Instant panic when trying to unload nvidia.ko

Hi,

I'm running -CURRENT from yesterday. In a futile attempt to make some
use of my rusty RivaTNT card, I installed the nvidia kernel port, but it
seems to be doing some weird stuff with memory. This appeared in the
messages log:

...
Oct 28 16:04:33 scienide kernel: malloc() of "32" with the following
non-sleepable locks held:
Oct 28 16:04:33 scienide kernel: exclusive sleep mutex dev.mtx_api r = 0
(0xc39d478c) locked _at_
/usr/ports/x11/nvidia-driver/work/NVIDIA-FreeBSD-x86-1.0-4365/src/nvidi
a_subr.c:753
Oct 28 16:04:33 scienide kernel: malloc() of "32" with the following
non-sleepable locks held:
Oct 28 16:04:33 scienide kernel: exclusive sleep mutex dev.mtx_api r = 0
(0xc39d478c) locked _at_
/usr/ports/x11/nvidia-driver/work/NVIDIA-FreeBSD-x86-1.0-4365/src/nvidi
a_subr.c:753
Oct 28 16:04:33 scienide kernel: malloc() of "4096" with the following
non-sleepable locks held:
Oct 28 16:04:33 scienide kernel: exclusive sleep mutex dev.mtx_api r = 0
(0xc39d478c) locked _at_
/usr/ports/x11/nvidia-driver/work/NVIDIA-FreeBSD-x86-1.0-4365/src/nvidi
a_subr.c:753
...

And now, the fun part. After booting, if I try to unload the module I
get the attached panic. I know the nvidia kernel is proprietary code and
yada yada, but the problem seems to be in the open source part of the
code, so I have some hope I can get to fix it with a little help :-)


Cheers,
-- 
	Miguel Mendez <flynn_at_energyhq.es.eu.org>
	http://www.energyhq.es.eu.org