Re: [mpf70-devel] GUI and 2.6

[Adrian R. <ad...@fh...>]

On Mon, Apr 26, 2004 at 10:05:43AM +0200, Jeremy Lain=E9 wrote:
> Cool, for the GUI, I'll have a look at it. I had taken care to separate
> MPF70 operations from the Qt-specific stuff, did you make use of the
> existing library? If so, it should be quite easy to add your source to
> the CVS repository.

Yes I did use your library to access the device.

> In the very near future, I am more interested in the work you did on th=
e
> 2.6 kernel, as it's something I've been meaning to do but never got
> round to! I took a look at how the USB & block device drivers had
> changed since 2.4 and there seemed to be a lot..

Some information about my 2.6 driver. As I already mentioned the driver
is still _very_ buggy. The first thing I have done during porting the
driver was to change block device stuff and I thing I have done it
pretty well. The next step was to port the usb related part. In the
beginning it seemed that this would not be very difficult and I quickly
had a version which compiled, loaded ans supported mounting the device.
But during mounting the kernel keeps printing call traces because the
driver does somehow call schedule although we are atomic.
After reading source code and mailinglists I think I do know now what
the problem is. During the mount operation we aren't any more in the
process context but in interrupt context and this is an atomic
situation. The usb call I am currently using (usb_bulk_msg()) does
however call schedule() somewhere. To avoid this situation I have to use
usb_submit_urb() because this call is not synchronous. I tired using
usb_submit_urb() but was not able to communicate with the player but I
am still trying to get this working.

Do you have any documentation of the hardware because the only thing I
have is your source :-)

		Adrian

--=20
Adrian Reber <ad...@li...>            http://lisas.de/~adrian/
intoxicated, adj.:
	When you feel sophisticated without being able to pronounce it.