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:   Wed, 16 Sep 2020 07:25:53 +0100
From:   Christoph Hellwig <hch@infradead.org>
To:     Rich Felker <dalias@libc.org>
Cc:     linux-api@vger.kernel.org,
        Alexander Viro <viro@zeniv.linux.org.uk>,
        Christoph Hellwig <hch@infradead.org>,
        linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH v2 1/2] vfs: block chmod of symlinks
Message-ID: <20200916062553.GB27867@infradead.org>
References: <20200916002157.GO3265@brightrain.aerifal.cx>
 <20200916002253.GP3265@brightrain.aerifal.cx>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20200916002253.GP3265@brightrain.aerifal.cx>
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Tue, Sep 15, 2020 at 08:22:54PM -0400, Rich Felker wrote:
> It was discovered while implementing userspace emulation of fchmodat
> AT_SYMLINK_NOFOLLOW (using O_PATH and procfs magic symlinks; otherwise
> it's not possible to target symlinks with chmod operations) that some
> filesystems erroneously allow access mode of symlinks to be changed,
> but return failure with EOPNOTSUPP (see glibc issue #14578 and commit
> a492b1e5ef). This inconsistency is non-conforming and wrong, and the
> consensus seems to be that it was unintentional to allow link modes to
> be changed in the first place.
> 
> Signed-off-by: Rich Felker <dalias@libc.org>
> ---
>  fs/open.c | 6 ++++++
>  1 file changed, 6 insertions(+)
> 
> diff --git a/fs/open.c b/fs/open.c
> index 9af548fb841b..cdb7964aaa6e 100644
> --- a/fs/open.c
> +++ b/fs/open.c
> @@ -570,6 +570,12 @@ int chmod_common(const struct path *path, umode_t mode)
>  	struct iattr newattrs;
>  	int error;
>  
> +	/* Block chmod from getting to fs layer. Ideally the fs would either
> +	 * allow it or fail with EOPNOTSUPP, but some are buggy and return
> +	 * an error but change the mode, which is non-conforming and wrong. */
> +	if (S_ISLNK(inode->i_mode))
> +		return -EOPNOTSUPP;

Our usualy place for this would be setattr_prepare.  Also the comment
style is off, and I don't think we should talk about buggy file systems
here, but a policy to not allow the chmod.  I also suspect the right
error value is EINVAL - EOPNOTSUPP isn't really used in normal posix
file system interfaces.