Message-ID: <b3a2d525-b30c-164c-705c-97fcb0016540@redhat.com>
Date:   Fri, 10 Feb 2023 17:07:13 +0100
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Thunderbird/102.7.1
Subject: Re: [RFC PATCH 9/9] drm/amdgpu: register a vga_switcheroo client for
 all GPUs that are not thunderbolt attached
Content-Language: en-US, nl
To:     Alex Deucher <alexdeucher@gmail.com>,
        Orlando Chamberlain <orlandoch.dev@gmail.com>
Cc:     platform-driver-x86@vger.kernel.org, amd-gfx@lists.freedesktop.org,
        linux-kernel@vger.kernel.org, alsa-devel@alsa-project.org,
        "Pan, Xinhui" <Xinhui.Pan@amd.com>,
        Lijo Lazar <lijo.lazar@amd.com>,
        Rander Wang <rander.wang@intel.com>,
        YiPeng Chai <YiPeng.Chai@amd.com>,
        Mario Limonciello <mario.limonciello@amd.com>,
        David Airlie <airlied@gmail.com>,
        Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>,
        Evan Quan <evan.quan@amd.com>,
        Ranjani Sridharan <ranjani.sridharan@linux.intel.com>,
        Yong Zhi <yong.zhi@intel.com>,
        Aun-Ali Zaidi <admin@kodeit.net>,
        Andrey Grodzovsky <andrey.grodzovsky@amd.com>,
        Bokun Zhang <Bokun.Zhang@amd.com>,
        Mark Gross <markgross@kernel.org>,
        Kerem Karabay <kekrby@gmail.com>,
        Jaroslav Kysela <perex@perex.cz>,
        Jack Xiao <Jack.Xiao@amd.com>,
        Kai Vehmanen <kai.vehmanen@linux.intel.com>,
        Somalapuram Amaranath <Amaranath.Somalapuram@amd.com>,
        Takashi Iwai <tiwai@suse.com>,
        Aditya Garg <gargaditya08@live.com>,
        Daniel Vetter <daniel@ffwll.ch>,
        =?UTF-8?Q?Amadeusz_S=c5=82awi=c5=84ski?= 
        <amadeuszx.slawinski@linux.intel.com>,
        Alex Deucher <alexander.deucher@amd.com>,
        =?UTF-8?Q?Christian_K=c3=b6nig?= <christian.koenig@amd.com>,
        Hawking Zhang <Hawking.Zhang@amd.com>
References: <20230210044826.9834-1-orlandoch.dev@gmail.com>
 <20230210044826.9834-10-orlandoch.dev@gmail.com>
 <CADnq5_PVsrk4+K45kVgU=r98B9GgC2CxbC8AvGUXRtv2sZAHPg@mail.gmail.com>
From:   Hans de Goede <hdegoede@redhat.com>
In-Reply-To: <CADnq5_PVsrk4+K45kVgU=r98B9GgC2CxbC8AvGUXRtv2sZAHPg@mail.gmail.com>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit
Precedence: bulk

Hi,

On 2/10/23 16:53, Alex Deucher wrote:
> On Fri, Feb 10, 2023 at 3:04 AM Orlando Chamberlain
> <orlandoch.dev@gmail.com> wrote:
>>
>> From: Kerem Karabay <kekrby@gmail.com>
>>
>> Commit 3840c5bcc245 ("drm/amdgpu: disentangle runtime pm and
>> vga_switcheroo") made amdgpu only register a vga_switcheroo client for
>> GPU's with PX, however AMD GPUs in dual gpu Apple Macbooks do need to
>> register, but don't have PX. Instead of AMD's PX, they use apple-gmux.
> 
> Is there a way to detect apple-gmux instead?  Otherwise, we register
> vga_switcheroo on any system with multiple GPUs which is not what we
> want.

Yes since 6.1.y (either stable series or just take 6.2.0) the apple-gmux
detect code has been factored out into a stand-alone
apple_gmux_detect() helper inside:

include/linux/apple-gmux.h

For usage outside of the actual apple-gmux driver you can simply
pass NULL for both arguments.

This was necessary to reliably check if the apple-gmux should be
used for backlight control.

Note there also is the older apple_gmux_present() helper, which is
already used in some drm code. That function is not reliable though
it detects if the ACPI tables contain an ACPI device describing
the presence of a gmux, but it turns out even Apple has buggy ACPI
tables and the mere presence of that ACPI device is not a reliable
indicator the gmux is actually there.

I have not changed over any of the existing apple_gmux_present()
users for fear of unwanted side effects...

Regards,

Hans


>> Revert to the old logic of registering for all non-thunderbolt gpus,
>> like radeon and nouveau.
>>
>> Fixes: 3840c5bcc245 ("drm/amdgpu: disentangle runtime pm and vga_switcheroo")
>> Signed-off-by: Kerem Karabay <kekrby@gmail.com>
>> [Orlando Chamberlain <orlandoch.dev@gmail.com>: add commit description]
>> Signed-off-by: Orlando Chamberlain <orlandoch.dev@gmail.com>
>> ---
>>  drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 18 +++++++++++-------
>>  1 file changed, 11 insertions(+), 7 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
>> index 2f28a8c02f64..0bb553a61552 100644
>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
>> @@ -3919,12 +3919,13 @@ int amdgpu_device_init(struct amdgpu_device *adev,
>>         if ((adev->pdev->class >> 8) == PCI_CLASS_DISPLAY_VGA)
>>                 vga_client_register(adev->pdev, amdgpu_device_vga_set_decode);
>>
>> -       if (amdgpu_device_supports_px(ddev)) {
>> -               px = true;
>> -               vga_switcheroo_register_client(adev->pdev,
>> -                                              &amdgpu_switcheroo_ops, px);
>> +       px = amdgpu_device_supports_px(ddev);
>> +
>> +       if (!pci_is_thunderbolt_attached(adev->pdev))
>> +               vga_switcheroo_register_client(adev->pdev, &amdgpu_switcheroo_ops, px);
>> +
>> +       if (px)
>>                 vga_switcheroo_init_domain_pm_ops(adev->dev, &adev->vga_pm_domain);
>> -       }
>>
>>         if (adev->gmc.xgmi.pending_reset)
>>                 queue_delayed_work(system_wq, &mgpu_info.delayed_reset_work,
>> @@ -4048,10 +4049,13 @@ void amdgpu_device_fini_sw(struct amdgpu_device *adev)
>>
>>         kfree(adev->bios);
>>         adev->bios = NULL;
>> -       if (amdgpu_device_supports_px(adev_to_drm(adev))) {
>> +
>> +       if (!pci_is_thunderbolt_attached(adev->pdev))
>>                 vga_switcheroo_unregister_client(adev->pdev);
>> +
>> +       if (amdgpu_device_supports_px(adev_to_drm(adev)))
>>                 vga_switcheroo_fini_domain_pm_ops(adev->dev);
>> -       }
>> +
>>         if ((adev->pdev->class >> 8) == PCI_CLASS_DISPLAY_VGA)
>>                 vga_client_unregister(adev->pdev);
>>
>> --
>> 2.39.1
>>
>