Patch to 2.3.99-pre5 kernel to get Linksys card working...

Donald Becker becker@scyld.com
Fri Apr 14 01:18:16 2000


On Thu, 13 Apr 2000, Christopher Smith wrote:

> > As you can see, a good part of select_media() is doing the operations
> > described by the EEPROM media table.  The EEPROM table format is more
> > complex than it needed to be, and the semantics different among chip types.
> > Thus the code cannot be simplified without removing a feature that some
> > board depends on.
> 
> Hmm... From where I'm standing it looks like the v0.91g-ppc code is
> already doing the same thing. Actually, it also looks like there's a
> fair bit of branching based on mleaf->type and tp->chip_id. It'd be
> nice if that could be pushed into the tulip_chip_table (much as the
> timer code was).

The EEPROM media table must be interpreted differently for different chip
types.  For instance, some fields are 8 bits wide on the 21140 and 16 bits
on the 21143.  The format has grown over time, and isn't regular enough to
make a table based solution feasible.

> Of course, I'm looking at this from the periphery
> with no experience working on the driver (other than yesterday ;-), so
> I am likely missing the point.

I've regularly flamed the bogus ill-considered design of Digital's EEPROM
table.  And Intel is making it even worse by introducing new entry types.

Compare it to the Speedo3 (eepro100), which has a clean EEPROM layout that
has remained mostly compatible over three design generations.

> > What compile problem are you having?
> 
> Basically it looks like problems with the kernel compatibility code
> (k_compat.h, pci-netif.h, and kern_compat.h). I even tried downloading
> those various packages from your site and playing around with the
> defines to try to get it to work right before I just gave up.

Hint:
  k_compat.h is just kernel compatibility defines, and requires
    additional pci-scan.[ch] (AKA pci-netif.[ch]) support.
  kern_compat.h includes an in-lined version of pci-scan.c 

The new driver pack on www.scyld.com has an updated Makefile to simplify
this.

Donald Becker
Scyld Computing Corporation, becker@scyld.com

-------------------------------------------------------------------
To unsubscribe send a message body containing "unsubscribe"
to linux-tulip-request@beowulf.org