Re: [PATCH] net: dsa: ocelot: add NET_VENDOR_MICROSEMI dependency

[Vladimir Oltean]

On Wed, 11 Dec 2019 at 00:04, Arnd Bergmann <arnd@xxxxxxxx> wrote:
>
> On Tue, Dec 10, 2019 at 10:37 PM Vladimir Oltean <olteanv@xxxxxxxxx> wrote:
> >
> > Hi Arnd,
> >
> > On Tue, 10 Dec 2019 at 22:37, Arnd Bergmann <arnd@xxxxxxxx> wrote:
> > >
> > > Selecting MSCC_OCELOT_SWITCH is not possible when NET_VENDOR_MICROSEMI
> > > is disabled:
> > >
> > > WARNING: unmet direct dependencies detected for MSCC_OCELOT_SWITCH
> > >   Depends on [n]: NETDEVICES [=y] && ETHERNET [=n] && NET_VENDOR_MICROSEMI [=n] && NET_SWITCHDEV [=y] && HAS_IOMEM [=y]
> > >   Selected by [m]:
> > >   - NET_DSA_MSCC_FELIX [=m] && NETDEVICES [=y] && HAVE_NET_DSA [=y] && NET_DSA [=y] && PCI [=y]
> > >
> > > Add a Kconfig dependency on NET_VENDOR_MICROSEMI, which also implies
> > > CONFIG_NETDEVICES.
> > >
> > > Fixes: 56051948773e ("net: dsa: ocelot: add driver for Felix switch family")
> > > Signed-off-by: Arnd Bergmann <arnd@xxxxxxxx>
> > > ---
> >
> > This has been submitted before, here [0].
> >
> > It isn't wrong, but in principle I agree with David that it is strange
> > to put a "depends" relationship between a driver in drivers/net/dsa
> > and the Kconfig vendor umbrella from drivers/net/ethernet/mscc ("why
> > would the user care/need to enable NET_VENDOR_MICROSEMI to see the DSA
> > driver" is a valid point to me). This is mainly because I don't
> > understand the point of CONFIG_NET_VENDOR_* options, they're a bit
> > tribalistic to my ears.
> >
> > Nonetheless, alternatives may be:
> > - Move MSCC_OCELOT_SWITCH core option outside of the
> > NET_VENDOR_MICROSEMI umbrella, and make it invisible to menuconfig,
> > just selectable from the 2 driver instances (MSCC_OCELOT_SWITCH_OCELOT
> > and NET_DSA_MSCC_FELIX). MSCC_OCELOT_SWITCH has no reason to be
> > selectable by the user anyway.
>
> You still need 'depends on NETDEVICES' in that case, otherwise this sounds
> like a good option.
>

I don't completely understand this. Looks like NETDEVICES is another
one of those options that don't enable any code. I would have expected
that NET_SWITCHDEV depended on it already? But anyway, it's still a
small compromise and not a problem.

> > - Remove NET_VENDOR_MICROSEMI altogether. There is a single driver
> > under drivers/net/ethernet/mscc and it's already causing problems,
> > it's ridiculous.
>
> It's only there for consistency with the other directories under
> drivers/net/ethernet/.
>
> > - Leave it as it is. I genuinely ask: if the build system tells you
> > that the build dependencies are not met, does it matter if it compiles
> > or not?
>
> We try very hard to allow all randconfig builds to complete without
> any output from the build process when building with 'make -s'.
> Random warnings like this just clutter up the output, even if it's
> harmless there is a risk of missing something important.
>
> Yet another option is
> - Change NET_DSA_MSCC_FELIX to use 'depends on
>   MSCC_OCELOT_SWITCH' instead of 'select NET_DSA_MSCC_FELIX'.
>

That's strange too. MSCC_OCELOT_SWITCH just enables a common driver
core which is used by both NET_DSA_MSCC_FELIX and
MSCC_OCELOT_SWITCH_OCELOT (possibly by more in the future). I don't
see any reason why the common library (purely an implementation
detail) should even be user-visible, let alone why it should hide a
DSA driver.

So, if you agree, I can take care of this tomorrow by reworking the
Kconfig in the 1st proposed way. I hope you don't mind that I'm
volunteering to do it, but the change will require a bit of explaining
which is non-trivial, and I don't expect that you really want to know
these details, just that it compiles with no issue from all angles.

>
>      Arnd

Thanks,
-Vladimir