Re: [ECOS] Where do I put ETH_DRV_SC?

[Hugo Tyson <hmt@redhat.com>]

David.Karlberg@combitechsystems.com writes:
> Im Trying to get TCP/IP working on a media Enging . I am using the
> nanoEnging template and I want to write a hardware driver using the
> ETH_DRV_SC()/NETDEVTAB_ENTRY(). 

They don't say what type of ethernet the mediaEngine has, but I see it's
not the same as the i82559(s) on the commEngine and nanoEngine.  It just
says 10BaseT - does that mean it's an SMSC91C96 or similar?  If so you
could re-use that driver?
 
> And my question is: where do I put the ETH_DRV_SC()/NETDEVTAB_ENTRY()
> code?

In the source ;-)  That sounds unhelpful, but it's not, really it's not.

Take a look at the generic i82559 driver that's used by the nano target.
It includes a header that sets up a load of context.  The header also does
the 
	ETH_DRV_SC(i82559_sc0, .........
	NETDEVTAB_ENTRY(i82559_netdev0, .........
invocations for as many devices as you want.

The compiled generic driver is linked in to libextras.a by the compile line
in its cdl - that's what makes those magic things be linked into the
system, just so long as the results of ETH_DRV_SC()/NETDEVTAB_ENTRY() are
in the image and no garbage collected.  Using libextras.a is what prevents
the garbage collection.

The other way we do generic drivers is to have the generic driver in a
header, and a .c sets up context, including ETH_DRV_SC()/NETDEVTAB_ENTRY()
and then includes the header to make all the functions.

Same difference; they're really just two ways to compile the catenation of
two C files, one from a different package selected according to the target
and one generic.

If you don't want to separate your driver into a generic part and a
specific part, you don't have to.  So just have one .c file and put the lot
in that.  The EBSA285 ethernet driver is like that - only because it's
rather old; the generic i82559 driver grew from it, but we didn't change
over to help keep the EBSA driver stable.

HTH,
	- Huge