Re: [PATCH] f2fs: prevent meta updates while checkpoint is in progress

From: Chao Yu
Date: Sat Mar 28 2020 - 04:38:19 EST


Hi all,

On 2020/3/28 3:24, Jaegeuk Kim wrote:
> Hi Sahitya,
>
> On 03/26, Sahitya Tummala wrote:
>> allocate_segment_for_resize() can cause metapage updates if
>> it requires to change the current node/data segments for resizing.
>> Stop these meta updates when there is a checkpoint already
>> in progress to prevent inconsistent CP data.
>
> Doesn't freeze|thaw_bdev(sbi->sb->s_bdev); work for you?

That can avoid foreground ops racing? rather than background ops like
balance_fs() from kworker?

BTW, I found that {freeze,thaw}_bdev is not enough to freeze all
foreground fs ops, it needs to use {freeze,thaw}_super instead.

---
fs/f2fs/gc.c | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/fs/f2fs/gc.c b/fs/f2fs/gc.c
index 26248c8936db..acdc8b99b543 100644
--- a/fs/f2fs/gc.c
+++ b/fs/f2fs/gc.c
@@ -1538,7 +1538,7 @@ int f2fs_resize_fs(struct f2fs_sb_info *sbi, __u64 block_count)
return -EINVAL;
}

- freeze_bdev(sbi->sb->s_bdev);
+ freeze_super(sbi->sb);

shrunk_blocks = old_block_count - block_count;
secs = div_u64(shrunk_blocks, BLKS_PER_SEC(sbi));
@@ -1551,7 +1551,7 @@ int f2fs_resize_fs(struct f2fs_sb_info *sbi, __u64 block_count)
sbi->user_block_count -= shrunk_blocks;
spin_unlock(&sbi->stat_lock);
if (err) {
- thaw_bdev(sbi->sb->s_bdev, sbi->sb);
+ thaw_super(sbi->sb);
return err;
}

@@ -1613,6 +1613,6 @@ int f2fs_resize_fs(struct f2fs_sb_info *sbi, __u64 block_count)
}
clear_sbi_flag(sbi, SBI_IS_RESIZEFS);
mutex_unlock(&sbi->resize_mutex);
- thaw_bdev(sbi->sb->s_bdev, sbi->sb);
+ thaw_super(sbi->sb);
return err;
}
--
2.18.0.rc1

>
>>
>> Signed-off-by: Sahitya Tummala <stummala@xxxxxxxxxxxxxx>
>> ---
>> fs/f2fs/gc.c | 2 ++
>> 1 file changed, 2 insertions(+)
>>
>> diff --git a/fs/f2fs/gc.c b/fs/f2fs/gc.c
>> index 5bca560..6122bad 100644
>> --- a/fs/f2fs/gc.c
>> +++ b/fs/f2fs/gc.c
>> @@ -1399,8 +1399,10 @@ static int free_segment_range(struct f2fs_sb_info *sbi, unsigned int start,
>> int err = 0;
>>
>> /* Move out cursegs from the target range */
>> + f2fs_lock_op(sbi);
>> for (type = CURSEG_HOT_DATA; type < NR_CURSEG_TYPE; type++)
>> allocate_segment_for_resize(sbi, type, start, end);
>> + f2fs_unlock_op(sbi);
>>
>> /* do GC to move out valid blocks in the range */
>> for (segno = start; segno <= end; segno += sbi->segs_per_sec) {
>> --
>> Qualcomm India Private Limited, on behalf of Qualcomm Innovation Center, Inc.
>> Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project.
> .
>