2.0.30 SMP lockup - not gone yet...

eyal (eyal@eyal.emu.id.au)
Thu, 17 Apr 1997 22:28:46 +1000


I was to hasty when I posted that the SMP lockup I was experiencing has
gone. It is still there, but not on every run of the test program.

This is a dual PPro200 on an ASUS mb (I had the same problem earlier
when
I used the Supermicro DualPPro mb which I replaced after getting bugger
all
support from the local supplier and the US company).

On the other hand, my real program (which was failing when using much
memory but was OK when using less) now fails consistently.

Here is a sample output from my test program. Normally you should
only see the 'got' message and then the 'read' prompt should run
from 0/8 to 7/8, then the 'done' message. As you see there are five
system messages embedded here:

t: got 8192 blocks
t: read 131072 write 1/8Couldn't get a free page.....
Couldn't get a free page.....
t: read 23000 write 2/8hda: dma_intr: status=0xd0 { Busy }
hda: disabled DMA
t: read 24000 write 2/8ide0: reset: success
t: read 131072 write 7/8
t: done.

The program (this time) finished OK (or so it thought). However, at
other times the system would freeze hard.

I do not know if the last messages are the result of the 1st/2nd. The
program was copying a large file on a SCSI disk. An IDE partition was
used
for swap.

The test program (t.c) was emailed to the list before.

BTW, I did try the post 2.0.30 patch for SMP lockup prevention but it
had no
effect on this problem.

My system is straight 2.0.30, except the SCSI driver (ncr BSD port) was
upgraded to 1.18c.

If anyone is interested I will post the full details of my system along
with
the program.

--
Eyal Lebedinsky		(eyal@eyal.emu.id.au)