Re: [PATCH v7 00/12] crypto: qce - Fix crypto self-test failures

[Bartosz Golaszewski]

On Wed, 16 Sep 2026 18:45:40 +0200, Demi Marie Obenour
<demiobenour@xxxxxxxxx> said:
> On 9/15/26 05:42, Bartosz Golaszewski wrote:
>> On Fri, 11 Sep 2026 20:34:39 +0200, Demi Marie Obenour
>> <demiobenour@xxxxxxxxx> said:
>>> On 9/10/26 09:00, Bartosz Golaszewski wrote:
>>>> This iteration - in addition to the previous fixes - proposes to split
>>>> the QCE driver into a core part necessary to bind to the QCE DT node and
>>>> enable runtime power management in order to allow to drop the
>>>> interconnect votes, and the crypto part registering the crypto
>>>> algorithms. The core module is then enabled in arm64 defconfig while the
>>>> crypto part stays disabled by default.
>>>>
>>>> The QCE hardware crypto engine has several limitations that cause it to
>>>> produce incorrect results or stall on certain inputs. This series fixes
>>>> several bugs and adds workaround allowing the deiver to pass crypto
>>>> self-tests.
>>>>
>>>> The failures addressed are:
>>>>
>>>> - HMAC self-test failures for empty messages
>>>> - AES-XTS returning success on zero-length input (should be -EINVAL)
>>>> - AES-CTR: partial final block causes the engine to stall, output IV
>>>>   derivation was incorrect
>>>> - AES-XTS with key1 == key2 is not supported by the CE
>>>> - AES-CCM: partial final block and fragmented payload both stall the
>>>>   engine
>>>>
>>>> All fixes were tested on an SM8650 QRD board with
>>>> CONFIG_CRYPTO_SELFTESTS=y and CONFIG_CRYPTO_SELFTESTS_FULL=y.
>>>>
>>>> Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@xxxxxxxxxxxxxxxx>
>>> For stable, would it be better to not register the algorithms, thus
>>> preventing their use?  Another option would be to replace the whole
>>> driver with an always-on stub driver that just does power management.
>>
>> A module parameter called: really_register_algos that's disabled by default?
>>
>> Bart
>
> Works for me!  If you add that and fix the tests, go ahead and revert
> my patch that marked it as BROKEN.  I suggest reverting the defconfig
> changes too, for power management reasons.
>
> While the driver is buggy, you might want to add a kernel taint if
> the option is set.  There's precedent for that in other parts of the
> kernel.

I'd prefer not to. I hope gating the algos behind a module parameter is enough.

Bart