[Cryptography] Randomness from network hardware?

[Jerry Leichter]

Ethernet is supposed to back off randomly after a collision.  There's no need for a strong notion of randomness, but the algorithm does have to introduce enough variation between stations that two of them will never end up following the same sequence of backoffs at the same time, or the algorithm won't work.

Is there any way to get access to whatever source of randomness drives this decision?  Old interfaces - I'm talking the original "yellow cable 10Mb/sec" stuff - used to have a test mode that would simulate a collision.  I think you could even force one.

I haven't looked at Ethernet hardware in many years/generations of the standards.  Do current interfaces, perhaps in some test mode (which a special driver could get at during boot), provide access to anything that could be used as (part of) a random seed?
                                                        -- Jerry