Peter Dunlap writes: > James Carlson wrote: > > Peter Dunlap writes: > > > >> http://cr.opensolaris.org/~pdunlap/iscsit-webrev/webrev/ > >> > > idm_so.c: [...] > > 1006,1010: what is this about? Why would you want "either" IPv6 or > > IPv4? Shouldn't you either know exactly which one you want, or need > > to construct *both*? [...] > I realize this is a bit of a delayed reaction but I don't really > understand this comment -- it sounds like this code doesn't do what we > think it does:
IPv6 and IPv4 sockets are different, so the idea that you could create either one and be equally happy here, or that you'd need to try them successively seems a bit strange to me. I was expecting to see two separate sockets -- one for v4 and the other for v6 -- if this service was intended to handle IPv4 and IPv6 at the same time. Using one IPv6 socket is possible, but generally limits the options you can use, as all of the network layer options are distinct, and neither applies to the other. (In other words, if you set an IPPROTO_IPV6 option, we don't translate that into an "equivalent" IPPROTO_IP option.) > The intent was to accept both IPv6 and IPv4 connections and I think the > logic is trying to "fall back to IPv4-only" if IPv6 is not available. > Are you saying that we can get by with just the PF_INET6 socket assuming > we always want to accept both IPv6 and IPv4 connections? Is there any > configuration where we would not be able to create the IPv6 listening > socket in Nevada? No. The integration of CR 6366093 changed the svc:/network/loopback service so that it always plumbs up IPv6 on lo0. A quick Perl demo on a system with no IPv6 configured: #!/usr/bin/perl use IO::Socket; socket(Handle,PF_INET6,SOCK_DGRAM,0) || die "no IPv6 socket"; print "that works\n"; Prints: that works -- James Carlson, Solaris Networking <[EMAIL PROTECTED]> Sun Microsystems / 35 Network Drive 71.232W Vox +1 781 442 2084 MS UBUR02-212 / Burlington MA 01803-2757 42.496N Fax +1 781 442 1677 _______________________________________________ networking-discuss mailing list [email protected]
