Forum Home
Press F1
 
Thread ID: 98107 2009-03-11 09:35:00 Occassional hard lock (Linux) Myth (110) Press F1
Post ID Timestamp Content User
755588 2009-03-12 10:34:00 Normally if you get a kernel locking up for some reason, it will tell you why in whatever log you are sending kernel messages to (or if you aren't logging them, you'll see them in tty1 by default) - hence the stuff I was saying before.

Myth, make sure you use a synchronous mount on your log filesystem- you'll lose a lot of performance, but it means that the final death throes of the system will be written to disk, rather than getting stuck in the write cache when the system hangs. I'll keep an eye on my im accounts- you're welcome to bug me for help on this one. Flick me a pm if I'm not online.

You say you're running gentoo, so I assume you have a custom kernel - can you post the .config? Also which kernel version & arch are you running, and what patches (if any) have you applied other than the default gentoo ones? Are you using vanilla or gentoo sources?
Erayd (23)
755589 2009-03-12 10:40:00 This is a question as I know naff all about linux, esp when it goes wrong - But does Linux have the equivalent of windows event Viewer ? Yep - almost all unix / Linux systems have logging that puts windows to shame! There's far, far more detail than event viewer would ever give you .


One thing that wouldn't change no matter what the OS is - does the HDD activity light on the case stay on solid when it locks - meaning the HDD is trying to work but cant ?
Usually true, but this can also occur on Linux if one of the critical filesystems is full, or if the filesystem is doing some re-organisation . Linux filesystems tend to do a lot more than their windows equivalents, much of it automatically .
Erayd (23)
755590 2009-03-12 17:22:00 Ok.. .config is here (pastebin.com)

Using gentoo kernel (gentoo-sources) no other patches applied
Kernel version is in .config
arch is amd64

synchronous mount?

CPU temp is probably not correct. I know when I was mucking round with conky a few years back there is a file with accurate temps that gets written to... just cant remember that file name (but I think that was mobo temps). I just did a quick install of lm_sensors, added sensor support into the kernel, and did 'sensors' (after sensor-detect of course). So unsure how accurate. But as said, air from CPU fan is quite cool, minimal warmth

Had an idea yesterday... might emerge genkernel and use that, see what happens
(Im no longer looking for uber-quick, just want stable and just like the gentoo way)
Myth (110)
755591 2009-03-12 22:01:00 ...synchronous mount?
mount -o sync (or you can put sync as a mount option on fstab). It basically means that everything gets flushed to disk immediately instead of queueing up in the write buffer.

Re your comments on genkernel - while it's still pretty good in terms of speed, it does have a substantially longer boot time, and I find the featureset / implementation of it annoys me (so I don't use it).
Erayd (23)
755592 2009-03-12 22:27:00 Comments on the kernel. It looks pretty good, although there are a few weird bits - these are what jumped out. Mostly just nitpicking, I didn't see anything particularly evil although it's impossible to say for sure without seeing your system.

Why is MCE (Machine Check Exception) disabled? Enable it, unless you have a very good reason not to. It should also help with diagnosing this problem.
You have AC / battery / dock ACPI options compiled in, but you're using a desktop???
CPUFreq is enabled, but you aren't using the PowerNow! interface
You are using PCMCIA on a desktop???
Are you a ham radio fan?
Your network driver config seems a bit wacky - lots of stuff enabled in there that I seriously doubt you use (e.g. token ring, 10G ethernet etc)
Why intel i2c on an AMD platform?
Strange ALSA driver config (things enabled that shouldn't be)
No reiserfs support (sigh.... ext3 just doesn't cut it these days - go get a proper filesystem:p)
x86_64 AES support disabled
No PRNG
Virtualisation is enabled, but none of the options under it (e.g. KVM)
No custom version string
Erayd (23)
1 2 3