Message-ID: <1495808645.12091.10.camel@tycho.nsa.gov>
Subject: Re: [PATCH RFC 2/2] nfs: update labeling behavior on a superblock
 when submounting
From: Stephen Smalley <sds@tycho.nsa.gov>
To: Scott Mayhew <smayhew@redhat.com>, selinux@tycho.nsa.gov,
        linux-nfs@vger.kernel.org
Cc: Paul Moore <paul@paul-moore.com>,
        Eric Paris <eparis@parisplace.org>,
        Trond Myklebust <trondmy@primarydata.com>,
        "J . Bruce Fields" <bfields@fieldses.org>
Date: Fri, 26 May 2017 10:24:05 -0400
In-Reply-To: <20170525210754.24265-3-smayhew@redhat.com>
References: <20170404232646.GB24146@parsley.fieldses.org>
         <20170525210754.24265-1-smayhew@redhat.com>
         <20170525210754.24265-3-smayhew@redhat.com>
Content-Type: text/plain; charset="UTF-8"
Mime-Version: 1.0
Sender: linux-nfs-owner@vger.kernel.org

On Thu, 2017-05-25 at 17:07 -0400, Scott Mayhew wrote:
> When the client traverses from filesystem exported without the
> "security_label" option to one exported with the "security_label"
> option, it needs to pass SECURITY_LSM_NATIVE_LABELS to
> security_sb_set_mnt_opts() so that the new superblock has SBLABEL_MNT
> set in its security mount options.  Otherwise, attempts to set
> security
> labels via setxattr over NFSv4.2 will fail.
> 
> Signed-off-by: Scott Mayhew <smayhew@redhat.com>
> ---
>  fs/nfs/super.c | 23 ++++++++++++++++++++++-
>  1 file changed, 22 insertions(+), 1 deletion(-)
> 
> diff --git a/fs/nfs/super.c b/fs/nfs/super.c
> index 2f3822a..d7a3b89 100644
> --- a/fs/nfs/super.c
> +++ b/fs/nfs/super.c
> @@ -2544,10 +2544,31 @@ EXPORT_SYMBOL_GPL(nfs_set_sb_security);
>  int nfs_clone_sb_security(struct super_block *s, struct dentry
> *mntroot,
>  			  struct nfs_mount_info *mount_info)
>  {
> +	int error;
> +	unsigned long kflags = 0, kflags_out = 0;
> +	struct security_mnt_opts opts;
> +
>  	/* clone any lsm security options from the parent to the new
> sb */
>  	if (d_inode(mntroot)->i_op != NFS_SB(s)->nfs_client-
> >rpc_ops->dir_inode_ops)
>  		return -ESTALE;
> -	return security_sb_clone_mnt_opts(mount_info->cloned->sb,
> s);
> +	error = security_sb_clone_mnt_opts(mount_info->cloned->sb,
> s);
> +	if (error)
> +		goto err;
> +
> +	if (NFS_SB(s)->caps & NFS_CAP_SECURITY_LABEL &&
> +		!(NFS_SB(mount_info->cloned->sb)->caps &
> NFS_CAP_SECURITY_LABEL)) {
> +		memset(&opts, 0, sizeof(opts));
> +		kflags |= SECURITY_LSM_NATIVE_LABELS;
> +
> +		error = security_sb_set_mnt_opts(s, &opts, kflags,
> &kflags_out);
> +		if (error)
> +			goto err;
> +
> +		if (!(kflags_out & SECURITY_LSM_NATIVE_LABELS))
> +			NFS_SB(s)->caps &= ~NFS_CAP_SECURITY_LABEL;
> +	}
> +err:
> +	return error;
>  }
>  EXPORT_SYMBOL_GPL(nfs_clone_sb_security);

Could this clobber a context set via context= mount option?