Date: Wed, 15 Nov 1995 21:57:59 +0100 (MET)
From: Tom I Helbekkmo <tih@Hamartun.Priv.NO>
To: Milo.Velimirovic@uwlax.edu
Cc: OldUnix MailingList <oldunix@minnie.cs.adfa.oz.au>
Subject: Re: mknod device numbers
In-Reply-To: <9511151722.AA02396@fingers.acs.uwlax.edu>
Message-Id: <Pine.NEB.3.91.951115213628.14251A-100000@barsoom.Hamartun.Priv.NO>
Mime-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII

On Wed, 15 Nov 1995, Milo Velimirovic 31 Wing 785-8030 wrote:

> All of this is very installation specific. The major/minor numbers in the
> device inodes need to correspond exactly to the entries in the devtab
> structure in the kernel.

True enough.  The example I gave was correct for the V6 distribution
that Ken Wellsch donated to the PUPS archive, and in which the file
/usr/sys/conf/c.c (generated by 'mkconf' in the same directory) ends
up with the following for the default "rkunix" with rk, tm and tc
drivers enabled, as per the script /usr/sys/run:

int     (*bdevsw[])()
{
        &nulldev,       &nulldev,       &rkstrategy,    &rktab, /* rk */
        &nodev,         &nodev,         &nodev,         0,      /* rp */
        &nodev,         &nodev,         &nodev,         0,      /* rf */
        &tmopen,        &tmclose,       &tmstrategy,    &tmtab, /* tm */
        &nulldev,       &tcclose,       &tcstrategy,    &tctab, /* tc */
        &nodev,         &nodev,         &nodev,         0,      /* hs */
        &nodev,         &nodev,         &nodev,         0,      /* hp */
        &nodev,         &nodev,         &nodev,         0,      /* ht */
        0
};

int     (*cdevsw[])()
{
        &klopen,   &klclose,  &klread,   &klwrite,  &klsgtty,   /* console */
        &nodev,    &nodev,    &nodev,    &nodev,    &nodev,     /* pc */
        &nodev,    &nodev,    &nodev,    &nodev,    &nodev,     /* lp */
        &nodev,    &nodev,    &nodev,    &nodev,    &nodev,     /* dc */
        &nodev,    &nodev,    &nodev,    &nodev,    &nodev,     /* dh */
        &nodev,    &nodev,    &nodev,    &nodev,    &nodev,     /* dp */
        &nodev,    &nodev,    &nodev,    &nodev,    &nodev,     /* dj */
        &nodev,    &nodev,    &nodev,    &nodev,    &nodev,     /* dn */
        &nulldev,  &nulldev,  &mmread,   &mmwrite,  &nodev,     /* mem */
        &nulldev,  &nulldev,  &rkread,   &rkwrite,  &nodev,     /* rk */
        &nodev,    &nodev,    &nodev,    &nodev,    &nodev,     /* rf */
        &nodev,    &nodev,    &nodev,    &nodev,    &nodev,     /* rp */
        &tmopen,   &tmclose,  &tmread,   &tmwrite,  &nodev,     /* tm */
        &nodev,    &nodev,    &nodev,    &nodev,    &nodev,     /* hs */
        &nodev,    &nodev,    &nodev,    &nodev,    &nodev,     /* hp */
        &nodev,    &nodev,    &nodev,    &nodev,    &nodev,     /* ht */
        0
};

The major device numbers are the offsets (counting from 0) in these
arrays, so rk has major 0 for the block device, 9 for the character
device.  I would assume that this holds for V6 in general -- if one
were to add device drivers, one would surely extend these arrays at
their ends, not insert anything into them...  :-)

-tih
-- 
Tom Ivar Helbekkmo
tih@Hamartun.Priv.NO