Oliver Hartkopp wrote:
>  
>
> 2. I got this inconsistent lock state, i've not seen before:
>

Hm - that was not completely correct 8->

I found some more of these lock state issues in my /var/log/kern.log but 
all of them are in the tg3 receive path ... (attached).

So this problem seems not to come from your request_firmware() 
conversion. Maybe there was some other change in the 2.6.28-git that 
causes these tg3 problems.

Regards,
Oliver