Gentoo Websites Logo
Go to: Gentoo Home Documentation Forums Lists Bugs Planet Store Wiki Get Gentoo!
Bug 146204 - i586 unsupported in 2006.1 install media
Summary: i586 unsupported in 2006.1 install media
Status: RESOLVED FIXED
Alias: None
Product: Gentoo Release Media
Classification: Unclassified
Component: InstallCD (show other bugs)
Hardware: All Linux
: High normal (vote)
Assignee: Gentoo Release Team
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2006-09-03 19:47 UTC by Wendall Cada
Modified: 2006-09-25 14:16 UTC (History)
1 user (show)

See Also:
Package list:
Runtime testing required: ---


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description Wendall Cada 2006-09-03 19:47:01 UTC
Install target AMD K6-2 500

Installing with install-x86-minimal-2006.1, installer indefinately hangs at initramfs. If using an earlier install cd 2005.1 for example, system boots fine. However, there is an additional problem with the 2006.1 stage1-x86 tarball. I get an illegal instruction when trying to chroot.

It would appear that everything is compiled i686 now. This effectively eliminates installation on anything not i686 or greater.

I did get a successful install up and running with a 1.4 stage1 tarball.

Wendall
Comment 1 Andrew Gaffney (RETIRED) gentoo-dev 2006-09-03 19:55:03 UTC
Ok, this is the second report we've gotten of this (other was on a K6-2 as well), so I'm guessing this is confirmed. I'm not sure why this happened, since the stage1-x86 was compiled with a i386-pc-linux-gnu CHOST and CFLAGS="-mcpu=i686", which just optimized for 686, but still allows it to run on 386.
Comment 2 Andrew Gaffney (RETIRED) gentoo-dev 2006-09-03 20:02:38 UTC
Is it possible that the meaning of -mcpu changed for x86 in gcc-4.1.1? Did it change to -mcpu/mtune instead of -march/mcpu, to match most other arches?
Comment 3 Andrew Gaffney (RETIRED) gentoo-dev 2006-09-03 22:23:39 UTC
Ok, I was wrong about -mcpu. The actual CFLAGS used in the build was:

CFLAGS="-O2 -mtune=i686 -pipe"
Comment 4 Wendall Cada 2006-09-04 13:11:06 UTC
Using 2005.1 install cd. I was able to get to the point of compiling a kernel.

make.conf:
CHOST="i586-pc-linux-gnu"
CFLAGS="-O2 -march=i586 -pipe"

error on kernel compile:
  CC      mm/page_alloc.o
{standard input}: Assembler messages:
{standard input}:0: Warning: end of file not at end of a line; newline inserted
gcc: Internal error: Segmentation fault (program cc1)
Please submit a full bug report.
See <URL:http://bugs.gentoo.org/> for instructions.
make[1]: *** [mm/page_alloc.o] Error 1
make: *** [mm] Error 2

Will reinstall gcc from working i586 compile from my binhost and retry. Just updating on my experiences for now.

Wendall
Comment 5 Wendall Cada 2006-09-04 13:22:26 UTC
This is likely a hardware issue, or the old 2005.1 kernel that's causing the issue. gcc is segfaulting at random places. I will get a working kernel and test on the system once I can boot.

Wendall
Comment 6 Wendall Cada 2006-09-04 13:43:24 UTC
make -j1 resolves the issue of segfaults. Maybe a gcc bug, dunno. More results to come.

Wendall
Comment 7 Andrew Gaffney (RETIRED) gentoo-dev 2006-09-04 14:07:38 UTC
Okay, I think we know that the stage3-586-legacy-nptl I gave you works fine.

Back to the actual issue of the bug, I've got a few leads on access to K6-2 boxes to try and reproduce this problem with -mtune=i686. I plan on create a chroot with the stage3-586-legacy-nptl and then recompiling bash and glibc with -mtune=i686 to see if they break. If so, we can probably point at gcc-4.1.1 for this.
Comment 8 Wendall Cada 2006-09-04 15:08:27 UTC
The segfaulting continued throughout the setup. Now that I have a properly built kernel, this behavior is gone. I'm moving on to testing CFLAGS="-mtune=i686" in a chroot for bash and glibc here shortly and will post my results.

Wendall
Comment 9 Andrew Gaffney (RETIRED) gentoo-dev 2006-09-04 17:27:55 UTC
I've got another confirmed illegal instruction on NeddySeagoon's K6 box with the stage1-x86-2006.1. It's not a kernel issue or anything, since this is from a 1.4_rc1 LiveCD running 2.4.20. I'm installing using my stage3-586-legacy and going to do a few tests to see if I can reproduce this with -mtune=i686.
Comment 10 Jiri Macha 2006-09-05 01:01:45 UTC
I can confirm the bug too.
If I am installing install-x86-minimal-2006.1 - selection 
gentoo or gentoo-nofb, installer indefinitely hangs at initramfs.

I am using old PC -  Pentium 200 MHz, MMX, Matrox Millenium II 

I tried to select options (disable SATA etc.) but nothing helped.
Only if I select option debug, I can get command prompt.

I have tested also minimal installation 2005.1 and there seems to be
everything OK - I've get command prompt for the default options.
Comment 11 Andrew Gaffney (RETIRED) gentoo-dev 2006-09-05 05:28:02 UTC
My results were inconclusive with CFLAGS="-O2 -mtune=i686 -pipe" and the i586 CHOST. I've switched the CHOST to i386-pc-linux-gnu and am re-bootstrapping. This is about as close to the broken stage1-x86 as I can get.
Comment 12 Chris Gianelloni (RETIRED) gentoo-dev 2006-09-05 07:00:20 UTC
There is no LiveCD for < i686, changing to InstallCD.

Exactly what point (what output) does it stop when booting the CD?
Comment 13 Wendall Cada 2006-09-05 10:45:34 UTC
As stated before, at the initramfs line during the boot sequence. Hangs indefinately.

As a side note, this box doesn't belong to me, realized it was overclocked which was causing segfaults and system instability. Issue with that is resolved, however, the illegal instruction problem still persists.

I tested with building for a chroot on this AMD k6-2 450 machine:
CFLAGS="-mtune=i686" emerge glibc bash
It chroots just fine, so it appears to not be a -mtune issue.

HTH

Wendall
Comment 14 Jiri Macha 2006-09-07 21:56:22 UTC
More detailed description:

If I let the installation CD in the CD ROM, 
reboot PC and do not interact with the installation 
process, this are the last messages I see on the screen:

.
.
.
>> Copying read-write image contents to tmpfs
>> No cdupdate.sh script found, skipping ...
>> Booting (initramfs) ...


Now there is pause approx. 1~2 seconds an then
LED diode of the CD ROM together with LED diode
of the hard drive blinks approx. three times.

This is the last activity of the PC. I let
it approx. 4 hours as it is, and nothing happened.

At this moment I can:

 - type commands via keyboard on the screen
 - reboot PC using keyboard - using CTRL + ALT + DELETE

I am sure that this old PC is OK. It was running
Red Hat Linux 7.3 before, memory test doesn't
report any problem, and if I use gentoo 2005.1 
minimal installation, everything is OK.

If I used the same installation CD (2006.1) to boot 
on the AMD 64 platform it boots gentoo successfully.
So there is also no problem on it.

Would it be possible to reproduce the problem
using virtual PC - using BOCHS ?
Comment 15 Chris Gianelloni (RETIRED) gentoo-dev 2006-09-25 14:16:22 UTC
This is fixed now on the mirrors.  You can verify that the mirror has the correct ISO by looking at the .DIGESTS file.  The MD5 should be: 560a09038504149992360140ad2b1793