From: "Jeff Kirsher" <redacted>
Date: Wed, 7 Jan 2009 10:23:44 -0800

The only testing that we were not able to do was the IPMI testing,
because of the lack of resources.  All other testing passed.

While all other testing passed, I am concerned about not being able to
test whether or not this change affects the ability to pass IPMI
traffic.  I am not sure if the "gain" of using request_firmware() out
weighs the potential risk that IPMI traffic may be broken with this
patch.  I guess I wondering what the gain is in using the
request_firmware() function?

quoted

From past experience with IPMI traffic and the e100, the loading of

the microcode in the correct manner greatly affected whether IPMI
traffic would pass or not.

Jeff, I've lost all of my patience.

All drivers are being converted this way.  I fought against doing it
to tg3 for various reasons, but the tide worked against me and I
accepted that.

We can't hold this patch up forever for a potential problem that you
don't have the resources to even test for more than a week.

I'm therefore adding this patch, and we'll fix or revert if the
"possible" IPMI problems do surface.