Subject: Re: marching unit numbers
To: Bill Sommerfeld <sommerfeld@orchard.arlington.ma.us>
From: Warner Losh <imp@village.org>
List: tech-kern
Date: 12/31/1999 16:24:53
In message <199912311723.MAA01091@thunk.hamachi.org> Bill Sommerfeld writes:
: It looks like it shouldn't be difficult for those PCMCIA network
: interfaces which have te MAC address in the CIS (Card Info String)
: area on the card (which is accessible in a driver-independant way ..
: I'm told that many early PCMCIA cards don't support it, but the couple
: of newer ones I've looked at do have it).  The pcmcia bus code could
: extract the MAC address and use it as a locator to match the
: appropriate device instance.

This is wrong.  A substantial number of network cards do not have
their MAC Address in the CIS.  The MAC Address extention is relatively
recent thing.  I know that about 1/2 of the cards I have knocking
around for FreeBSD pccard work do not have this CIS tuple in them.  I
see from the PAO work that a large number of cards do this in a
non-standard way.

I think that this is going to be hard to do in general.  You have a
chicken an egg problem.  The driver will know how to get a unique
identifier from the device, if possible, but would have no place to
store that identifier that would be persistant accross a detach
(assuming that the softc goes away with the detach).

Warner