LinuxQuestions.org
Share your knowledge at the LQ Wiki.
Go Back   LinuxQuestions.org > Forums > Linux Forums > Linux - Distributions > Linux From Scratch
User Name
Password
Linux From Scratch This Forum is for the discussion of LFS.
LFS is a project that provides you with the steps necessary to build your own custom Linux system.

Notices


Reply
  Search this Thread
Old 05-15-2005, 04:18 PM   #1
Yerp
Member
 
Registered: Jun 2004
Distribution: Turtle Kevux 0.8.8
Posts: 129

Rep: Reputation: 15
2.6.11 kernel or glibc segfault everywhere..?


Okay, this issue is very .. vague .. to say the least.

The issues themselves, I am unable to specifically locate where the problems are and these are my observations/suspicions from working with and around these problems.

The Situations
The original systems themselves were purely lfs based, built from many successful builds. I had felt it was time to start experimenting with 2.6 kernel.
I had originally used linuxthreads and not nptl, so this seems to rule out any problems with nptl. System was successfully built with glibc-2.3.4 (stable/non-cvs) and linux-2.6.11.4. But had serious problems with tst-cancel2.out failures, in which everyone seems to recommend going around, so i skipped the test, which resulted in binutils make check failures that were also recommended to skip around. (Recommended by researching the LFS support lists, google, and linuxquestions)

I then went into, pretty much, the blfs stage of installation with no problems.
Type: startx. So far so good. run any program, 50% will Segfault: SIGV. Non-X related programs also silently segfault (witha few that openly spitout a segfault): openssh seemed to segfault when a proper connection is established and user gets disconnected immediately on a successfully login.

Okay...SO I bootstrap the OS. My unstable LFSHOST -> builds new LFS -> Builds new LFS. For about 6 Operating System installations.

Received around 33% Seggfaults this time...nice improvement, helped me run the Anime organization pretty interesting...couldn't tell it apart from a stable Windows - Assuming you know what I mean.

Frustrated, I rebuild entirely off the LFS 6.0 live-cd, running a 2.6.9 kernel. Only differences between my code at the tome and the LFS-6.0 CD were the kernels. 2.6.11 and 2.6.9. Building off of the 2.6.9 returned no glibc-errors whatsoever. - interesting enough.. But I am insanely stubborn so I build the 2.6.11 system again from the 2.6.9 live-cd.

Getting the same problems during make checks as before, I decide not to go on and start hacking the system. 1 Major Hack: nptl hacked to build off of a linuxthreads system -> Build successful, with the glibc tst-cancel2 error, but without binutils make check errors! Theres some success. (15% segfaults)

By now I have been playing bleeding edge, and ended up with a glibc-2.3.5(+nptl) as well as the linux-2.6.11.8. (then I bootstrap it two times)

Now I copy the core binary (built under a Pentium4m) and use that to built an LFS under an athlon-xp. Not one Make check failure...glibc went straight through...Aside from other hardware problems that resulted in an MBR corruption, system worked well, but still showed signs of segfaults: but I never got a chance to test the programs that segfault the most: X programs. As the system was lost due to the MBR problem..corrupted the system anyway!! - damn my luck...

So I get lazy and copy my intel binary that has the 15% segfault to the AMD. System has yet to crash, but I've been to busy elsewhere to test the X progs.

I also placed the binary onto 3 different other intel systems: Pentium2, another Pentium4m, and Pentium4. Each one segfaults consistently in different programs, the OTHER pentium4m segfaults the lest, while the pentium4 seems to have jumped up to 30% segfaults....WTF? Its a binary...

Bottom Line
So, I have reason to believe that something with the 2.6.11 kernel and random intel architectures seem to have SIGV segfault problems left in right, for reasons that are beyond me. But, I do not believe I have enough evidence to support this. But recently, I found some numerous discussions pertaining to random segfaulting and the 2.6.11 kernel.

here is an example: http://forums.gentoo.org/viewtopic-t-313030.html <-- These segfault issues are identical to mine.
If any of problems there are identical to the problems I have, then this suggests multiple problems in regard to segfaults....if thats the case who starting making the kernel sloppy after 2.6.9? - ah well shit happens

help..?
I'm not sure where I need the help..but its obvious that I need some. Or maybe I should just wait for 2.6.12 kernel? I am hoping there is somebody reading these who has at least an idea of how to solve this (and the time to read this massive post)

Current Status
I used IBM's Secure Data Disposal a few times on my IBM Thinkpad (the lfs host system), installed SUSE 9.3, and am using it as the host at the moment.
The tst-cancel2 patch has now been applied, and I get an error with tst-cancel4.out in make check....! I bypassed it and binutils-2.16 went along fine. I will post any results from this installation when its done.
 
Old 05-18-2005, 10:43 AM   #2
cate@d.o
LQ Newbie
 
Registered: Mar 2005
Distribution: Debian GNU/Linux
Posts: 9

Rep: Reputation: 0
SEGV are usually not related to kernel. I think there is a problem in some library/programs.

But maybe you should check your memory with memtest86+. Random SEGV are also caused by corrupted memory.
 
Old 05-18-2005, 12:46 PM   #3
Yerp
Member
 
Registered: Jun 2004
Distribution: Turtle Kevux 0.8.8
Posts: 129

Original Poster
Rep: Reputation: 15
yeah, I;ve been pretty sure that its not the kernel, but having spent so much time on this problem, I had to spit out my worries.

Anyway, I checked my RAM using memtest86+

No problems there.

However, I am running ont the system as its building itself at the moment, and so far the programs are no longer segfaulting.

I am pretty sure now that it is a harddrive problem that was solved by the governmental wipe..including the MBR.

But what had confused me the most was that the programs didn't fail as much under 2.6.9 kernel...which threw me into a loop.

Perhaps the 2.6.11 kernel was better at detecting these problems??
 
Old 05-19-2005, 12:58 PM   #4
Yerp
Member
 
Registered: Jun 2004
Distribution: Turtle Kevux 0.8.8
Posts: 129

Original Poster
Rep: Reputation: 15
Damn, still got a 5% random segfault problem and memtest86+ shows no problems whatsoever....at least it's down to 5%...
 
Old 05-19-2005, 08:13 PM   #5
madluther
Member
 
Registered: Aug 2004
Distribution: LFS
Posts: 350

Rep: Reputation: 31
I'm seeing a similar problem. I have been cross-compiling LFS-6.0 for an ancient laptop (an i586) and it failed on the make check, specifically on nptl/tst-cancel17. Googling for a reason reveals that kernel 2.6.11 is to blame which I was running on the host build system, I have built many LFS systems as you have and only had problems with glibc on hosts running 2.6.11 kernels, using an older kernel 2.6.9 resulted in clean builds. Unfortunately none of this explains your segfaults.

Mad.
 
Old 05-20-2005, 03:42 PM   #6
Yerp
Member
 
Registered: Jun 2004
Distribution: Turtle Kevux 0.8.8
Posts: 129

Original Poster
Rep: Reputation: 15
okay, I recomplied the system AGAIN; except this time I took the following notes in mind:

1) I need to only make 1 major change to find this problem (or 1 change at all as my programming ethics go..)

2) I have had to make many patches to programs to get them to compile under glibc-2.3.5 <-- only the blfs part of the LFS install is subject to any of these problems...at least directly.

3) I was going to build off of the almost identical system.

So, my change was: nows using glibc-2.3.4

I will make further notes about my problems and hopefully have none related.

Side Note
I am beginning to think I was having a string of seperate problems and now that I have wired out most, dropping back to glibc-2.3.4 may resolve my segfault problems.

The most popular patch I made to numerous programs was fixing strcasestr
I am thinking that a major amount of changes in glibc-2.3.5 may result in these random inconsitancies as most programs aren't written to work with the glibc-2.3.5 library..at least not yet.
 
  


Reply


Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is Off
HTML code is Off



Similar Threads
Thread Thread Starter Forum Replies Last Post
Kernel version and glibc ashley75 Linux - General 1 03-22-2005 02:54 AM
glibc recompile for kernel 2.6.10 dhave Slackware 2 01-24-2005 04:59 AM
Alsa + new kernel = segfault JohnGalt00 Red Hat 0 09-03-2003 06:16 PM
Segfault on glibc AxeZ Linux From Scratch 5 02-07-2003 02:32 PM
kernel doc (glibc) don1 Programming 0 01-25-2003 03:16 PM

LinuxQuestions.org > Forums > Linux Forums > Linux - Distributions > Linux From Scratch

All times are GMT -5. The time now is 07:35 AM.

Main Menu
Advertisement
My LQ
Write for LQ
LinuxQuestions.org is looking for people interested in writing Editorials, Articles, Reviews, and more. If you'd like to contribute content, let us know.
Main Menu
Syndicate
RSS1  Latest Threads
RSS1  LQ News
Twitter: @linuxquestions
Open Source Consulting | Domain Registration