[PATCH] drm/amdgpu: move some release handles into fail labels

[Christian König]

Thanks for that info, I was busy most of the day bisecting this 
regression as well.

Yeah, the bug is rather obvious, there is a "return 0;" in the good case 
missing here.

So after loading the firmware we release it again. I will hack together 
a patch, just give me a minute.

Christian.

Am 14.09.2016 um 16:21 schrieb Marek Olšák:
> Hi,
>
> This patch causes a massive performance drop (by a factor of 100)
> followed by GPU hangs with F1 2015 (it needs more than 3GB of memory,
> the card has 2GB). The game normally works really well (30+ FPS on
> Polaris11 2GB, 40+ FPS on Tonga 2GB).
>
> It should be fixed or the commit should be reverted really quickly.
>
> Marek
>
>
> On Mon, Sep 12, 2016 at 4:16 AM, Huang Rui <ray.huang at amd.com> wrote:
>> Clean up the codes to move the release handles into fail labels.
>>
>> Signed-off-by: Huang Rui <ray.huang at amd.com>
>> ---
>>   drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c | 21 ++++++++++-----------
>>   1 file changed, 10 insertions(+), 11 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c
>> index 5cc95f1..4a5c211 100644
>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c
>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c
>> @@ -247,35 +247,28 @@ int amdgpu_ucode_init_bo(struct amdgpu_device *adev)
>>          const struct common_firmware_header *header = NULL;
>>
>>          err = amdgpu_bo_create(adev, adev->firmware.fw_size, PAGE_SIZE, true,
>> -                       AMDGPU_GEM_DOMAIN_GTT, 0, NULL, NULL, bo);
>> +                              AMDGPU_GEM_DOMAIN_GTT, 0, NULL, NULL, bo);
>>          if (err) {
>>                  dev_err(adev->dev, "(%d) Firmware buffer allocate failed\n", err);
>> -               err = -ENOMEM;
>>                  goto failed;
>>          }
>>
>>          err = amdgpu_bo_reserve(*bo, false);
>>          if (err) {
>> -               amdgpu_bo_unref(bo);
>>                  dev_err(adev->dev, "(%d) Firmware buffer reserve failed\n", err);
>> -               goto failed;
>> +               goto failed_reserve;
>>          }
>>
>>          err = amdgpu_bo_pin(*bo, AMDGPU_GEM_DOMAIN_GTT, &fw_mc_addr);
>>          if (err) {
>> -               amdgpu_bo_unreserve(*bo);
>> -               amdgpu_bo_unref(bo);
>>                  dev_err(adev->dev, "(%d) Firmware buffer pin failed\n", err);
>> -               goto failed;
>> +               goto failed_pin;
>>          }
>>
>>          err = amdgpu_bo_kmap(*bo, &fw_buf_ptr);
>>          if (err) {
>>                  dev_err(adev->dev, "(%d) Firmware buffer kmap failed\n", err);
>> -               amdgpu_bo_unpin(*bo);
>> -               amdgpu_bo_unreserve(*bo);
>> -               amdgpu_bo_unref(bo);
>> -               goto failed;
>> +               goto failed_kmap;
>>          }
>>
>>          amdgpu_bo_unreserve(*bo);
>> @@ -291,6 +284,12 @@ int amdgpu_ucode_init_bo(struct amdgpu_device *adev)
>>                  }
>>          }
>>
>> +failed_kmap:
>> +       amdgpu_bo_unpin(*bo);
>> +failed_pin:
>> +       amdgpu_bo_unreserve(*bo);
>> +failed_reserve:
>> +       amdgpu_bo_unref(bo);
>>   failed:
>>          if (err)
>>                  adev->firmware.smu_load = false;
>> --
>> 2.7.4
>>
>> _______________________________________________
>> amd-gfx mailing list
>> amd-gfx at lists.freedesktop.org
>> https://lists.freedesktop.org/mailman/listinfo/amd-gfx