USB2 - umass problem

[M. Warner Losh]

In message: <200902091035.26738.nick at anywi.com>
            Nick Hibma <nick at anywi.com> writes:
: > : > By some reason devfs semantic was changed:
: > : > Instead of /dev/cuaU0.[0-2] and /dev/ttyU0.[0-2], I've get
: > : > /dev/cuaU[0-2] /dev/ttyU[0-2] and! /dev/cuau1 /dev/ttyu1
: > : > What is reason for such change (additional port with lowercase 'u'
: > : > and U[0-2] instead of more logical U0.[0-2]) ?
: > :
: > : It is because we are attaching drivers per interface instead of per
: > : device. A new modem unit is allocated every time we find a modem,
: > : simply put. If the modem has multiple instances in an interface,
: > : /dev/cuaU0.[0...] will be created. Else /dev/cuaU... .
: >
: > Generally, we try not to change the details of how a device attaches
: > /dev entries from release to release.  Why the change?
: 
: The USB1 u3g driver also attaches to interfaces, but collects all interfaces 
: in one go, leaving all unused interfaces available for other drivers (e.g. 
: umass) or claims them (to hide the 'driver disks'). It is the main reason 
: why I wrote a separate driver in the first place. Otherwise the UMTS cards 
: could be treated as serial ports without any port singalling.
: 
: It is important to be able to determine in an automated way the 2 or more 
: serial ports that belong together. As an example: If you create a router 
: box that automatically configures itself depending on the hardware it 
: finds, we somehow need to find out which two serial ports are found on each 
: GPRS/UMTS card, so we can assign the first one to PPP and the other one to 
: our control application.

If the devices have serial numbers, then I think that's published in
the sysctl tree...

Warner