On Fri, 18 Aug 2006, Havard Eidnes wrote:

> > Yes I am working on that just now - the ubtbcmfw device will apparently
> > work if you fix it to compile with gcc4 -
>
> It's not so much gcc4, as e.g. post-timecounter adaptation and a
> couple of other interface changes, among them compensating for
> usbd_bulk_transfer() both being able to read and write, so for
> the write case it can't take a const buffer...  The attached diff
> should fix it to at least build.

I have a similar patch but I've been holding off fixing it because its
still not quite right. I dont have a BCM2033 device but apparently
ubtbcmfw panics if the device is present at boot time, but works ok if the
device is inserted after the filesystem is up and running.

Even if this part was fixed (and presumably it didnt do that when it was
written), I'm still not quite happy that it just keeps trying every N
seconds until it gets the firmware. It could be possible (eg on a faster
system) that it will miss the rc.d/btconfig script and the bluetooth
controller will remain unconfigured.

Unfortunately it can't work like the bt3c(4) driver (which loads the
firmware automatically when the device is enabled) because after you load
the firmware the device detaches and reattaches to the USB bus with a
different product ID which would be a little confusing for btconfig.

So, the only way I can see to get firmware is loaded without trouble is to
do it manually before btconfig happens. I'll make this into a package
which will also install the firmware files, and ubtbcmfw(4) will
disappear.

iain