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:   ebiederm@xmission.com (Eric W. Biederman)
To:     Bernd Edlinger <bernd.edlinger@hotmail.de>
Cc:     Alexander Viro <viro@zeniv.linux.org.uk>,
        Kees Cook <keescook@chromium.org>,
        "linux-fsdevel\@vger.kernel.org" <linux-fsdevel@vger.kernel.org>,
        "linux-kernel\@vger.kernel.org" <linux-kernel@vger.kernel.org>
References: <AM8PR10MB47081071E64EAAB343196D5AE4399@AM8PR10MB4708.EURPRD10.PROD.OUTLOOK.COM>
Date:   Sun, 06 Jun 2021 14:34:53 -0500
In-Reply-To: <AM8PR10MB47081071E64EAAB343196D5AE4399@AM8PR10MB4708.EURPRD10.PROD.OUTLOOK.COM>
        (Bernd Edlinger's message of "Sun, 6 Jun 2021 12:41:18 +0200")
Message-ID: <87mts2kcrm.fsf@disp2133>
User-Agent: Gnus/5.13 (Gnus v5.13) Emacs/26.1 (gnu/linux)
MIME-Version: 1.0
Content-Type: text/plain
Subject: Re: [PATCH] Fix error handling in begin_new_exec
Precedence: bulk

Bernd Edlinger <bernd.edlinger@hotmail.de> writes:

> If get_unused_fd_flags() fails, the error handling is incomplete
> because bprm->cred is already set to NULL, and therefore
> free_bprm will not unlock the cred_guard_mutex.
> Note there are two error conditions which end up here,
> one before and one after bprm->cred is cleared.

Acked-by: "Eric W. Biederman" <ebiederm@xmission.com>

Yuck.  I wonder if there is a less error prone idiom we could be using
here than testing bprm->cred in free_bprm.  Especially as this lock is
expected to stay held through setup_new_exec.

Something feels too clever here.

> Fixes: b8a61c9e7b4 ("exec: Generic execfd support")
>
> Signed-off-by: Bernd Edlinger <bernd.edlinger@hotmail.de>
> ---
>  fs/exec.c | 3 +++
>  1 file changed, 3 insertions(+)
>
> diff --git a/fs/exec.c b/fs/exec.c
> index 18594f1..d8af85f 100644
> --- a/fs/exec.c
> +++ b/fs/exec.c
> @@ -1396,6 +1396,9 @@ int begin_new_exec(struct linux_binprm * bprm)
>  
>  out_unlock:
>  	up_write(&me->signal->exec_update_lock);
> +	if (!bprm->cred)
> +		mutex_unlock(&me->signal->cred_guard_mutex);
> +
>  out:
>  	return retval;
>  }