[1088] in linux-net channel archive
Re: NE2000 probe broken in 1.3.XX?
daemon@ATHENA.MIT.EDU (Paul Gortmaker)
Sun Sep 10 14:54:59 1995
From: Paul Gortmaker <gpg109@rsphy1.anu.edu.au>
To: gilbert@inviso.com (Gilbert Callaghan)
Date: Mon, 11 Sep 1995 00:38:37 +1000 (EST)
Cc: linux-net@vger.rutgers.edu
In-Reply-To: <m0sr6CZ-0005o8C@inviso.com> from "Gilbert Callaghan" at Sep 8, 95 11:22:50 am
> I just recently upgraded from 1.2.8 to 1.3.24 to play with some of the
> new toys in the kernel. 1.3.24 locks up at the NE*000 probe:
>
> NE*000 ethercard probe at 0x280:
>
> and it's over. I'm using a Cabletron E2000 (yes I know, but it's worked fine
> since 0.99.6).
>
> Has anything changed recently ?
Yes. A full reset was inserted before trying to read the SA PROM. This
helped poor sods that had crummy ne2k cards that would hang at the PROM
read (just like yours does now...)
> Do I need to #def CONFIG_NE_BAD_CLONES?
No, that is the default. That has to do with yet *another* compatibility
problem... :-(
> FYI Here is the successful probe from 1.2.8:
>
> NE*000 ethercard probe at 0x280: 00 00 1d 03 11 94
> eth0: Cabletron found at 0x280, using IRQ 9.
Unfortunately, without knowing what card is there in advance, it is
impossible to determine whether to reset or not before trying to
read the SA PROM. As you can see, we use the 00 00 1d of the SA PROM
to actually identify the C'tron cards. So the question is, "Do we break the
C'tron cards or the crummy ne2k clones?".
Please comment out the recently added reset code between
printk("NE*000 ethercard probe at %#3x:", ioaddr);
and the
/* Read the 16 bytes of station address PROM.
and then verify if things work like they used to for the dain-bramaged
C'tron card. In the meantime, I will try and think of a sensible way
to handle both bits of ugly hardware without breaking the other....
Don't hold your breath.
Paul.
