Adaptec Fast-Wide & IBM DHFS Drive

[Daniel M. Eischen]

> > Perhaps it doesn't support tagged queueing?
>
> In that case, shouldn't it simply refuse negotiation for it?  This is one
> of their top-of-the-line units, and I'd find it quite hard to believe that
> it didn't support this feature.

The driver will catch a message reject from the drive if it issues
one.  In that case, tagged queueing will be disabled.  This is not
happening in your case, though.  It might be that it's responding
with target busy, or check condition.

> > You can turn off tagged queueing by on a device-by-device basis,
> > if you wish to enable tagged queueing for other devices on the bus.
>
> Is that in the mainstream driver currently?  I'll have another look. 

Yes, it should be.  In order to make this work, though, you have
to edit aic7xxx.c.  See the comments in there.

> > Force a panic so that you can capture the first few messages?
>
> Ah, yes, but _where_?  I'm not familiar at all with the driver anymore.

You could do it upon the first time aic7xxx_reset or aic7xxx_abort
were called.  Another place could be in aic7xxx_queue(), where you
could force a panic after N commands are queued to your device (cmd->
target).  The interrupt routine would be another place you could
do it, probably after N BAD_STATUS received from your device.

Do you still have problems if you disable wide negotiation on
your drive?

Dan Eischen
deischen at iworks.InterWorks.org