[0/1] mxs-dcp: Fix support for UNIQUE_KEY?

Message

Sven Schwermer Feb. 24, 2025, 7:42 a.m. UTC

Hi there,

I'm not 100% certain about this patch but trial and error seems to
confirm that this patch makes it indeed possible to use UNIQUE_KEY which
I was not able to do with the current implementation.

I would appreciate if somebody with access to this hardware could test
this independently, e.g. the folks at sigma star who authored the
original patch (3d16af0b4cfac).

Best regards,
Sven

Sven Schwermer (1):
  crypto: mxs-dcp: Only set OTP_KEY bit for OTP key

 drivers/crypto/mxs-dcp.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

Comments

David Gstir Feb. 27, 2025, 9:55 a.m. UTC | #1

Hi Sven,

> On 24.02.2025, at 08:42, Sven Schwermer <[e-mail]> wrote:
> 
> Hi there,
> 
> I'm not 100% certain about this patch but trial and error seems to
> confirm that this patch makes it indeed possible to use UNIQUE_KEY which
> I was not able to do with the current implementation.
> 
> I would appreciate if somebody with access to this hardware could test
> this independently, e.g. the folks at sigma star who authored the
> original patch (3d16af0b4cfac).

thanks for the patch! I’ll test this on my end and will report back ASAP.

- David

Sven Schwermer Feb. 27, 2025, 10:48 a.m. UTC | #2

Hi David,

I actually found documentation supporting my findings in the mean time:

 From the i.MX6ULL Security Reference Manual (section 5.3.4):

The OTP key (CRYPTO KEY) can be selected using the DCP_Control1[OTP_KEY] 
bit in the control field of the packet descriptor _or_ by using the key 
select 0xFF in the CTRL1 field of the descriptor.

However, this seems to be contradict by what's later stated in section 
5.3.6.4.2:

If the OTP_KEY value is set, the KEY_SELECT field from the Control1 
register indicates which OTP key is to be used.

After your test, I'll send a v2 with a reworded commit message 
incorporating this.

Sven

On 2/27/25 10:55 AM, David Gstir wrote:
> Hi Sven,
> 
>> On 24.02.2025, at 08:42, Sven Schwermer <[e-mail]> wrote:
>>
>> Hi there,
>>
>> I'm not 100% certain about this patch but trial and error seems to
>> confirm that this patch makes it indeed possible to use UNIQUE_KEY which
>> I was not able to do with the current implementation.
>>
>> I would appreciate if somebody with access to this hardware could test
>> this independently, e.g. the folks at sigma star who authored the
>> original patch (3d16af0b4cfac).
> 
> thanks for the patch! I’ll test this on my end and will report back ASAP.
> 
> - David
> 
>