Re: [linux-next:master 5946/11253] ld.lld: error: undefined symbol: reset_controller_register

[Krzysztof Kozlowski <krzk@kernel.org>]

On 18/06/2021 17:11, Thierry Reding wrote:
> On Fri, Jun 18, 2021 at 05:05:30PM +0200, Krzysztof Kozlowski wrote:
>> On 18/06/2021 17:00, Matthew Wilcox wrote:
>>> On Fri, Jun 18, 2021 at 04:55:23PM +0200, Krzysztof Kozlowski wrote:
>>>> These are randconfigs, so I don't think Linus will hit them when
>>>> merging. Therefore assuming the reset-stubs are applied (and tree is in
>>>> next), we should be fine, shouldn't we?
>>>
>>> It's a bisection hazard.  Please fix it before the merge.  This is
>>> exactly the kind of thing that linux-next is designed to detect.
>>
>> Still hazard for a compile testing configuration, not a real bisection
>> for troubleshooting bugs. You cannot hit the bug in any real case (so
>> !COMPILE_TEST).
> 
> I think you could hit this in legit configurations where people want to
> build ARM_SMMU but not TEGRA_MC. So it's either a dependency that we
> need or the stubs. But I think you could squash the stub patch that I
> sent into your tree if you can rebase that. The patch that the stub
> should go into is this:
> 
> 	393d66fd2cac ("memory: tegra: Implement SID override programming")
> 

I think you refer here to the probe() stub, but I meant reset controller
stubs. These can happen only on COMPILE_TEST && !ARCH_TEGRA &&
!RESET_CONTROLLER.


Best regards,
Krzysztof