@carlos your observations are very useful because it shows it happens in other 
brands, and with AMD (not just intel). I think we should try to figure out what 
all these notebooks have in common. 
Your kernel parameter acpi_osi=Linux won't change anything because the kernel 
already pass to the BIOS that and many Windows versions: the default for the 
kernel is to pass something like this to the BIOS: acpi_osi=Linux 
acpi_osi=Windows acpi_osi="Windows 2011" acpi_osi="Windows 2012" 
acpi_osi="Windows 2014" acpi_osi="Windows 2015" acpi_osi="Windows 2020"; so if 
you want to test something different, first you need to "delete" all these ids 
with acpi_osi=! and then add the id you want to test (for example acpi_osi=! 
acpi_osi=Linux). But I've tried all those, none helped. 
I've tried acpi_enforce_resources=lax, it didn't help either. 

Summary of major tests that don't fix:
1. None of the i8042 quirks alone or combined, or combined with pnpacpi=off or 
combined with atkbd.reset or atkbd.softrepeat=1,
2. Compiling the kernel with any of the different preempt models,
3. Any combination of acpi_osi=! acpi_osi="Windows XXXX",
4. Moving irq1 to the performance core and increasing the priority of ksoftirqd 
(echo 0 > /proc/irq/1/smp_affinity_list + chrt -f -p 99 $(pgrep -f 
"ksoftirqd/0"),
5. Completely disabling C-States or combining it with disabling P-States,
6. Disabling Bluetooth or the Wifi radio or both.
7. This one seems to make things even worse: threadirqs
8. pcie_aspm=off
9. changing Power > Thermal from "ultra-performance" to "cool" (either alone or 
combined with disabling SpeedStep)

What I started to test: disabling the turbo. My theory: turbo -> temperature 
spike managed by the firmware (so the OS isn't even aware there was a spike) -> 
firmware throttle and "blocks" the cpu -> miss irq1 signal (either press or 
release). 
If disabling turbo fix it, the plan is to re-enable it and test limiting the 
maximum frequency. 

Other tests I din't make yet because they don't seem promising:
acpi_irq_nobalance or notsc or clocksource=acpi_pm or isolcpu=0 and move
irq1 to that cpu or a different distro not based on Debian like Solus or
CachyOS or acpi_mask_gpe= or acpi=rsdt or pci=noacpi

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2093376

Title:
  Random keys "stick" repeating until any key is pressed

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/linux-signed-hwe-5.15/+bug/2093376/+subscriptions


-- 
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Reply via email to