@@ -480,6 +480,7 @@ bool dss_div_calc(unsigned long fck_min, dss_div_calc_func func, void *data)
unsigned long fck_hw_max;
unsigned long fckd_hw_max;
unsigned long prate;
+ unsigned m;
if (dss.dpll4_m4_ck == NULL) {
/*
@@ -495,15 +496,16 @@ bool dss_div_calc(unsigned long fck_min, dss_div_calc_func func, void *data)
fck_hw_max = dss_feat_get_param_max(FEAT_PARAM_DSS_FCK);
fckd_hw_max = dss.feat->fck_div_max;
- prate = dss_get_dpll4_rate() * dss.feat->dss_fck_multiplier;
+ m = dss.feat->dss_fck_multiplier;
+ prate = dss_get_dpll4_rate();
fck_min = fck_min ? fck_min : 1;
- fckd_start = min(prate / fck_min, fckd_hw_max);
- fckd_stop = max(DIV_ROUND_UP(prate, fck_hw_max), 1ul);
+ fckd_start = min(prate * m / fck_min, fckd_hw_max);
+ fckd_stop = max(DIV_ROUND_UP(prate * m, fck_hw_max), 1ul);
for (fckd = fckd_start; fckd >= fckd_stop; --fckd) {
- fck = prate / fckd;
+ fck = prate / fckd * m;
if (func(fckd, fck, data))
return true;
@@ -521,7 +523,8 @@ int dss_set_clock_div(struct dss_clock_info *cinfo)
prate = clk_get_rate(clk_get_parent(dss.dpll4_m4_ck));
DSSDBG("dpll4_m4 = %ld\n", prate);
- r = clk_set_rate(dss.dpll4_m4_ck, prate / cinfo->fck_div);
+ r = clk_set_rate(dss.dpll4_m4_ck,
+ DIV_ROUND_UP(prate, cinfo->fck_div));
if (r)
return r;
} else {
@@ -531,7 +534,9 @@ int dss_set_clock_div(struct dss_clock_info *cinfo)
dss.dss_clk_rate = clk_get_rate(dss.dss_clk);
- WARN_ONCE(dss.dss_clk_rate != cinfo->fck, "clk rate mismatch");
+ WARN_ONCE(dss.dss_clk_rate != cinfo->fck,
+ "clk rate mismatch: %lu != %lu", dss.dss_clk_rate,
+ cinfo->fck);
DSSDBG("fck = %ld (%d)\n", cinfo->fck, cinfo->fck_div);
DSS func clock is calculated with prate / div * m. However, the current omapdss code calculates it with prate * m / div, which yields a slightly different result when there's a remainder. For example, 432000000 / 14 * 2 = 61714284, but 432000000 * 2 / 14 = 61714285. In addition to that, the clock framework wants the clock rate given with clk_set_rate to be higher than the actual (truncated) end result. So, if prate is 432000000, and div is 14, the real result is 30857142.8571... We need to call clk_set_rate with 30857143, which gives us a clock of 30857142. That's why we need to use DIV_ROUND_UP() when calling clk_set_rate. This patch fixes the clock calculation. Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ti.com> --- drivers/video/omap2/dss/dss.c | 17 +++++++++++------ 1 file changed, 11 insertions(+), 6 deletions(-)