On Mon, Aug 28, 2000 at 09:42:06AM -0700, Matthew Jacob wrote:
> That's *sort of* the rub. I'd suggest that you try (and, Manuel, it's a heck
> of a lot easier for you to do in PSIOP than it is for me to do in the QLogic
> driver, let me tell you!) to do the message dance for the first command when
> you're at NIL state. You'll end up with 4 outcomes:
> 
> a) The target won't even go into MESSAGE OUT phase (not possible to be a wide
> or sync capable device- don't worry about it then).
> 
> b) The target rejects the wide and sync data xfr messages. Don't worry
> about it- it's not a wide/sync capable device.
> 
> c) The target accepts the wide and sync data xfr messages, but lies and
> hangs up. After a brief period, you time out, issue a BUS RESET and you're now
> no worse off than you are in the current model.
> 
> d) The target accepts the wide and sync data xfr messages and everyone's
> happy.
> 
> [ this is, btw, what the Solaris glm driver does for the LSI parts. This
> indeed is somewhat a pain because the PROM has negotiated sync/wide so
> you have to indeed not assume you're at any state but get there ]
> 
> 
> > Also the reset can help is case of a hard reboot of the system, leaving the
> > targets with pending commands.
> 
> Aieee! No, no, no.... That's like saying, "I refuse to wait for my sync to
> complete....". If something's hung, yes, issue a reset if you have to.

Ok, I'll try to clean this out when I have time :)

--
Manuel Bouyer, LIP6, Universite Paris VI.           Manuel.Bouyer@lip6.fr
--