From: Gerald Heim (heim@peanuts.informatik.uni-tuebingen.de)
Date: 07/15/92


From: heim@peanuts.informatik.uni-tuebingen.de (Gerald Heim)
Subject: Re: System crash with SCSI and 16M
Date: 15 Jul 1992 14:27:40

In article <1992Jul14.172812.189@athena.mit.edu> Philip Balister <balister@pdsrvr.salem.ge.com> writes:

   Here's the problem -

   A colleague running linux has upgraded his linux box to 16Meg of Ram.

   The box is a 386-DX with AMI bios. The disk controller is a ST-01 with
   48 Meg SCSI disk. The system works with 4 meg but dies during boot
   when it has 16M ram.

   It dies as follows:
   .
   .
   .
   16384/38912 free blocks
   10851/12970 free inodes
   general protection: 0000
   EIP: 0008:00002ac46
   EFLAGS: 00010246
   FS: 0017
   base: 04000000, limit 000A0000
   pid: 1, process nr:1
   8a 13 88 10 ff 45 0c eb d2 90

   Anyone have any suggestions?
   Has anyone else tried 16M of ram with a SCSI disk? I remember a thread
   regarding the 16M limit on usable ram but don't remeber this being a
   symptom. BTW, the kernel version is 0.96b and 0.96bpl2, both barf.
   Philip

   balister@salem.salem.ge.com

This is nearly the problem I had when I upgraded my 386sx20 (32k cache)
from 4 MB to 12 MB using two 4MB SIMMS
It seems that linux still works in the 4MB, but whenever some part
of the new memory is invoked, the programs would crash.
The kernel and parts of the buffer cache still worked, it seemed...
(yes, parts of the buffer cache, once a sync destroyed my whole root
partition) getty and update worked, too. But no way to get a shell running.
1st solution: turning off the cache fixed this problem
But this may be a side-effect....
Could there be a bug in the virtual memory management that occurs only
if not all banks of memory are of the same size???

btw: msdos :( and minix :(( still work WITH the cache...

bye
(wondering) gjh