[epic-bug] cannot initialize SMC9432TX card under Linux after Windows is booted?

[Donald Becker]

On Wed, 22 Nov 2000, James Ralston wrote:

> On Tue, 14 Nov 2000, Donald Becker wrote:
> 
> > > Oct 31 10:58:03 myhost kernel: epic100.c:v1.07h 8/18/99 Donald Becker http://cesdis.gsfc.nasa.gov/linux/drivers/epic100.html
> > > Oct 31 10:58:03 myhost kernel:   The PCI BIOS has not enabled Ethernet device 10b8-0005.  Updating PCI command 0010->0015.
> >
> > The card was left in ACPI D3 state.  The problem is that the old
> > driver structure does not restore the ACPI power state of the card.
> 
> Just out of curiosity, do you agree with Linus' suggestion on how this
> problem should be addressed within the kernel's driver structure?
> 
> http://www.uwsg.indiana.edu/hypermail/linux/kernel/9910.2/0510.html

The suggestion from Linus was that he preferred "the incremental approach".
That's hardly a technical solution.

> But the current philosophical disagreement over the state and
> maintenance methods of the kernel networking drivers is a big PITA for
> me, and many other people.  For me, it's not so much a practical issue
> for me as a political one: if I can't pull out a recent Linux distro
> CD (e.g., Red Hat 7.0), slap it on a machine, and have it *work*, then
> it makes it difficult for me to convince my managers (and their
> managers above them) that Linux is a viable OS.

Yes, Red Hat is broken.  There was a working solution to the ACPI-D3
problem over a year ago.

I don't have any influence over or responsibility for the Red Hat distribution.
I am responsible for the Scyld Beowulf distribution, which does use it.

I've worked on Linux since 1992.  Last year Linus decided that he would
rather take someone else's (IMNSHO, flawed) untested driver interface changes
rather than my documented and tested changes.  He believed, based on the
number of people excited after Red Hat's IPO, that there was an army of
experienced people ready to take over maintenance.

>  Yeah, I can rip out
> the networking code and put in your latest stuff.  But when they ask
> *why* I have to do that, and I explain to them that it's because the
> networking modules that ship with the Red Hat distro are the
> "official" kernel versions, which are outdated and buggy, and the
> *reason* they're outdated and buggy is because the main networking
> driver maintainer and the kernel developers (Linus, et. al.) can't
> agree on how those drivers should be maintained and included in the
> kernel, then it reflects badly on Linux as an OS.
> Again, I appreciate your efforts, as do many other people.  But
> please, consider renewing/redoubling your efforts to at least reach
> some compromise concerning the kernel integration issue.

Hmmm, compromise?
"We spent a few weeks doing the fun part of making the interface changes.
Now we are bored by all of the work.  Why don't you go back to the time
consuming work of maintaining the drivers.  Oh, but don't change the
interface."

Instead we've spent the past year finishing up a revolutionary new approach
to clustered computing, and putting together a new distribution so that we
can deploy this new approach.


Donald Becker				becker@scyld.com
Scyld Computing Corporation		http://www.scyld.com
410 Severn Ave. Suite 210		Second Generation Beowulf Clusters
Annapolis MD 21403			410-990-9993