[StarLite Mk V] poweroff / suspend broken with certain PCI ASPM states

[BlackIkeEagle]

I don't know if I wrecked something while testing last week, but during the weekend I noticed my StarLite Mk V was no longer going in standby/poweroff properly.

I had reset the firmware configuration to test and I think it started happening after that but not sure.

I tested with the following firmware versions:

• 24.10
• 25.01
• 25.02

These are upgraded downgraded with fwupdmgr install and then manually select the device and the version you want to install.

Once I change PCI ASPM to Disabled or L0s suspend and poweroff work properly again.

If I change PCI ASPM to L1s or L0sL1 (which I think it set on pressing F9=Reset to Defaults) my standby and poweroff keep me stuck on display has gone out, but the power led keeps on and when no power is attached it does not go further to the 12 seconds -> quick flashing of powerled)

While I was trying things I also noticed some kworker threads staying in D state causing my load indication to stay pretty high without doing anything.

ps aux | grep ' D '
root          40  0.0  0.0      0     0 ?        D    16:56   0:00 [kworker/3:0+events]
root          83  0.0  0.0      0     0 ?        D    16:56   0:00 [kworker/3:2+events_freezable]

I can't provide any logs it seems since my journal states it actually finished in a shutdown properly, and it looks the same on first glance as a successfull shutdown.

[BlackIkeEagle]

Ok the kworker issue is this: #183 since I'm on 6.12 lts which does not yet have torvalds/linux@235b630

[BlackIkeEagle]

And torvalds/linux@235b630 got reverted again in 6.13.7 it seems since it introduced delays: torvalds/linux@853736d

Anyway, even without the patch and the sd-card slot dustcover removed the original problem remains

When PCI ASPM is set to L1s or L0sL1 I have issues with suspend/shutdown/reboot not happening

[MrChromebox]

distro? kernel?

not seeing any issues with suspend/shutdown here on the Lite w/firmware 25.02 under Ubuntu 24.10 (kernel 6.11.0-19) and PopOS 24.04 (kernel 6.12.10)

[BlackIkeEagle]

Arch Linux, kernels 6.1, 6.12, 6.13

also with the iso of 2024-06 running kernel 6.9

changing to L0s fixes it for me

[MrChromebox]

sounds like this is something Arch-specific then. we've been seeing a lot of that lately =/

[BlackIkeEagle]

Its strange it worked fine with these before I reset the settings in the firmware, I also restarted with a snapshot which was working before. But I cant get proper insight in why the power led stays on after everything else stopped.

[BlackIkeEagle]

I just tested it with the ubuntu live version, same problem as with Arch, so much for being Arch specific. So quickly dismissing it with "it is again something Arch specific" is IMO not really helpfull. I'm also ok with the statement that L0sL1 or L1s don't work. I would like to be able to gain some insight in what is not working or blocking the Starlite from actually powering off when the earlier mentioned PCI ASPM states are used, versus when it is disabled or L0s is used.

So it is confirmed for me on my device with ubuntu live (24.04.2) as well, L0sL1 set;

• poweroff
• screen goes dark
• keyboard backlight goes out (the keyboard was attached for the test)
• powerled stays on
• after ~4 minutes in that state I gave up and pressed the power button around 8 seconds
• the power button blinked
• wait ~12 seconds
• the fast blinking of the firmware reset happens
• can start again

[MrChromebox]

I'm also ok with the statement that L0sL1 or L1s don't work.

Everything is working just fine with them enabled on the 2 lites I have here

I would like to be able to gain some insight in what is not working or blocking the Starlite from actually powering off when the earlier mentioned PCI ASPM states are used, versus when it is disabled or L0s is used.

I have no idea offhand

So quickly dismissing it with "it is again something Arch specific" is IMO not really helpful.

better if I call it device specific then? or setup-specific?