🚀 go-pugleaf

david@ods.com (David Engel) writes:

>Is anyone out there using a Buslogic 946c (or any other model) SCSI
>controller with a Micropolis 4110 disk drive under Linux?  If so,
>please get in touch with me.

>I've started seeing disk drive timeouts since upgrading to the newer,
>higher-performance, Buslogic driver.  These timeouts always occur when
>accessing the tape drive (a Sony SDT-5000), occasionally occur when
>accessing the CD-ROM (a Plextor 3028) and occur every once in a while
>only accessing the disk.

>The author of the Buslogic driver for Linux, Leonard Zubkoff, has been
>very cooperative in trying to diagnose the problem but suspects that
>it is due to a faulty tagged-queuing implementation in the disk
>firmware.  The problem goes away when the tagged-queuing option in the
>Buslogic driver is disabled.  I'd rather not do that all the time

I have a hunch that once Leonard's new driver gets used on a lot more
sites (eg. shipped with a distribution) then we will uncover a whole
flock of drives with poorly tested and buggy tagged queueing 
firmware support.  So much so, that TQ will need to be disabled by
default, and an ioctl() will be used to turn it on (akin to hdparm
for IDE disks). The IDE situation was/is similar, in that there
are lots of IDE drives with buggy "block mode" firmware, and so Mark
had to turn it off by default in the Linux driver.

Leonard is more optimistic and is hoping the drive manufacturers
have done their homework, and performed rigourous testing on the
TQ firmware, in addition to the "normal" firmware. I hope he is 
right and I am wrong in this case.

>since it has a significant impact on performance.  So far, Micropolis
>technical support has not shown any interest in pursuing the problem.
>I'm hoping they'll show more interest if I can find others who are
>seeing it also.

Heh, I am not surprised. Look how long it took Conner to own up to
the buggy firmware on the 1060S drive. And that was broken under
*normal* use (i.e. no TQ or anything fancy), with probably hundreds
of linux users reporting it.

Assuming the Micropolis is broken, and it isn't a driver issue, the 
best thing for users to do is form a list of drives with broken TQ 
firmware, so that others can avoid them. Of course, contacting the 
manufacturers is still the Right Thing To Do.

Good luck,
Paul.