Re: [v5, PATCH 4/5] cpufreq: mediatek: add opp notification for SVS support

[andrew-sh.cheng <andrew-sh.cheng@mediatek.com>]

On Wed, 2020-03-11 at 11:36 +0530, Viresh Kumar wrote:
> On 10-03-20, 16:11, andrew-sh.cheng wrote:
> > On Tue, 2019-12-10 at 14:43 +0800, Viresh Kumar wrote:
> > > On 09-12-19, 14:56, andrew-sh.cheng wrote:
> > > > On Wed, 2019-11-27 at 14:06 +0530, Viresh Kumar wrote:
> > > > > On 26-11-19, 19:50, Andrew-sh.Cheng wrote:
> > > > > > +		if (!IS_ERR(opp_item))
> > > > > > +			dev_pm_opp_put(opp_item);
> > > > > > +		else
> > > > > > +			freq = 0;
> > > > > > +
> > > > > 
> > > > > What is the purpose of the above code ?
> > > > When dev_pm_opp_find_freq_ceil() doesn't find matching opp item, freq
> > > > value won't be set.
> > > > Set it as 0 for below checking
> > > > > 
> > > > > > +		/* case of current opp is disabled */
> > > > > > +		if (freq == 0 || freq != info->opp_freq) {
> > > > > > +			// find an enable opp item
> > > > > > +			freq = 1;
> > > > > > +			opp_item = dev_pm_opp_find_freq_ceil(info->cpu_dev,
> > > > > > +							     &freq);
> > > > > > +			if (!IS_ERR(opp_item)) {
> > > > > > +				dev_pm_opp_put(opp_item);
> > > > > > +				policy = cpufreq_cpu_get(info->opp_cpu);
> > > > > > +				if (policy) {
> > > > > > +					cpufreq_driver_target(policy,
> > > > > > +						freq / 1000,
> > > > > > +						CPUFREQ_RELATION_L);
> > > > > 
> > > > > Why don't you simply call this instead of all the code in the else
> > > > > block ?
> > > > These else code is used to check "current opp item is disabled or not".
> > > > If not, do nothing.
> > > > If current opp item is disabled, need to find an not-disabled opp item,
> > > > and set frequency to it.
> > > 
> > > Right. So this notifier helper of yours receive the opp which is getting
> > > disabled, why don't you compare its frequency directly to see if the current OPP
> > > is getting disabled ?
> > Sorry to overlook your question.
> > This is because when the opp is disabled,
> > we cannot use dev_pm_opp_get_freq() to get frequency of that opp.
> > There is a check:
> > 	if (IS_ERR_OR_NULL(opp) || !opp->available) {
> 
> I think we can remove the available check here, as we are jut trying
> to find frequency of an OPP we already have. Send a patch for that
> please.
> 

Hi Viresh,
I have something want to consult you.
For your previous comment, you suggest use read-write lock to replace
mutex lock.
Will it be more efficiently even when all are write lock?
(all lock region are "setting VProc voltage")

> > 		pr_err("%s: Invalid parameters\n", __func__);
> > 		return 0;
> > 
> > > 
> > 
> 

_______________________________________________
Linux-mediatek mailing list
Linux-mediatek@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-mediatek