tlug: Remove some memory and Linux won't boot

[Jonathan Q <jq@example.com>]

Here's my weird effect of the day  :-)

I planned to move some memory from one machine to another, since the machine
I primarily use has less memory than the one that's mostly just used as my
NAT box for the rest of my network.

The memory donor machine is a Pentium 100 with six SIMM sockets. The sockets
are filled as follows:

0-a:  8 meg 60 ns EDO
0-b:  8 meg 60 ns EDO
1-a:  16 meg 60 ns EDO
1-b:  16 meg 60 ns EDO
2-a:  16 meg 60 ns EDO
2-b:  16 meg 60 ns EDO

Total: 80 MB

I had an append=80M line in lilo.conf so all the memory would be recognized.
I removed the line, shut down the system, removed the memory in 2-a and 2-b,
and powered up again.

Everything looked normal, 48 meg detected at POST, the LILO prompt came up,
and the boot started.  It got as far as starting fsck (using fsck -a
/dev/hda) and then barfed with a GPF and stopped right there.

Putting back the memory back in 2-a and 2-b solved the problem.  Shuffling
the memory around so that which two sticks were in 2-a and 2-b made no
difference to the problem.  I used to run this machine with 48 meg in it for
a long time before I put the two extra sticks in, but now I can't boot
without them.

Does Linux store memory size info from an append=80M line in lilo.conf
somewhere else and it's biting me, or is something totally unrelated to that
behind this?

Thanks in advance,

Jonathan

-------------------------------------------------------------------
Next Technical Meeting: July 10 (Sat), 13:00  place: Temple Univ.
*** Topic: High Performance JAVA, by Matt Welsh
Next Nomikai: August 20 (Fri), 19:30 Tengu TokyoEkiMae 03-3275-3691
-------------------------------------------------------------------
more info: http://www.tlug.gr.jp        Sponsor: Global Online Japan