Re: [PATCH] pci: controller: tegra194: remove dependency on Tegra194 SoC

[Jon Hunter <jonathanh@nvidia.com>]

On 05/01/2026 12:21, Manivannan Sadhasivam wrote:
> On Mon, Jan 05, 2026 at 12:18:35PM +0000, Jon Hunter wrote:
>>
>> On 05/01/2026 12:13, Niklas Cassel wrote:
>>> On Mon, Jan 05, 2026 at 05:38:34PM +0530, Manivannan Sadhasivam wrote:
>>>> On Mon, Jan 05, 2026 at 11:56:37AM +0000, Jon Hunter wrote:
>>>>>
>>>>>
>>>>> On 05/01/2026 09:09, Niklas Cassel wrote:
>>>>>> On Mon, Jan 05, 2026 at 02:09:34PM +0530, Manivannan Sadhasivam wrote:
>>>>>>> On Fri, Jan 02, 2026 at 10:58:15AM +0000, Jon Hunter wrote:
>>>>>>>>
>>>>>>>> On 23/12/2025 06:45, Manivannan Sadhasivam wrote:
>>>>>>>>> On Wed, Nov 26, 2025 at 11:25:30AM +0100, Francesco Lavra wrote:
>>>>>>>>>
>>>>>>>>> + Tegra maintainers
>>>>>>>>>
>>>>>>>>>> This driver runs (for both host and endpoint operation) also on other Tegra
>>>>>>>>>> SoCs (e.g. Tegra234).
>>>>>>>>>> Replace the Kconfig dependency on ARCH_TEGRA_194_SOC with a more generic
>>>>>>>>>> dependency on ARCH_TEGRA; in addition, amend the Kconfig help text to
>>>>>>>>>> reflect the fact that this driver is no longer exclusive to Tegra194.
>>>>>>>>>>
>>>>>>>>>
>>>>>>>>> I vaguely remember asking about this a while back during some other patch review
>>>>>>>>> and I don't remember what we concluded.
>>>>>>>>>> Thierry, Jon, thoughts?
>>>>>>>>
>>>>>>>> So ARCH_TEGRA means that this can be enabled for the legacy 32-bit Tegra
>>>>>>>> devices as well as the current 64-bit Tegra devices (such as Tegra194).
>>>>>>>> Given that this driver is only used for Tegra194 and Tegra234, it seems it
>>>>>>>> would be better to only enable this for Tegra194 and Tegra234 instead of any
>>>>>>>> Tegra.
>>>>>>>>
>>>>>>>
>>>>>>> The dependency means, whenever someone tries to enable PCIE_TEGRA194_HOST,
>>>>>>> ARCH_TEGRA should be enabled. So as long as someone not trying to enable
>>>>>>> PCIE_TEGRA194_HOST for 32bit Tegra systems, ARCH_TEGRA should work fine and
>>>>>>> PCIE_TEGRA194_HOST is not selected by arch/arm/configs/tegra_defconfig. So I
>>>>>>> don't see any blocker with this patch. In fact, many other archs do the same.
>>>>>>>
>>>>>>> But I don't like extending the Kconfig with per SoC dependency as it won't
>>>>>>> scale.
>>>>>>
>>>>>> We already have a patch from Vidya:
>>>>>> [PATCH V4] PCI: dwc: tegra194: Broaden architecture dependency
>>>>>> that was sent 2025-05-08
>>>>>>
>>>>>> Back then, the reason why it wasn't merged was because it required a
>>>>>> similar change to the PHY driver to go in first:
>>>>>> https://lore.kernel.org/linux-pci/174722268141.85510.14696275121588719556.b4-ty@kernel.org/
>>>>>>
>>>>>> The PHY driver change was merged in v6.16:
>>>>>> 0c2228731974 ("phy: tegra: p2u: Broaden architecture dependency")
>>>>>>
>>>>>> So, I think we could just merge:
>>>>>> https://lore.kernel.org/linux-pci/20250508051922.4134041-1-vidyas@nvidia.com/
>>>>>>
>>>>>> (Assuming it still applies.)
>>>>>
>>>>> Yes it does and applying Sagar's patch is fine with me. So it you want to
>>>>> apply Sagar's patch please add my ...
>>>>>
>>>>
>>>> Don't we need:
>>>>
>>>> 	depends on (ARCH_TEGRA && ARM64) || COMPILE_TEST
>>>
>>> This is exactly what I originally suggested to Vidya:
>>> https://lore.kernel.org/linux-pci/Z6XjWJd9jm0HHNXW@ryzen/
>>>
>>>
>>>>
>>>> in the above patch?
>>>
>>> The above patch instead has:
>>>
>>> depends on ARCH_TEGRA && (ARM64 || COMPILE_TEST)
>>>
>>> I don't know why Vidya did not use my suggestion exactly, but I guess I
>>> assumed that he had a reason not to use my suggestion exactly.
>>
>> Looking 0c2228731974 ("phy: tegra: p2u: Broaden architecture
>> dependency") we ended up just merging ...
>>
>> diff --git a/drivers/phy/tegra/Kconfig b/drivers/phy/tegra/Kconfig
>> index f30cfb42b210..342fb736da4b 100644
>> --- a/drivers/phy/tegra/Kconfig
>> +++ b/drivers/phy/tegra/Kconfig
>> @@ -13,7 +13,7 @@ config PHY_TEGRA_XUSB
>>   config PHY_TEGRA194_P2U
>>          tristate "NVIDIA Tegra194 PIPE2UPHY PHY driver"
>> -       depends on ARCH_TEGRA_194_SOC || ARCH_TEGRA_234_SOC || COMPILE_TEST
>> +       depends on ARCH_TEGRA || COMPILE_TEST
>>          select GENERIC_PHY
>>          help
>>
>> So I guess we could just merge Francesco's original suggestion for
>> consistency. Otherwise I would be happy with ...
>>
>>   depends on (ARCH_TEGRA && ARM64) || COMPILE_TEST
>>
> 
> I made the above change applied Vidya's patch. Thanks Jon and Niklas!

Thanks! Feel free to add my ...

Acked-by: Jon Hunter <jonathanh@nvidia.com>

Jon

-- 
nvpublic