Hi Stephen,
On Wed, Sep 4, 2019 at 12:09 AM Stephen Boyd <sboyd@xxxxxxxxxx> wrote:
> Quoting Geert Uytterhoeven (2019-08-30 06:45:07)
> > As the .round_rate() callback returns a long clock rate, it cannot
> > return clock rates that do not fit in signed long, but do fit in
> > unsigned long. The newer .determine_rate() callback does not suffer
> > from this limitation. In addition, .determine_rate() provides the
> > ability to specify a rate range.
> >
> > This patch series performs the customary preparatory cleanups, and
> > switches the Z (CPU) and SD clocks in the R-Car Gen2 and Gen3 clock
> > drivers from the .round_rate() to the .determine_rate() callback.
> > Note that the "div6" clock driver hasn't been converted yet, so div6
> > clocks still use .round_rate().
> >
> > Changes compared to v1[1]:
> > - Add preparatory arithmetic division improvements
> > - Split off cpg_sd_clock_calc_div() absorption and SD clock best rate
> > calculation,
> > - Use div_u64() for division by unsigned long,
> >
> > This has been tested on R-Car M2-W and various R-Car Gen3, and should
> > have no behavioral impact.
>
> From what I recall the rate range code is broken but I can't remember
> how. Anyway, I was just curious if you ran into any issues with that
> code.
I didn't ran into any issues. But please note that in all tested cases, the
limits were 0 and ULONG_MAX anyway, so probably it didn't trigger the
broken cases in the rate range code.
So, is it good to have .determine_rate() support in individual clock drivers
now, or do you want me to postpone the last 3 patches of my series until the
rate range code is fixed?
Thanks!
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
