[Top] [All Lists]

RE: [PATCH] netdev_ops

To: "David S. Miller" <davem@xxxxxxxxxx>
Subject: RE: [PATCH] netdev_ops
From: "Feldman, Scott" <scott.feldman@xxxxxxxxx>
Date: Thu, 10 Jul 2003 01:18:50 -0700
Cc: <willy@xxxxxxxxxx>, <netdev@xxxxxxxxxxx>
Sender: netdev-bounce@xxxxxxxxxxx
Thread-index: AcNGt+/LXquM9ebEQEO0AjBtV3zQ1gAAU8Vg
Thread-topic: [PATCH] netdev_ops
> Don't tell me you're seriously considering having _TWO_ 
> copies of all this code sitting around?
> At that point backwards compat becomes absolutely 
> rediculious.  If it's important to you, just stick to the 
> current scheme.  You gain nothing by maintaining two copies 
> of the same code.

Either way we end up with duplicated code.  If I stick with the current
scheme (no netdev_ops), I duplicate in the driver all of the wrapper
code that Matt has pulled into netdev_ops.  Each driver that sticks to
the current scheme duplicates Matt's code.  With HAVE_NETDEV_OPS, you're
right, we're maintaining the wrapper code outside the kernel.  But, it
does leave the possibility of having a shared backwards compatibility
code for multiple (all?) drivers for those stuck with supporting kernels
without netdev_ops. 

<Prev in Thread] Current Thread [Next in Thread>