Received-SPF: pass (google.com: domain of linux-ext4-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Wed, 2 Feb 2022 02:03:59 +0530
From:   Ritesh Harjani <riteshh@linux.ibm.com>
To:     Xin Yin <yinxin.x@bytedance.com>
Cc:     harshadshirwadkar@gmail.com, tytso@mit.edu,
        adilger.kernel@dilger.ca, linux-ext4@vger.kernel.org,
        linux-kernel@vger.kernel.org
Subject: Re: [PATCH 1/2] ext4: use ext4_ext_remove_space() for fast commit
 replay delete range
Message-ID: <20220201203359.owrnrfqydjloy7oq@riteshh-domain>
References: <20211223032337.5198-1-yinxin.x@bytedance.com>
 <20211223032337.5198-2-yinxin.x@bytedance.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20211223032337.5198-2-yinxin.x@bytedance.com>
Precedence: bulk

Hello Xin,

Sorry about revisiting this thread so late :(
Recently when I was working on one of the fast_commit issue, I got interested
in looking into some of those recent fast_commit fixes.

Hence some of these queries.

On 21/12/23 11:23AM, Xin Yin wrote:
> For now ,we use ext4_punch_hole() during fast commit replay delete range
> procedure. But it will be affected by inode->i_size, which may not
> correct during fast commit replay procedure. The following test will
> failed.
>
> -create & write foo (len 1000K)
> -falloc FALLOC_FL_ZERO_RANGE foo (range 400K - 600K)
> -create & fsync bar
^^^^ do you mean "fsync foo" or is this actually a new file create and fsync
bar?


> -falloc FALLOC_FL_PUNCH_HOLE foo (range 300K-500K)
> -fsync foo
> -crash before a full commit
>
> After the fast_commit reply procedure, the range 400K-500K will not be
> removed. Because in this case, when calling ext4_punch_hole() the
> inode->i_size is 0, and it just retruns with doing nothing.

I tried looking into this, but I am not able to put my head around that when
will the inode->i_size will be 0?

So, what I think should happen is when you are doing falocate/fsync foo in your
above list of operations then, anyways the inode i_disksize will be updated
using ext4_mark_inode_dirty() and during replay phase inode->i_size will hold
the right value no?

Could you please help understand when, where and how will inode->i_size will be
0?

Also - it would be helpful if you have some easy reproducer of this issue you
mentioned.

-ritesh

>
> Change to use ext4_ext_remove_space() instead of ext4_punch_hole()
> to remove blocks of inode directly.
>
> Signed-off-by: Xin Yin <yinxin.x@bytedance.com>
> ---
>  fs/ext4/fast_commit.c | 13 ++++++++-----
>  1 file changed, 8 insertions(+), 5 deletions(-)
>
> diff --git a/fs/ext4/fast_commit.c b/fs/ext4/fast_commit.c
> index aa05b23f9c14..3deb97b22ca4 100644
> --- a/fs/ext4/fast_commit.c
> +++ b/fs/ext4/fast_commit.c
> @@ -1708,11 +1708,14 @@ ext4_fc_replay_del_range(struct super_block *sb, struct ext4_fc_tl *tl,
>  		}
>  	}
>
> -	ret = ext4_punch_hole(inode,
> -		le32_to_cpu(lrange.fc_lblk) << sb->s_blocksize_bits,
> -		le32_to_cpu(lrange.fc_len) <<  sb->s_blocksize_bits);
> -	if (ret)
> -		jbd_debug(1, "ext4_punch_hole returned %d", ret);
> +	down_write(&EXT4_I(inode)->i_data_sem);
> +	ret = ext4_ext_remove_space(inode, lrange.fc_lblk,
> +				lrange.fc_lblk + lrange.fc_len - 1);
> +	up_write(&EXT4_I(inode)->i_data_sem);
> +	if (ret) {
> +		iput(inode);
> +		return 0;
> +	}
>  	ext4_ext_replay_shrink_inode(inode,
>  		i_size_read(inode) >> sb->s_blocksize_bits);
>  	ext4_mark_inode_dirty(NULL, inode);
> --
> 2.20.1
>