retitle 175855 ping6: document -r (DONTROUTE) option breakage severity 175855 minor thanks
Le lundi 20 janvier 2003 21:38, vous avez écrit : > On Wednesday 8 January 2003, at 13 h 6, > > "Noah L. Meyerhans" <[EMAIL PROTECTED]> wrote: > > The documentation is wrong. The Linux kernel does not implement > > the DONTROUTE socket option. I don't think this is a bug in Linux, > > either; I think this is more fundamental to IPv6. > > I do not see why. IMHO, this is the same problem as with the "default on-link assumption" that is now discouraged because it was so annoying. By the way, I don't think printing a warning at ping6 startup is such a good idea. Afterall, ping6 *does* the SO_DONTROUTE option properly, and setsockopt returns no error... so if the option ever gets implemented by the kernel, there will be no way to determine whether the warning should be printed or not. This should rather be mentioned in the documentation. > > I notice that the FreeBSD > > kernel also does not implement that sockopt. > > May be a possible workaround would be to send packets with a TTL of > 1? The ICMP Time Exceeded would indicate the failure. It ensures the packet won't be routed, but as far as layer 2 is concerned, it will still send the packet to the gateway instead of trying neighbor discovery against the destination. As such, it might replace -r in some case, but not all of them. On the other hand, if you want to check whether a node is on-link, it is probably more appropriate to use ndisc6 or even "ip -6 neigh list". Otherwise, you can of course invoke "ping6 -t 1" manually. Regards, -- Rémi Denis-Courmont http://www.remlab.net/
pgp4920X6F7dL.pgp
Description: PGP signature
