Date: Mon, 11 Dec 1995 12:56:10 +0800 (CST)
From: Pai-Hsiang Hsiao <shawn@iis0.MIT.EDU>
To: Andy Valencia <vandys@cisco.com>
Cc: VSTa mailing list <vsta@cisco.com>
Subject: Re: Interrupts arrive to fast? 
In-Reply-To: <199512101821.KAA04338@vandys-lap.cisco.com>
Message-Id: <Pine.SUN.3.90.951211123714.16176B-100000@iis0>
Mime-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII

On Sun, 10 Dec 1995, Andy Valencia wrote:

> >In cam server, after receiving M_ISR, driver speciic interrupts handling
> >routines are called (in cam_ccb_wait()).
> 
> So you have seen at least one M_ISR message?

No, in cam; yes, in kernel.

> If you have IDE as well, it can be really nice to boot the system up
> entirely and then run cam under the debugger.

Yeap, I run cam under debugger, and find that it never returns on
msg_receive().

> Mike Larson can comment on the SCSI aspect.  Overall, it sounds like the
> sequence of interrupts and card interactions might be subtly different from
> the 1542, and you're getting out of step with the controller.  Or, the
> interrupts for the 2940 are getting directed to the wrong port.

I make a simple test after posting the message. I clear the interrupt
status on adaptor, which will deasset the interrupt# line on PCI, in kernel
space (in delivery_isr() actually). The machine hangs again after 
reporting one interrupt.

I will work on this. But, it seems that if I do not clear the interrupt
status on the adaptor (on PCI bus), the interrupt receves continuously.
If it's true, should VSTa add some kernel facility to help this? Just a
guess, I need to check out where the problem is.

// Shawn