Re: [PATCH V2 1/2] opp: Allow dev_pm_opp_get_opp_table() to return -EPROBE_DEFER

From: Viresh Kumar <viresh.kumar@linaro.org>
To: Sudeep Holla <sudeep.holla@arm.com>
Cc: ulf.hansson@linaro.org, "Rafael J. Wysocki" <rjw@rjwysocki.net>,
	Kevin Hilman <khilman@kernel.org>, Pavel Machek <pavel@ucw.cz>,
	Len Brown <len.brown@intel.com>,
	Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
	Viresh Kumar <vireshk@kernel.org>, Nishanth Menon <nm@ti.com>,
	Stephen Boyd <sboyd@kernel.org>, Kukjin Kim <kgene@kernel.org>,
	Krzysztof Kozlowski <krzk@kernel.org>,
	linux-pm@vger.kernel.org,
	Vincent Guittot <vincent.guittot@linaro.org>,
	nks@flawful.org, georgi.djakov@linaro.org,
	Stephan Gerhold <stephan@gerhold.net>,
	linux-kernel@vger.kernel.org,
	linux-arm-kernel@lists.infradead.org,
	linux-samsung-soc@vger.kernel.org
Subject: Re: [PATCH V2 1/2] opp: Allow dev_pm_opp_get_opp_table() to return -EPROBE_DEFER
Date: Tue, 20 Oct 2020 15:29:35 +0530	[thread overview]
Message-ID: <20201020095935.2ttoked5tozvcr24@vireshk-i7> (raw)
In-Reply-To: <20201020095205.GB10604@bogus>

On 20-10-20, 10:52, Sudeep Holla wrote:
> On Tue, Oct 20, 2020 at 03:11:34PM +0530, Viresh Kumar wrote:
> > On 20-10-20, 10:37, Sudeep Holla wrote:
> > > On Tue, Oct 20, 2020 at 11:24:32AM +0530, Viresh Kumar wrote:
> > > > On 20-10-20, 10:35, Viresh Kumar wrote:
> > > > > On 19-10-20, 15:10, Sudeep Holla wrote:
> > > > > > On Mon, Oct 19, 2020 at 04:05:35PM +0530, Viresh Kumar wrote:
> > > > > > > On 19-10-20, 11:12, Sudeep Holla wrote:
> > > > > > > > Yes it has clocks property but used by SCMI(for CPUFreq/DevFreq) and not
> > > > > > > > by any clock provider driver. E.g. the issue you will see if "clocks"
> > > > > > > > property is used instead of "qcom,freq-domain" on Qcom parts.
> > > > > > > 
> > > > > > > Okay, I understand. But what I still don't understand is why it fails
> > > > > > > for you. You have a clocks property in DT for the CPU, the OPP core
> > > > > > > tries to get it and will get deferred-probed, which will try probing
> > > > > > > at a later point of time and it shall work then. Isn't it ?
> > > > > > >
> > > > > > 
> > > > > > Nope unfortunately. We don't have clock provider, so clk_get will
> > > > > > never succeed and always return -EPROBE_DEFER.
> > > > > 
> > > > > Now this is really bad, you have a fake clocks property, how is the
> > > > > OPP core supposed to know it ? Damn.
> > > > 
> > > > What about instead of fixing the OPP core, which really is doing the
> > > > right thing, we fix your driver (as you can't fix the DT) and add a
> > > > dummy CPU clk to make it all work ?
> > > >
> > > 
> > > I really would avoid that. I would rather change the binding as there is
> > > no single official users of that binding in the upstream tree.
> > 
> > But how will you solve backward compatibility thing then ?
> > 
> 
> I am just betting on the fact that no users upstream means no backward
> compatibility needed. If someone raises issue we need to add backward
> compatibility with dummy clk as you suggested.

Okay. I would have done a change in the OPP core to fix the issue, but
the current code looks correct and we shouldn't change it to satisfy
buggy users. I hope that makes sense.

-- 
viresh