ACPI Kernel issue

[pebenito]

I would appreciate if you could test the max_cstate=3 since my setup don't allow me to push huge files at high speed.

It seems to be ok. Twice over the weekend I pushed about 1GB over to it, without hanging. I'll continue to keep an eye on it for the next couple weeks to be sure.

[prj]

Nice! I haven't had any problems either (been running since my previous post). When I get some free time I'll send a bug report to the LKML. Perhaps C6 for SBC-FITPC2 could be blacklisted in the kernel so future vanilla installs are stable.

[pebenito]

When I get some free time I'll send a bug report to the LKML. Perhaps C6 for SBC-FITPC2 could be blacklisted in the kernel so future vanilla installs are stable.

Frankly, I think blacklisting in the kernel should be a last resort. It would be nice to have an official response from CompuLab on this issue. I'd like to know if this is a hardware or software(BIOS) issue. If its the latter, a BIOS update should be done. I certainly want to squeeze as much power savings as possible out of this system as I can (not that its bad when constraining the C states).

[prj]

Nobody seems to be working on it and I cannot do any further investigation without having a extra Fit-PC2 for debugging (will get one when I get my tax refund). I do think that blacklisting is good enough for the average user at this point. If it gets fixed later on the blacklist can be removed. Nothing is permanent in kernel-space

If you like to squeeze more out of your unit you might wanna try compiling your own distro with better compile-time optimizations. Look at this article.

I'm currently writing a GMA 500 2D driver from scratch based on what has previously been released as GPL. This new driver has the potential to save even more power in environments that don't need fancy graphics.

[pebenito]

If you like to squeeze more out of your unit you might wanna try compiling your own distro with better compile-time optimizations. Look at this article.

I do compile my own distro; I use Gentoo

[prj]

Have you tried any compiler optimizations? What arch and settings are you using?

[raldi]

I'm having a problem on a brand-new fitpc2 that's doing a lot of heavy disk / network operations. The kernel dumps a bunch of trace beginning with:

Code: Select all

BUG: unable to handle kernel paging request at ff09dfc0
IP: [<c0150300>] kthread_should_stop+0x10/0x20

or

Code: Select all

BUG: unable to handle kernel paging request at ff09dfc8
IP: [<c0521bc8>] _spin_lock_irqsave+0x18/0x30

...and then networking dies. I posted about it at http://www.fit-pc2.com/forum/viewtopic. ... 6078#p6078 and was directed to this thread.

Switching to "GV3 Only" didn't help, or if it did, it only made the problem happen less often. It still seems to occur once every 24-36 hours.

Does this seem like the bug that this thread is discussing? Could someone who is still plagued by it run this command the next time it happens and post the results?

Code: Select all

grep -A 1 'unable to handle kernel paging request' /var/log/kern.log

I am using Gigabit ethernet. I'll try disabling it in the meantime and see if that makes the problem go away.

[prj]

Hi Raidi, what you are describing sounds like a completely different problem. What you could do is check your RAM with memtest. Other than that I have no idea.

[Frans]

I also had the problem with ramdom freezing of the fit-PC2. Switching to [GV3 Only] worked perfectly.

[raldi]

I switched to GV3 Only, and even plugged the machine into a non-gigE switch, in case it was somehow related to gigabit ethernet. Still happens. I'll try memtest next.

If it turns out my machine is defective, how much time do I have from the purchase date to discover and report it?