Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
References: <20190531013350.GA4642@zhanggen-UX430UQ>
In-Reply-To: <20190531013350.GA4642@zhanggen-UX430UQ>
From:   Paul Moore <paul@paul-moore.com>
Date:   Fri, 31 May 2019 11:45:28 -0400
Message-ID: <CAHC9VhTmj8b9jYMaXd=ORhBgTAWUgF=srgqAXkECe7MFkDXOmg@mail.gmail.com>
Subject: Re: [PATCH v2] hooks: fix a missing-check bug in selinux_sb_eat_lsm_opts()
To:     Gen Zhang <blackgod016574@gmail.com>
Cc:     Stephen Smalley <sds@tycho.nsa.gov>,
        Eric Paris <eparis@parisplace.org>, selinux@vger.kernel.org,
        linux-kernel@vger.kernel.org, netdev@vger.kernel.org,
        bpf@vger.kernel.org, omosnace@redhat.com
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Thu, May 30, 2019 at 9:34 PM Gen Zhang <blackgod016574@gmail.com> wrote:
>
> In selinux_sb_eat_lsm_opts(), 'arg' is allocated by kmemdup_nul(). It
> returns NULL when fails. So 'arg' should be checked.
>
> Signed-off-by: Gen Zhang <blackgod016574@gmail.com>
> Reviewed-by: Ondrej Mosnacek <omosnace@redhat.com>
> Fixes: 99dbbb593fe6 ("selinux: rewrite selinux_sb_eat_lsm_opts()")

One quick note about the subject line, instead of using "hooks:" you
should use "selinux:" since this is specific to SELinux.  If the patch
did apply to the LSM framework as a whole, I would suggest using
"lsm:" instead of "hooks:" as "hooks" is too ambiguous of a prefix.

> diff --git a/security/selinux/hooks.c b/security/selinux/hooks.c
> index 3ec702c..5a9e959 100644
> --- a/security/selinux/hooks.c
> +++ b/security/selinux/hooks.c
> @@ -2635,6 +2635,8 @@ static int selinux_sb_eat_lsm_opts(char *options, void **mnt_opts)
>                                                 *q++ = c;
>                                 }
>                                 arg = kmemdup_nul(arg, q - arg, GFP_KERNEL);
> +                               if (!arg)
> +                                       return -ENOMEM;

It seems like we should also check for, and potentially free *mnt_opts
as the selinux_add_opt() error handling does just below this change,
yes?  If that is the case we might want to move that error handling
code to the bottom of the function and jump there on error.

>                         }
>                         rc = selinux_add_opt(token, arg, mnt_opts);
>                         if (unlikely(rc)) {

-- 
paul moore
www.paul-moore.com