Re: [PATCH] ARM: tegra: Enable CPUFreq support for Tegra124 Chromebooks

[Jon Hunter <jonathanh-DDmLM1+adcrQT0dZR+AlfA@public.gmane.org>]

On 04/11/15 13:23, Jon Hunter wrote:
> 
> On 04/11/15 12:19, Jon Hunter wrote:
>> Correcting Mikko's email ...
>>
>> On 04/11/15 12:16, Jon Hunter wrote:
>>>
>>> On 15/09/15 10:00, Tomeu Vizoso wrote:
>>>> On 15 September 2015 at 10:12, Thierry Reding <thierry.reding-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org> wrote:
>>>>> On Thu, Sep 03, 2015 at 03:40:45PM +0200, Tomeu Vizoso wrote:
>>>>>> On 13 July 2015 at 15:08, Jon Hunter <jonathanh-DDmLM1+adcrQT0dZR+AlfA@public.gmane.org> wrote:
>>>>>>> On 03/06/15 12:44, Jon Hunter wrote:
>>>>>>>> Adding LAKML. Jon
>>>>>>>>
>>>>>>>> On 03/06/15 12:43, Jon Hunter wrote:
>>>>>>>>> Add the device-tree DFLL clock node and CPU regulator phandle for
>>>>>>>>> tegra124 chromebooks to enable CPUFreq support on these boards.
>>>>>>
>>>>>> Ping.
>>>>>>
>>>>>> Thanks,
>>>>>
>>>>> Sorry that this has gone unnoticed for so long. I've applied it now to
>>>>> the for-4.4/dt branch.
>>>>>
>>>>> Tomeu, do you want me to add your Tested-by, Reviewed-by or Acked-by
>>>>> before I push this out?
>>>>
>>>> I haven't done proper tests, but I have tested for several weeks a
>>>> branch containing these changes on a nyan-big and have found no
>>>> issues.
>>>>
>>>> You can add my Reviewed-by though. Sorry for not having made this clear.
>>>
>>> I have noticed that system suspend to LP1 is not working on the nyan-big
>>> with linux-next and never exits suspend. It appears that this patch is
>>> the culprit. I tested cpufreq was changing the frequency as expected but
>>> I did not test suspend.
> 
> By the way, LP2 works fine with this change, it is just when LP1 is
> enabled (which is the default).

Quick update on this ... it appears that the cause of the lock-up is
related to pll_x (default pll that clocks the cpu cluster before
switching to the dfll). On the jetson-tk1 the pll_x is running at 696MHz
where as on the nyan-big it is running at 2.1GHz. Although the pll is
disabled when switching to dfll, having it configured for 2.1GHz on the
nyan-big is causing a hang during suspend if the dfll is running at much
lower frequencies (>1.5GHz) than the pll_x. So seems like the CPU
voltage is too low for the pll_x, however, it should be disabled?!?

If I hack the cpufreq code to set the pll_x to 696MHz after switching to
the dfll on the nyan-big then the problem goes away ...

diff --git a/drivers/cpufreq/tegra124-cpufreq.c
b/drivers/cpufreq/tegra124-cpufreq.c
index 20bcceb58ccc..a87cef4071e8 100644
--- a/drivers/cpufreq/tegra124-cpufreq.c
+++ b/drivers/cpufreq/tegra124-cpufreq.c
@@ -53,6 +53,10 @@ static int tegra124_cpu_switch_to_dfll(struct
tegra124_cpufreq_priv *priv)

        clk_set_parent(priv->cpu_clk, priv->dfll_clk);

+       ret = clk_set_rate(priv->pllx_clk, 696000000);
+       if (ret)
+               return ret;
+
        return 0;

Need to figure out if the suspend code is touching the pll_x.

Jon