[(partially?) SOLVED] HP dm1z (4000 model) freezes under heavy load

Question

Alright, well I've had this laptop for a few weeks now, and it's been pretty good. But I've noticed somewhat random freezes when in Gnome 3 (as I explain below, I've been unable to reproduce this on demand, so I do not know if it would happen were I to use something else or not use X), in which the system is totally unresponsive (pointer doesn't move, unable to switch to a virtual console, no sign of I/O) and has to be turned off manually. The laptop has an AMD E-450 APU, with Radeon HD 6320M graphics built in.
Magic sysrq stuff sort of works. The last time it happened, I was able to reboot using sysrq, however, I was not able to restore the framebuffer console and thus was unable to get any data from it. And I do know that Alt + printScreen + v works when the system is not in a crashy state.
The problem has occurred since I got it, and I've kept the system fairly up to date, so the issue has occurred with every kernel version that's been available from three weeks or so ago to now. The problem occurs when using the stock kernel as well as the K8-optimized linux-ck kernel.
I have read every single log in /var/log, and I have not noticed any messages that seem like they could be associated with the problem. A couple do pop out at me, however:
Jan 14 10:25:13 localhost logger: ACPI group/action undefined: thermal_zone / LNXTHERM:00
This has often shown up around the time of the freezes, but also shows up a lot when the system is operating normally. I've got sensord running, however, and the CPU is almost always in the range of 60-70°C, which as far as I can tell is safely below the critical threshold.
[ 292.154] (II) RADEON(0): radeon_dri2_schedule_flip:670 fevent[0x17eb6d0]
[ 292.175] (II) RADEON(0): radeon_dri2_flip_event_handler:1067 fevent[0x17eb6d0] width 1366 pitch 5632 (/4 1408)
Xorg.log gets spammed with this quite a bit, with it showing up anywhere from two or three times a second to once every 30 seconds. Twice so far, I've noticed (after examining Xorg.log.old) post-freeze that only part of the message was written to the log (e.g. just the timestamp, and a handful of characters from the start of the message, or just the first message of the pair).
Jan 14 10:49:17 localhost kernel: [ 6.170288] [Firmware Bug]: Invalid critical threshold (0)
This gets written on every boot up.
I've been unable to find anyone experiencing a similar issue. I am also unable to intentionally reproduce the problem. I've used stress-testing tools to hammer the CPU with at least 3-4 threads, as well as hammer the memory and I/O, but running it for around an hour has not worked.
Also, it seems to be much more likely to freeze after having just rebooted from a freeze. This suggests to me that it is a thermal issue, but nothing I've seen corroborates that.
I've really got no idea how to troubleshoot this, so any help at all would be appreciated. And if you need any additional information, just ask. Thanks in advance.
In case it helps:
rc.conf
lsmod output
lspci output
Last edited by Guff (2012-07-17 16:36:01)

Answer 1

@dimath: thanks for the link. There was a link there to another bug report, which mentioned some kernel parameters to try out. Didn't help.
Since I was having issues with getting my wireless to work properly (which I ultimately was not able to do, so just got a tiny USB dongle for the time being), I did a lot of mucking around in the system. Mucking around with many things that I did not understand. So, I figured well maybe I did something stupid that somehow brought this curse down upon me, so last week I did a total reinstall.
It didn't work.
I've tried KDE, and the problem persists, so it's not gnome-specific for me. I'm trying the catalyst driver now, but of course it's fairly unusable with gnome3 at the moment, so I'm only testing it in KDE. It seems like it might be okay, because I've been doing my best to try and trigger a crash/freeze.
Then again, it took a long while for the issue to pop up once I started using KDE. I still haven't quite worked out how to trigger the damn thing.
It often shows up when building some larger packages, installing packages, and when launching Firefox (I often have an ungodly amount of tabs open, so it is fairly resource-intensive). As mentioned, however, it's not at all consistent.
dimath, have you tried another DE yet? And if the problem persisted, have you tried catalyst? Obviously, I don't expect you to use it with gnome given that it's buggy as hell, but I'd be interested to see if it helps for you at all.