Hi Alan, On Wednesday 16 November 2005 21:07, Alan Stern wrote: > On Wed, 16 Nov 2005, Duncan Sands wrote: > > > The problem here is the way ATM connections are opened: you specify > > a device number, called the interface number, as well as some other info. > > Alternatively, you can specify ATM_ITF_ANY instead of a device number, > > and all devices will be queried until one is found that is prepared to > > open the connection. > > > > With the speedtouch USB modems, the problem people usually see is: > > > > (A) for normal people, the modem is the only ATM device (and there's > > only one modem). Thus this is ATM device zero. > > > > (B) people connect using pppd; pppd doesn't support ATM_ITF_ANY, so > > most people specify that the connection should be to device number 0. > > pppd is usually run from a hotplug or init script. > > > > (C) at some point the modem spontaneously disconnects and reconnects > > (surprisingly common with some motherboards). pppd takes several seconds > > to understand that ATM device 0 has died; until it understands this, it > > keeps a connection open, which means that the ATM device hangs around. > > The "new" modem (= old one reconnected) creates a new ATM device, which > > is given device number 1, because number 0 is still being used by the > > old device. > > > > (D) a new instance of pppd is launched from a hotplug script, or the > > old instance of pppd tries to open a new connection once it understands > > that the old connection has died. It tries to connect to device 0, but > > there is no longer any device 0, only device 1, so the connection attempt > > fails. > > > > (E) the user notices that there is no internet connection and wonders why. > > The user runs pppd by hand - and it still fails to connect (no device 0). > > Eventually the user gives up and reboots the machine. > > This sounds like an application problem. Isn't it possible to add support > for ATM_ITF_ANY to pppd?
all the needed code is already in pppd, but it's turned off. Staszek found out why when he turned it on: it crashed the kernel at once! He sent a patch to fix it up (and cure a race while he was there), at which point Mitchell Blank Jr objected, saying that ATM_ITF_ANY was a bad idea in the first place, and he would rather get rid of it (I'm curious to hear his reasoning). This thread was the result. > It's a mistake to used fixed identification numbers in a situation where > devices can be added or removed at random and the numbers changed > accordingly. The ATM layer was designed in the good old days when devices were there to stay (and, I sometimes get the impression, when SMP was a twinkle in Alan Cox's eye)... All the best, Duncan. ------------------------------------------------------- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today Register for a JBoss Training Course. Free Certification Exam for All Training Attendees Through End of 2005. For more info visit: http://ads.osdn.com/?ad_id=7628&alloc_id=16845&op=click _______________________________________________ linux-usb-devel@lists.sourceforge.net To unsubscribe, use the last form field at: https://lists.sourceforge.net/lists/listinfo/linux-usb-devel
