we have some troubles with kernel 2.0.3[67] and IBM Netfinity 5000 server
with more than 64mbytes of RAM.
Whe are moving fast to 2.2.x series kernel (that doesn't show this problem)
but I think that this report will be usefull for all the people that still
use the good "old" 2.0.3x kernel...
The hardware
------------
IBM Netfinity 5000 Pentium III 500mhz (no RAID) 256M Ram
The Software
------------
Caldera Linux 1.3 (libc5) with self compiled Kernel 2.0.37
The problem
-----------
After 15~18 minutes the "Warning" light turned on, after that
this kernel message come out:
kernel: Problem: block on freelist at 0109e7b0 isn't free.
and machine become unusable. The memory address showed above is always the
same and the problem is 100% reproducible.
The Netfinity hardware error log reports as follows:
DMI TYPE : 00
Source : SMI Hdlr
Error Code: MCA: Unrecoverable Error detected
Error Data: ProcID=1
Some of Test that I did (use FIXED font here)
---------------------------------------------
------+-------------+---------------+------+---------
ERROR |8p ISA Serial|RAM |Kernel|rc.serial
|Board | | |
------+-------------+---------------+------+---------
YES |YES |256M=2x128Dimm | B |YES
------+-------------+---------------+------+---------
YES |YES |256M=2x128Dimm | B |NO
------+-------------+---------------+------+---------
YES |NO |256M=2x128Dimm | B |YES
------+-------------+---------------+------+---------
YES |YES |128M=1x128Dimm | B |YES
------+-------------+---------------+------+---------
NO |YES |64M=1x64Dimm | B |YES
------+-------------+---------------+------+---------
YES |NO |256M=2x128Dimm | B |NO
------+-------------+---------------+------+---------
YES |YES |64M=2x128Dimm | B |YES
| |mem=64M on lilo| |
------+-------------+---------------+------+---------
NO |YES |63M=2x128Dimm | B |YES
| |mem=63M on lilo| |
------+-------------+---------------+------+---------
NO |YES |62M=2x128Dimm | B |YES
| |mem=62M on lilo| |
------+-------------+---------------+------+---------
NO |* |* | A |*
------+-------------+---------------+------+---------
A = monolithic kernel 2.0.37 (828k kernel code, 384k reserved, 3936k data)
B = monolithic kernel 2.0.37 (772k kernel code, 384k reserved, 3916k data)
* = everything
Conclusion
----------
The problem is apparently (clear for me ;) connect to the "kernel code size"
and the 64M barrier....
Kernel "A" works fine:
Memory: 256932k/262080k available (828k kernel code, 384k reserved, 3936k
data)
Kernel "B" got the error:
Memory: 257008k/262080k available (772k kernel code, 384k reserved, 3916k
data)
You can download kernel "A" and "B" image with relative dmesg e .config from
http://www.teamfab.it/linux/netfinity/problem/
Solution
--------
Just made a bigger kernel, including other ,not needed, drivers.
The Bug
-------
Write the patch here ;)
Thanks for your time,
Luca Montecchiani
-- Office Email: luca.montecchiani@teamfab.it Home Email..: m.luca@iname.com Home Web....: http://i.am/m.luca
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/