Date: Tue, 5 May 1998 08:29:36 -0700 From: Parag Patel <parag@cgt.com> Cc: <hardware@FreeBSD.ORG> Subject: Re: [Target Mode] [short aside on NCR chips] Message-ID: <199805051527.IAA07559@mail1.sirius.com>
next in thread | raw e-mail | index | archive | help
On 5/4/98 10:28 PM, Matthew Jacob (mjacob@feral.com) said:
>[...]
>or doing the whole NCR/Symbios SCRIPTS foo (which, as best as I
>can tell, works fitfully and somewhat unpredictably in the at
>least 20 different OS and target device implementations I've
>seen it in)
Funny you mention this - here's something we discovered about the NCR
chips when debugging what we though was a PCI problem on an evaluation
system.
We ended up getting a PCI bus analyzer to watch every PCI transaction to
debug the unrelated problem. While debugging, we noticed some
interesting PCI bus cycles that the CPU was not initiating - nor was it
the target.
Turns out that when the NCR chip is running its SCRIPT, and that SCRIPT
needs to access a PCI register on the NCR chip, then it actually
performes PCI bus transactions to read/write those registers. Yes,
that's right, it actually negotiates for control of the PCI bus then
talks to itself over it!
(These were *not* transactions to access SCRIPTS memory. It was easy to
figure out what was going on when we realized the extra PCI transactions
were doing exactly what we told the SCRIPTS engine to do and the accesses
were to the NCR register mapped space.)
It looks as if someone took a raw SCSI PCI engine, a completely separate
SCRIPTS PCI engine, and slammed the two designs together using the PCI
bus as an intermediary.
-- Parag Patel <parag@cgt.com>
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-hardware" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199805051527.IAA07559>