nvidia-driver-470 makes fans spin at ~50% w/o any load

[Vladislav K. Valtchev]

After upgrading from Ubuntu 21.04 to Ubuntu 21.10 a few days ago, I noticed that the fans of my laptop (Lenovo ideapad 700-15ISK) were spinning all the time at about ~50%, no matter the CPU load.

I forced all my cores to run at 100% with a simple while(1) loop to see if the fan speed would increase but nothing. Apparently, no matter the CPU load and temperature, with kernel 5.13.0-19 the fans of my laptop spin at the same RPM. From their noise, I believe that's about 50%.

Therefore, I decided to boot the same Ubuntu 21.10 with the older kernel 5.11.0-37 and the problem disappeared. Looks like in the newer kernel there is a problem with the control of fans for my HW. To complete the picture, I built and booted kernel v5.15-rc5 to see if with a bleeding edge kernel the problem would disappear: it didn't. Therefore, the problem appears to be introduced somewhere between kernel 5.11 and kernel 5.13.

The output of `sensors`, unfortunately doesn't show any fans and that's the same no matter the kernel version:

pch_skylake-virtual-0
Adapter: Virtual device
temp1: +54.5°C

BAT0-acpi-0
Adapter: ACPI interface
in0: 12.25 V

coretemp-isa-0000
Adapter: ISA adapter
Package id 0: +39.0°C (high = +100.0°C, crit = +100.0°C)
Core 0: +37.0°C (high = +100.0°C, crit = +100.0°C)
Core 1: +38.0°C (high = +100.0°C, crit = +100.0°C)
Core 2: +39.0°C (high = +100.0°C, crit = +100.0°C)
Core 3: +38.0°C (high = +100.0°C, crit = +100.0°C)

acpitz-acpi-0
Adapter: ACPI interface
temp1: +39.0°C (crit = +99.0°C)
temp2: +42.0°C (crit = +85.0°C)

ProblemType: Bug
DistroRelease: Ubuntu 21.10
Package: linux-image-5.13.0-19-generic 5.13.0-19.19
ProcVersionSignature: Ubuntu 5.13.0-19.19-generic 5.13.14
Uname: Linux 5.13.0-19-generic x86_64
NonfreeKernelModules: nvidia_modeset nvidia
ApportVersion: 2.20.11-0ubuntu70
Architecture: amd64
AudioDevicesInUse:
 USER PID ACCESS COMMAND
 /dev/snd/controlC1: vlad 5972 F.... pulseaudio
 /dev/snd/controlC0: vlad 5972 F.... pulseaudio
CasperMD5CheckResult: unknown
CurrentDesktop: ubuntu:GNOME
Date: Mon Oct 18 19:04:03 2021
InstallationDate: Installed on 2020-03-17 (579 days ago)
InstallationMedia: Ubuntu 19.10 "Eoan Ermine" - Release amd64 (20191017)
MachineType: LENOVO 80RU
ProcFB: 0 i915drmfb
ProcKernelCmdLine: BOOT_IMAGE=/boot/vmlinuz-5.13.0-19-generic root=UUID=0e855d19-79da-402d-9586-b7bfe9a9622a ro
RelatedPackageVersions:
 linux-restricted-modules-5.13.0-19-generic N/A
 linux-backports-modules-5.13.0-19-generic N/A
 linux-firmware 1.201
SourcePackage: linux
UpgradeStatus: Upgraded to impish on 2021-10-16 (1 days ago)
dmi.bios.date: 08/09/2017
dmi.bios.release: 1.61
dmi.bios.vendor: LENOVO
dmi.bios.version: E5CN61WW
dmi.board.asset.tag: No Asset Tag
dmi.board.name: Lenovo ideapad 700-15ISK
dmi.board.vendor: LENOVO
dmi.board.version: NO DPK
dmi.chassis.asset.tag: No Asset Tag
dmi.chassis.type: 10
dmi.chassis.vendor: LENOVO
dmi.chassis.version: Lenovo ideapad 700-15ISK
dmi.ec.firmware.release: 1.22
dmi.modalias: dmi:bvnLENOVO:bvrE5CN61WW:bd08/09/2017:br1.61:efr1.22:svnLENOVO:pn80RU:pvrLenovoideapad700-15ISK:skuLENOVO_MT_80RU_BU_idea_FM_Lenovoideapad700-15ISK:rvnLENOVO:rnLenovoideapad700-15ISK:rvrNODPK:cvnLENOVO:ct10:cvrLenovoideapad700-15ISK:
dmi.product.family: IDEAPAD
dmi.product.name: 80RU
dmi.product.sku: LENOVO_MT_80RU_BU_idea_FM_Lenovo ideapad 700-15ISK
dmi.product.version: Lenovo ideapad 700-15ISK
dmi.sys.vendor: LENOVO

[Ubuntu Kernel Bot (ubuntu-kernel-bot)]

This change was made by a bot.

[Vladislav K. Valtchev (vvaltchev)]

Given that the problem exist with kernel 5.13.0-19 and newer but not with kernel 5.11.0-37, I tried to find which the change triggered this weird behavior using git bisect. I used bisect in the range [v5.11, v5.13] (official tags) and I've built the standard non-ubuntu kernel(s) using:

    make -j4 bindeb-pkg LOCALVERSION=-custom

As config, I used the Ubuntu one from: /boot/config-5.11.0-37-generic and accepted all the new config options as default.

But, before finishing to test all the bisect steps, I found that even with the first kernel, v5.11, commit: f40ddce88593482919761f74910f42f4b84c004b, the problem already exists.

Therefore, it looks like there were some Ubuntu-specific patches in 5.11.0-37 that prevented the problem to happen. In the original v5.11 kernel, the fans spin at ~50% apparently no matter the CPU load.

Apparently, those Ubuntu patches that prevented the problem from happening, got changed in 5.13.0-19. Does anybody have a clue what could that be?

Thanks,
Vlad

[Vladislav K. Valtchev (vvaltchev)]

To check the theory about the Ubuntu-specific patches, I tried to boot another Linux distro (Fedora Live 35) in order to see if the problem would exist: it did not.

The kernel used was: 5.14.10-300.fc35.x86_64.

Which is more recent that the kernel 5.13.0-20, currently used by Ubuntu 21.10. Therefore, it looks that both the older Ubuntu-specific kernel and the newest Fedora kernel do _not_ have the problem, thanks to some custom patches that (apparently) do _not_ exist in 5.13.0-19 nor in the mainstream Linux kernel.

It's getting even more tricky.

An alternative long-shot theory is that thermal management policies changed in the newest Ubuntu, but does not activate with kernel 5.11 because of the lack of some kernel feature. That would explain why the problem exists with the mainstream kernel as well, but not on another distro.

[Vladislav K. Valtchev (vvaltchev)]

Interesting update. I've followed the build instructions for the Ubuntu kernel: https://wiki.ubuntu.com/Kernel/BuildYourOwnKernel

By cloning the repo: git://kernel.ubuntu.com/ubuntu/ubuntu-impish
And building the kernel tagged as: Ubuntu-5.11.0-20.21+21.10.1

Which is *older* than the last _good_ Ubuntu kernel I have (5.11.0-37) left installed from the previous version of Ubuntu (21.04, Hirsute) but the problem is *reproducible*.

This means that the issue is a combination between the configuration in Ubuntu 21.10 and some Ubuntu kernels built from the ubuntu-impish repo. I have to try building another Ubuntu kernel from the Hirsute repo and see what happens. Apparently all mainstream kernels >= 5.11 and all Ubuntu Impish kernels have this issue, when running Ubuntu 21.10 (desktop).

[Vladislav K. Valtchev (vvaltchev)]

Major update
--------------

I've discovered that by switching to the open source Nouveau display driver the problem disappears, even with the newest Ubuntu 5.13.0-20 kernel.

The problem was related to the "nvidia-driver-470" for my video card: GM107M [GeForce GTX 950M].
The older 5.11.0-37 Ubuntu did not had this problem, because simply there were no Nvidia drivers built by DKMS for it. Probably, I switched to the Nvidia proprietary driver after upgrading to Ubuntu 21.10, so the older kernel remained without the Nvidia modules.

I'm not sure at the moment if this still a bug for this place. I'll leave this decision to you guys. At least, it will be useful for other people with the same issue.