AMD Segfaults in Bootsrap

jtza7 · n00b Joined: 23 Jan 2003 Posts: 7

I have seen many people in the forums with the exact problem that I have. All their CPUs where AMD by make, and nearly all where told that they have faulty hardware.

The mear fact that I'm writing this sujests that it's not the hardware but the software.

Situation:

You are happely and without any deviation performing a standard Gentoo installation. Nothing was skipped, you even zeroed your entier 30+ GB HD. After partitioning and formating and mounting and extracting the stage 1 tarball on the CD, and after you did an emerge sync, you proceed to bootstrapping.

All goes well - for a short while - you even get past a few packages. Then you hit gcc and presto! Segmentation Fault!

You go to the forum on your trusty Knoppix CD (or beeing a Gentoo Finatic, you probably have more than 2 PCs anyway

), and search the forums for seg faults (Well I did). After finding what seems to be an answer, you find yourself doing a tests on the hardware, in such detail that you are the prime bug reporter for AMD! 48 sleepless hours later, you frustratingly give up. I hope for your part, you figured out by now, it's NOT YOUR FAULT!!!!!

I suggest that the base system has some fatal flaw and that some sick person submited a bug as part of the new glibc (2.3.2-rc*). OK, I know, it is an unstable release. I guess that GCC might also be to blame. It might not be glibc, but mabe, since there is no build for an AMD older than the Athlon-XP, the build was performed on an Intel CPU and here we are. To quote the Linux From Scratch howto:

This package is known to behave badly when you have changed its default optimization flags (including the -march and -mcpu options). Glibc is best left alone. Therefore, if you have defined any environment variables that override default optimizations, such as CFLAGS and CXXFLAGS, we recommend unsetting or modifying them when building Glibc. You have been warned.

Also, don't pass the --enable-kernel option to the configure script. It's known to cause segmentation faults when other packages like fileutils, make and tar are linked against it.

I would like to put glibc 2.2.5-* into my new base directory, I know how, but how do I merge it into the portage database?

Any help, testimonials or anything else would help.

Jens. (South Africa)

bonsaikitten · Posted: Sat Aug 30, 2003 12:39 pm Post subject:

Maybe I misread you, but did you install an experimental version of glibc with athlon optimizations on an intel machine :?:

To · Posted: Sat Aug 30, 2003 12:46 pm Post subject:

I think you downloaded the wrong live cd version.

Tó
_________________

------------------------------------------------
Linux Gandalf 3.2.35-grsec
Gentoo Base System version 2.2
------------------------------------------------

jtza7 · n00b Joined: 23 Jan 2003 Posts: 7

Sorry for the sketchy details, here they are:

cyrillic · Posted: Sat Aug 30, 2003 7:22 pm Post subject:

NicholasDWolfwood · Apprentice Joined: 19 Jan 2003 Posts: 235

hrm, strange.

I've got an Athlon XP 1700+ running fine with no segfaults on bootstrap, and I have an Athlon Tbird 1Ghz upstairs running fine, but there were errors on bootstrap, specific segfaults on baselayout. What I did to solve it? I ran a stage3 setup and then when the base system was up and running (IE: I could login without the LiveCD) I set my new CFlags and CXXFlags and changed the CHOST setting from i486 to i686, then did an emerge -e system && emerge -e world overnight
_________________
AMD Athlon XP 1700+
Abit KG7-RAID
512MB PC2100 DDR266 Corsair/Kingston
40GB WD 7200RPM (37.1GB)
120GB WD 8MB 7200RPM (111GB)
160GB WD 8MB 7200RPM (149GB)
40GB WD 7200RPM (37.1GB)
Pioneer DVR-A06 4x DVD±RW
HP CD-RW 9500i (12x8x32x)

jtza7 · n00b Joined: 23 Jan 2003 Posts: 7

I Kapp'en Cyrillic, it was the 'ardware!

kyron

mangobrain · Tux's lil' helper Joined: 11 Sep 2003 Posts: 109

What are those of us with laptops supposed to do? Take them back? I don't fancy that, or opening it up. My bootstrap was segfaulting all over the place. I ruled out faulty RAM, ruled out corrupt sectors in the swap partition, didn't do anything other than follow installation instructions to the letter...
in the end, I just gave up and unpacked a stage1 tarball. I've since managed to do a perfectly good "emerge system" and compile 2 2.6.0-test5 kernels (I left devfs out first time round, and it can't be modularised

); as I type this, XFree86 is compiling away with seemingly no problems.

It's possible that it's just heat issues - whether you're overclocking or not. When a system's working hard, it can still be pushed over the edge, regardless of whether it's a desktop or a laptop. During the summer I get all manner of video card problems, from minor visual artefacts to BSODs, on a stock GeForce3 in my desktop machine (I'm probably making it sound worse than it is - this is not an everyday thing, and will only happen after hours of gaming

so don't go telling me I need to replace or clean my card).
I don't know for sure, but I suspect heat was causing the bootstrap failures on my laptop - the ACPI kernel on the install CD isn't functional (no acpid, no acpi kernel modules), so the fan speed never increased regardless of CPU temp. On the currently running 2.6 kernel, ACPI seems to work like a charm - I've listened to the fan sound change as extra cooling kicks in during heavy workloads.

Compiling gcc and glibc one after the other with virtually no other processes to get in the way is going to push any system's cooling to the limits, and sometimes it just isn't up to it. Whilst cleaning all your contacts can't hurt, it was probably removal of the dust that collects and clogs heatsinks and fans that did the trick.

Chickpea · l33t Joined: 03 Jun 2002 Posts: 846 Location: Vancouver WA

I have to chime in with the bootstrapping issue too. I also have an AMD athlon thunderbird 1.1ghz and I have not been able to bootstrap until today. and even that took a couple of times.

I dont know if you consider a computer freeze as a segfault but that is what happened.

It is really hard for me to pinpoint what the problem is however because I have tested the ram and the cpu and the video card. I can only guess that perhaps the motherboard is faulty somehow (ecs k7vzm kt133 chipset) or it is just getting too hot. About 3 or 4 months ago, I had a bad powersupply and because it still worked, it took a while before I figured it out....I had a different motherboard then so when it gave out, I tried the curren mobo (the ecs) and it worked for a while. It took a lot of googling before I came up with the bad powersupply...I had no idea that a bad power supply would still turn on......

So, because I used this bad power supply with current mobo, it is hard to tell if the random freezes were caused by the cpu going out or the mobo whackin' out?

Now that the weather is cooler, I have been able to get things up and running. I did have an issue the first bootstrap right after gcc. I thought I was home free and then it crapped out.

I wish there was a way of gradually bootstrapping rather than the all or nothing approach.

I could go on forever, but I should end this now before it gets way too long.

cyrillic · Posted: Fri Sep 12, 2003 2:06 am Post subject:

mangobrain · Tux's lil' helper Joined: 11 Sep 2003 Posts: 109

don't despair!
if you have to give up and go for a stage2 install, do so. you can always rebuild everything later if you like. I really don't know why the bootstrap process is so unstable - but if your machine can unpack a stage2 tarball and make it through an "emerge system", then just keep going, the kernel compiled for the install might turn out to be much more stable. I didn't bother rebuilding everything after install, but even so the only thing not specifically optimized for my machine is the compiler chain, and I can live with that. Right now I'm running a 2.6.0-test5 kernel, with ACPI working, using XFree 4.3 with NVidia's accelerated drivers, posting this via "links -g"... got a way to go yet if I want to build GNOME or similar, but it seems as if the instabilities do NOT continue once you move away from the livecd's kernel. I just finished verifying that ALSA is working, too.

taskara · Posted: Fri Sep 12, 2003 5:29 am Post subject:

I had this problem on a number of AMD machines.

I was getting segfaults at different times, like even extracting the stage tarball

I tested the cd on my AMD machine at home and it worked fine

I mem tested the machines, prime95 stress tested the machines, passed every time.

so I booted back to the gentoo live cd and passed

Roguelazer · Posted: Fri Sep 12, 2003 10:40 am Post subject:

Um. As you can see, I have an Athlon Thunderbird. I bootstrapped without any problems, have compiled kde four (yes, 4) times, have compiled gcc three times since firstboot and have compiled gnome several times too. No random segfaults here, and I have a 250W power supply, bargain basement fan and RAM that cost next to nothing.
_________________
Registered Linux User #263260

Chickpea · l33t Joined: 03 Jun 2002 Posts: 846 Location: Vancouver WA

kyron · Posted: Sat Sep 13, 2003 4:55 pm Post subject:

Chickpea · l33t Joined: 03 Jun 2002 Posts: 846 Location: Vancouver WA