CAM layer gets stuck when sim->refcount > 0 at UMASS detach (10-current)

[Hans Petter Selasky]

Hi,

When running a stress test on a USB mass storage device, the CAM layer will 
eventually block at detach, because the refcount is not going towards zero on 
the SIM structure.

Test script: ugenX.Y is a mass storage device

#!/bin/sh

while true
do
usbconfig -d ugenX.Y reset
echo "."
done

After a minute or two the printouts above stops because the CAM layer is not 
releasing the reference count, indicating a bug somewhere during probe of the 
newly attached device.

Any patches, ideas or things you want me to test?

Please CC me hence I'm not subscribed to this list.

--HPS