Re: [PATCH v16 6/9] clk: realtek: Add support for MMC-tuned PLL clocks

From: Brian Masney

Date: Wed Nov 18 2026 - 21:24:03 EST


Hi Yu-Chun,

On Fri, Nov 13, 2026 at 02:12:59PM +0000, Yu-Chun Lin [林祐君] wrote:
> > > +static const struct rtk_clk_pll_mmc_rate *
> > > +rtk_clk_pll_mmc_find_closest_rate(const struct rtk_clk_pll_mmc_rate *tbl,
> > u32 num,
> > > + unsigned long rate, unsigned long
> > min_rate,
> > > + unsigned long max_rate) {
> > > + const struct rtk_clk_pll_mmc_rate *best = NULL;
> > > + unsigned long best_delta = ~0UL;
> > > +
> > > + for (unsigned int i = 0; i < num; i++) {
> > > + const struct rtk_clk_pll_mmc_rate *entry = &tbl[i];
> > > + unsigned long delta;
> > > +
> > > + if (entry->rate < min_rate || entry->rate > max_rate)
> > > + continue;
> > > +
> > > + delta = entry->rate >= rate ? entry->rate - rate : rate -
> > entry->rate;
> > > + if (delta < best_delta) {
> > > + best = entry;
> > > + best_delta = delta;
> > > + }
> > > + }
> > > +
> > > + return best;
> > > +}
> >
> > This duplicates the functionality in ftbl_find_closest_by_rate() from patch 3.
> >
>
> I considered sharing the logic, but the two tables have different entry
> layouts.
>
> Unifying them would need a macro or (void *) (both lose type checking), or
> embedding rtk_freq_table with container_of() (a larger rework for little
> gain). I prefer keeping the two type-safe functions. Is that acceptable?

What about this? This is completely untested.

#define ftbl_find_closest_by_rate(_tbl, _rate, _min, _max) ({ \
typeof(&(_tbl)[0]) __best = NULL, __e; \
unsigned long __r = (_rate), __lo = (_min), __hi = (_max); \
unsigned long __best_delta = ULONG_MAX; \
\
for (__e = (_tbl); __e->rate; __e++) { \
unsigned long __d; \
\
if (__e->rate < __lo || __e->rate > __hi) \
continue; \
\
__d = abs_diff(__e->rate, __r); \
if (__d < __best_delta) { \
__best_delta = __d; \
__best = __e; \
} \
} \
__best; \
})

Brian