Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
From:   Jeff Moyer <jmoyer@redhat.com>
To:     Yang Li <yang.lee@linux.alibaba.com>
Cc:     viro@zeniv.linux.org.uk, linux-fsdevel@vger.kernel.org,
        linux-kernel@vger.kernel.org
Subject: Re: [PATCH] direct-io: Using kmem_cache_zalloc() instead of kmem_cache_alloc() and memset()
References: <1614243581-50870-1-git-send-email-yang.lee@linux.alibaba.com>
Date:   Thu, 25 Feb 2021 10:46:15 -0500
In-Reply-To: <1614243581-50870-1-git-send-email-yang.lee@linux.alibaba.com>
        (Yang Li's message of "Thu, 25 Feb 2021 16:59:41 +0800")
Message-ID: <x4935xknoag.fsf@segfault.boston.devel.redhat.com>
User-Agent: Gnus/5.13 (Gnus v5.13) Emacs/26.1 (gnu/linux)
MIME-Version: 1.0
Content-Type: text/plain
Precedence: bulk

Y, Yang,

Yang Li <yang.lee@linux.alibaba.com> writes:

> Fix the following coccicheck warning:
> ./fs/direct-io.c:1155:7-23: WARNING: kmem_cache_zalloc should be used
> for dio, instead of kmem_cache_alloc/memset
>
> Reported-by: Abaci Robot <abaci@linux.alibaba.com>
> Signed-off-by: Yang Li <yang.lee@linux.alibaba.com>
> ---
>  fs/direct-io.c | 4 +---
>  1 file changed, 1 insertion(+), 3 deletions(-)
>
> diff --git a/fs/direct-io.c b/fs/direct-io.c
> index 0957e1b..6ec2935 100644
> --- a/fs/direct-io.c
> +++ b/fs/direct-io.c
> @@ -1152,7 +1152,7 @@ static inline int drop_refcount(struct dio *dio)
>  	if (iov_iter_rw(iter) == READ && !count)
>  		return 0;
>  
> -	dio = kmem_cache_alloc(dio_cache, GFP_KERNEL);
> +	dio = kmem_cache_zalloc(dio_cache, GFP_KERNEL);
>  	if (!dio)
>  		return -ENOMEM;
>  	/*
> @@ -1160,8 +1160,6 @@ static inline int drop_refcount(struct dio *dio)
>  	 * performance regression in a database benchmark.  So, we take
>  	 * care to only zero out what's needed.
>  	 */
> -	memset(dio, 0, offsetof(struct dio, pages));
> -

You must have missed the comment just above this memset:

        /*
         * Believe it or not, zeroing out the page array caused a .5%
         * performance regression in a database benchmark.  So, we take
         * care to only zero out what's needed.
         */

That's referring to this part of the dio struct:

        /*
         * pages[] (and any fields placed after it) are not zeroed out at
         * allocation time.  Don't add new fields after pages[] unless you
         * wish that they not be zeroed.
         */
        union {
                struct page *pages[DIO_PAGES];  /* page buffer */
                struct work_struct complete_work;/* deferred AIO completion */
        };
} ____cacheline_aligned_in_smp;

Nacked-by: Jeff Moyer <jmoyer@redhat.com>