Re: [PATCH v3 2/2] proc: add Kconfigs to restrict /proc/pid/mem access

From: Kees Cook <keescook@chromium.org>
To: Adrian Ratiu <adrian.ratiu@collabora.com>
Cc: linux-fsdevel@vger.kernel.org,
	linux-security-module@vger.kernel.org,
	linux-kernel@vger.kernel.org, linux-hardening@vger.kernel.org,
	linux-doc@vger.kernel.org, kernel@collabora.com, gbiv@google.com,
	ryanbeltran@google.com, inglorion@google.com,
	ajordanr@google.com, jorgelo@chromium.org,
	Guenter Roeck <groeck@chromium.org>,
	Doug Anderson <dianders@chromium.org>,
	Jann Horn <jannh@google.com>,
	Andrew Morton <akpm@linux-foundation.org>,
	Randy Dunlap <rdunlap@infradead.org>,
	Christian Brauner <brauner@kernel.org>
Subject: Re: [PATCH v3 2/2] proc: add Kconfigs to restrict /proc/pid/mem access
Date: Fri, 26 Apr 2024 16:16:34 -0700	[thread overview]
Message-ID: <202404261611.958FEB4@keescook> (raw)
In-Reply-To: <20240409175750.206445-2-adrian.ratiu@collabora.com>

On Tue, Apr 09, 2024 at 08:57:50PM +0300, Adrian Ratiu wrote:
> Some systems might have difficulty changing their bootloaders
> to enable the newly added restrict_proc_mem* params, for e.g.
> remote embedded doing OTA updates, so this provides a set of
> Kconfigs to set /proc/pid/mem restrictions at build-time.
> 
> The boot params take precedence over the Kconfig values. This
> can be reversed, but doing it this way I think makes sense.
> 
> Another idea is to have a global bool Kconfig which can enable
> or disable this mechanism in its entirety, however it does not
> seem necessary since all three knobs default to off, the branch
> logic overhead is rather minimal and I assume most of systems
> will want to restrict at least the use of FOLL_FORCE.
> 
> Cc: Guenter Roeck <groeck@chromium.org>
> Cc: Doug Anderson <dianders@chromium.org>
> Cc: Kees Cook <keescook@chromium.org>
> Cc: Jann Horn <jannh@google.com>
> Cc: Andrew Morton <akpm@linux-foundation.org>
> Cc: Randy Dunlap <rdunlap@infradead.org>
> Cc: Christian Brauner <brauner@kernel.org>
> Signed-off-by: Adrian Ratiu <adrian.ratiu@collabora.com>
> ---
>  fs/proc/base.c   | 33 +++++++++++++++++++++++++++++++++
>  security/Kconfig | 42 ++++++++++++++++++++++++++++++++++++++++++
>  2 files changed, 75 insertions(+)
> 
> diff --git a/fs/proc/base.c b/fs/proc/base.c
> index c733836c42a65..e8ee848fc4a98 100644
> --- a/fs/proc/base.c
> +++ b/fs/proc/base.c
> @@ -889,6 +889,17 @@ static int __mem_open_check_access_restriction(struct file *file)
>  		    !__mem_open_current_is_ptracer(file))
>  			return -EACCES;
>  
> +#ifdef CONFIG_SECURITY_PROC_MEM_WRITE_RESTRICT

No, please. :)

Just use use the _MAYBE/_maybe variants of the static branch DECLAREs and
branches, and make Kconfigs for:

CONFIG_PROC_MEM_RESTRICT_READ_DEFAULT
CONFIG_PROC_MEM_RESTRICT_WRITE_DEFAULT
CONFIG_PROC_MEM_RESTRICT_FOLL_FORCE_DEFAULT

Like:

DECLARE_STATIC_KEY_MAYBE(CONFIG_PROC_MEM_RESTRICT_READ_DEFAULT, proc_mem_restrict_read);

and then later:

	if (static_branch_maybe(CONFIG_PROC_MEM_RESTRICT_READ_DEFAULT,
				&proc_mem_restrict_read))
		...

Then all builds of the kernel will have it available, but system
builders who want it enabled by default will get a slightly more
optimized "if".

-Kees

-- 
Kees Cook