Hi Wolfram,
On Tue, Jun 30, 2020 at 9:45 PM Wolfram Sang <wsa@xxxxxxxxxxxxx> wrote:
> On Thu, Jun 25, 2020 at 05:16:58PM +0200, Wolfram Sang wrote:
> > I spend some more thoughts on this.
> >
> > > > > In general, pm_runtime_get_sync() is not safe to call from atomic
> > > > > context.
> > > > > For Renesas SoCs, I think both the power and clock domains are safe, as
> > > > > the respective drivers don't sleep. The PM core might, though.
> > > >
> > > > Still, that sounds to me like we should protect these calls as in V1?
> >
> > I still think we should guard these calls just because it is not safe to
> > call them from atomic contexts.
> >
> > > And talk to the i2c controller while it is disabled?
> >
> > Is there maybe some "always-on" property which we could add to the
> > respective IIC clock?
>
> Ping to this question...
You mean in DT? DT describes hardware, not software policy.
Anyway, won't help on R-Mobile A1, as there's a real power domain.
> > > That does seem to work on R-Car Gen2 (similar to SMP bringup accessing
> > > registers of a disabled WDT?), though.
> >
> > Yes. Uli's patch will not cause a regression because we are already
> > calling i2c_transfer very late. And we do call the runtime_pm functions
> > currently. So, it will improve the situation there.
> >
> > > Needs testing on R-Mobile A1....
> >
> > That's armadillo, right? I don't have that, sadly.
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
