search

StarLabsLtd / firmware

68 stars 4 forks source link

[Starbook MK VI - Intel] Bluetooth USB drops at high load #152

Open mysfitt opened 4 months ago

mysfitt commented 4 months ago

I've been noticing for some time now that my i7-1260p has been having disconnects with the AX210 bluetooth chipset. I recently bought a trackball that uses a USB dongle with 2.4ghz RF instead of bluetooth, and it also disconnected today, and I think I've figured out the connection. When the machine is under high load, I will get this in the kernel log:

Mar 11 12:47:55 gryphon kernel: usb 1-6: USB disconnect, device number 5
Mar 11 12:47:56 gryphon kernel: usb usb1-port7: over-current condition
Mar 11 12:47:56 gryphon kernel: usb 1-7: USB disconnect, device number 6
Mar 11 12:47:57 gryphon kernel: usb usb1-port6: over-current condition

In this particular example, I was running a build for OpenWRT, which is multi-threaded and involves not only a kernel build, but a full system build with an arm cross-compiler. Other examples of disconnects have been when playing a game on Steam with a bluetooth controller, and when playing music to a bluetooth speaker on Spotify, which is a crappy Electron app that consumes way too many system resources. I know that the Bluetooth chipset is connected through the USB bus, so when I had the failure today (that did not involve bluetooth) I realized that it's not Bluetooth being flaky, it's the USB bus. Note that I do have my machine set to "performance" in the BIOS. I'm running the latest version as of this writing, Coreboot 24.02.

Is this something that can be fixed in firmware, or am I actually seeing a hardware failure?

Steps to Reproduce

  1. connect a bluetooth device, such as a mouse or speaker
  2. Put a high load on the system. I'd recommend a large compile, such as the kernel. A benchmarking suite should also do the trick.
  3. Monitor the kernel log for the over-current condition message
Sean-StarLabs commented 4 months ago

Are you outside the EU by any chance?

mysfitt commented 3 months ago

Yes, I'm in the US. I'm same guy that's been bugging you about flashrom not working on Arch :) What would country code have to do with it? Wifi regulatory domain and Radar detection stuff?

Sean-StarLabs commented 3 months ago

It's something to do with certain channels shouldn't be used at max power. I'm pretty sure we can limit it from the firmware so I'll take a look

mysfitt commented 3 months ago

@Sean-StarLabs Cool. Let me know if there's useful logs I can provide. I keep a pretty good history buffer in my journal. Is the wifi chip also on the USB bus? I know that the bluetooth is since it's connected through btusb, but I thought that the wifi 6e bit was a discrete chip on the PCI bus.

Sean-StarLabs commented 2 months ago

Can you try 24.04 (once it's jumped through the LVFS's hoops), with PCI Clock Power Management, PCI ASPM and PCI L1 Substates all disabled?