linux-kernel - Re: [PATCH 21/31] soc/tegra: Remove the call to devm_pm_opp_set

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20220624002805.anv62ufihdrncwus@vireshk-i7>
Date:   Fri, 24 Jun 2022 05:58:05 +0530
From:   Viresh Kumar <viresh.kumar@...aro.org>
To:     Dmitry Osipenko <digetx@...il.com>,
        Jon Hunter <jonathanh@...dia.com>
Cc:     Thierry Reding <thierry.reding@...il.com>,
        linux-pm@...r.kernel.org,
        Vincent Guittot <vincent.guittot@...aro.org>,
        Rafael Wysocki <rjw@...ysocki.net>,
        Stephen Boyd <sboyd@...nel.org>, Nishanth Menon <nm@...com>,
        Krzysztof Kozlowski <krzysztof.kozlowski@...aro.org>,
        linux-tegra@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH 21/31] soc/tegra: Remove the call to
 devm_pm_opp_set_clkname()

On 23-06-22, 22:15, Jon Hunter wrote:
> On 26/05/2022 12:42, Viresh Kumar wrote:
> > diff --git a/drivers/soc/tegra/common.c b/drivers/soc/tegra/common.c
> > index 32c346b72635..49a5360f4507 100644
> > --- a/drivers/soc/tegra/common.c
> > +++ b/drivers/soc/tegra/common.c
> > @@ -108,12 +108,6 @@ int devm_tegra_core_dev_init_opp_table(struct device *dev,
> >   	u32 hw_version;
> >   	int err;
> > -	err = devm_pm_opp_set_clkname(dev, NULL);
> > -	if (err) {
> > -		dev_err(dev, "failed to set OPP clk: %d\n", err);
> > -		return err;
> > -	}
> > -
> >   	/* Tegra114+ doesn't support OPP yet */
> >   	if (!of_machine_is_compatible("nvidia,tegra20") &&
> >   	    !of_machine_is_compatible("nvidia,tegra30"))
> 
> 
> This appears to be breaking a few Tegra drivers. For example, on Tegra210
> Jetson TX1 I am seeing the following and the eMMC is no longer working ...
> 
> [    0.526729] sdhci-tegra 700b0600.mmc: dev_pm_opp_set_rate: device's opp table doesn't exist
> [    0.526733] sdhci-tegra 700b0600.mmc: failed to set clk rate to 400000Hz: -19
> [    0.528830] sdhci-tegra 700b0600.mmc: dev_pm_opp_set_rate: device's opp table doesn't exist
> [    0.528833] sdhci-tegra 700b0600.mmc: failed to set clk rate to 400000Hz: -19
> 
> I have seen another instance of this on Jetson Xavier NX ...
> 
> [   12.301336] tegra-pwm 32d0000.pwm: dev_pm_opp_set_rate: device's opp table doesn't exist
> [   12.301350] tegra-pwm 32d0000.pwm: Failed to set max frequency: -19
> 
> Bisect is point to this commit and so something is not working as
> expected.

Thanks again Jon.

This is what happens when the special code doesn't have a comment
attached with it. Neither the reviewer, nor the author remember why
the special piece was required :)

I had to go through the whole sequence, along with DT to understand
what might have broken this stuff :)

I will drop this patch and add this comment in its place:

diff --git a/drivers/soc/tegra/common.c b/drivers/soc/tegra/common.c
index 32c346b72635..9f3fdeb1a11c 100644
--- a/drivers/soc/tegra/common.c
+++ b/drivers/soc/tegra/common.c
@@ -108,6 +108,13 @@ int devm_tegra_core_dev_init_opp_table(struct device *dev,
        u32 hw_version;
        int err;

+       /*
+        * For some devices we don't have any OPP table in the DT, and in order
+        * to use the same code path for all the devices, we create a dummy OPP
+        * table for them via this call. The dummy OPP table is only capable of
+        * doing clk_set_rate() on invocation of dev_pm_opp_set_rate() and
+        * doesn't provide any other functionality.
+        */
        err = devm_pm_opp_set_clkname(dev, NULL);
        if (err) {
                dev_err(dev, "failed to set OPP clk: %d\n", err);


Though there will still be a problem here with my changes, we don't
accept NULL clkname anymore for the set-clkname API. And tegra does
this to pick the first clock available in DT (at index 0) I think.
Other drivers (mostly qcom) who need such dummy OPP table, provide a
real clock name instead. Will it be possible to pass that here somehow
?

-- 
viresh