Re: [PATCH v6] hw/misc/aspeed_hace: Fix SG Accumulative hashing

["Cédric Le Goater" <clg@redhat.com>]

On 10/22/24 13:54, Joel Stanley wrote:
> On Wed, 16 Oct 2024 at 01:23, Jamin Lin <jamin_lin@aspeedtech.com> wrote:
> 
>> 3. Test HACE model with u-boot hash command
>> a. load test file to address 83000000 via tftp
>> ast# tftp 83000000 jamin_lin/32MB
>> b. get sha256
>> ast# hash sha256 83000000 2000000
>> sha256 for 83000000 ... 84ffffff ==> 1ddcccdba742d762e2b8da0bceaf4778727c5eba54a24d7ae0c573c65414f736
>> c. get sha384
>> ast# hash sha384 83000000 2000000
>> sha384 for 83000000 ... 84ffffff ==> 825d9b24bb797695545b3cbd2f373b9738627c7a1878e620415570a57c7faed77916d47084c954254f101fc0f10c0591
>> d. get sha512
>> ast# hash sha512 83000000 2000000
>> sha512 for 83000000 ... 84ffffff ==> b5ae725b2dc1e521f48eae37dd82c3d5fc94f7acb5fff3dabf1caa4bb4b5bcfb498e7cc1fbaa97dda2664bff99f9f8e778f823e95afaf76fbf0985181522e478
> 
> I attempted this same test and noticed that the 'hash' command was not
> using the hardware. You can see this by putting some printf or
> breakpoint in eg hw/misc/aspeed_hace.c do_hash_operation. There's some
> missing work on the u-boot side to move the "hash" command over to the
> hash uclass, so it can be used to test this code path (or add support
> for the old API to the hace driver).
> 
> Separately, I attempted to test with u-boot by enabling hash
> verification of the FIT image, and it fails to calculate the correct
> SHA.
> 
> I think to have any confidence that this model works, we need to add
> some testing to qemu. I did this for the initial version of the model
> in tests/qtest/aspeed_hace-test.c.

There are "accumulative mode" tests in QEMU, which were added by commit
e0c371a0d23b ("tests/qtest: Add test for Aspeed HACE accumulative mode")
They pass today with this patch. Are you suggesting we should add more?

Thanks,

C.


  
> The upstream u-boot situation is a mess, and cannot be used to
> exercise the qemu model at this stage.
> 
> Cheers,
> 
> Joel
>