Re: [PATCH v3] ASoC: atmel: ATMEL drivers don't need HAS_DMA

[Geert Uytterhoeven]

Hi Randy,

On Fri, Sep 3, 2021 at 9:53 PM Randy Dunlap <rdunlap@xxxxxxxxxxxxx> wrote:
> On 9/2/21 9:44 AM, Geert Uytterhoeven wrote:
> > On Thu, Jul 8, 2021 at 6:51 PM Randy Dunlap <rdunlap@xxxxxxxxxxxxx> wrote:
> >> On 7/8/21 1:19 AM, Codrin.Ciubotariu@xxxxxxxxxxxxx wrote:
> >>> On 08.07.2021 00:47, Randy Dunlap wrote:
> >>>> EXTERNAL EMAIL: Do not click links or open attachments unless you know the content is safe
> >>>>
> >>>> On a config (such as arch/sh/) which does not set HAS_DMA when MMU
> >>>> is not set, several ATMEL ASoC drivers select symbols that cause
> >>>> kconfig warnings. There is one "depends on HAS_DMA" which is no longer
> >>>> needed. Dropping it eliminates the kconfig warnings and still builds
> >>>> with no problems reported.
> >>>>
> >>>> Fix the following kconfig warnings:
> >>>>
> >>>> WARNING: unmet direct dependencies detected for SND_ATMEL_SOC_PDC
> >>>>     Depends on [n]: SOUND [=m] && !UML && SND [=m] && SND_SOC [=m] && SND_ATMEL_SOC [=m] && HAS_DMA [=n]
> >>>>     Selected by [m]:
> >>>>     - SND_ATMEL_SOC_SSC [=m] && SOUND [=m] && !UML && SND [=m] && SND_SOC [=m] && SND_ATMEL_SOC [=m]
> >>>>     - SND_ATMEL_SOC_SSC_PDC [=m] && SOUND [=m] && !UML && SND [=m] && SND_SOC [=m] && SND_ATMEL_SOC [=m] && ATMEL_SSC [=m]
> >>>>
> >>>> WARNING: unmet direct dependencies detected for SND_ATMEL_SOC_SSC_PDC
> >>>>     Depends on [n]: SOUND [=m] && !UML && SND [=m] && SND_SOC [=m] && SND_ATMEL_SOC [=m] && ATMEL_SSC [=m] && HAS_DMA [=n]
> >>>>     Selected by [m]:
> >>>>     - SND_AT91_SOC_SAM9G20_WM8731 [=m] && SOUND [=m] && !UML && SND [=m] && SND_SOC [=m] && SND_ATMEL_SOC [=m] && (ARCH_AT91 || COMPILE_TEST [=y]) && ATMEL_SSC [=m] && SND_SOC_I2C_AND_SPI [=m]
> >>>>
> >>>> WARNING: unmet direct dependencies detected for SND_ATMEL_SOC_SSC
> >>>>     Depends on [n]: SOUND [=m] && !UML && SND [=m] && SND_SOC [=m] && SND_ATMEL_SOC [=m] && HAS_DMA [=n]
> >>>>     Selected by [m]:
> >>>>     - SND_ATMEL_SOC_SSC_DMA [=m] && SOUND [=m] && !UML && SND [=m] && SND_SOC [=m] && SND_ATMEL_SOC [=m] && ATMEL_SSC [=m]
> >>>>
> >>>> WARNING: unmet direct dependencies detected for SND_ATMEL_SOC_SSC_DMA
> >>>>     Depends on [n]: SOUND [=m] && !UML && SND [=m] && SND_SOC [=m] && SND_ATMEL_SOC [=m] && ATMEL_SSC [=m] && HAS_DMA [=n]
> >>>>     Selected by [m]:
> >>>>     - SND_ATMEL_SOC_WM8904 [=m] && SOUND [=m] && !UML && SND [=m] && SND_SOC [=m] && SND_ATMEL_SOC [=m] && (ARCH_AT91 || COMPILE_TEST [=y]) && ATMEL_SSC [=m] && I2C [=m]
> >>>>     - SND_AT91_SOC_SAM9X5_WM8731 [=m] && SOUND [=m] && !UML && SND [=m] && SND_SOC [=m] && SND_ATMEL_SOC [=m] && (ARCH_AT91 || COMPILE_TEST [=y]) && ATMEL_SSC [=m] && SND_SOC_I2C_AND_SPI [=m]
> >>>>
> >>>> Fixes: 3951e4aae2ce ("ASoC: atmel-pcm: dma support based on pcm dmaengine")
> >>>> Fixes: 18291410557f ("ASoC: atmel: enable SOC_SSC_PDC and SOC_SSC_DMA in Kconfig")
> >>>> Fixes: 061981ff8cc8 ("ASoC: atmel: properly select dma driver state")
> >>>
> >>> I am not sure about these fixes tags. As Alexandre mentioned, it looks
> >>> like the reason for HAS_DMA in the first place was the COMPILE_TEST with
> >>> m32r arch. I dig a bit, and, if any, I think we should use:
> >>> Fixes: eb17726b00b3 ("m32r: add simple dma")
> >>> since this commit adds dummy DMA support for m32r and seems to fix the
> >>> HAS_DMA dependency.
> >>
> >> Ah, I forgot to update the Fixes: tag(s).
> >>
> >> I won't disagree with your Fixes: suggestion (good digging) but
> >> I would probably have used 8d7d11005e930:
> >>    ASoC: atmel: fix build failure
> >> which is the commit that added "depends on HAS_DMA".
> >
> > M32r was not the only platform NO_DMA, so I guess the build would
> > have failed for the others, too (e.g. Sun-3).
> >
> > So the real fix was probably commit f29ab49b5388b2f8 ("dma-mapping:
> > Convert NO_DMA get_dma_ops() into a real dummy"), or one of the
> > related commits adding dummies to subsystems.
>
> Hi Geert,
> Does this mean that some other actions are needed here?
> E.g. revert + a different kind of fix?

While we can now compile drivers using DMA features on NO_DMA
platforms, thanks to the dummies, it does mean many of these drivers
cannot work on such platforms.  So I think it makes sense to replace
"depends on HAS_DMA" by "depends on HAS_DMA || COMPILE_TEST" if DMA
is not optional to the driver.

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds