RE: [ECOS] Default eth hardware address is an eCos bug

[Doug Fraser <dfraser at photuris dot com>]

Actually, for a default MAC address,
the locally administered address bit
should be set, since it isn't really
a globally administered address.
I believe you should mask in an 0x02
on the first octet. (check the standards docs...)
That way, you don't conflict with any
globally administered addresses, ever.

But yes, you certainly should not set
the broadcast address bit. To take that
example:

0x12, 0x13, 0x14, 0x15, 0x16, 0x17

would be okay, no broadcast, locally administered....

Doug

> -----Original Message-----
> From: Martin Buck [mailto:martin.buck@ascom.ch]
> Sent: Monday, December 10, 2001 3:47 AM
> To: ecos-discuss@sources.redhat.com
> Subject: Re: [ECOS] Default eth hardware address is an eCos bug
> 
> 
> Paul Sheer wrote:
> > ./devs/eth/arm/flexanet/current/cdl/flexanet_eth_drivers.cdl
> > contains a default hardware address as follows:
> >         default_value {"{0x11, 0x12, 0x13, 0x14, 0x15, 0x16}"}
> 
> IMNSHO, having a default hardware address at all is a bug. If 
> you don't
> have a proper hardware address for an interface, you shouldn't be able
> to bring it up.
> 
> > The hardware address should be:
> >         default_value {"{0x0, 0x12, 0x13, 0x14, 0x15, 0x16}"}
> > since most (all?) ethernet cards have the first byte zero.
> 
> There are quite a few starting with 08:.. these days. But at least the
> LSB of the first Byte should be 0 - otherwise, it's a
> multicast/broadcast address.
> 
> Martin
>