[PATCH] drm/amdgpu: defer test IBs on the rings at boot (V3)

[Christian König]

Am 16.04.2018 um 11:04 schrieb Zhang, Jerry (Junwei):
> On 04/16/2018 03:17 PM, Shirish S wrote:
>> amdgpu_ib_ring_tests() runs test IB's on rings at boot
>> contributes to ~500 ms of amdgpu driver's boot time.
>>
>> This patch defers it and ensures that its executed
>> in amdgpu_info_ioctl() if it wasn't scheduled.
>>
>> V2: Use queue_delayed_work() & flush_delayed_work().
>> V3: removed usage of separate wq, ensure ib tests is
>>      run before enabling clockgating.
>>
>> Signed-off-by: Shirish S <shirish.s at amd.com>
>> ---
>>   drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 17 ++++++-----------
>>   drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c    |  3 +++
>>   2 files changed, 9 insertions(+), 11 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c 
>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
>> index 1762eb4..f225840 100644
>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
>> @@ -1656,6 +1656,10 @@ static int 
>> amdgpu_device_ip_late_set_cg_state(struct amdgpu_device *adev)
>>       if (amdgpu_emu_mode == 1)
>>           return 0;
>>
>> +    r = amdgpu_ib_ring_tests(adev);
>> +    if (r)
>> +        DRM_ERROR("ib ring test failed (%d).\n", r);
>> +
>
> Just confirm:
> IIRC, ib may not run immediately via scheduler thread.
> Is there any possible that the actual ib ring test interferes cg 
> enablement?

No, the IB tests are run directly on the hardware ring and not through 
the scheduler.

But even when we run them through the scheduler we wait for the to 
complete, so that shouldn't be an issue.

Christian.

>
> Jerry
>
>>       for (i = 0; i < adev->num_ip_blocks; i++) {
>>           if (!adev->ip_blocks[i].status.valid)
>>               continue;
>> @@ -1706,8 +1710,8 @@ static int amdgpu_device_ip_late_init(struct 
>> amdgpu_device *adev)
>>           }
>>       }
>>
>> -    mod_delayed_work(system_wq, &adev->late_init_work,
>> -            msecs_to_jiffies(AMDGPU_RESUME_MS));
>> +    queue_delayed_work(system_wq, &adev->late_init_work,
>> +               msecs_to_jiffies(AMDGPU_RESUME_MS));
>>
>>       amdgpu_device_fill_reset_magic(adev);
>>
>> @@ -2374,10 +2378,6 @@ int amdgpu_device_init(struct amdgpu_device 
>> *adev,
>>           goto failed;
>>       }
>>
>> -    r = amdgpu_ib_ring_tests(adev);
>> -    if (r)
>> -        DRM_ERROR("ib ring test failed (%d).\n", r);
>> -
>>       if (amdgpu_sriov_vf(adev))
>>           amdgpu_virt_init_data_exchange(adev);
>>
>> @@ -2640,11 +2640,6 @@ int amdgpu_device_resume(struct drm_device 
>> *dev, bool resume, bool fbcon)
>>
>>       amdgpu_fence_driver_resume(adev);
>>
>> -    if (resume) {
>> -        r = amdgpu_ib_ring_tests(adev);
>> -        if (r)
>> -            DRM_ERROR("ib ring test failed (%d).\n", r);
>> -    }
>>
>>       r = amdgpu_device_ip_late_init(adev);
>>       if (r)
>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c 
>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c
>> index 487d39e..83d7160 100644
>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c
>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c
>> @@ -279,6 +279,9 @@ static int amdgpu_info_ioctl(struct drm_device 
>> *dev, void *data, struct drm_file
>>       if (!info->return_size || !info->return_pointer)
>>           return -EINVAL;
>>
>> +    /* Ensure IB tests are run on ring */
>> +    flush_delayed_work(&adev->late_init_work);
>> +
>>       switch (info->query) {
>>       case AMDGPU_INFO_ACCEL_WORKING:
>>           ui32 = adev->accel_working;
>>