Hi Neil,

On Fri, 30 Aug 2019 at 13:01, Neil Armstrong [off-list ref] wrote:

On 29/08/2019 20:35, Anand Moon wrote:

quoted

Hi Neil,

On Thu, 29 Aug 2019 at 13:58, Neil Armstrong [off-list ref] wrote:

quoted

On 28/08/2019 22:27, Anand Moon wrote:

quoted

Below small changes help re-configure or fix missing inter linking
of regulator node.

Changes based top on my prevoius series.

For the serie:
Reviewed-by: Neil Armstrong <redacted>

Thanks for your review.

quoted

quoted

[0] https://patchwork.kernel.org/cover/11113091/

TOOD: Add support for DVFS GXBB odroid board in next series.

I'm curious how you will do this !

I was just studying you previous series on how you have implemented
this feature for C1, N2 and VIM3 boards.

[0] https://patchwork.kernel.org/cover/11114125/

I started gathering key inputs needed for this ie *clk / pwm*
like VDDCPU and VDDE clk changes.

But it looks like of the complex clk framework needed, so I leave this to the
expert like your team of developers to do this much quick and efficiently.

On GXBB, GXL, GXM and AXG SoCs, CPU Frequency setting and PWM Regulator setup is
done by the SCPI Co-processor via the SCPI protocol.

Thus, we should not handle it from Linux, and even if we could, we don't have the
registers documentation of the CPU clusters clock tree.

Ok thanks.

SCPI works fine on all tested devices, except Odroid-C2, because Hardkernel left
the > 1.5GHz freq in the initial SCPI tables loaded by the BL2, i.e. packed with U-Boot.
Nowadays they have removed the bad frequencies, but still some devices uses the old
bootloader.

But in the SCPI case we trust the table returned by the firmware and use it as-in,
and there is no (simple ?) way to override the table and set a max frequency.

This is why we disabled SCPI.

See https://patchwork.kernel.org/patch/9500175/

I have quickly enable this on my board and here the cpufreq info

[alarm@alarm ~]$  cpupower frequency-info
analyzing CPU 0:
  driver: scpi-cpufreq
  CPUs which run at the same hardware frequency: 0 1 2 3
  CPUs which need to have their frequency coordinated by software: 0 1 2 3
  maximum transition latency: 200 us
  hardware limits: 100.0 MHz - 1.54 GHz
  available frequency steps:  100.0 MHz, 250 MHz, 500 MHz, 1000 MHz,
1.30 GHz, 1.54 GHz
  available cpufreq governors: conservative ondemand userspace
powersave performance schedutil
  current policy: frequency should be within 100.0 MHz and 1.54 GHz.
                  The governor "ondemand" may decide which speed to use
                  within this range.
  current CPU frequency: Unable to call hardware
  current CPU frequency: 250 MHz (asserted by call to kernel)

I did some simple stress testing and observed the freq scaling is
working fine when cpufreq governor is set to ondemand.

Powertop output.
            Package |            CPU 0
 100 MHz     5.2%   |  100 MHz     1.6%
 250 MHz     4.4%   |  250 MHz     4.3%
 500 MHz     2.6%   |  500 MHz     2.4%
1000 MHz     0.5%   | 1000 MHz     0.3%
1296 MHz     0.2%   | 1296 MHz     0.1%
1.54 GHz     0.2%   | 1.54 GHz     0.1%
Idle        86.9%   | Idle        91.2%

Here the output on the linaro's pm-qa testing for cpufreq.

[1] https://pastebin.com/h880WATn
Almost all the test case pass with this one as off now.

Best Regards
-Anand

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel