Porteus

Hello! I have a HP 20-c210 system (Celeron J3355-based) which is looping on CPU soft lockup errors when attempting to boot from the Porteus Kiosk 5.5.0 installation media:

c210 boot screen

Note that everything before the lockup messages seems to be standard behavior for this machine. The machine has BIOS updated to the most-recent available version (2021).

The installation media has worked (and continues to work) properly on various systems.

Noting that 5.5.0 is based on the 6.1.20 kernel, I installed Ubuntu on this c210 and compiled the 6.1.20 kernel on it with default configuration and booted into it. No problems.

Here is a file containing output of "uname -r", "dmidecode -t baseboard", "cat /proc/cpuinfo", and the default configuration file used to build the working 6.1.20 kernel:

c210 info

The most similar machine I have is a HP 20-c013w system (Celeron J3060-based) which has run 5.5.0 just fine. For comparison, here is a file containing the output of "dmidecode -t baseboard" and "cat /proc/cpuinfo" on it:

c013w info

Note that the c210 BIOS settings are nearly identical to those of the c013w.

I only have the one c210 available, but it successfully runs Ubuntu, Kali, and Windows so I tend to think it is fine.

My next step was to try to create PK installation media with the vanilla 6.1.20 kernel and/or run Ubuntu with the PK kernel, but I couldn't find much support for this. But I am happy to test custom ISOs or kernel builds as requested and made possible.

Thanks!

Hello brunch,

1) Is the system working OK except for logging the lockup messages?
This is not clear from your post.

2) Did you try to deactivate watchdog?
Maybe its the watchdog driver fault which "thinks" the first CPU freezes.

Thanks

Hello!

1) whoops, no, it is printing a lockup message every 26 seconds or so, apparently forever.

2) the BIOS doesn't give me an option, unfortunately. I poked around the install ISO trying to disable it on the kernel side via GRUB option, but no luck. I also turned on some extra watchdog stuff in my 6.1.20 kernel under Ubuntu trying to get soft lockups there, but also no luck.

I tend to believe this is an upstream kernel bug tickled by the particular PK config options, but haven't been able to isolate.

Thanks for the response!

Update! I was able to add "nowatchdog" to GRUB options in the on-SSD version, which demonstrates same behavior (actually with a lot more debugging information). Now instead of the watchdog errors, I get:

[<time>] rcu: INFO: rcu_preempt self-detected stall on CPU
[<time>] rcu: o1-....: (330019 ticks this GP) idle=141c/1/0x4000000000000000 softirq=63/63 fqa=165013
[<time>] o(t=330056 jiffies g=-1063 q=16 ncpus=2)

... followed by debugging information apparently similar to that printed when the kernel watchdog timer is on. And also looping forever, but at a slower rate than 26s.

Soooo yeah, 6.1.20 kernel preemption bug??

Strange that its affecting a single PC model only.

I have no solution to this problem.

Welllll adding "acpi=off" to the kernel options seems to have fixed it, so it looks like I'm good. Still haven't duplicated it in the mainstream 6.1.20 kernel but that's ongoing. Thanks for looking into it!