[2114] in linux-scsi channel archive
Tagged Queueing Problem w/ Buslogic and Quantum
daemon@ATHENA.MIT.EDU (John Taylor)
Sun Jul 6 22:05:52 1997
From: John Taylor <john@verdi.cviog.uga.edu>
To: linux-scsi@vger.rutgers.edu
Date: Sun, 6 Jul 1997 22:03:17 -0400 (EDT)
scsi0: Configuring BusLogic Model BT-948 PCI Ultra SCSI Host Adapter
scsi0: Firmware Version: 5.06J, I/O Address: 0xD800, IRQ Channel: 11/Level
scsi0: Target 0: Synchronous at 20.0 mega-transfers/second, offset 15
Vendor: QUANTUM Model: FIREBALL_TM2110S Rev: 300X
SCSI device sda: hdwr sector= 512 bytes. Sectors= 4124736 [2014 MB] [2.0 GB]
PROBLEM:
Tagged Queuing does not work with this adapter/drive combination. I tried
the factory standards, womper88 and womper89 firmwares, all with no luck.
On another system, I had a "Vendor: SEAGATE Model: ST31055N" in which random
scsi timeouts would occur. Eventually the system would reset the scsi bus,
and things would resume to normal after a 20 second pause. The fix for
this was to upgrade to womper89 (Firmware Version: 5.06J). This fixed the
problem and I have not had any problems since.
I had the same timeout problem occur with the Quantum. So I though that
upgrading the firmware would solve the problem. This did not fix the
problem for Linux 2.0.30 (or pre 31). I do not know a whole lot about
tagged queuing, except that when I disabled this feature for the Quantum,
the timeouts stopped occurring. Problem solved at a big performance loss.
Does Tagged Queuing come in just enabled/disabled flavors? Or is it possible
for me to fine tune the queueing depth (or something else) in order to have
crippled Tagged Queuing? Ideally I would want to have some lower level of
Tagged Queuing while still retaining stability (ie no timeouts whatsoever).
This would allow me to have not as big of a performance loss while still
remaining stable.
Is this possible?
Thanks
John Taylor Institute of Government
System Administrator University of Georgia
john@www.cviog.uga.edu http://www.cviog.uga.edu/
