bge IPMI patch final review

[Doug Ambrisko]

Julian Elischer writes:
| Doug Ambrisko wrote:
| 
| >I'd like to commit my final bge version of the IPMI patch to keep IPMI
| >working when FreeBSD attaches and also when the NIC is downed.
| >I also have patches for the bce to do the same that I sent to David.
| >This has various reviews and testing in the wild.  I want to give
| >a heads up since there have been changes in the bge driver recently.
| >I'd like to not hold off on this anymore.
| 
| code looks good but could you add a bunch of comments explaining the 
| handshake and how this works. (be verbose please)

You tell me and we'll both know :-(  It's inspired from Linux and
that doesn't explain it.  Their doc. doesn't say much except do this 
then don't touch the PHY which is what Linux does.  Via emperical 
evidence I found that if I was careful I could talk to the PHY initially.
Now if you want a more interesting thing to look at then try the bce version.
There is absolutely no doc's.  that I could find.

About the only thing I can say is that we have to play with is the
BGE_MODECTL_STACKUP that says the OS driver is running.  I could mention 
that.

If we really had insight into a potential firmware handshake then we could 
try to co-operate and access the PHY in a safe way to watch link state
changes.  Via emperical means I could not figure anything out :-(  It's
even worse that we can't tell if IPMI was actually enabled or not since
some implementations required a reset before talking to the chip
to see what it was doing :-(  I don't see any clues in the kernel.org
or BroadCom's driver.

Currently it does more then what Linux and Solaris do :-)
 
Doug A.