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;
Subject: Re: [PATCH 0/4] Fix misused kernel_read_file() enums
To:     Luis Chamberlain <mcgrof@kernel.org>
Cc:     Kees Cook <keescook@chromium.org>,
        James Morris <jmorris@namei.org>,
        Mimi Zohar <zohar@linux.ibm.com>,
        Scott Branden <scott.branden@broadcom.com>,
        Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
        "Rafael J. Wysocki" <rafael@kernel.org>,
        Alexander Viro <viro@zeniv.linux.org.uk>,
        Jessica Yu <jeyu@kernel.org>,
        Dmitry Kasatkin <dmitry.kasatkin@gmail.com>,
        "Serge E. Hallyn" <serge@hallyn.com>,
        Casey Schaufler <casey@schaufler-ca.com>,
        "Eric W. Biederman" <ebiederm@xmission.com>,
        Peter Zijlstra <peterz@infradead.org>,
        Matthew Garrett <matthewgarrett@google.com>,
        David Howells <dhowells@redhat.com>,
        Mauro Carvalho Chehab <mchehab+huawei@kernel.org>,
        Randy Dunlap <rdunlap@infradead.org>,
        "Joel Fernandes (Google)" <joel@joelfernandes.org>,
        KP Singh <kpsingh@google.com>, Dave Olsthoorn <dave@bewaar.me>,
        Peter Jones <pjones@redhat.com>,
        Andrew Morton <akpm@linux-foundation.org>,
        Stephen Boyd <stephen.boyd@linaro.org>,
        Paul Moore <paul@paul-moore.com>, linux-kernel@vger.kernel.org,
        linux-fsdevel@vger.kernel.org, linux-integrity@vger.kernel.org,
        linux-security-module@vger.kernel.org
References: <20200707081926.3688096-1-keescook@chromium.org>
 <3c01073b-c422-dd97-0677-c16fe1158907@redhat.com>
 <f5e65f73-2c94-3614-2479-69b2bfda9775@redhat.com>
 <20200708115517.GF4332@42.do-not-panic.com>
From:   Hans de Goede <hdegoede@redhat.com>
Message-ID: <8766279d-0ebe-1f64-c590-4a71a733609b@redhat.com>
Date:   Wed, 8 Jul 2020 13:58:47 +0200
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:68.0) Gecko/20100101
 Thunderbird/68.9.0
MIME-Version: 1.0
In-Reply-To: <20200708115517.GF4332@42.do-not-panic.com>
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

Hi,

On 7/8/20 1:55 PM, Luis Chamberlain wrote:
> On Wed, Jul 08, 2020 at 01:37:41PM +0200, Hans de Goede wrote:
>> Hi,
>>
>> On 7/8/20 1:01 PM, Hans de Goede wrote:
>>> Hi,
>>>
>>> On 7/7/20 10:19 AM, Kees Cook wrote:
>>>> Hi,
>>>>
>>>> In looking for closely at the additions that got made to the
>>>> kernel_read_file() enums, I noticed that FIRMWARE_PREALLOC_BUFFER
>>>> and FIRMWARE_EFI_EMBEDDED were added, but they are not appropriate
>>>> *kinds* of files for the LSM to reason about. They are a "how" and
>>>> "where", respectively. Remove these improper aliases and refactor the
>>>> code to adapt to the changes.
>>>>
>>>> Additionally adds in missing calls to security_kernel_post_read_file()
>>>> in the platform firmware fallback path (to match the sysfs firmware
>>>> fallback path) and in module loading. I considered entirely removing
>>>> security_kernel_post_read_file() hook since it is technically unused,
>>>> but IMA probably wants to be able to measure EFI-stored firmware images,
>>>> so I wired it up and matched it for modules, in case anyone wants to
>>>> move the module signature checks out of the module core and into an LSM
>>>> to avoid the current layering violations.
>>>>
>>>> This touches several trees, and I suspect it would be best to go through
>>>> James's LSM tree.
>>>>
>>>> Thanks!
>>>
>>>
>>> I've done some quick tests on this series to make sure that
>>> the efi embedded-firmware support did not regress.
>>> That still works fine, so this series is;
>>>
>>> Tested-by: Hans de Goede <hdegoede@redhat.com>
>>
>> I made a mistake during testing I was not actually running the
>> kernel with the patches added.
>>
>> After fixing that I did find a problem, patch 4/4:
>> "module: Add hook for security_kernel_post_read_file()"
>>
>> Breaks module-loading for me. This is with the 4 patches
>> on top of 5.8.0-rc4, so this might just be because I'm
>> not using the right base.
>>
>> With patch 4/4 reverted things work fine for me.
>>
>> So, please only add my Tested-by to patches 1-3.
> 
> BTW is there any testing covered by the selftests for the firmware
> laoder which would have caputured this? If not can you extend
> it with something to capture this case you ran into?

This was not a firmware-loading issue. For me in my tests,
which were limited to 1 device, patch 4/4, which only touches
the module-loading code, stopped module loading from working.

Since my test device has / on an eMMC and the kernel config
I'm using has mmc-block as a module, things just hung in the
initrd since no modules could be loaded, so I did not debug
this any further. Dropping  patch 4/4 from my local tree
solved this.

Regards,

Hans