Message-ID: <72d566d3-d677-8d0a-612d-fe9c5f3c94cb@linux.ibm.com>
Date:   Wed, 1 Dec 2021 09:52:19 -0500
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:91.0) Gecko/20100101
 Thunderbird/91.2.0
Subject: Re: [PATCH v4 2/2] selftests: tpm2: Reset the dictionary attack lock
Content-Language: en-US
To:     Jarkko Sakkinen <jarkko@kernel.org>
Cc:     Stefan Berger <stefanb@linux.vnet.ibm.com>,
        linux-integrity@vger.kernel.org, peterhuewe@gmx.de,
        linux-kernel@vger.kernel.org, linux-kselftest@vger.kernel.org,
        skhan@linuxfoundation.org
References: <20211128041052.1395504-1-stefanb@linux.vnet.ibm.com>
 <20211128041052.1395504-3-stefanb@linux.vnet.ibm.com>
 <YaVljk1vLRZ/TDJ/@iki.fi>
 <e569444c-e0cd-52bc-308f-7fa457dbf086@linux.ibm.com>
 <YadLaHB0oJZYTMbh@iki.fi>
From:   Stefan Berger <stefanb@linux.ibm.com>
In-Reply-To: <YadLaHB0oJZYTMbh@iki.fi>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit
Precedence: bulk

On 12/1/21 05:16, Jarkko Sakkinen wrote:

> On Mon, Nov 29, 2021 at 07:26:12PM -0500, Stefan Berger wrote:
>> On 11/29/21 18:43, Jarkko Sakkinen wrote:
>>> On Sat, Nov 27, 2021 at 11:10:52PM -0500, Stefan Berger wrote:
>>>> From: Stefan Berger <stefanb@linux.ibm.com>
>>>>
>>>> Reset the dictionary attack lock to avoid the following types of test
>>>> failures after running the test 2 times:
>>>>
>>>> ======================================================================
>>>> ERROR: test_unseal_with_wrong_policy (tpm2_tests.SmokeTest)
>>>> ----------------------------------------------------------------------
>>>> Traceback (most recent call last):
>>>>     File "/root/linux-ima-namespaces/tools/testing/selftests/tpm2/tpm2_tests.py", line 105, in test_unseal_with_wrong_policy
>>>>       blob = self.client.seal(self.root_key, data, auth, policy_dig)
>>>>     File "/root/linux-ima-namespaces/tools/testing/selftests/tpm2/tpm2.py", line 620, in seal
>>>>       rsp = self.send_cmd(cmd)
>>>>     File "/root/linux-ima-namespaces/tools/testing/selftests/tpm2/tpm2.py", line 397, in send_cmd
>>>>       raise ProtocolError(cc, rc)
>>>> tpm2.ProtocolError: TPM_RC_LOCKOUT: cc=0x00000153, rc=0x00000921
>>>>
>>>> Signed-off-by: Stefan Berger <stefanb@linux.ibm.com>
>>>> ---
>>>>    tools/testing/selftests/tpm2/tpm2_tests.py | 2 ++
>>>>    1 file changed, 2 insertions(+)
>>>>
>>>> diff --git a/tools/testing/selftests/tpm2/tpm2_tests.py b/tools/testing/selftests/tpm2/tpm2_tests.py
>>>> index e63a37819978..ad6f54c01adf 100644
>>>> --- a/tools/testing/selftests/tpm2/tpm2_tests.py
>>>> +++ b/tools/testing/selftests/tpm2/tpm2_tests.py
>>>> @@ -139,6 +139,8 @@ class SmokeTest(unittest.TestCase):
>>>>            except:
>>>>                self.client.flush_context(handle)
>>>>                raise
>>>> +        finally:
>>>> +            self.client.reset_da_lock()
>>>>            self.assertEqual(rc, tpm2.TPM2_RC_POLICY_FAIL)
>>>> -- 
>>>> 2.31.1
>>>>
>>> I don't agree with this as a DA lock has legit use. This would be adequate
>>> for systems dedicated for kernel testing only.
>> The problem is this particular test case I am patching here causes the above
>> test failures upon rerun. We are testing the driver here presumably and not
>> the TPM2, so I think we should leave the TPM2 as cleaned up as possible,
>> thus my suggestion is to reset the DA lock and we won't hear any complaints
>> after that.
>
>> The tss packages also have command line tools to reset the DA lock, but it
>> shouldn't be necessary to use them after running a **driver** test case.
> If you speak about TSS, please alway say which one :-)

packages : both

tpm2_dictionarylockout -c [ -p password ]

tssdictionaryattacklockreset [-pwd password]


>
> Adding non-volatile state changes explicitly is to a test case is both
> intrusive and wrong. These type of choices are not to be done in the
> test case implementation for sure.

I drop this patch because if someone changed the password the lock reset 
will not work as expected.

One can also argue with having a test case that triggers a failure 
condition in the TPM2 is both intrusive and wrong. That said, the 
offending test cases should probably not even exist, especially since 
they test a user's knowledge about the device afterwards...


>
> An improvement that does not add extra side-effect, would be to read the
> TPM_PT_LOCKOUT_COUNTER and roll back with a proper error message for the
> lockout condition.
>
> You can also configure the maximum number of tries up to (2 << 31) - 1
> = 4294967295 with TPM2_DictionaryAttackParameters...

... which I think the user of a device driver test should not have to do.


   Stefan

>
> /Jarkko