Qcom-5.4 cpu frequency is very low

In this release qcom-5.4 cpu freqency is very low

This path itself not available. Please suggest.
sys/devices/system/cpu/cpu0/cpufreq/scaling_governor

we are not planning to support DB820c (at this point) in the release/qcomlt-5.4 branch. With regards to 820c, what you have in this branch, is pretty much what is available in the vanilla 5.4. The problem with cpu frequency is that all drivers required for cpufreq (and CPR, and voltage scaling, …) are still missing…

we will be using the 5.4 branch for releases for 410c and 845c only.

i should have added that… if anyone is willing to contribute patches to this 5.4 branch, or even better to mainline to help fix these issues, please let us know, we would happily review code and support such developers…

@ndec: snapshots for db820c are with 5.4 now;
http://snapshots.linaro.org/96boards/dragonboard820c/linaro/debian/latest/

And further, vanilla 5.4+ doesn’t work on db820c.

i see the confusion… i think we should remove 5.4 from the snapshots then…

Before doing that, its actually a lot more stable on 5.4 than on 4.14, and the work needed to get it into shape is not appearing to be all that much.

release/5.4 is ‘a lot more’ stable than release/4.14? really? i am a bit surprised here… do you have examples?

the problem is that without cpufreq it will be very soon unusable, no? and that’s some serious work to bring that into 5.4…

4.14 crashes about 9 times out of 10 boots. There is a patch in 5.4 that fixes the problem, but I haven’t been able to get it working successfully with 4.14.

https://lore.kernel.org/linux-arm-msm/20200106080546.3192125-1-bjorn.andersson@linaro.org/

I’ve achieved a PARTIAL control over the CPU frequency.
See all of the commits on the top of here;

*** NOTE: The db820c_defconfig is tailored for ANDROID, not Debian.

Also note that I have the available frequencies restricted to one speed ~800 MHz for all cores. It was crashing for me without locking it there. It may be related to the commit that I included and then reverted (wouldn’t compile).

I’m also using the wrong driver… cpufreq-dt rather than nvmem.

Still working on it, obviously, but this makes a HUGE difference in usability.

I seem to have it working in a pretty good shape now.

@ndec : What’s the process for getting these in to qcomlt-5.4?