Subject: port-i386/2284: No warning DMA capable ISA scsi controllers with > 16Mb.
To: None <gnats-bugs@NetBSD.ORG>
From: None <david@mono.org>
List: netbsd-bugs
Date: 04/01/1996 05:58:58
>Number:         2284
>Category:       port-i386
>Synopsis:       No warning for DMA capable ISA scsi controllers with > 16Mb.
>Confidential:   no
>Severity:       serious
>Priority:       high
>Responsible:    gnats-admin (GNATS administrator)
>State:          open
>Class:          doc-bug
>Submitter-Id:   net
>Arrival-Date:   Mon Apr  1 09:35:01 1996
>Last-Modified:
>Originator:     David Brownlee
>Organization:
Monochrome
>Release:        NetBSD-1.1B
>Environment:

>Description:
	People using DMA capable ISA scsi controllers with > 16Mb in their PC
	(or 16Mb with bios remapping) have no warning as to why they are being
	toasted.
	Bounce buffers will fix this when they come, but in the meantime
	there should be mention of it in the docs.

>How-To-Repeat:
	Boot a machine with > 16Mb with a 1542 or similar, grow frustrated,
	mail mailing list, start another debate about bounce buffers, consume
	more time from many people on the lists, gain a bad impression of a
	system that not only doesnt support your hardware (yet), but cant
	even give you any reason _why_. 
>Fix:
	The installation notes (+supported hardware?) should mention the
	problem, plus a quick check in the drivers' attach routines to panic
	with an appropriate message if >16Mb is detected would be a _real_
	win.

	Of course taking the existing available 'hack' to permit bounce
	buffers and adding it into the i386 port as an acknowledged stopgap
	measure until cgd has sufficient time to fix bounce buffers in a MI
	fashion would be a much 'better' solution from the viewpoint of users.

	What was it Chris said about what researchers _really_ want from an
	OS? :)
>Audit-Trail:
>Unformatted: