lkml.org 
[lkml]   [2017]   [Apr]   [24]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRe: [PATCH 3/3] sound: ASoC: tegra: Select tegra30 i2s and ahub for tegra124 SoC
From
Date
Le lundi 24 avril 2017 à 09:35 -0600, Stephen Warren a écrit :
> On 04/24/2017 09:07 AM, Paul Kocialkowski wrote:
> > Le mercredi 19 avril 2017 à 16:00 -0600, Stephen Warren a écrit :
> > > On 04/18/2017 10:38 AM, Paul Kocialkowski wrote:
> > > > Le mardi 18 avril 2017 à 10:15 -0600, Stephen Warren a écrit :
> > > > > On 04/18/2017 09:11 AM, Paul Kocialkowski wrote:
> > > > > > This selects the tegra30 i2s and ahub controllers for the tegra124
> > > > > > SoC.
> > > > > > These are needed when building without ARCH_TEGRA_3x_SOC set.
> > > > > > diff --git a/sound/soc/tegra/Kconfig b/sound/soc/tegra/Kconfig
> > > > > > index efbe8d4c019e..bcd18d2cf7a7 100644
> > > > > > --- a/sound/soc/tegra/Kconfig
> > > > > > +++ b/sound/soc/tegra/Kconfig
> > > > > > @@ -46,7 +46,7 @@ config SND_SOC_TEGRA20_SPDIF
> > > > > >
> > > > > >  config SND_SOC_TEGRA30_AHUB
> > > > > >   tristate
> > > > > > - depends on SND_SOC_TEGRA && ARCH_TEGRA_3x_SOC
> > > > > > + depends on SND_SOC_TEGRA && (ARCH_TEGRA_3x_SOC ||
> > > > > > ARCH_TEGRA_124_SOC)
> > > > >
> > > > > Is this really a compile-time dependency?
> > > >
> > > > From a quick look at the code, I doubt this is really a build
> > > > dependency.
> > > >
> > > > > If so, don't we need to add T210 and T186 entries into that ||
> > > > > condition
> > > > > too,
> > > > > since we could be building a kernel with just T210/T186 support and no
> > > > > T124
> > > > > support?
> > > >
> > > > In the spirit of this patch, adding entries for other tegra platforms
> > > > would
> > > > make
> > > > sense. Would you prefer that we leave out the dependency from
> > > > SND_SOC_TEGRA30_*
> > > > and only select the right I2S driver to use in each codec driver?
> > > >
> > > > If not, we'd have to list all relevant platforms both in the I2S/AHUB
> > > > drivers
> > > > and in each codec's rules (which is not necessarily and issue, but
> > > > there's
> > > > no
> > > > need to have artificial platform dependencies).
> > > >
> > > > What do you think?
> > >
> > > I think we should just remove most of these "depends on" since they're
> > > mostly set up to reflect runtime requirements rather than build time
> > > requirements. The only points I'd make are:
> >
> > I definitely agree we should do that for all the codec Kconfig options.
> >
> > > 1)
> > >
> > > Everything should "depends on SND_SOC_TEGRA" simply so the options don't
> > > show up and clutter menuconfig menus unless SND_SOC_TEGRA is enabled.
> >
> > Agreed.
> >
> > > 2)
> > >
> > > SND_SOC_TEGRA30_I2S does need the Tegra30 AHUB driver in order to
> > > compile/link, since it directly calls functions in that driver. This is
> > > already handled by SND_SOC_TEGRA30_I2S doing "select
> > > SND_SOC_TEGRA30_AHUB".
> >
> > Agreed.
> >
> > > 3)
> > >
> > > The machine drivers all do e.g. "select SND_SOC_TEGRA30_I2S if
> > > ARCH_TEGRA_3x_SOC". This was an attempt to make the machine drivers only
> > > pull in the relevant drivers for the SoC(s) being compiled for. I'm not
> > > sure this still makes sense; this won't work on kernels that only
> > > support T124/T210/T186 since ARCH_TEGRA_3x_SOC isn't enabled then.
> > > Should we just remove all those and make sure the defconfigs are updated
> > > to make sure the relevant I2S/AHUB/SPDIF/AC97 drivers are explicitly
> > > enabled? Perhaps we should default all the I2S/AHUB/SPDIF/AC97 to y
> > > (which will only apply if SND_SOC_TEGRA is enabled)?
> >
> > I think it would be easier for everyone to just auto-select the machine
> > drivers
> > automatically based on the architecture (so we could have the list of
> > ARCH_TEGRA_*_SOC here) when SND_SOC_TEGRA is selected.
>
> I don't think selecting the machine drivers is the correct approach, 
> since then they can't be disabled.
>
> Making certain machine drivers "default y if ARCH_TEGRA_nn_SOC" would 
> address that,

That's right, my mistake. Let's take that as the solution I'm backing then.

> but still isn't very scalable since we need to go back and 
> edit the Kconfig every time we define a new SoC, in order to add that 
> SoC into the default statement.

Well, that's what platform bringup is all about, isn't it? I think it makes a
lot more sense to have to add a new platform once (and it's not like one will
forget to look at the sound part when adding a new platform) rather than
requiring users to hand-pick the option.

> > Not only does this preserve existing configs (including external ones that
> > aren't part of the kernel tree), it also clearly maps which machine driver
> > to use for which SoC instead of having users do it by hand.
>
> The machine drivers aren't terribly tied to SoCs by design; most of them 
> would work on pretty much any SoC. They're only tied to SoCs as a 
> side-effect of a machine driver being tied to a certain CODEC, and 
> certain CODECS just by chance are only used (so far) on specific boards, 
> which have specific SoCs.

I'm a bit confused: aren't the machine driver (i2s/ahub/spdif/ac97) tied to
specific hardware blocks that are found in specific SoCs and not in others? I
can see these blocks haven't evolved much across generations, but they're still
either part of a specific SoC or not, aren't they?

The compatible strings in the common SoC dts seem to indicate that only one of
these blocks is found at a time.

--
Paul Kocialkowski, developer of free digital technology and hardware support

Website: https://www.paulk.fr/
Coding blog: https://code.paulk.fr/
Git repositories: https://git.paulk.fr/ https://git.code.paulk.fr/[unhandled content-type:application/pgp-signature]
\
 
 \ /
  Last update: 2017-04-24 20:44    [W:0.085 / U:5.252 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site