On Tue, Dec 06, 2005 at 01:58:02AM -0500, Allen Briggs wrote:
> On Mon, Dec 05, 2005 at 05:29:57AM +0000, Matthias Scheler wrote:
> > OpenBSD's sk(4) driver supports this variant. As NetBSD's driver is based
> > on OpenBSD's one (which is based on FreeBSD's one) it shouldn't be
> > difficult to pull that change to NetBSD.
> 
> Does it work properly with it?  Just glancing at the head revision
> of if_sk.c, it looks like it matches the card, but doesn't handle
> it properly.  The queueing mechanism is different for the Yukon-2-
> based cards.

This is one of those in-between chips, isn't it (Yukon 2 EC?)?  It can be
extremely confusing to tell which chips work which way using the Marvell
driver as a reference, but as I recall there are four cases:

1) The original XMAC + "Genesis" cards

2) The Yukon chips

3) The chips that use the code that's #ifdeffed as "Yukon 2" in the
   file in the Marvell driver that handles the plain Yukon chips

4) The chips (Yukon 2 XL and maybe Yukon 2 FE) that use the alternate
   versions of several functions that are kept in a separate source file
   in the Marvell driver.

It's early in the morning, and I'm buried in unrelated work, so I can
neither remember nor look up what the differences between cases 2 and 3
(or 3 and 4) are, but I know they were somewhat smaller than the total
set of differences between cases 2 and 4.

Thor