[igt-dev] [PATCH i-g-t v4] tests/i915/gem_huc_copy: Enable a HuC copy test

[Ye, Tony]

There are two issues with the HUC_STATUS query ioctl.

1. on a platform with huc supported, i915.enable_guc=0, got unexpected 
return value:
ret=0, param.value=0, errno=ENODEV

2. the return value doesn't differentiate these two conditions:
a. i915.enable_guc=0/1, huc is disabled, skip the huc_copy test;
b. i915.enable_guc=2/3, huc load/authentication failed, fail huc_copy.
A patch like this is required:
https://patchwork.freedesktop.org/patch/306419/?series=61001&rev=1

Regards,
Tony
On 1/10/2020 5:05 PM, Ye, Tony wrote:
> 
> 
> On 1/10/2020 11:54 AM, Chris Wilson wrote:
>> Quoting Ye, Tony (2020-01-10 19:51:58)
>>>
>>>
>>> On 1/10/2020 10:28 AM, Chris Wilson wrote:
>>>> Quoting Robert M. Fosha (2020-01-10 17:41:46)
>>>>> +static bool
>>>>> +check_huc_enabled(int fd)
>>>>> +{
>>>>> +       int dir;
>>>>> +       dir = igt_sysfs_open_parameters(fd);
>>>>> +
>>>>> +       if (dir < 0)
>>>>> +               return 0;
>>>>> +
>>>>> +       if (igt_sysfs_get_u32(dir, "enable_guc") & 2)
>>>>> +               return 1;
>>>>> +       else
>>>>> +               return 0;
>>>>
>>>> No. If you have this in your userspace, please immediately remove it,
>>>> and backport that to any and all stable branches and releases.
>>>> -Chris
>>>
>>> The code is to skip the test when HuC is disabled by the enable_guc
>>> module parameter. If HuC is disabled by the enable_guc param, then skip
>>> the test. If the HuC failed to load due to a real error, then fail the
>>> test.
>>>
>>> Could you suggest how to handle HUC_STATUS failure properly without this
>>> code?
>>
>> Look at the error code from querying HUC_STATUS.
>> -Chris
> 
> Agree that the check should be done via uAPI instead of the sysfs. 
> Thanks for the feedback.
> Tony
>>