'Llo All,

I have been using CDCEther recently, a device driver that is to be 
driving CDC- devices (over USB, that is). It should be working (as a 
design goal) with a standard Linux host. However, up to now I only get 
it working in around 5 percent of the cases. At first, I thought that it 
was due to the CDCEther- module that my device didn't get started. But 
in the running of my tests, I've started suspecting my hotplug 
configuration. When it works OK (like I said, in around 5 percent of the 
cases), I get a log like in attachment 1 (log_OK). When it does not 
work, I get a log like in attachment 2 (log_NOT_OK).

What is happening, is that my Linux CDC host driver (i.e. CDCEther) does 
(most of the times) not come to the point of sending the USB- command 
'Set Interface 1, Alternate 1'. This is needed before my CDC- device can 
start sending/receiving Ethernet frames. I see that CDCEther correctly 
claims the USB- device (see the logs), but that's all. Because all 
descriptors are fetched by the host (I believe that's done by the USB- 
generic part of the kernel, not so ?), and they are correctly answered 
by my device. But at the point where the host should send 'Set Intf 1 
Alt 0' which should after some time be succeeded by 'Set Intf 1 Alt 1', 
nothing happens. I have started thinking that the host only does this 
once he has been able to create an eth1- device (see the logs).

IF the system comes to the point of finding the needed hotplug 
information (i.e., if there is not that bunch of hotplug- errors in the 
log), it looks like there is a valid eth1 created. But in the other 
case, I get a whole lot of errors in the log, and there is NO eth1- 
device created. In the good case, the CDCEther- driver selects Intf 1 
Alt 1, and thus there can be network traffic (which I am still to debug, 
but then I'd at least have a working test configuration...) What is 
bothering me, is that it SOMETIMES works. Are there reasons to believe 
there are some timing issues involved here ?

I am using SuSE linux 8.2, so kernel 2.4.20. I already tried the same 
with the new 2.6 standard kernel (using usbnet, the replacing module for 
the 2.4- kernel CDCEther module). But apart from having less debug 
information, the phenomenon remains the same.

Thanks for any helpful answer,

Kind regards,

Philippe Bertin.

P.S. I think that this newsgroup is the most feasable for my problem, 
unless ... linux-usb-devel or -users ?