Re: [PATCH V6 14/21] clk: tegra210: Add suspend and resume support

From: Stephen Boyd <sboyd@kernel.org>
To: Dmitry Osipenko <digetx@gmail.com>,
	Sowjanya Komatineni <skomatineni@nvidia.com>,
	jason@lakedaemon.net, jonathanh@nvidia.com,
	linus.walleij@linaro.org, marc.zyngier@arm.com,
	mark.rutland@arm.com, stefan@agner.ch, tglx@linutronix.de,
	thierry.reding@gmail.com
Cc: pdeschrijver@nvidia.com, pgaikwad@nvidia.com,
	linux-clk@vger.kernel.org, linux-gpio@vger.kernel.org,
	jckuo@nvidia.com, josephl@nvidia.com, talho@nvidia.com,
	linux-tegra@vger.kernel.org, linux-kernel@vger.kernel.org,
	mperttunen@nvidia.com, spatra@nvidia.com, robh+dt@kernel.org,
	devicetree@vger.kernel.org
Subject: Re: [PATCH V6 14/21] clk: tegra210: Add suspend and resume support
Date: Fri, 02 Aug 2019 10:51:18 -0700	[thread overview]
Message-ID: <20190802175119.1E401217F5@mail.kernel.org> (raw)
In-Reply-To: <8c259511-d8ea-51b2-0b1d-c85b964bc44c@gmail.com>

Quoting Dmitry Osipenko (2019-07-22 00:12:17)
> 22.07.2019 10:09, Dmitry Osipenko пишет:
> > 22.07.2019 9:52, Sowjanya Komatineni пишет:
> >>
> >> On 7/21/19 11:10 PM, Dmitry Osipenko wrote:
> >>> 22.07.2019 1:45, Sowjanya Komatineni пишет:
> >>>> On 7/21/19 2:38 PM, Dmitry Osipenko wrote:
> >>>>> 21.07.2019 22:40, Sowjanya Komatineni пишет:
> >>>>>> @@ -2853,9 +2859,8 @@ static int tegra210_enable_pllu(void)
> >>>>>>        reg |= PLL_ENABLE;
> >>>>>>        writel(reg, clk_base + PLLU_BASE);
> >>>>>>    -    readl_relaxed_poll_timeout_atomic(clk_base + PLLU_BASE, reg,
> >>>>>> -                      reg & PLL_BASE_LOCK, 2, 1000);
> >>>>>> -    if (!(reg & PLL_BASE_LOCK)) {
> >>>>>> +    ret = tegra210_wait_for_mask(&pllu, PLLU_BASE, PLL_BASE_LOCK);
> >>>>>> +    if (ret) {
> >>>>> Why this is needed? Was there a bug?
> >>>>>
> >>>> during resume pllu init is needed and to use same terga210_init_pllu,
> >>>> poll_timeout_atomic can't be used as its ony for atomic context.
> >>>>
> >>>> So changed to use wait_for_mask which should work in both cases.
> >>> Atomic variant could be used from any context, not sure what do you
> >>> mean. The 'atomic' part only means that function won't cause scheduling
> >>> and that's it.
> >>
> >> Sorry, replied incorrect. readx_poll_timeout_atomic uses ktime_get() and
> >> during resume timekeeping suspend/resume happens later than clock
> >> suspend/resume. So using tegra210_wait_for_mask.
> >>
> >> both timekeeping and clk-tegra210 drivers are registered as syscore but
> >> not ordered.
> > 
> > Okay, thank you for the clarification.
> > 
> > [snip]
> > 
> 
> You should remove the 'iopoll.h' then, since it's not used anymore.

And also add a comment to this location in the code because it's
non-obvious that we can't use iopoll here.