Re: [PATCH] xen/arm: optee: Fix arm_smccc_smc's a0 for OPTEE_SMC_DISABLE_SHM_CACHE

[Volodymyr Babchuk <Volodymyr_Babchuk@xxxxxxxx>]

Hi Oleksandr, Stefano,

Oleksandr <olekstysh@xxxxxxxxx> writes:

> On 07.10.21 01:42, Stefano Stabellini wrote:
>
> Hi Stefano, Julien.
>
>> On Wed, 6 Oct 2021, Julien Grall wrote:
>>> Hi Stefano,
>>>
>>> On 28/09/2021 06:52, Stefano Stabellini wrote:
>>>> On Mon, 27 Sep 2021, Oleksandr Tyshchenko wrote:
>>>>> From: Oleksandr Tyshchenko <oleksandr_tyshchenko@xxxxxxxx>
>>>>>
>>>>> Fix a possible copy-paste error in arm_smccc_smc's first argument (a0)
>>>>> for OPTEE_SMC_DISABLE_SHM_CACHE case.
>>>>>
>>>>> This error causes Linux > v5.14-rc5
>>>>> (b5c10dd04b7418793517e3286cde5c04759a86de
>>>>> optee: Clear stale cache entries during initialization) to stuck
>>>>> repeatedly issuing OPTEE_SMC_DISABLE_SHM_CACHE call and waiting for
>>>>> the result to be OPTEE_SMC_RETURN_ENOTAVAIL which will never happen.
>>>>>
>>>>> Signed-off-by: Oleksandr Tyshchenko <oleksandr_tyshchenko@xxxxxxxx>
>>>> Acked-by: Stefano Stabellini <sstabellini@xxxxxxxxxx>
>>>>
>>>> I added Fixes: and Backport: tags to the commit
>>> Per SUPPORT.MD, OP-TEE is still a technical preview. So I would argue that 
>>> we
>>> should not do any backport because the feature itself is not officially
>>> considered supported.
>> Good point!
>>
>>
>>> That said, what's missing to make the feature officially supported?
>> If Oleksandr is also happy to make OP-TEE support in Xen "Supported" in
>> SUPPORT.md I'd be happy with that too. Specifically I suggest to change
>> it to:
>>
>> Status: Supported, not security supported
>>
>> Security Support is a bit of a heavy process and I am thinking that
>> "Supported, not security supported" would be an excellent next step.
>
> I would be happy, and can send a formal patch. But I am not an expert
> in this code.

I'm will be happy with this too. We are using this mediator in our
projects and I know that OP-TEE community adopted tests for
virtualization in theirs CI stack. So this is kind of official now.

Also, I helped other people to bring up virtualization on theirs
platforms, so there are other users for this feature besides EPAM and
Linaro.

> (looks like there are some TODO left in the code and I have no idea
> what are the implications)

Well, there were a lot of TODOs when I submitted initial
implementation. At that time it indeed wasn't ready for production. But
I eventually fixed almost all of them. Only one left now. It is about
very unlikely situation when one of guest pages in mapped at PA=0. I'm
not sure that is even possible at all.

-- 
Volodymyr Babchuk at EPAM