Re: [PATCH] vga-switcheroo: Allow VGA_SWITCHEROO to be enabled on LoongArch

[Thomas Zimmermann <tzimmermann@suse.de>]

Hi

Am 30.10.26 um 09:39 schrieb zhaotianrui:
>
> 在 2026/10/30 下午4:17, Thomas Zimmermann 写道:
>> Hi
>>
>> Am 30.10.26 um 08:04 schrieb zhaotianrui:
>>>
>>> 在 2026/10/29 下午4:16, Thomas Zimmermann 写道:
>>>> Hi
>>>>
>>>> Am 29.10.26 um 02:36 schrieb Tianrui Zhao:
>>>>> LoongArch systems with hybrid graphics (integrated +
>>>>> discrete GPU) can benefit from the same GPU switching
>>>>> mechanism that x86 laptops use. Relax the dependency
>>>>> to allow VGA_SWITCHEROO to be built on LoongArch,
>>>>> consistent with the existing dependencies on ACPI,
>>>>> PCI, and FB=y.
>>>>>
>>>>> Change-Id: I84d38fdfe5e1591495fb3681474a7f1f64281fa8
>>>>
>>>> Internal change ids should be remove before submitting patches.
>>>>
>>> Thanks for your review, I will fix this.
>>>
>>>
>>> Best regards
>>>
>>> Tianrui Zhao
>>>
>>>>> Signed-off-by: Tianrui Zhao <zhaotianrui@loongson.cn>
>>>>> ---
>>>>>   drivers/gpu/vga/Kconfig | 2 +-
>>>>>   1 file changed, 1 insertion(+), 1 deletion(-)
>>>>>
>>>>> diff --git a/drivers/gpu/vga/Kconfig b/drivers/gpu/vga/Kconfig
>>>>> index bad4bcee313..89edd1211aa 100644
>>>>> --- a/drivers/gpu/vga/Kconfig
>>>>> +++ b/drivers/gpu/vga/Kconfig
>>>>> @@ -1,7 +1,7 @@
>>>>>   # SPDX-License-Identifier: GPL-2.0-only
>>>>>   config VGA_SWITCHEROO
>>>>>       bool "Laptop Hybrid Graphics - GPU switching support"
>>>>> -    depends on X86
>>>>> +    depends on X86 || LOONGARCH
>>>>
>>>> I'm a bit skeptical if that's all there is to do.  Does loongarch 
>>>> really work with the existing handlers in the graphics drivers 
>>>> as-is?  See [1] for the interface that implements the actual 
>>>> switching.
>>>>
>>>> [1] 
>>>> https://elixir.bootlin.com/linux/v7.3/source/include/linux/vga_switcheroo.h#L93
>>>>
>>>> Best regards
>>>> Thomas
>>>>
>>> Thanks,  I will carefully check whether all these 
>>> vga_switcheroo_handler callbacks are supported on the LoongArch 
>>> platform.
>>
>> Switcheroo used to be required for x86 notebooks from ~2010. Those 
>> systems had two graphics cards, but only one set of output 
>> connectors. That design is become rare, I think.
>>
>> So I'm a bit surprised to see it come back. Which loongarch systems 
>> are there?
>>
>> Best regards
>> Thomas
>>
>>
> Hi Thomas,
>
> Thanks for your reply.
>
> On the Loongson platform, we don't have the "two GPUs but only one set 
> of display outputs" scenario you described. However, we do have 
> systems where an integrated GPU and a discrete GPU coexist.
> We would like to use the vga_switcheroo interface to switch between 
> the discrete GPU (DIS) and the integrated GPU (IGD), and to configure 
> the default VGA device.
> Would this approach be feasible? Do you have any suggestions or 
> recommendations?

Oh, that's not what vga_switcheroo is for. For your scenario, simply 
provide a DRM driver for each GPU. Selection of the boot-up display 
happens in the UEFI/BIOS.  Picking the correct display hardware at 
runtime is left to clients in user space (or fbcon). DRM itself has not 
preference.

Best regards
Thomas

>
> Best regards
> Tianrui Zhao
>>>
>>>
>>> Best regards
>>>
>>> Tianrui Zhao
>>>
>>>>>       depends on ACPI
>>>>>       depends on PCI
>>>>>       depends on FB=y
>>>>
>>>
>>
>

-- 
--
Thomas Zimmermann
Graphics Driver Developer
SUSE Software Solutions Germany GmbH
Frankenstr. 146, 90461 Nürnberg, Germany, www.suse.com
GF: Stefan Gaiser, Jochen Jaser, Abhinav Puri, (HRB 36809, AG Nürnberg)