Made another attempt to install Redhat 5.0 (kernel 2.0.31 on Intel).
The attempt ended in
Kernel panic: scsi_free: Trying to free unused memory
In swapper task - not syncing
and the machine was dead (no console switch, CapsLock, CtrlAltDel).
Now this machine has good SIMMs (never any crash, and always
in active use), so this means a kernel bug. I'll cc to the kernel list.
I have now seen this machine crash 3 times during a RedHat 5.0 install,
and at least twice this was a kernel bug. Maybe 2.0.31 is no good
(on my hardware). The third time things ended in an infinite chain
scsi: aborting command due to timeout - resetting - Sent BUS DEVICE RESET -
Sending DID RESET
(repeated hundreds of times).
[Such a nonsense: if fifty resets do not help, the fifty-first one
will be no good either. Resetting the SCSI bus (for Adaptec 1542CF)
has never helped, and has always led to a panic shortly afterwards.
I vaguely recall some correspondence about this, and hoped that it
had been fixed, but it has not been fixed in 2.0.31.]
So, so far I do not succeed in installing Redhat 5.0.
The CD reader, disk, and SCSI adapter are in reasonable shape;
for example, a SuSE install a week or so ago went flawlessly.
Apparently the Redhat access pattern triggers a kernel bug.
(32MB memory, 100MB swap, installing from SCSI CD-ROM onto IDE disk)
Andries
cc: lnz, linux-kernel
[PS - 2.0.31 is otherwise just fine on my hardware - I've used
this kernel the past 6 weeks without any problems - only this
install fails.]
I don't know why you are hitting these SCSI problems. It's strange that one
distribution would have problems and another one not.
As for the infinite resets, I believe that the Bus Device Reset code in the
Adaptec 1542 driver has been broken since day one. If you ever trigger that
code, you're going to lose badly, since it doesn't actually send a Bus Device
Reset even though it claims to. I've explained my reasoning to the driver
author on a couple of occasions (which is based on my understanding of the
mailbox protocols on BusLogic, which started with Adaptec compatibility), but
nothing has come of it. Your best bet would be to comment out the Bus Device
Reset code and take the hard reset path, unless you're interested in actually
trying to get the Bus Device Reset code working correctly.
Leonard