----- Original Message -----
From: "Ben Harris" <bjh21@netbsd.org>
To: <jufi@nerdnet.de>
Cc: <port-acorn32@netbsd.org>
Sent: Monday, August 05, 2002 9:14 PM
Subject: Re: SCSI in 1.6 ?


> In article <20020805193325.H29963@nerdnet.de> you write:
> >On Sun, Aug 04, 2002 at 08:55:26PM +0100, Chris Gilbert wrote:
> >> > It's stable, I've tested it again this evening, copying my whole /usr
> >> > on it, and doing a 'cvs up' in /usr/pkgsrc worked fine.
> >> > That should be enough..
> >>
> >>
> >> Hmm, is this with or without diagnostics on?  I note from the bug
report
> >> that you had to disable DIAGNOSTIC at one point.  the diff in the PR
hasn't
> >> been put into -current (assuming we're talking about 8924)
> >> http://www.netbsd.org/cgi-bin/query-pr-single.pl?number=8924
> >
> >Well, strange, I tested it without DIAGNOSTIC in the kernel, and
> >everything was fine and stable.
> >
> >But including DIAGNOSTIC into the kernel leads to a panic when I tried
> >to mount my SCSI disc/CDROM.
> >Are any details useful to anybody?
> >I don't have them at hand, but could produce them tonight..
> >
> >They start with
> >panic: scsipi_done: ASYNC and POLL
> >Stopped in pid 201 ...
> >
> >Hope that makes sense to somebody.. :-)
>
> It means that the card driver is buggy, and is forcing the XS_POLL bit in
> SCSI requests rather than setting SCSIPI_ADAPT_POLL_ONLY in adapt_flags
and
> getting the scsipi layer to do it.  Fixing this should be pretty trivial;
> I'll try to look over this thread and work out what needs doing this week.

The fix appears to be in the PR, it was just never committed.  That's where
I got the question about DIAGNOSTIC.  I'd do it, but lack the card to do so
(and my acorn32 isn't really a dev box as it runs my mail and web server)

Chris