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;
Date:   Sun, 20 Sep 2020 11:49:20 -0700 (PDT)
From:   Hugh Dickins <hughd@google.com>
To:     Tianjia Zhang <tianjia.zhang@linux.alibaba.com>
cc:     Hugh Dickins <hughd@google.com>,
        Andrew Morton <akpm@linux-foundation.org>,
        Al Viro <viro@zeniv.linux.org.uk>, linux-mm@kvack.org,
        linux-kernel@vger.kernel.org
Subject: Re: [PATCH] mm/shmem.c: Fix the missing unaccount on the failed
 path
In-Reply-To: <20200920163005.97079-1-tianjia.zhang@linux.alibaba.com>
Message-ID: <alpine.LSU.2.11.2009201141510.4318@eggly.anvils>
References: <20200920163005.97079-1-tianjia.zhang@linux.alibaba.com>
User-Agent: Alpine 2.11 (LSU 23 2013-08-11)
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Precedence: bulk

On Mon, 21 Sep 2020, Tianjia Zhang wrote:

> In function __shmem_file_setup(), shmem_unacct_size() is forgotten
> on the failed path, so add it.
> 
> Fixes: 93dec2da7b234 ("... and switch shmem_file_setup() to alloc_file_pseudo()")
> Cc: Al Viro <viro@zeniv.linux.org.uk>
> Signed-off-by: Tianjia Zhang <tianjia.zhang@linux.alibaba.com>
> ---
>  mm/shmem.c | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/mm/shmem.c b/mm/shmem.c
> index 8e2b35ba93ad..591410dc3541 100644
> --- a/mm/shmem.c
> +++ b/mm/shmem.c
> @@ -4200,8 +4200,10 @@ static struct file *__shmem_file_setup(struct vfsmount *mnt, const char *name, l
>  	if (!IS_ERR(res))
>  		res = alloc_file_pseudo(inode, mnt, name, O_RDWR,
>  				&shmem_file_operations);
> -	if (IS_ERR(res))
> +	if (IS_ERR(res)) {
>  		iput(inode);
> +		shmem_unacct_size(flags, size);
> +	}
>  	return res;
>  }
>  
> -- 
> 2.19.1.3.ge56e4f7

Looks mistaken to me.

Is this something you noticed by source inspection,
or something you have observed in practice?

I haven't tried exercising this path while injecting errors into
alloc_file_pseudo(); but what I'd expect to happen is that the
iput(inode), which you see already on that error path, will get
to evict the inode, which will entail calling shmem_evict_inode(),
which does that shmem_unacct_size() itself.

Hugh