From mboxrd@z Thu Jan 1 00:00:00 1970 From: Ian Walters Date: Tue, 27 Aug 2002 23:50:14 +0000 Subject: Re: config questions Message-Id: List-Id: References: In-Reply-To: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: linux-hotplug@vger.kernel.org On Wed, 28 Aug 2002 12:53 am, David Brownell wrote: > > I just finished (successfully after much pain) get hotplug to initialize > > a device for me when it was plugged in. But to do this I had to bypass > > all of the existing net.agent script. > > Hmm, you didn't say what Linux distro you're using. They differ > in how they manage network interfaces ... though there does seem > to be a common failing of requring interfaces to be predefined, > combined with inadequate (or even incorrect) documentation on > how to predefine them. > Using SuSE 8.0 at the moment. > But then it's complicated by the fact that different network devices > all have different initialization models. At least some of those > issues are already noted in the net.agent script. > > > I am guessing I didn't find the right documentation. I will tell you > > what I did and hopefully someone can tell me what I should have done. > > > > Have usbnet loaded as a module. > > Not necessary, it'll be modprobed when needed. Did you look at > > http://www.linux-usb.org/usbnet/ admittedly not. It was already part of my kernel source code and worked without me having to do anything, so I didn't do anything :) > > for information about this driver? Your question made me push out > some pending updates, which should get synced to the mirrors sometime > tonight (document date 28 August). More examples, including zcip, > but not yet a complete "out of the box" config example (except for a > simple laptop/desktop setup). > > The configuration I'd like to see work is basically combining the > bridge tools (brctl etc) and zeroconf (zcip, plus ...) so those > interfaces never require preconfiguration. (It'd be an option, > but not the only or preferred one.) > Pre-configuration will always need to be possible. It might be I want a alert to go somwhere else when the device is plugged in, or a certain program to start. But I understand what you mean here, its best if the 'end' guy only has to tack on something, and then only if he needs to do something out of the ordinary. > > in /etc/hotplug/net.agent before anything, do > > > > > > mesg Checking for subscript for $INTERFACE $ACTION > > if [ -x "/etc/hotplug/interfaces/$INTERFACE" ] ; then > > mesg Calling subscript for $INTERFACE $ACTION > > /etc/hotplug/interfaces/$INTERFACE $ACTION > > exit 0 > > fi > > > > and in /etc/hotplug/interfaces reacted to the arguments registered and > > derigistered to do what I needed for /dev/usb0 > > There's no such thing as /dev/usb0! :) Um, yeah. my bad. What I was trying to say was 'unique ident for my device', which for me is INTERFACE=usb0. > > But what you're doing there is defining a new network administration > policy, which is something hotplug has so far avoided doing. Though > I'm pretty much convinced it'll be essential for IP-over-USB, since > otherwise distros will continue requiring pre-configuration (yeech) > instead of just making that a (less desirable) policy option. > > The /etc/sysconfig/network-scripts/ifcfg-usb0 script is how at least > some distros handle that problem, not that there's good documentation > on how to make them work. Not interfaces/usb0. > > > Ideally I would always like to say... for devX put file into > > interfaces/devX and in that file do X for plugged in and Y for unplugged. > > Just like the init.d scripts. > > Whereas, the networking scripts currently invoke "ifup" on interfaces > that are known to the distro's network admin tools. They can't often > do anything useful on "unregister", like "ifdown", since the interface > is gone by then. (Though ISTR that either SuSE or Debian had a version > of "ifup" that updated some user mode database, so "ifdown" had to purge > that data...) And sysadmin tools don't often handle "usb0" (etc) yet. > > > Information on why my approach is wrong would help, the page > > http://linux-hotplug.sourceforge.net/ > > talked a lot about what hotplug without ever getting to the one question > > most people would have... > > "I have new device X and I want Y to happen when I plug it in. What do I > > do?" > > If you're asking as a user, the answer is: "It's probably supposed to > happen automatically, without needing pre-configuration. But not all the > tools are there yet." > > If you're asking as a developer, then "UTSL, and good patches always help" > is closer to the answer. Most of those pages are directed towards > developers right now. > > And it'd be a big help if someone packaged a general "hotplug event > triggers user dialog" facility, so that when sysadmin attention is needed, > the system can at least initiate a conversation with the user. This sounds exactly what I am thinking about. Something you could run, plug in your device, and then tell it what to do on that trigger, after all the hotplug stuff has been done. Guess its source code that I would be diving into and not Documentation at this point. Might have a look on the weekend. > - Dave Thanks for the information. I definately have a much clear idea of how its set up now and whats going on. Ian. ------------------------------------------------------- This sf.net email is sponsored by: Jabber - The world's fastest growing real-time communications platform! Don't just IM. Build it in! http://www.jabber.com/osdn/xim _______________________________________________ Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net Linux-hotplug-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel