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

[Arnd Bergmann]

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.

> - 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'.


     Arnd