rkagan at mail.ru
Sat Apr 9 12:32:14 EDT 2005
On Sat, Apr 09, 2005 at 06:08:23PM +0200, Duncan Sands wrote:
> > The following trivial patch makes it behave somewhat better. The
> > endpoint number of 1 is arbitrary, the important thing is supposedly
> > that it differs from 0. Can't outline the exact scenario which led to
> > the failures, though...
> Hi Roman, I'm confused... shouldn't there only be trouble if you
> try to send packets or open a connection (that starts the
> receive tasklet)?
Ah, stupid me! I've tried with my usbatm patch applied, which submits
the reading urbs and starts reading data from the modem immediately in
usbatm_atm_init(). I'll check with the CVS usbatm.
BTW did excluding of the currently disconnected interface from the list
of unbound interfaces fix the problem?
FWIW, with a single-interface device (and my usbatm patch, with the
above mentioned consequences), I get:
- CVS testatm: hang on loading
- CVS testatm + the patch to exclude the current interface: testatm loads
/ unloads OK, but interferes with cat /proc/bus/usb/devices and
various files in device's directory in sysfs
- CVS testatm + the patch to use endpoint 1: OK
More information about the Usbatm