From owner-freebsd-hackers  Tue May  9 09:11:57 1995
Return-Path: hackers-owner
Received: (from majordom@localhost)
          by freefall.cdrom.com (8.6.10/8.6.6) id JAA18203
          for hackers-outgoing; Tue, 9 May 1995 09:11:57 -0700
Received: from egeo.unipg.it (egeo.unipg.it [141.250.1.4])
          by freefall.cdrom.com (8.6.10/8.6.6) with SMTP id JAA18189
          for <freebsd-hackers@FreeBSD.org>; Tue, 9 May 1995 09:11:51 -0700
Received: by egeo.unipg.it (AIX 3.2/UCB 5.64/MH-1.08)
          id AA34090; Tue, 9 May 1995 18:11:06 +0200
Date: Tue, 9 May 1995 18:11:06 +0200
From: peppe@unipg.it (Giuseppe Vitillaro)
Message-Id: <9505091611.AA34090@egeo.unipg.it>
To: freebsd-hackers@FreeBSD.org
Subject: Intel Plato Premiere PCI/II and edge/level triggered IRQs
Sender: hackers-owner@FreeBSD.org
Precedence: bulk

I have problem with SCSI timeouts under FreeBSD 2.0
SNAP950412 in a system composed basically of an
Intel Plato P90 Premier PCI II (Neptune chipset)
and a PCI Adaptec 2940.

For what I read until now under freebsd,linux groups
seems that the MB I'm testing use edge-triggered IRQ
rather than level-sensitive on the PCI bus.

This is confirmed too from an online manual for
this board I found on gatekeeper.dec.com. It claims
that the version of BIOS I use now setup the card for edge-triggerd IRQ.
This was not true evidentily for previous versions.

This seems confirmed too from the fact that using
FreeBSD-current aic7* drivers that report IRQ status as edge-triggered
the system behave correctly.

Questions:

  (1) Anyone may confirm this is correct for this
      particular MB?

  (2) Are there any way to setup the MB for use
      level-sensitive IRQS?

  (3) In the case answer to question (2) is "impossible"
      it would be "safe" to buy a mother in this conditions?

  (4) Should the kernel for this MB compiled with PCI_EDGE_INT
      and what would be the perfomance penalization?

Thanks, Peppe.