[SOLVED] updatedb freeze on ThinkPad R52 - overheating?

[emuller]

I have two machines both installed with gentoo weeks apart. One is a IBM thinkpad r52, one is a home build sempron 2600.

The problem is with the thinkpad. It freezes during updatedb (reproducible), makewhatis (sometimes), emerge -s (sometimes)

No ssh access, requires poweroff

- updatedb can work if I run it in console mode.

I thought it might be the memory upgrade so I removed it and still problems. Memtest86+ and memtester say everythings ok

The only thing I can think is different is the kernel options needed for the thinkpad. It has SATA drives.

I dunno.

The symptoms sound similar to this:
https://forums.gentoo.org/viewtopic-t-195310-highlight-updatedb+freeze.html

but it persisted without the memory upgrade.

I've run heavy simulations overnight no problem now that the updatedb, makewhatis crons are disabled. Last crash was the emerge -s

The simulations are memory intensive ...

I've heard updatedb allocates memory in a more direct way from the kernel:

Bradeeoh wrote in topic:

https://forums.gentoo.org/viewtopic.php?t=155348

"Everytime the kernel reads an inode off any filesystem, it caches it internally so the next time a reference to that inode is made, it doesn't need to actually perform the read. Well, when updatedb runs, it accesses a WHOLE BUTTLOAD (yes, technical term) of inodes, and they make the kernels internal memory usages skyrocket."

Could it be a SATA memory bug? Like I said the other machine works fine.

[emuller]

OK an update:

I booted with one of the first kernels I compiled after installing gentoo and it seems to be stable. updatedb, etc work fine.

I was messing around with distcc and crossdev but never got it to work. Maybe I broke something like gcc then and now have a dirty kernel.

or... one of the kernel options I enabled since then isn't good for my machine.

I'll try compiling the old kernel again... I think I still have the config. If it becomes unstable then I'll suspect something with gcc

e.

[emuller]

Yes! OK I used the old kernel settings and recompiled the kernel. Everything is stable.

Great!!! so I didn't mess something up with crossdev. That would have been a chore to fix.

Now I just gotta find out which kernel option is the culprit.

[emuller]

It seems it was CONFIG_CPU_FREQ. Disabled that for now and the machine is stable... so far.

[emuller]

No It's still not stable. Better, but no.

I ran a simulation over night and it was alive but crashed pretty quick after I logged in.

It always wondered me that the fan doesn't speed up on when I run the simulations. Sure enough the cpu temp was over 80 C

I could get it to crash regularily with this cpu temp. emerge -s, kernel compile, anything loady. I let it sit for a while and it seems stable again.

And I've been looking all around: lots of people having this problem. Installed ibm-acpi, which doesn't seem to support the r52 fan.
The basic acpi thermal_zones don't support active cooling and have crazy default trips like 95 C for passive cooling...

So everything is supported for my thinkpad *EXCEPT* the cooling fan? Oh man this ia saddening and frustrating.

PLEASE HELP: Does anybody know how to get the fan running on a R52 ThinkPad?

I mean get the fan to speed up when the CPU gets hot. My laptop is overheating.

[emuller]

OK so it's a half a year since I reported the problem. I've been dealing with occasional crashes...

It helped to change BIOS CPU settings form MAX to THROTTLED (keeps CPU at 60C) or something to that effect.

But I have a what I believe is a fix now: upgrade from kernel 2.6.12 to kernel 2.6.16. System is stable for a week. So I reenabled MAX CPU in the BIOS and it goes up to 80C and still stable. Due to the correlation of the crashes with heavy drive usage, I think it is a disk driver issue rather than a CPU temp issue which was fixed in the new kernel.

Rejoice, I don't have to fear a lockup during a presentation ! For a linux user that's really embarassing.