From owner-freebsd-hackers  Tue Jul  4 02:38:00 1995
Return-Path: hackers-owner
Received: (from majordom@localhost)
          by freefall.cdrom.com (8.6.10/8.6.6) id CAA20642
          for hackers-outgoing; Tue, 4 Jul 1995 02:38:00 -0700
Received: from hq.icb.chel.su (icb-rich-gw.icb.chel.su [193.125.10.34])
          by freefall.cdrom.com (8.6.10/8.6.6) with ESMTP id CAA20366
          for <hackers@freebsd.org>; Tue, 4 Jul 1995 02:33:38 -0700
Received: from localhost (babkin@localhost) by hq.icb.chel.su (8.6.5/8.6.5) id OAA03586; Tue, 4 Jul 1995 14:55:29 +0500
From: "Serge A. Babkin" <babkin@hq.icb.chel.su>
Message-Id: <199507040955.OAA03586@hq.icb.chel.su>
Subject: Re: Minor for Digiboard driver
To: bde@zeta.org.au (Bruce Evans)
Date: Tue, 4 Jul 1995 14:55:29 +0500 (GMT+0500)
Cc: hackers@freebsd.org
In-Reply-To: <199507040555.PAA11661@godzilla.zeta.org.au> from "Bruce Evans" at Jul 4, 95 03:55:36 pm
X-Mailer: ELM [version 2.4 PL23]
Content-Type: text
Content-Length: 1753      
Sender: hackers-owner@freebsd.org
Precedence: bulk

> 
> >>My developing of Digiboard driver came to the point where I need
> >>a major number. Give me one please.
> 
> >   Just pick the next available one. If we need to re-assign it before
> >committing the driver (very unlikely), then that's okay.
> 
> We're close to needing a single major number for all ttys (like sd has
> a single major number for all scsi controllers).  You need a new major
> now, however.  Please be careful not to implement any ioctl or minor
> number magic that is inconsistent with the other drivers (sio, cy and
> rc; cx and ity are already inconsistent).

I took the scheme for minor numbers from sio, the only thing I added
is 2 bits of card number :

	not_used{14} card_number{2} major{8} sio_like_minor{8}

The only problem is that I don't know how to name the special files to
be consistent with existing conventions. I know Peter Wemm who wrote
Specialix card driver had the same problem and I'll ask him about his
decision.

> Did you use any "intelligent" features of the Digiboard?  Minor design

I use its big buffers and flow control features. It has some ability
to do the canonical processing but I don't know how to use it because
Linux driver doesn't uses it.

> characters caused overrun errors.  I thought had the cd1400 working
> slightly more efficiently than a 16550, but that was only with an
> input fifo size that was too large to avoid overruns.  To avoid
> overruns for just 2 active ports, the input fifo size has to be 6,
> while 14 works for 16550's and is considerably more efficient.

I think Digi should be more efficient due to its big buffers in
dual-ported memory.


		Serge Babkin

! (babkin@hq.icb.chel.su)
! Headquarter of Joint Stock Commercial Bank "Chelindbank"
! Chelyabinsk, Russia