Arch Linux

tuxfusion

Member

Registered: 2007-12-31

Posts: 98

[SOLVED]Compiling kernel fails with segfault

Hi guys,

I'm out of answers.

The situation :
Compiling kernel randomly stops with : "segmentation fault" and never runs through.
Reissuing make will make it continue until it will fail at some other point.

Background:

The system feels completely unstable to say the least, but nothing 100 % reproduceable ( vuze(java) crash, firefox/chromuim flash sites crash every 120 seconds, X crashes once a day , for instance when moving tvtime around, kernel oops daily, sometimes hard lockups, logs of yesterdays oops : http://pastie.org/private/imrqdzdbejdb3jxn4czuw )

But only the kernel compile is 100 % reproduceable so this must be fixed asap.

Specs:
intel e8400,nvidia gtx260, 2 G ddr2,Gigabyte EP45-UD3R, "gigs of space", off. arch kernel 2.6.33-ARCH , x86_64,  "bigtime" cooling :

cpu temps 50 C° max  during compiling

What configs/packages are important ? I got paranoid and ditched my old makgepkg.conf and used a newer *.pacnew file , I used to have  :MAKEFLAGS="-j3" but not in this config :
http://pastie.org/private/4ryshivi2hgyqenjlkicg

local packages:
local/gcc 4.5.0-1 (base-devel)
    The GNU Compiler Collection
local/gcc-libs 4.5.0-1 (base)
    Runtime libraries shipped by GCC for C and C++ languages

Anything else updated daily.

The blowup:

http://pastie.org/private/hoamev0sli7zapbr0xvpbq

What I tried:
fsck -f: it fixed two inodes, but still the same behavior
badblocks -nvs : three hours of badblock testing, 0 hits
memtest : two passed runs, no errors
stress from AUR, 15 minutes :
stress --cpu 8 --io 4 --vm 2 --vm-bytes 128M --timeout 15m
stress: info: [26841] dispatching hogs: 8 cpu, 4 io, 2 vm, 0 hdd

stress: info: [26841] successful run completed in 900s

Max coretemp was 52 C° during stress.

The usual footnote statement ^_^ :
On w7 _everything_ runs stable.
And not only stable , I'm able to OC this box to 4.5 GHz and game under w7 without a prop for hours, well this is what it was build for at some point.

Conclusion:
So I'd say hardware wise this box kicks ass and now software wise my Arch kicks me in the a**.
I doubt there is a faulty hardware part, but what else could it be ?

Last edited by tuxfusion (2010-05-17 22:48:06)

tuxfusion

Member

Registered: 2007-12-31

Posts: 98

Re: [SOLVED]Compiling kernel fails with segfault

Update:
Trying to compile another big package just to have another result, gcc itself now :
- segfault after 4 m18 s
- segfault after 5 m50 s

Update²:

I think I found a faulty BIOS setting in my non-OC profile concerning DRAM termination, which said "normal" where "auto" would be correct.
My OC-profile has this setting correct and since I run this profile in Arch only it might explain that w7 did not fail.
Still have to verify this but i makes perfect sense, only gcc seems to trigger this. Compiling gcc still running with over 10 minutes in the game now.

After hitting 4.5 GHz with OC I must have reset the non OC-profile wrong.

*slams head on desk*

Last edited by tuxfusion (2010-05-17 13:10:04)

tuxfusion

Member

Registered: 2007-12-31

Posts: 98

Re: [SOLVED]Compiling kernel fails with segfault

Okay after ~30? kernel compile tests, here is where I'm at, it was of course not just setting something from "normal" to "auto" in BIOS. :
default bios setting, no OC : compiling passes
mild OC 3.8GHz : was a heavy fight , reduced some MGhz, lowered CAS latency, did not use manufactures ram voltage but bios "auto" , compiling now passes . God knows why system ( gcc, kernel, new power outlet is so picky. It worked flawlessly in the past)
heavy OC from 3 GHz to 4.5 GHz : compiling always worked. No change needed

Last edited by tuxfusion (2010-05-17 22:54:24)