Re: forcedeth - WOL [SOLVED]

[Andrew Morton <akpm@osdl.org>]

On Thu, 28 Sep 2006 04:24:47 +0200
Björn Steinbrink <B.Steinbrink@gmx.de> wrote:

> On 2006.09.28 04:04:38 +0200, Björn Steinbrink wrote:
> > On 2006.09.27 18:36:25 -0700, Andrew Morton wrote:
> > > On Thu, 28 Sep 2006 03:01:33 +0200
> > > Björn Steinbrink <B.Steinbrink@gmx.de> wrote:
> > > 
> > > > > > > Do we know if this reversal *always* happens with this driver, or only
> > > > > > > sometimes?
> > > > 
> > > > I only tried 2.6.18 twice this time, but when I wrote my own tool to do
> > > > it, I had probably 20-30 power on -> ethtool -> poweroff cycles before I
> > > > decided to look into Bugzilla. As it looked like being fixed already and
> > > > I did use the nForce NIC for testing only, I didn't spend any further
> > > > time on it back then.
> > > 
> > > What I'm angling towards is: "is this just a driver bug"?
> > 
> > I just took a peek at the code.
> > 
> > The version on bugzilla (last attachment, comment #22), which was
> > reported to work correctly, has the MAC address reversal hardcoded.
> > The driver in 2.6.18 has some logic to detect if it should reverse the
> > MAC address. So it looks like a hardware oddity/bug that the driver
> > wants to fix but fails. I'll see what happens if I force address
> > reversal and if I can decipher anything, but probably someone else will
> > have to cast the runes...
> 
> OK, please excuse me wasting your time, it's late over here... I've
> actually been looking at Linus' git tree (pulled yesterday) while
> writing that mail, not 2.6.18.
> 2.6.18 does _not_ contain the address reversal detection.
> Using the git tree instead of 2.6.18 WOL works as expected, without
> having to reverse the MAC address.
> 

hm, OK, thanks.  Ayaz, do you think 5070d3408405ae1941f259acac7a9882045c3be4 is
a suitable thing for 2.6.18.x?