[PATCH v2 0/1] clk: nuvoton: ma35d1-divider: use direct formula instead of divider table

[Joey Lu]

The MA35D1 ADC clock divider register implements a simple closed-form
relation, rate = parent_rate / (2 * (N + 1)), but the driver models it
by building a clk_div_table with one entry per possible divider value
(up to 2^width entries) and feeding it through the generic
divider_recalc_rate()/divider_determine_rate()/divider_get_val() helpers.
Replace the table with direct arithmetic in recalc_rate()/determine_rate()/
set_rate(), and drop the unused mask_bit mechanism, whose only call site
passed a bitmask instead of a bit index, causing an out-of-range BIT()
shift.

Changes in v2 (addressing the Sashiko AI review of v1):
- Fix a possible division by zero: 2 * rate in determine_rate() and
  set_rate() could overflow for very large requested rates, making the
  divisor zero. Rates above parent_rate / 2 now short-circuit to a
  divider of 1 before the multiplication.
- Fix the mismatch between determine_rate() and set_rate(): both used
  a ceiling to pick the divider, yet determine_rate() reported a
  closest-rounded rate, so set_rate() could program a different divider
  than the one promised. Both now use closest rounding, as the original
  CLK_DIVIDER_ROUND_CLOSEST code did, via a single shared helper
  (ma35d1_clkdiv_get_val()) that also clamps the divider to its valid
  range.
- Not changed: min_rate/max_rate are not checked in the driver, since
  the clk core already applies them and the previous
  divider_determine_rate() based code did not either.

Joey Lu (1):
  clk: nuvoton: ma35d1-divider: use direct formula instead of divider
    table

 drivers/clk/nuvoton/clk-ma35d1-divider.c | 79 ++++++++++++------------
 drivers/clk/nuvoton/clk-ma35d1.c         |  2 +-
 drivers/clk/nuvoton/clk-ma35d1.h         |  2 +-
 3 files changed, 42 insertions(+), 41 deletions(-)

-- 
2.43.0