Bug 125136 - anaconda crashes w/ invalid operand
Summary: anaconda crashes w/ invalid operand
Keywords:
Status: CLOSED NEXTRELEASE
Alias: None
Product: Fedora
Classification: Fedora
Component: kernel
Version: 2
Hardware: i686
OS: Linux
medium
medium
Target Milestone: ---
Assignee: Dave Jones
QA Contact:
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2004-06-03 00:31 UTC by Joe Harrington
Modified: 2015-01-04 22:06 UTC (History)
2 users (show)

Fixed In Version:
Clone Of:
Environment:
Last Closed: 2005-04-16 05:45:52 UTC
Type: ---
Embargoed:


Attachments (Terms of Use)
logs, etc., from /tmp (25.99 KB, text/plain)
2004-06-03 00:32 UTC, Joe Harrington
no flags Details

Description Joe Harrington 2004-06-03 00:31:01 UTC
From Bugzilla Helper:
User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.4.1)
Gecko/20031114 Galeon/1.3.11a

Description of problem:
I am doing a new install on an old machine (Intel 440LX chipset, dual
PII-300).

Anaconda exits abnormally and does not ask to save an anacdump.txt
file.  This follows immediately after a kernel-issued error (see
attachment).  It happens consistently after filesystems are made and
before the first package is loaded.  Switching to the shell prompt, I
can navigate around and save /tmp/syslog and friends, but if I ls the
mounted partitions, the ls never returns.  I can still switch screens,
so I assume the ls is in diskwait.  However, if I boot into rescue
mode, I can make filesystems on the partitions, mount them, and
list them without trouble.  I make those filesystems with
mke2fs -j -v /dev/hda1
mke2fs -j -v /dev/hda2

The problem happens consistently for the installaton type I am using:
The FTP install (from rescue CD, to a site serving the contents of the
FC2 DVD) errors out *right* after the mkfs.  This is what the
attachment refers to.  The CD install errors out a short time later,
after filesystems are made and one or two other progress bars
complete, but before the actual package loading.  I tried the FTP
install about half a dozen times with identical results.  I tried the
CDs only once.

The disk passes
badblocks -c 65536 -w /dev/hda

The disk held an FC1 installation before.  All media pass muster.
Memory passes (at least) 15 minutes of memtest86+.  The CDs installed
an even older uniprocessor PII-266 machine previously with an
"everything" install.


Version-Release number of selected component (if applicable):
FC2 Anaconda

How reproducible:
Always

Steps to Reproduce:
1.See above
2.
3.
    

Actual Results:  See above

Expected Results:  Clean install.

Additional info:

--jh--

Comment 1 Joe Harrington 2004-06-03 00:32:16 UTC
Created attachment 100809 [details]
logs, etc., from /tmp

Comment 2 Joe Harrington 2004-06-03 20:03:58 UTC
An FC1 CD install succeeds.

When doing an FC2 CD install, it fails right after putting up the
first RPM install screen, after "Preparing to install..."  Screen
switching is still possible with control-alt-F#.  The shell prompt
exists but does not respond.  I tried this three times and it was
consistent (I didn't check the shell status the first of those three
times).

When doing an FC2 FTP install from the rescue CD, it fails a little
earlier, just before the in the pause just before the "Formatting
/boot filesystem..."  status bar reaches the right end.  Screen
switching is still possible with control-alt-F#.  I still have a
shell.  In that shell, commands that do not access /dev/hda work,
those that do never return or print anything, but the keyboard and
Ctrl-Alt-F# screen switching still work.  For example, df works, but
ls /mnt/sysimage goes into the diskwait-like state (which I'll call
diskwait below).

works:
df
fdisk -l /dev/hda
dd if=/dev/hda of=/dev/null bs=10240 count=1
dd if=/dev/hda of=/dev/null bs=10240k count=1
dd if=/dev/hda of=/dev/null bs=10240k
 (I quit after ~1900 of ~2500 blocks)
dd if=/dev/zero of=/dev/hda bs=10240 count=1
dd if=/dev/zero of=/dev/hda bs=10240k count=1
dd if=/dev/zero of=/dev/hda bs=10240k count=100
 (I quit after 74 of 100 blocks)

fails:
fdisk /dev/hda
 (interaction works, but it diskwaits on write)
badblocks -c 65536 -w /dev/hda
 (diskwaits, later the partition table is found intact, so it didn't
    write much)
ls /mnt/sysimage
  (diskwaits)

Previously, I reported that 
badblocks -c 65536 -w /dev/hda
worked.  It didn't, but it didn't report any bad blocks.  I had
assumed that after an hour or so of run with no bad blocks reported,
it had checked the top quarter of the disk, more than enough to cover
/ and /boot.  In fact it had diskwaited.

Note that if I boot into the FC2 rescue mode, fdisk (including write),
mke2fs, mount, and ls all work on the disk.  The diskwait failures all
happen after the kernel error is reported.

--jh--


Comment 3 Joe Harrington 2004-06-03 20:39:04 UTC
If I install FC1 from the CD and then try to upgrade to FC2, it gets
the same error when the upgrade is reading package headers, or
immediately after that.

--jh--


Comment 4 Joe Harrington 2004-06-04 21:29:18 UTC
One more failed test.  I tried it on a different disk.  The first was
an IBM 25GB disk.  The new one is a Maxtor 160GB.  Same problem.
  
Since it works on FC1 and fails on FC2, and since hardware parts are
working fine under FC1 and pass the tests I can throw at them, I think
this is a real kernel bug.  Perhaps it has to do with the IDE driver
for the aging chipset.

Folks, I'm out of ideas.  I'm willing to try more things if someone
tells me what else to try.  Otherwise, I guess I'm stuck running FC1
on this machine.

--jh--

Comment 5 Dave Jones 2005-04-16 05:45:52 UTC
Fedora Core 2 has now reached end of life, and no further updates will be
provided by Red Hat.  The Fedora legacy project will be producing further kernel
updates for security problems only.

If this bug has not been fixed in the latest Fedora Core 2 update kernel, please
try to reproduce it under Fedora Core 3, and reopen if necessary, changing the
product version accordingly.

Thank you.



Note You need to log in before you can comment on or make changes to this bug.