[SLOF] [PATCH v4 00/33] Add vTPM support to SLOF

[Stefan Berger]

On 12/27/19 11:11 PM, Alexey Kardashevskiy wrote:
>
>
> On 28/12/2019 08:13, Stefan Berger wrote:
>> On 12/23/19 6:41 PM, Alexey Kardashevskiy wrote:
>>>
>>> On 24/12/2019 08:40, Stefan Berger wrote:
>>>> On 12/18/19 6:05 PM, Alexey Kardashevskiy wrote:
>>>>> Hey,
>>>>>
>>>>> It's been a while since the last attempt :) 33 is a lot! Comments
>>>>> below...
>>>> I have a series of patches now here with some of the concerns 
>>>> addressed.
>>>> I would say it's bisectable but my concern is that reshuffling more 
>>>> code
>>>> will cause issues that aren't there right now. So I would appreciate
>>>> some tolerance for refactoring while building up TPM 2 support.
>>>>
>>>>
>>>> https://github.com/stefanberger/SLOF-tpm/commits/SLOF-tpm.Dec2019.23
>>> imho you are overthinking this. This needs to be several prerequisite
>>> patches (one per lib/directory/driver/whatever), one patch for vtpm v1
>>
>> What are prerequisites? Are these functions needed by lib/libtpm but 
>> are located in other directories?
>
> Yes.
>
>> I have certainly not put the patches in random order, so they do 
>> compile at every step and functions needed in one step are introduced 
>> before it.
>
> Good.
>
>>> (and I still eagerly want to hear why we want v1 at all), one patch for
>>
>> The application stack for TSS 1.2 is (still) being packaged for ppc64le:
>
>
> Ok. But what about v2.0? "trousers" does not support it?


There is different stack for TPM 2.0, actually there are two:

- tss2 (IBM): 
https://koji.fedoraproject.org/koji/packageinfo?packageID=23163

- tpm2-tss (Intel): 
https://koji.fedoraproject.org/koji/buildinfo?buildID=1422031


>
>
>>
>> https://packages.ubuntu.com/eoan/trousers
>>
>> https://koji.fedoraproject.org/koji/packageinfo?packageID=5471
>>
>> https://software.opensuse.org/package/trousers
>>
>> https://packages.debian.org/search?keywords=trousers
>>
>>
>>
>>
>>> vtpm v2, one patch for the menu. "One patch" could be more if patch 1
>>> adds some function on its own and consequent patches enhance patch 1
>>> (such as add vtpm with 1 algo enabled in patch 1 and then add other
>>> hashing algorithms later) but I do not see this happening or very
>>> useful. Thanks,
>>
>> I tried to make it small patches for easier digestion...
>
>
> Not seeing how helpers are actually used does not help much and 
> bisectability suffers as well. Thanks,


So then changing it to a series of patches that add prerequisites, then 
the low lever PAPR driver, then the SHA1 implementation, and then TPM 
1.2 support followed by TPM 2.0 support, and then the combined menu for 
TPM 1.2 and TPM 2.0 will improve it for the review? The prerequisites 
are like two-liners for invoking FORTH functions from C.


>
>
>