View previous topic :: View next topic |
Author |
Message |
Remillard Apprentice
Joined: 07 Mar 2003 Posts: 200 Location: Irvine, CA
|
Posted: Wed Feb 11, 2004 9:15 am Post subject: Stage1 Install - Stuck with Segfaults - [SOLVED sortof] |
|
|
Howdy folks,
Well I'm in the midst of trying a few different things but I'm now officially stuck.
I have the Pentium 4 LiveCD. I DID check the md5sum, so I'm assuming it's correct. I also have Knoppix 3.3.
Installation History:
Attempt #1: Got /dev/hda partitioned for / (reiserfs), /boot (ext3), /home (reiserfs), /media (reiserfs). Last two aren't mounted. stage1 tarball extraction fails, tar segfaults or EOF's. Can't remember precisely.
Attempt #2: Started with Knoppix. Mounted /mnt/gentoo/ and /mnt/gentoo/boot. Got stage1 from website. Extracted successfully. emerge sync freezes entire system after first site can't be reached.
Attempt #3: Reboot with Knoppix, put it in failsafe mode, start eth0. Everything okay. Put the SYNC server thing in make.conf. Make sure there is a SYNC_RETRY field listed. emerge sync now works fine. Attempt stage1 bootstrap. Segfaults at gettext.
Attempt #3.1: Restart bootstrap. Redoes portage. Redoes texinfo, and gettext, and then starts binutils. Segfaults during binutils.
Attempt #4: Reboot with Gentoo LiveCD again. Do the whole mounting, chrooting and miscellaneous. Try bootstrap again. Segfaults. Error (roughly)
Code: |
Unable to handle kernel NULL pointer dereference at virtual address 00000040
printing eip:
0021e07f
*pde = 00000000
Oops: 0000
sis900 floppy serial isa-pnp cloop usb-storage hid usb-ohci ehci-hcd usbcore
.......
|
So I'm sort of stuck here. I can't seem to get bootstrap to work completely under Knoppix. I can't seem to get bootstrap to work under Gentoo LiveCD (1.4). I'm hoping someone can suggest an out here, or something to check. Further information:
Code: |
CFLAGS="-march=pentium4 -O3 -pipe -fomit-frame-pointer -funroll-loops -fprefetch-loop-arrays"
CCFLAGS="${CFLAGS}"
CHOST="i686-pc-linux-gnu"
|
Other than the divergence to Knoppix because of the tarball problem, I've gone by the book. I think the CFLAGS are high end, but not unsafe. I'm just not sure about all the segfaulting. Knoppix seems to run fine, sort of did a little burn-in with it, so I'm reasonably certain the hardware is solid, no overheating or whatnot. It just seems sort of random. Attempts 3 and 3.1 kind of bug me, segfaults one time through, but the 2nd time through it's okay?
Anyhow, here's hoping this looks familiar to someone.
Oh, hardware:
Pentium 4 2.66 GHz
512 M RAM
120 G HD
SiS chipset with embedded ethernet (this explains the sis900 in the segfault code above).
Regards,
Remillard _________________ This signature is printed with 100% post-consumer recycled electrons.
Last edited by Remillard on Thu Feb 12, 2004 6:30 pm; edited 1 time in total |
|
Back to top |
|
|
Remillard Apprentice
Joined: 07 Mar 2003 Posts: 200 Location: Irvine, CA
|
Posted: Wed Feb 11, 2004 4:59 pm Post subject: |
|
|
I remembered this morning that at one point, Knoppix had a segfault during boot as well. Though I tried again with failsafe and had no problems.
This must be a sticky one though. The Gentoo community usually has some idea about all sorts of these install issues .
Remi _________________ This signature is printed with 100% post-consumer recycled electrons. |
|
Back to top |
|
|
Panther_7 n00b
Joined: 04 Jun 2003 Posts: 46 Location: FL
|
Posted: Wed Feb 11, 2004 6:27 pm Post subject: |
|
|
Doesn't pentium4 in GCC produce some kind of incorrect code? I think somewhere in the make.conf or make.conf.sample it says that the pentium4 option is bugged. They say to compile with pentium3 instead. Hope that helps. |
|
Back to top |
|
|
pixie Guru
Joined: 29 Dec 2002 Posts: 305
|
Posted: Wed Feb 11, 2004 6:44 pm Post subject: |
|
|
-march=pentium4 is fixed now I think. I believe it was only early releases of GCC3.2 that had a problem. |
|
Back to top |
|
|
Remillard Apprentice
Joined: 07 Mar 2003 Posts: 200 Location: Irvine, CA
|
Posted: Wed Feb 11, 2004 7:50 pm Post subject: |
|
|
Thanks, I'll keep that in my hip pocket. It's possible that the issue is mounting with bind (see the Argh!! ... Hardware? post nearby). If I can't get that to work consistently, I'll try shifting down to -march=pentium3 (can't hurt!). _________________ This signature is printed with 100% post-consumer recycled electrons. |
|
Back to top |
|
|
pixie Guru
Joined: 29 Dec 2002 Posts: 305
|
Posted: Wed Feb 11, 2004 8:05 pm Post subject: |
|
|
You may want to test your memory. The problems seem very random. |
|
Back to top |
|
|
Remillard Apprentice
Joined: 07 Mar 2003 Posts: 200 Location: Irvine, CA
|
Posted: Wed Feb 11, 2004 9:01 pm Post subject: |
|
|
Do you have any suggestions for a program to do this with? I wouldn't mind finding something that can thrash the processor, memory, and harddrive a bit, a hardstyle burn-in. _________________ This signature is printed with 100% post-consumer recycled electrons. |
|
Back to top |
|
|
pixie Guru
Joined: 29 Dec 2002 Posts: 305
|
Posted: Wed Feb 11, 2004 9:41 pm Post subject: |
|
|
There should be an option to boot into memtest86 or whatever it's called on the live cd.
never done one myself though. but then I've never had symtomps of dodgy memory sticks since I started using Linux a few years back. |
|
Back to top |
|
|
Remillard Apprentice
Joined: 07 Mar 2003 Posts: 200 Location: Irvine, CA
|
Posted: Thu Feb 12, 2004 6:30 pm Post subject: |
|
|
Success!! (sort of)
The memtest kernel option is fantastic. I ran that and within about 10-15 I got the bad news that I had a bad stick of memory. I let it run overnight and it just kept racking them up on a particular test.
The cool thing is that this is on the Gentoo LiveCD. Just about everything you need is really on that CD! While I'm not thrilled that there was a memory error (because it slows down my install to wait for the overnight reship) I'm VERY glad I was able to find it.
And I have to admit the http://www.xtechnology.com/ folks were really cool about it. Actually they used the memtest-86 program so there was no issue about having to do retesting or further benchmarking. I think they're still mostly a MS house, but it was refreshing not to have to deal with the dreaded "let's-go-down-the-checklist" type of technical support.
Well here's hoping tomorrow evening goes far better than things have gone so far .
Regards,
Remillard _________________ This signature is printed with 100% post-consumer recycled electrons. |
|
Back to top |
|
|
irf2003 Veteran
Joined: 10 Sep 2003 Posts: 1078
|
Posted: Thu Feb 12, 2004 9:09 pm Post subject: |
|
|
watch your CFLAGS, tone them down.
I build about 2 to 3 gentoo boxens a week, and my
findings as far as PIV boxes are concerned, is that the following
compiler flags seem to be the safest and most high yield in terms of perfrmance gains
Code: |
CFLAGS="-march=pentium4 -mcpu=pentium4 -mfpmath=sse -msse2 -O2 -fomit-frame-pointer"
|
I would stay away from "-O3" and "Os", as I consider them unsafe.
with some of the current packages in the portage tree, you
may have to force "-mno-sse2" (xine-lib)
but otherwise, you will have no problems
HTH |
|
Back to top |
|
|
michaelb l33t
Joined: 06 Jun 2002 Posts: 686 Location: Ann Arbor, MI
|
Posted: Thu Feb 12, 2004 9:25 pm Post subject: |
|
|
So far as I understand GCC flags, having both an -march as well as an -mcpu set will end up doing the same thing as just the -mcpu. _________________ Behold, The power of SEARCH! |
|
Back to top |
|
|
irf2003 Veteran
Joined: 10 Sep 2003 Posts: 1078
|
Posted: Fri Feb 13, 2004 4:33 pm Post subject: |
|
|
michaelb wrote: | So far as I understand GCC flags, having both an -march as well as an -mcpu set will end up doing the same thing as just the -mcpu. |
some ebuilds in the portage tree filter "-march" flag, that
why I have "-mcpu" flag there
HTH |
|
Back to top |
|
|
|