Subject: Re: Removing iTE IT87xxF code from the lm(4) driver (new it(4) driver)
To: Thor Lancelot Simon <firstname.lastname@example.org>
From: Bill Studenmund <email@example.com>
Date: 05/08/2006 19:39:53
Content-Type: text/plain; charset=us-ascii
On Mon, May 08, 2006 at 09:38:10PM +0100, David Laight wrote:
> On Sun, May 07, 2006 at 04:43:40PM -0400, Thor Lancelot Simon wrote:
> > On Sun, May 07, 2006 at 09:43:10PM +0200, Juan RP wrote:
> > >=20
> > > Hi,
> > >=20
> > > I've adapted the it(4) driver from OpenBSD to work on NetBSD/envsys.
> > > There is one improvement as well, for example this driver does not
> > > need to specify the port address on the kernel config anymore.
> > >=20
> > > The IT87xxF driver can be located at the 0x280, 0x290, 0xc00 and 0xd00
> > > port addresses, so I've added code to check in these addresses.
> > Generally, it's not a great idea to poke (or even peek) around on the
> > ISA bus looking for a device, because other things at those addresses
> > might object. In particular, 0x280 is a common address for ISA
> > Ethernet cards. Are you *sure* the probe is nondestructive?
> Be warned, there are no non-destructive ISA probes....
> So, enabling such a probe in any kernel (other than one on a system where
> expect the device to actualy be present) is a bad, bad idea.
> Even the 'grope' code done by manufacturer supplied DOS configuration
> programs has been know to lock systems solid....
While I hate to say it, I really really agree with David here(*). It's far=
far too easy to wipe out systems or devices with probes. :-( We have a=20
famous example of an ethernet card whose MAC gets re-written by the probe=
for another device.
ISA's horrid tack record at this is why subsequent bus systems have gone=20
for PnP and hardware that identifies itself. :-|
(*) Not that there's anything wrong with agreeing with David. I am just=20
feeling a little sad at saying, "NO NO DON'T DO THAT." :-)
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.3 (NetBSD)
-----END PGP SIGNATURE-----