Re: NE2000 PCI problems FIXED!

Paul Gortmaker (
Fri, 22 May 1998 17:46:34 +1000 (EST)

On Thu, 21 May 1998 Linus Torvalds wrote:

> The stupid driver initializes the "dev->priv" field _after_ allocating the
> In short, it's a crap driver,

While the ne2000 driver may not be rocket science, it probably supports
the most varied and broken group of hardware out of *all* the linux
drivers. Many hours have been spent on trying to make it work on all
the various $0.02 clones.

Granted the dev->priv is a genuine bug, but calling the whole driver
"stupid" and "crap" is not what I'd call constructive feedback and shows
zero respect for the people who have spent hours of volunteer effort on it.
Nor does it encourage those same people to continue to do volunteer work.

If you want to call the hardware stupid crap then I'd agree 100% - the
PCI ne2000 clones are horrible things that should not even exist, and
the ISA cards aren't much better.

> On Thu, 21 May 1998, Richard B. Johnson wrote:
>> Is anybody going to fix it? Does Donald still work on these drivers?
>> If not, I could hack it. I just don't like submitting patches then
>> receiving replies like; "Dumb sh... it's already been fixed... look in

The driver is still being worked on, but not everybody can afford the time
to follow linux-kernel religiously and reply within 24hrs. For most of the
new drivers from Donald, there is a driver specific mailing list hosted at
cesdis which allows Donald to keep track of things without having to wade
through pages of e-mail cruft on boot logos, C precedence, unicode, etc.
Second to that is the linux-net@vger mailing list, which doesn't get 1/10th
the crap that appears on linux-kernel.

WRT ne2000 cards, the PCI support will be removed from ne.c (superceeded by
ne2k-pci.c) to make maintenance simpler. I have a 2.1.10x patch with
the ne2k-pci.c support ready for testing after I look into the "kmod autoload
8390 support" a bit more. The 2.0.x version of this driver can be found in
Alan's 2.0.34preX and further driver details are at:

It is worth stressing that even the simplest changes to ne.c/8390.c need
to be well thought out or they will inevitably break some lame clone part.
For instance, what looked like a sensible change in the recent 2.1.101 patch
to 8390.c broke all cards with the SMC83c690 chip because it has buggy
multicast registers.

>As far as I know, nobody is working on the ne driver, so feel free.
> Linus

I don't suppose anybody thought to check the MAINTAINERS file and cc a
copy of the bug report there before assuming that things were abandoned.
Or are the sensible guidelines Alan laid out at the top of the MAINTAINERS
file no longer considered relevant?

Anyway, Richard, thanks for the gdb output pointing out the problem. I
will look over a few of the other drivers to ensure they don't make the
same mistake and fix them where required.


To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to