9

Long story short: We have one of those little $100 Chinese mini PCs running 18.04 LTS. ("Wintel Pro" / "Wintel Box") It thinks it has a battery, possibly because it thinks it is a laptop. The "battery" level notifications - that I have tried a million ways to kill - are driving me nuts because I'm trying to use this machine to run a looping slideshow while the workplace is open.

(Update: It wasn't just the battery notifications on the desktop, either. Sometimes it would just shut off randomly. I have a nearly identical model without a desktop and it exhibited random shut-offs until I applied the same fix detailed in my own answer below.)

Thought I might have had this fixed but HAHAHA NOPE

Related question: Why does Ubuntu power management set a battery as the default power supply on a desktop machine?

This is where someone asked "What is the output of laptop-detect -v?"

It is: We're a laptop (non device ACPI batteries found)

But we are emphatically not a laptop, and we don't want to be a laptop. How can we not be a laptop?

CPU is Intel Atom x5-Z8350. Do let me know if additional information is required. The fault may have nothing to do with Ubuntu but I live in hope Ubuntu might have the tools to fix it.

(Aside: LibreOffice Impress slides run fine, but I would not recommend these machines for high-resolution video playback.)

Top side of mini PC

Bottom of mini PC

LSHW output: https://pastebin.com/e5XXp8A0

Also, thank you to @Sri for reminding me that I'd already tried turning off "Notification Popups" and it did as much good as a piece of paper towel against a flood. The "battery" notifications seem to cheerfully disregard this setting - even with notification popups ostensibly OFF, the nonexistent battery still breaks through.

Update: Apparently there is an "axp288" device inside this miniPC, and apparently there is some discussion somewhere on the LKML about this: https://lore.kernel.org/lkml/20190520115247.828904395@linuxfoundation.org/

Is there a way to get this code into my instance of Ubuntu? Would it even do any good? (Maybe it's already there?)

muru
  • 207,970
Will Matheson
  • 453
  • 2
    Suspect this is a hardware issue, since other mini- and micro-PCs (i.e., Intel NUCs and Supermicro Atom SBCs) don't freak out because of a non-existent battery. However, I also hope I am wrong. Would you mind, please, finding the parts model number from the serial number sticker, and telling us the make and that parts model number? – K7AAY Jan 29 '20 at 23:48
  • Sure, I've added pictures of the mini PC to the post as well as a pastebin of the output of sudo lshw.

    Despite the model being W8 Pro, it shipped with Windows 10 (probably Home, I don't remember). I don't recall any weirdness with a "battery" there, but I also wasn't using it for long. I have an identical unit at home that is still on W10, and I don't observe this kind of problem.

    The 32GB of "ROM" refers to the eMMC memory used for the OS and storage. It is not a lot and it is not fast, but it works.

     – Will Matheson Jan 30 '20 at 00:54
  • 3
    Have you checked for any BIOS/firmware updates from the vendor? That may make a big difference in how things work. – ubfan1 Jan 30 '20 at 01:00
  • "Lifetime technical support from store vendor" advertised on Amazon site. You may need to take them up on that. – K7AAY Jan 30 '20 at 16:25
  • @ubfan1 Well, American Megatrends makes it so that all of their BIOS updating is done through a Windows-only tool. So unless I want to try running their tools in Wine and risk bricking the whole system, then basically I might as well just go back to Windows. – Will Matheson Jan 30 '20 at 19:04
  • And getting to @K7AAY's comment, this would make me sincerely doubt that they do much of anything to support Linux. I am here looking for ways to make Ubuntu work with the hardware, not looking to ask vendors of ultra-budget throwaway systems why they don't have developers contributing to the Linux kernel. – Will Matheson Jan 30 '20 at 19:06
  • You may be pleasantly surprised... more overseas folks support Linux and Ubuntu every day. – K7AAY Jan 30 '20 at 19:08
  • this might very well be the result of a too weak power suply 5V 2,5A=15W this might be too little,and it might make the system think there is a power drop and a power drop=battery low.try a larger powe suply and see if it solves the problem.do not expect miracles from the box 2gig ram is in the lower end of things. – trond hansen Feb 12 '20 at 05:37
  • Do you mean giving it 5V, but more than 2.5A? I'm using the AC adapter that came with it, and it says 5V 2500mA. I'm not sure I have any other would-fit AC adapters that supply the same voltage and even the same amperage, much less more. Also the CPU (I mean there is more than the CPU, but just for the sake of saying) seems to be designed for ~2W.

    https://ark.intel.com/content/www/us/en/ark/products/93361/intel-atom-x5-z8350-processor-2m-cache-up-to-1-92-ghz.html

     – Will Matheson Feb 12 '20 at 21:26
  • Dunno if this means anything, but I ran something like this:

    echo - | awk "{printf \"%.1f\", \ $(( \ $(cat /sys/class/power_supply/axp288_fuel_gauge/charge_now) * \ $(cat /sys/class/power_supply/axp288_fuel_gauge/voltage_now) \ )) / 1000000000000 }" ; echo " W " and got "19.5 W"

    based on this: https://unix.stackexchange.com/a/78437/337130

     – Will Matheson Feb 12 '20 at 22:05

3 Answers3

7

I'm an upstream kernel developer and the author of the kernel-patch linked from the original post.

I just discovered this thread today, this is a known issue with some of these Intel Atom x5-Z83xx based boxes, they use an AXP288 PMIC and often the BIOS does not turn off the fuel-gauge (battery monitoring) part of this chip.

The driver for the AXP288 fuel-gauge contains a list of devices like this: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/power/supply/axp288_fuel_gauge.c#n679

And it will automatically disable itself for devices on this list.

So the proper fix for this, which will make future Linux versions work out of the box, without needing manual configuration, is to add your device to this list.

I can do this for you, but I need some information about your model to be able to do this.

Please as a regular user in a terminal run:

grep . /sys/class/dmi/id/* 2> /dev/null

On the device and then copy and paste the output here, or send me an email with it at "Hans de Goede hdegoede@redhat.com".

That will give me the info which I need to add your device to the no-fuel-gauge list.

Kulfy
  • 18,163
Hans de Goede
  • 71
  • Right on! Here you go:
    • The output from the 2GB model mini PC featured in this post: https://pastebin.com/03R32ScV
    • The output from a very similar but 4GB model mini PC that I think has the same issue - making the same fix of blacklisting the fuel gauge stopped a spate of mysterious shutdowns (even though it wasn't running a desktop environment, so no visible-to-me gauge to also complain about): https://pastebin.com/2H3aW1CZ

    Thanks very much!

     – Will Matheson Mar 19 '21 at 16:27
  • 1
    Sorry for being so slow to respond, I somehow missed the notification that there were new comments on this thread.

    Thank you for providing the DMI info for the 2 models, I've submitted a set of patches upstream to automatically disable the axp288-fuel-guage on these devices without needing to disable the driver: https://lore.kernel.org/linux-pm/20210617192702.77106-2-hdegoede@redhat.com/T/#u

     – Hans de Goede Jun 17 '21 at 19:27
2

I think I fixed it. At the very least, the battery gauge is gone from GNOME.

Finally, no battery gauge!

In responding to @trond hansen's comment, and trying to find my wattage, I found I had an "axp288" something-or-other, and searching more about that, I found that the axp288 could be blacklisted in the kernel. (Perhaps check under /sys/class/power_supply to see what you have, if you are experiencing a problem like this.) It seems to be on a model-by-model basis, though - perhaps the axp288 is part of a legitimate battery sometimes. Anyway, I wanted to see if I could somehow add my own mini PC to that list, ideally without having to modify and compile the kernel myself!

But for now, you can just disable the battery gauge on your one machine:

sudo nano /etc/modprobe.d/blacklist-axp288.conf

Type in blacklist axp288_fuel_gauge then Ctrl+X to exit, Y to save changes, and Enter to confirm the filename. Then reboot. I hope this is the end of the misery. If I never come back to this thread, you can assume it was.

Shoutouts to the openSUSE forums for the fix! https://forums.opensuse.org/showthread.php/531934-Leap-15-on-Intel-Atom-power-supply-fuel-gauge-driver-failed-endless-errors

user68186
  • 37,795
Will Matheson
  • 453
0

While hardware experts figure out who and why we are ;), your issue with presentation interruptions can be resolved perhaps by turning off the notification popups in settings.

enter image description here

Yeah, we tend to overlook simple things, with existence becoming complicated day by day.

Edit: My bad, I assumed the desktop interface. The battery notification settings can be handled in terminal with gsettings. Please refer the answer here: https://askubuntu.com/a/572159/37991

Sri
  • 1,743
  • 2
  • 20
  • 39
  • Thank you, but I have the "Notification Popups" setting OFF and it is blithely disregarded inasmuch as the "battery" ones come up anyway. Maybe I'll turn "Lock Screen Notifications" OFF too just for the heck of it. Also there isn't an application specific setting for battery or power that I can see.

    I've also already used dconf Editor to do something similar with the battery notification settings - that is, crank the thresholds down to 0% so it's less likely that I'll see them. I haven't seen them since, so I'm cautiously optimistic. But I think there ought to be a real fix.

     – Will Matheson Feb 04 '20 at 23:24