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:   Thu, 3 Aug 2023 16:38:25 +0200
From:   Jan Kara <jack@suse.cz>
To:     Zhang Yi <yi.zhang@huaweicloud.com>
Cc:     linux-ext4@vger.kernel.org, tytso@mit.edu,
        adilger.kernel@dilger.ca, jack@suse.cz, yi.zhang@huawei.com,
        chengzhihao1@huawei.com, yukuai3@huawei.com
Subject: Re: [PATCH 07/12] jbd2: add fast_commit space check
Message-ID: <20230803143825.f364hmpsgqbzvjwo@quack3>
References: <20230704134233.110812-1-yi.zhang@huaweicloud.com>
 <20230704134233.110812-8-yi.zhang@huaweicloud.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20230704134233.110812-8-yi.zhang@huaweicloud.com>
Precedence: bulk

On Tue 04-07-23 21:42:28, Zhang Yi wrote:
> From: Zhang Yi <yi.zhang@huawei.com>
> 
> If JBD2_FEATURE_INCOMPAT_FAST_COMMIT bit is set, it means the journal
> have fast commit records need to recover, so the fast commit size
> should not be zero, and also the leftover normal journal size should
> never less than JBD2_MIN_JOURNAL_BLOCKS. Add a check into the
> journal_check_superblock() and drop the pointless branch when
> initializing in-memory fastcommit parameters.
> 
> Signed-off-by: Zhang Yi <yi.zhang@huawei.com>

Some comments below.


> diff --git a/fs/jbd2/journal.c b/fs/jbd2/journal.c
> index efdb8db3c06e..210b532a3673 100644
> --- a/fs/jbd2/journal.c
> +++ b/fs/jbd2/journal.c
> @@ -1392,6 +1392,18 @@ static int journal_check_superblock(journal_t *journal)
>  		return err;
>  	}
>  
> +	if (jbd2_has_feature_fast_commit(journal)) {
> +		int num_fc_blks = be32_to_cpu(sb->s_num_fc_blks);
> +
> +		if (!num_fc_blks ||
> +		    (be32_to_cpu(sb->s_maxlen) - num_fc_blks <
> +		     JBD2_MIN_JOURNAL_BLOCKS)) {
> +			printk(KERN_ERR "JBD2: Invalid fast commit size %d\n",
> +			       num_fc_blks);
> +			return err;
> +		}

This is wrong sb->s_num_fc_blks == 0 means that the fast-commit area should
have the default size of 256 blocks. At least that's how it behaves
currently and we should not change the behavior.

Similarly if the number of fastcommit blocks was too big (i.e. there was
not enough space left for ordinary journal), we effectively silently
disable fastcommit and you break this behavior in this patch.

								Honza

> +	}
> +
>  	if (jbd2_has_feature_csum2(journal) &&
>  	    jbd2_has_feature_csum3(journal)) {
>  		/* Can't have checksum v2 and v3 at the same time! */
> @@ -1460,7 +1472,6 @@ static int journal_load_superblock(journal_t *journal)
>  	int err;
>  	struct buffer_head *bh;
>  	journal_superblock_t *sb;
> -	int num_fc_blocks;
>  
>  	bh = getblk_unmovable(journal->j_dev, journal->j_blk_offset,
>  			      journal->j_blocksize);
> @@ -1498,9 +1509,8 @@ static int journal_load_superblock(journal_t *journal)
>  
>  	if (jbd2_has_feature_fast_commit(journal)) {
>  		journal->j_fc_last = be32_to_cpu(sb->s_maxlen);
> -		num_fc_blocks = jbd2_journal_get_num_fc_blks(sb);
> -		if (journal->j_last - num_fc_blocks >= JBD2_MIN_JOURNAL_BLOCKS)
> -			journal->j_last = journal->j_fc_last - num_fc_blocks;
> +		journal->j_last = journal->j_fc_last -
> +				  be32_to_cpu(sb->s_num_fc_blks);
>  		journal->j_fc_first = journal->j_last + 1;
>  		journal->j_fc_off = 0;
>  	}
> -- 
> 2.39.2
> 
-- 
Jan Kara <jack@suse.com>
SUSE Labs, CR