search

nberlee / talos

Friendly fork for Turing RK1 on Talos
https://www.talos.dev
Mozilla Public License 2.0
66 stars 0 forks source link

Missing CPU frequency #4

Open pl4nty opened 7 months ago

pl4nty commented 7 months ago

Bug Report

Description

talosctl dashboard reports 8x0MHz for RK1 nodes. Might be related to the big.LITTLE arch, but I see the same issue on my homogenous SOQuartz node. Not sure if upstream cpufreq issues are relevant either.

Logs

Nothing in dmesg, is there somewhere else to look?

Environment

nberlee commented 7 months ago

Mhz comes from /proc/cpuinfo, see https://github.com/siderolabs/talos/blob/main/internal/app/machined/internal/server/v1alpha1/v1alpha1_monitoring.go#L153

/proc/cpuinfo is kind of empty the RK3588, even the PI4 does not report the Mhz in /proc/cpuinfo. So i am affraid this is a kernel limition, or upstream Talos should support cpufreq /proc/ sources.

When it comes to the cpufreq. I am eagerly watching patchwork for the v3 version of Alexey Charkov (which is mentioned in the gitlab issue of collabora), The v2 version has some optimalisation in voltage (lower) and more higher 2.4 Ghz max freqs compared to the current patch sets which are 1:1 copy of the rockchip Kernel 5.1.

However the v3 seems to be to a standstill as Dragan Simic joined testing: https://lore.kernel.org/all/d2f17fdc111bba89a0e107c058d4fca0@manjaro.org/

This will not help at all with this issue, but was related :)

nberlee commented 7 months ago

Hi @pl4nty ,

Ive created an image with version 2 over Alexey Charkov patches, would like to test it, use tag v1.6.5-1-gf5b2ba84c-rk3588. I am running it now and will ask a few others, as collabora deemed it unstable, but I did not find any problems yet.

bhuism commented 7 months ago

@nberlee v1.6.5-1-gf5b2ba84c runs perfectly fine for a week now

nberlee commented 7 months ago

Great.Yesterday I've released 1.6.6 with these patches incorporated.

bplein commented 4 months ago

Running v1.7.4 and getting:

user: warning: [2024-06-03T14:00:29.714604753Z]: [talos] controller failed {"component": "controller-
 runtime", "controller": "runtime.KernelModuleSpecController", "error": "error loading module
 \"rockchip-cpufreq\": module not found"}
nberlee commented 4 months ago

@bplein please use the -rk3588 image or if you have other extentension, ADD the rk3588 extension.

bplein commented 4 months ago

Oh, I think I broke things.

I'm working with the uboot on eMMC and /var on NVMe, and I think I messed something up along the way. My machine configs now show the generic talos image (not yours, not -rk3588) in the install section.

I am adding iscsi-tools, should I be pointing at your image or the generic Talos? (I am following the iscsi-tools installation notes at https://github.com/democratic-csi/democratic-csi)

bplein commented 4 months ago

My most recent actions were to run talosctl upgrade -i ghcr.io/nberlee/installer:v1.7.4-rk3588 --talosconfig=./talosconfig -n rk1-3 -e rk1-3 to force a reload of the iscsi tools.

nberlee commented 4 months ago

@bplein Please create a new issue. You are notifying other people when adding to this issue who do not have this problem.

You need to add to your machine config (talosctl edit mc):

machine:
    install:
        extensions:
           - image: ghcr.io/nberlee/rk3588:v1.7.4

if you have already added iscisi and do the upgrade again