Re: [PATCH] drm: pl111: add CONFIG_VEXPRESS_CONFIG dependency

From: Arnd Bergmann
Date: Wed May 27 2020 - 15:03:53 EST


On Wed, May 27, 2020 at 7:50 PM Sam Ravnborg <sam@xxxxxxxxxxxx> wrote:
> On Wed, May 27, 2020 at 05:47:21PM +0200, Arnd Bergmann wrote:
> > On Wed, May 27, 2020 at 4:52 PM Sam Ravnborg <sam@xxxxxxxxxxxx> wrote:
> > >
> > > Hi Arnd.
> > >
> > > On Wed, May 27, 2020 at 03:31:42PM +0200, Arnd Bergmann wrote:
> > > > The vexpress_config code fails to link in some configurations:
> > > >
> > > > drivers/gpu/drm/pl111/pl111_versatile.o: in function `pl111_versatile_init':
> > > > (.text+0x1f0): undefined reference to `devm_regmap_init_vexpress_config'
> > > >
> > > > Add a dependency that links to this only if the dependency is there,
> > > > and prevent the configuration where the drm driver is built-in but
> > > > the config is a loadable module.
> > > >
> > > > Fixes: 826fc86b5903 ("drm: pl111: Move VExpress setup into versatile init")
> > > > Signed-off-by: Arnd Bergmann <arnd@xxxxxxxx>
> > >
> > > Could this be another way to fix it:
> > >
> > > diff --git a/drivers/gpu/drm/pl111/pl111_versatile.c b/drivers/gpu/drm/pl111/pl111_versatile.c
> > > index 64f01a4e6767..1c38d3bd2e84 100644
> > > --- a/drivers/gpu/drm/pl111/pl111_versatile.c
> > > +++ b/drivers/gpu/drm/pl111/pl111_versatile.c
> > > @@ -379,7 +379,7 @@ static int pl111_vexpress_clcd_init(struct device *dev, struct device_node *np,
> > > u32 val;
> > > int ret;
> > >
> > > - if (!IS_ENABLED(CONFIG_VEXPRESS_CONFIG))
> > > + if (!IS_REACHABLE(CONFIG_VEXPRESS_CONFIG))
> > > return -ENODEV;
> > >
> > > /*
> > >
> > >
> > > Then we no longer have the whole driver depending on
> > > the value of VEXPRESS_CONFIG.
> > > Not that I like IS_REACHABLE() but we already had
> > > IS_ENABLED() to cover up here, and that was not enough.
> > >
> > > With your patch would we then need the IS_ENABLED()
> > > check?
> >
> > The IS_ENABLED() check is what I'm adding, not removing. I'd still
> > the Kconfig dependency combined with that check over
> > IS_REACHABLE(), which is more likely to silently not work.
>
> Then the now redundant IS_ENABLED() check should go.
> With you patch it looks like this:
>
> ...
> if (IS_ENABLED(CONFIG_VEXPRESS_CONFIG) && ...)
> pl111_vexpress_clcd_init()
>
>
> And in pl111_vexpress_clcd_init() we have:
>
> {
> if (!IS_ENABLED(CONFIG_VEXPRESS_CONFIG))
> return -ENODEV;
>
> The IS_ENABLED() in pl111_vexpress_clcd_init() is redundant
> and the patch should drop it.

Ah I see your point now, sorry I missed the double IS_ENABLED()
check at first. I'll remove the second one from my patch and
resubmit after some more build testing then.

Arnd