Re: [GIT PULL]: dmaengine updates for v5.7-rc1

[Peter Ujfalusi]

Hi Linus,

On 03/04/2020 2.28, Linus Torvalds wrote:
> On Thu, Apr 2, 2020 at 4:25 AM Vinod Koul <vkoul@xxxxxxxxxx> wrote:
>>
>> Here are the changes for this cycle. SFR has told me that you might see
>> a merge conflict, but I am sure you would be okay with it :)
> 
> It looked trivial enough. That said, it's in the TI_K3_UDMA driver,
> which I can't build. The driver is marked as COMPILE_TEST, but it also
> has
> 
>         depends on TI_SCI_PROTOCOL
>         depends on TI_SCI_INTA_IRQCHIP

Right.

> which means that it depends on TI_MESSAGE_MANAGER, which in turn has a
> 
>         depends on ARCH_KEYSTONE || ARCH_K3

And the INTA_IRQCHIP needs INTA_MSI_DOMAIN.
and the UDMA driver actually needs the ringacc.
It goes pretty deep it looks like.

> so it may be *marked* for build testing, but it doesn't actually get
> any outside of those builds.

Yep, sadly this is true.

> So I did the resolution that looked trivial, but mistakes happen, and
> I can't even build-test that driver..
> 
> Just a heads-up. It does look like it was _meant_ to be build-tested,
> but that intent didn't work out.

The merge was correct, thank you!

> Adding a COMPILE_TEST option to TI_MESSAGE_MANAGER gets things a bit
> further, but even then it doesn't actually build that driver because
> that TI_SCI_INTA_IRQCHIP dependency needs to be enabled too.
> 
> And that one doesn't even have a question, it's just a plain bool, and
> expects to be selected. Which the arm64 platform does.

Yes, I'll sort this out. I prefer my drivers to be compile tested.

> Anyway, to make a long story short: "the COMPILE_TEST marker is a lie".

I'll remove the lie for now and fix things up so I will have legal
grounds to put it back.

> So somebody should actually test my merge.

Tested, thank you again!

> 
>                   Linus
> 

- PÃter

Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki.
Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki