Message-ID: <bdefbbd3-c652-2d6f-bd38-b7de6b1b9155@semihalf.com>
Date:   Fri, 24 Feb 2023 21:06:37 +0100
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Thunderbird/102.7.1
Subject: Re: [PATCH 1/1] ext4: Don't show commit interval if it is zero
Content-Language: en-US
To:     Wang Jianjian <wangjianjian3@huawei.com>,
        linux-ext4@vger.kernel.org
Cc:     zhangzhikang1@huawei.com, wangqiang62@huawei.com,
        zhengbowen7@huawei.com, Theodore Ts'o <tytso@mit.edu>,
        Lukas Czerner <lczerner@redhat.com>,
        Sarthak Kukreti <sarthakkukreti@google.com>,
        Dmitry Torokhov <dtor@google.com>,
        Tomasz Nowicki <tn@semihalf.com>,
        Grzegorz Jaszczyk <jaz@semihalf.com>
References: <20221219015128.876717-1-wangjianjian3@huawei.com>
From:   Dmytro Maluka <dmy@semihalf.com>
In-Reply-To: <20221219015128.876717-1-wangjianjian3@huawei.com>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 12/19/22 02:51, Wang Jianjian wrote:
> If commit interval is 0, it means using default value.
> 
> Fixes: 6e47a3cc68fc ("ext4: get rid of super block and sbi from handle_mount_ops()")
> Signed-off-by: Wang Jianjian <wangjianjian3@huawei.com>
> ---
>  fs/ext4/super.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/fs/ext4/super.c b/fs/ext4/super.c
> index 16a343e8047d..b93911d80cd9 100644
> --- a/fs/ext4/super.c
> +++ b/fs/ext4/super.c
> @@ -2146,7 +2146,7 @@ static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param)
>  		return 0;
>  	case Opt_commit:
>  		if (result.uint_32 == 0)
> -			ctx->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE;
> +			result.uint_32 = JBD2_DEFAULT_MAX_COMMIT_AGE;
>  		else if (result.uint_32 > INT_MAX / HZ) {
>  			ext4_msg(NULL, KERN_ERR,
>  				 "Invalid commit interval %d, "

Hi,

Thanks for this fix, I just have some remarks about its description in
the commit log.

I think the bug fixed by this patch is more severe than the description
suggests: it's not just about showing or not showing commit interval
(btw I'm not sure what does "show" mean here), it's about setting
incorrect commit interval (0 instead of JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
when mounting ext4 with default settings.

Also I believe the "Fixes:" tag is misleading: it actually fixes commit
461c3af045d3 ("ext4: Change handle_mount_opt() to use fs_parameter").

This bug introduced in 461c3af045d3 causes a significant performance
degradation at least in certain setups, for instance I'm observing ~10x
performance drop in direct I/O write tests with PSEDN128GA17DC0 SSD disk
and PS5013 E13 NVMe controller.

Ted, I see you've already merged this fix into ext4 dev branch, but I
guess it's not too late to amend the commit message before it gets into
mainline?

Thanks,
Dmytro