Bug 17873
| Summary: | Can not routing IPV6 packet with tun-device | ||||||
|---|---|---|---|---|---|---|---|
| Product: | Base System | Reporter: | greg <greg> | ||||
| Component: | kern | Assignee: | Brian Somers <brian> | ||||
| Status: | Closed FIXED | ||||||
| Severity: | Affects Only Me | ||||||
| Priority: | Normal | ||||||
| Version: | 4.0-STABLE | ||||||
| Hardware: | Any | ||||||
| OS: | Any | ||||||
| Attachments: |
|
||||||
> > >Number: 17873 > >Category: kern > >Synopsis: Can not routing IPV6 packet with tun-device [.....] > This is a diffs for net/if_tun.c This breaks the if_tun module. If the module is built with INET6 defined, it'll fail to load on a non-INET6 kernel. If it's not, it can't call in6_ifattach. if_tun.c is already broken by calling if_attach() here. We need to abstract the interface creation in a way similar to the way data's dispatched via sys/net/intrq.c. Maybe there should be a all_ifattach() that's unconditionally built into the kernel and does something like #ifdef INET if_attach(ifp); #endif #ifdef INET6 in6_ifattach(ifp, IN6_IFT_P2P, 0, 0); #endif and so on ? There'd need to be a more generic way of telling it to use IN6_IFT_P2P, and I'm not sure why the if_type = IFT_RS232 bit is necessary.... -- Brian <brian@Awfulhak.org> <brian@[uk.]FreeBSD.org> <http://www.Awfulhak.org> <brian@[uk.]OpenBSD.org> Don't _EVER_ lose your sense of humour ! Responsible Changed From-To: freebsd-bugs->brian I've got an interest in keeping if_tun.ko as functional as the built in version. > <200004091030.LAA02368@hak.lan.Awfulhak.org>$B$N5-;v$K$*$$$F(B > brian@Awfulhak.org$B$5$s$O=q$-$^$7$?!#(B > > >This breaks the if_tun module. > > Oh, Sorry, this is kernel built-in case. if_tun is built as a module too. Any changes to it should work for /sys/modules. > > If the module is built with INET6 > >defined, it'll fail to load on a non-INET6 kernel. If it's not, > > In this case. tun device was staticaly linked in the kernel. > > >if_tun.c is already broken by calling if_attach() here. > > Where do I call in6_ifattach() collectly ? In if_attach() ? > I think it seems to be not wrong. > > >all_ifattach() that's unconditionally built into the kernel and does > >something like > > Look like little dirty, if all_ifattach called at opening tun device. > But these code has no problem basicly for me. The idea here is to move the ``#ifdef INET6'' decisions out of if_tun.c as this decision cannot be made at module build time. if there was a global all_ifattach() function that if_tun calls at softc creation time, it would make the problem go away. > >#ifdef INET > > if_attach(ifp); > >#endif > >#ifdef INET6 > > in6_ifattach(ifp, IN6_IFT_P2P, 0, 0); > >#endif > > So, if quated lines will be changed like this. > > in6_ifattach(ifp, ifp->if_type, 0, 0); > > and insert into some where. Ok, but that still needs to be done somewhere in *every* compiled kernel so that the module doesn't get built with #ifdefs that will fail in some environments. > --- > Fingerprint16 = 4F CC 44 F8 54 BE 45 3A 4F 9F 1C 4E 5E 3B 91 E9 > Fingerprint20 = 12CA 6B2D DC50 8248 A636 992B 0292 F548 D65F 4D5B > -----------------+-----------------------------------$BA}ED!!7r:n(B---+ > $B;0!&O;$r<i$m$&!*(B | greg@greg.rim.or.jp | > $B!!$*2H$X5"$m$&!*(B | greg@fxis.fujixerox.co.jp | > | http://www.st.rim.or.jp/~greg/ | > -----------------+------------------------------------------------+ -- Brian <brian@Awfulhak.org> <brian@[uk.]FreeBSD.org> <http://www.Awfulhak.org> <brian@[uk.]OpenBSD.org> Don't _EVER_ lose your sense of humour ! State Changed From-To: open->feedback Could the submitter confirm if just setting ifp->if_type in if_tun.c solves the problem (as discussed in April) ? State Changed From-To: feedback->closed if_type set in if_tun.c as per discussions. I'll MFC on Wednesday |
V6 enviroment was completed. And after setting up tun device for V6. creating a static routeing information by od -h /dev/tun0 & route add -inet6 default -interface tun0 And execute ping6 -I tun0 fe80::1 At normal case. V6 packet was routed into interface tun0. But packet was rejected before tun device. So problem dose exist on initializing routine. if tun device is staticaly network interface, Seem to be no problem. But it is not. Must execute a initializing routine "in6_ifattach" directly. Fix: This is a diffs for net/if_tun.c How-To-Repeat: Open and configure some tun for v6. And send some v6 packet which is routed into tun device.