On Tue, 07 Jul 2009 15:34:16 -0700, "Garrett D'Amore" <[email protected]>
wrote:
> Generally, +1.  The old dnet code is really crufty, and a PITA to work
> with.
> 
> Actually, in retrospect (I didn't know this when I wrote the code), the 
> Macronix devices are fairly faithful clones of the Digital tulip 
> devices.  Starting with my mxfe driver might not be a bad approach.

Ahh, excellent. I'll make sure to nose around a bit.

> However, there are some tricky issues involved for dnet, which my mxfe 
> driver doesn't deal with:
> 
>     1) ROM parsing.  Some NICs use a ROM to support alternative media.  
> That said, its probably not unreasonable to ditch support for dnet 
> devices that don't support either MII or NWay style 100Base-T/10Base-T.  
> (I.e. I think BNC and AUI support could go bye-bye.)

Fortunately I've been able to find the PROM specification for the device,
but you are absolutely right, there are a number of different PHY's
supported.

>     2) interrupt hackery.  It turns out that some multiport dnet cards 
> exist that use non-compliant interrupt routing.  This is something that 
> would probably need to be maintained, since I believe that most folks 
> still using dnet cards are doing so with multiport cards.  (There is 
> little need for the dnet driver otherwise.)

I actually have one of these (its an older full-length Adaptec). I'll make
sure to keep the interrupt hacks intact (cleaned up though - they still use
the older interrupt function calls).

> All that said, I'd just assume see dnet go bye-bye altogether.  These 
> devices are *really* ancient; I don't think a genuine tulip part has 
> been sold in a *very* long time.   Apart from the multiport variants, 
> there is no good reason anyone should be still using these devices, 
> IMO.   There are cheap and readily available 10/100 and even gigabit 
> alternatives.

I go back and forth on this all of the time - it is nice to keep legacy
drivers around just for those users who use legacy hardware, but by the
same token, at *some* point these devices need to be dropped. IMHO, I had
always thought that the bus should be the limiting factor (i.e. once 32-bit
PCI goes by the wayside, so should those drivers which rely on it).

Just my 0.02.

-- 
Yet magic and hierarchy
arise from the same source,
and this source has a null pointer.

Reference the NULL within NULL,
it is the gateway to all wizardry.
_______________________________________________
driver-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/driver-discuss

Reply via email to