The lowest closest multiplier/divisor combination should be used for the PLL, not the largest. Reverse the search order. Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
@@ -681,9 +681,9 @@ cyber2000fb_decode_clock(struct par_info *hw, struct cfb_info *cfb,
* pll_ps_calc = best_div1 / (ref_ps * best_mult)
*/
best_diff = 0x7fffffff;
- best_mult = 32;
- best_div1 = 255;
- for (t_div1 = 32; t_div1 > 1; t_div1 -= 1) {
+ best_mult = 2;
+ best_div1 = 32;
+ for (t_div1 = 2; t_div1 < 32; t_div1 += 1) {
u_int rr, t_mult, t_pll_ps;
int diff;