im trying to figure out why proper c-states not getting set with the default debian 11 kernel.
when using the default one, powertop reports up to c3 state on package level. when using backports 6.0 kernel it goes up to c8 like it is configured in bios. (no xmp, no enhanched turbo etc.)
how can i debug such a “mess” by myself? im not a real tech professional
i came so far that ASPM is not enabled for the pch according to lspci -vvv, but when booting the 6.0 kernel via backports repo, it es enabled. i installed the non-free intel firmware and realtek firmware aswell misc-nonfree. i did update-init -cv -k all after it did this, package manager doenst report any missing firmware. i installed the headers aswell.
for grub boot i tried to add to force aspm for 5.10, but no effect.
im flabbergasted why backports kernel does c8 state on package level but not the original one.
What is the hardware and how do you know that the hardware supports all of this. I am assuming that this is the first foray into GNU/Linux? Sometimes there are quirks in the hardware that are supported in MS Windows but are not in other OSes because the vendor has not exposed this information.
not really the first time, i fiddled a couple of times around with linux before.
its a old kaby lake 7600k and a simple z270a pro (msi) mainboard, (3 toshiba x300 hdd’s and one 840 evo ssd) so simply the most basic stuff. na when using the backports kernel und look into powertop, it reports c8 state in package level. but not the original debian kernel. looks like aspm is disabled when using the debian kernel. but its activated when using the backports kernel. #1 SMP PREEMPT_DYNAMIC Debian 6.0.3-1~bpo11+1 (2022-10-29)
but the “old” debian kernel does not switch to c8 on package level.
powertop --auto-tune is performed at startup, so dont need to hassle about it. #1 SMP Debian 5.10.158-2 (2022-12-13)
i installed the i915 firmware and the module for mainboard sensors,aswell as the realtek firmware out of the non-free branch.
the only difference i see atm is that aspm is not enable in the old kernel.
with c8 on package level, it idles at bout ~15w out of the wall, with disabled aspm its about +21w (hdd spun down in both cases)
I wonder if it is a hardware acceleration issue so the machine cannot hit the right C-state because the VPU decode hardware is still waiting for instructions. I know that there were some issues back in the day with Kaby Lake and Intel never fully addressed all issues with the i915 stuff.
when ive googled my specific problem some ppl in arch and ubuntu complained about this in their laptops for kaby lake. im wondering because with the new kernel it works and it the same modul, if im not completly wrong, i should mention that the igp gpu intel non free media driver is installed aswell, because i was fiddling around with jellyfin server, but i scrap jellyfin, samba and vlc is more versatile for me.
how shall i look for that vpu decode hardware? intel_gpu_top didnt showed anything in that.
when ive looked into the lscpi, aspm was disabled for chipset and the realtek gbit device with the original kernel. using the backports kernel solved those problems. dunno how should to reproduce it properly.
Unfortunately, I don’t know. The common way of troubleshooting this is to bisect. What kernel was it last working on… what kernel did you notice it not working on. Then you upgrade from last working until it stops working and that would be where to troubleshoot. If it has never worked and now it does, then I could not tell you how to go about it. Maybe reach out to your distro maintainer or the upstream maintainer?!
but on my system it seems to be broken with the integrated stuff. nvm.
dunno if i should compile my own kernel now and test it your does your recommendation and testing allday long. is there any kind of automated “testing” like. grabbing all kernels and reboot and save some output results?
hm, when i looked a bit further seems like its more like a kernel bug. according to unraid some ppl got the same problem even with “nearly” identical system. looks like this is somehow a mess in total. guess ive to live with the backports kernel.
