linux-f2fs-devel mailing list
>>> Linux-f2fs-devel@lists.sourceforge.net
>>> https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
>>> .
>>>
> .
>
_______________________________________________
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:[~2020-05-27 2:35 UTC|newest]
Thread overview: 12+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-05-22 14:47 [f2fs-dev] [PATCH] f2fs: avoid inifinite loop to wait for flushing node pages at cp_error Jaegeuk Kim
2020-05-22 23:32 ` Jaegeuk Kim
2020-05-25 3:56 ` [f2fs-dev] [PATCH v3] " Jaegeuk Kim
2020-05-25 6:30 ` Chao Yu
2020-05-25 15:06 ` Jaegeuk Kim
2020-05-26 1:11 ` Chao Yu
2020-05-26 1:34 ` Chao Yu
2020-05-26 1:56 ` Jaegeuk Kim
2020-05-27 2:35 ` Chao Yu [this message]
2020-05-27 20:56 ` Jaegeuk Kim
2020-05-28 1:20 ` Chao Yu
2020-05-25 2:16 ` [f2fs-dev] [PATCH] " 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=765a1ac5-a318-14d6-666f-eab46f892d01@huawei.com \
--to=yuchao0@huawei.com \
--cc=jaegeuk@kernel.org \
--cc=kernel-team@android.com \
--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 NNTP newsgroup(s).