Re: [PATCH v6 2/3] clk: qcom: clk-rcg2: add support for rcg2 freq multi ops

[Christian Marangi]

On Tue, Nov 21, 2023 at 12:01:15PM +0530, Devi Priya wrote:
> 
> 
> On 11/20/2023 11:44 PM, Christian Marangi wrote:
> > On Mon, Nov 20, 2023 at 03:51:50PM +0530, Devi Priya wrote:
> > > 
> > > 
> > > On 9/16/2023 7:30 PM, Christian Marangi wrote:
> > > > Some RCG frequency can be reached by multiple configuration.
> > > > 
> > > > Add clk_rcg2_fm_ops ops to support these special RCG configurations.
> > > > 
> > > > These alternative ops will select the frequency using a CEIL policy.
> > > > 
> > > > When the correct frequency is found, the correct config is selected by
> > > > calculating the final rate (by checking the defined parent and values
> > > > in the config that is being checked) and deciding based on the one that
> > > > is less different than the requested one.
> > > > 
> > > > These check are skipped if there is just on config for the requested
> > > > freq.
> > > > 
> > > > qcom_find_freq_multi is added to search the freq with the new struct
> > > > freq_multi_tbl.
> > > > __clk_rcg2_select_conf is used to select the correct conf by simulating
> > > > the final clock.
> > > > If a conf can't be found due to parent not reachable, a WARN is printed
> > > > and -EINVAL is returned.
> > > > 
> > > > Signed-off-by: Christian Marangi <ansuelsmth@xxxxxxxxx>
> > > > ---
> > > >    drivers/clk/qcom/clk-rcg.h  |   1 +
> > > >    drivers/clk/qcom/clk-rcg2.c | 167 ++++++++++++++++++++++++++++++++++++
> > > >    drivers/clk/qcom/common.c   |  18 ++++
> > > >    drivers/clk/qcom/common.h   |   2 +
> > > >    4 files changed, 188 insertions(+)
> > > > 
> > > > diff --git a/drivers/clk/qcom/clk-rcg.h b/drivers/clk/qcom/clk-rcg.h
> > > > index c81458db6ce4..dc9a77965e68 100644
> > > > --- a/drivers/clk/qcom/clk-rcg.h
> > > > +++ b/drivers/clk/qcom/clk-rcg.h
> > > > @@ -190,6 +190,7 @@ struct clk_rcg2_gfx3d {
> > > >    extern const struct clk_ops clk_rcg2_ops;
> > > >    extern const struct clk_ops clk_rcg2_floor_ops;
> > > > +extern const struct clk_ops clk_rcg2_fm_ops;
> > > >    extern const struct clk_ops clk_rcg2_mux_closest_ops;
> > > >    extern const struct clk_ops clk_edp_pixel_ops;
> > > >    extern const struct clk_ops clk_byte_ops;
> > > > diff --git a/drivers/clk/qcom/clk-rcg2.c b/drivers/clk/qcom/clk-rcg2.c
> > > > index e22baf3a7112..617e7ff0f6a3 100644
> > > > --- a/drivers/clk/qcom/clk-rcg2.c
> > > > +++ b/drivers/clk/qcom/clk-rcg2.c
> > > > @@ -266,6 +266,116 @@ static int _freq_tbl_determine_rate(struct clk_hw *hw, const struct freq_tbl *f,
> > > >    	return 0;
> > > >    }
> > > > +static const struct freq_conf *
> > > > +__clk_rcg2_select_conf(struct clk_hw *hw, const struct freq_multi_tbl *f,
> > > > +		       unsigned long req_rate)
> > > > +{
> > > > +	unsigned long rate_diff, best_rate_diff = ULONG_MAX;
> > > > +	const struct freq_conf *conf, *best_conf;
> > > > +	struct clk_rcg2 *rcg = to_clk_rcg2(hw);
> > > > +	const char *name = clk_hw_get_name(hw);
> > > > +	unsigned long parent_rate, rate;
> > > > +	struct clk_hw *p;
> > > > +	int index, i;
> > > > +
> > > > +	/* Init best_conf to the first conf */
> > > > +	best_conf = f->confs;
> > > > +
> > > > +	/* Exit early if only one config is defined */
> > > > +	if (f->num_confs == 1)
> > > > +		goto exit;
> > > > +
> > > > +	/* Search in each provided config the one that is near the wanted rate */
> > > > +	for (i = 0, conf = f->confs; i < f->num_confs; i++, conf++) {
> > > > +		index = qcom_find_src_index(hw, rcg->parent_map, conf->src);
> > > > +		if (index < 0)
> > > > +			continue;
> > > > +
> > > > +		p = clk_hw_get_parent_by_index(hw, index);
> > > > +		if (!p)
> > > > +			continue;
> > > > +
> > > > +		parent_rate =  clk_hw_get_rate(p);
> > > > +		rate = calc_rate(parent_rate, conf->n, conf->m, conf->n, conf->pre_div);
> > > > +
> > > > +		if (rate == req_rate) {
> > > > +			best_conf = conf;
> > > > +			goto exit;
> > > > +		}
> > > > +
> > > > +		rate_diff = abs(req_rate - rate);
> > > > +		if (rate_diff < best_rate_diff) {
> > > > +			best_rate_diff = rate_diff;
> > > > +			best_conf = conf;
> > > > +		}
> > > > +	}
> > > > +
> > > > +	/*
> > > > +	 * Very unlikely. Warn if we couldn't find a correct config
> > > > +	 * due to parent not found in every config.
> > > > +	 */
> > > > +	if (unlikely(i == f->num_confs)) {
> > > > +		WARN(1, "%s: can't find a configuration for rate %lu.",
> > > > +		     name, req_rate);
> > > > +		return ERR_PTR(-EINVAL);
> > > > +	}
> > > Hi Christian,
> > > 
> > > Thanks a lot for the patch!
> > > We have incorporated these changes along with the corresponding clock driver
> > > changes & tested it on IPQ9574 & IPQ5332 targets.
> > > 
> > > When setting the clk rate for the nss port clocks, for the requested
> > > frequency the correct config gets selected and the
> > > clk rate is set properly.
> > > We see the WARN getting printed for other frequencies (rate * i where
> > > i=2 to maxdiv) that is requested by the clk_hw_round_rate function.
> > > 
> > > Upon analysis, we see that the for loop in clk_divider_bestdiv iterates
> > > until the maxdiv value and requests (rate*i) via the clk_hw_round_rate
> > > API to find the bestdiv and best_parent_rate. For frequencies which are
> > > multiples of the requested frequency (rate*i where i=2 to maxdiv), it
> > > seems unlikely to see the WARN being printed.
> > > 
> > > Can you please help us understand when the WARN is likely to be printed
> > > & Looking forward to your suggestions on how this WARN could
> > > be suppressed in the afore mentioned scenario!
> > > 
> > 
> > Hi,
> > 
> > thanks a lot for testing this. Maybe was a small oversight by me.
> > 
> > I attached an alternative patch. Can you test it and tell me if the WARN
> > is still printed? (the WARN must be printed only when the parent is not
> > found, I don't think it's your case)
> > 
> Hi Christian,
> 
> WARN does not get printed with the attached patchset.
> Thanks much!
> 

Can you also confirm that the code correctly works with your div
scenario?

Would love to have some Tested-by tag to move this further!

(since it seems newer SoC after ipq807x will use this implementation
even more)

-- 
	Ansuel