Do 31. Dez 15:32:40 CET 2009 Gert Doering * Basic IPv6 p2mp functionality implemented * new options: - server-ipv6 - ifconfig-ipv6 - ifconfig-ipv6-pool - route-ipv6 - iroute-ipv6 * modules touched: - init.c: init & setup IPv6 route list & add/delete IPv6 routes - tun.c: add "ifconfig" and "route" handling for IPv6 - multi.c: IPv6 ifconfig-pool assignments put to route-hash table push to client - pool.c: extend pools to handle IPv4+IPv6, and also return IPv6 address IPv6 address saved to file if ifconfig-pool-persist is set (but ignored on read due to the way pools work) - mroute.c: handle reading src/dst addresses from IPv6 packets (so multi.c can check against route-hash table) handle printing of IPv6 mroute_addr structure - helper.c: implement "server-ipv6" macro (->ifconfig-ipv6, pool, ...) - options.c: implement all the new options add helper functions for IPv6 address handling - forward.c: tell do_route() about IPv6 routes - route.c: handle IPv6 route lists + route option lists extend add_routes() to do IPv4 + IPv6 route lists extend delete_routes() to do IPv4 + IPv6 route lists implement add_route_ipv6(), delete_route_ipv6() to call system-dependend external program to do the work - push.c: handle pushing of "ifconfig-ipv6" option - socket.c: helper function to check & print IPv6 address strings * known issues: - operating system support on all but Linux (ifconfig, route) - route-ipv6 gateway handling - iroute-ipv6 not implemented - TAP support: ifconfig, routing (route needs gateway!) * release as patch 20091231-1 Thu Dec 31 17:02:08 CET 2009 * NetBSD port (NetBSD 3.1 on Sparc64) * mroute.c, socket.c: make byte/word access to in6_addr more portable * tun.c: fix IPv6 ifconfig arguments on NetBSD still doesn't work on NetBSD 3.1, "ifconfig tun0 inet6..." errors with ifconfig: SIOCAIFADDR: Address family not supported by protocol family (sys/net/if_tun.c, needs to be revision 1.80 or later, NetBSD PR 32944, included in NetBSD 4.0 and up) Fri Jan 1 14:07:15 CET 2010 * FreeBSD port (FreeBSD 6.3-p12 on i386) * tun.c: implement IPv6 ifconfig setting for FreeBSD * route.c: fix %s/%s argument to IPv6 route add/delete command for *BSD * TEST SUCCESS: FreeBSD 6.3-p12, server-ipv6, route-ipv6, ccd/iroute-ipv6 * multi.c: implement setting and deleting of iroute-ipv6 (multi_add_iroutes(), multi_del_iroutes()) * mroute.c: add mroute_helper_add_iroute6(), mroute_helper_del_iroute6() * mroute.h: add prototypes, increase MR_HELPER_NET_LEN to 129 (/0.../128) * multi.c: zeroize host part of IPv6 iroutes in multi_learn_in6_addr() * mroute.c: implement mroute_addr_mask_host_bits() for IPv6 * TEST SUCCESS: Linux 2.6.30 (Gentoo)/iproute2, server-ipv6, ccd/iroute-ipv6 * TEST SUCCESS: Linux 2.6.30 (Gentoo)/ifconfig, client-ipv6 * TEST FAIL: NetBSD 5.0, IPv6 client - "ifconfig tun0 .../64" does not create a "connected" route - adding routes fails --> more work to do here. * release as patch 20100101-1 * TEST FAIL: FreeBSD 6.3-p12 server "--topology subnet" Linux/ifconfig client - BSD sends ICMP6 neighbor solicitations, which are ignored by Linux - server tun interface is not in p2p mode, client tun interface *is* * TEST SUCCESS: non-ipv6 enabled client -> "--server-ipv6" server (warnings in the log file, but no malfunctions) Sat Jan 2 19:48:35 CET 2010 * tun.c: change "ipv6_support()", do not turn off tt->ipv6 unconditionally if we don't know about OS IPv6 support - just log warning * tun.c: implement "ifconfig inet6" setting for MacOS X / Darwin * route.c: split *BSD system dependent part of add/delete_route_ipv6() into FreeBSD/Dragonfly and NetBSD/Darwin/OpenBSD variants ("2001:db8::/64" vs. "2001:db8:: --prefixlen 64"). * tun.c: on MacOS X, NetBSD and OpenBSD, explicitely set on-link route * TEST SUCCESS: MacOS X, client-ipv6 with route-ipv6 Sun Jan 3 10:55:31 CET 2010 * route.c: NetBSD fails with "-iface tun0", needs gateway address (assume that the same syntax is needed for OpenBSD) * route.h: introduce "remote_endpoint_ipv6" into "struct route_ipv6_list" * init.c: pass "ifconfig_ipv6_remote" as gateway to init_route_ipv6_list() * route.c: - init_route_ipv6(): use "remote_endpoint_ipv6" as IPv6 gateway address if no gateway was specified explicitely - init_route_ipv6_list(): fill in "remote_endpoint_ipv6", if parseable - get rid of "GATEWAY-LESS ROUTE6" warning * route.c, add_route_ipv6() - explicitely clear host bits of base address, to be able to more easily set up "connected" /64 routes on NetBSD+Darwin - split system-dependent part between Darwin and NetBSD/OpenBSD (Darwin can use "-iface tun0", NetBSD/OpenBSD get gateway address) - change Solaris comments from "known-broken" to "unknown" * tun.c: rework NetBSD tunnel initialization and tun_read() / tun_write() to work the same way OpenBSD and NetBSD do - tunnel is put into "multi-af" mode, and all packet read/write activity is prepended by a 32 bit value specifying the address family. * TEST SUCCESS: NetBSD 5.0/Sparc64: client-ipv6 with route-ipv6 * TEST SUCCESS: MacOS X 10.5: client-ipv6 with route-ipv6 * (RE-)TEST SUCCESS: Linux/iproute2: server-ipv6 Linux/ifconfig: client-ipv6 FreeBSD 6.3: server-ipv6 * release as patch 20100103-1 * options.c: document all new options in "--help" * tun.c: fix typo in Solaris-specific section * socket.h, socket.c: change u_int32_t to uint32_t (Solaris - and all the rest of the code uses "uintNN" anyway) Mon Jan 4 17:46:58 CET 2010 * socket.c: rework add_in6_addr() to use 32-bit access to struct in6_addr (Solaris has no 16-bit values in union, but this is more elegant as well) * tun.c: fix "ifconfig inet6" command for Solaris * tun.c: make sure "tun0 inet6" is unplumbed first, cleanup leftovers * route.c: add routes with "metric 0" on solaris, otherwise they just don't work (someone who understands Solaris might want to fix this). * Solaris "sort of" works now - ifconfig works, route add does not give errors, "netstat -rn" looks right, but packets are discarded unless the routes are installed with "metric 0". So we just use "metric 0"... * CAVEAT: Solaris "ifconfig ... preferred" interferes with source address selection. So if there are any active IPv6 interfaces configured with "preferred", packets leaving out the tunnel will use the wrong source IPv6 address. Not fixable from within OpenVPN. * CAVEAT2: Solaris insists on doing DHCPv6 on tun0 interfaces by default, so DHCPv6 solicitation packets will be seen. Since the server end has no idea what to do with them, they are a harmless nuisance. Fixable on the Solaris side via "ndpd.conf" (see ``man ifconfig''). * release as patch 20100104-1 Fri Jan 8 10:00:50 CET 2010 * import into git repository * options.c: add sanity checks for most typical error cases (--ifconfig-ipv6-pool configured with no --ifconfig-ipv6, etc) * options.c: modify get_ipv6_addr() to be more flexible about netbits (optional now, default to /64) and to return the address-without-netbits string now (-> for options that want the IPv6 address in printable form, but without /nn) * options.c: modify --ifconfig-ipv6 to optionally accept /netbits, you can do now "ifconfig-ipv6 2001:df8::1/64 2001:df8::2" or just "ifconfig-ipv6 2001:df8::5 2001:df8::7", defaulting to /64 * options.h: add necessary structure elements for --ifconfig-ipv6-push * options.c: implement "parse options" side of --ifconfig-ipv6-push Tue Jan 12 22:42:09 CET 2010 * tun.c: in TARGET_NETBSD #ifdef, distinguish between "old" code (IPv4 only, but unmodified read/write) and "new" code (multi-af, extra 32 bit AF on read/write of the tun interface) - pre-4.0 NetBSD systems don't have TUNSIFHEAD, no way to have common code. * TEST SUCCESS: NetBSD 5.0/Sparc64: client-ipv6 with route-ipv6 (v4+v6) * TEST SUCCESS: NetBSD 3.1/Sparc64: client-ipv6 with route-ipv6 (v4-only) Thu Jan 14 15:41:50 CET 2010 * multi.c: if "--ifconfig-push" is used together with "--ifconfig-ipv6-pool" and no "--ifconfig-ipv6-push" is seen, issue warning - the current implementation of pools has IPv6 tied to IPv4, so if v4 does not use the pool, it breaks for IPv6. Not a *big* problem (since there is enough v6, just give those users a static v6 address as well), but needs to be pointed out clearly. * release as patch 20100114-1 Tue Feb 16 14:43:28 CET 2010 * options.c: print "IPv6 payload patch" release date in "--version" * tun.c: undo change to init_tun() (moving "bool tun" and call to "is_tun_p2p()" further up) - it wasn't needed and breaks "make check" * git stuff: rebase on David Sommerseth's openvpn-testing git tree * release as patch 20100216-1 Fri Feb 26 19:59:01 CET 2010 * init.c: initialize tuntap->ipv6 in do_init_tun() (to make sure it's always initialized early-enough, independent of the sequence of do_ifconfig()/open_tun() [see ifconfig_order() in tun.h]) * tun.c, init.c: remove "bool ipv6" argument to tuncfg(), open_tun() and open_tun_generic() - obsoleted by previous change * tun.c: remove ipv6_support() - original purpose was unclear, and all current platforms (except linux-very-old) fully support IPv6 now :-) * tun.c: initial implementation of "netsh" IPv6-ifconfig for Win32 * RE-TEST SUCCESS: Linux/i386/ifconfig, client-tun/net30, v4+v6 Sun Feb 28 17:05:57 CET 2010 * tun.c: NetBSD dependent part: correct destroying/re-creation of tun dev * tun.c: move adding of "connected" IPv6 prefix to new helper function, add_route_connected_v6_net() * RE-TEST SUCCESS: NetBSD 5.0/Sparc64, client-tun/net30, v4+v6 * RE-TEST SUCCESS: NetBSD 3.1/Sparc64: client-tun/net30, v4-only * RE-TEST SUCCESS: Linux/i386/iproute2: server-tun/net30, v4+v6 * tun.c: add #ifdef TARGET_DARWIN block for *_tun() functions, to be able to modify close_tun() for unconfiguring IPv6 * tun.c: on close_tun() on MacOS X, need to de-configure "lo0" route for configured IPv6 address * RE-TEST SUCCESS: MacOS X (10.5)/i386: client-tun/net30, v4+v6 * route.c: implement ipv6 route adding / deletion via "netsh" for WIN32 * TEST FAIL: Windows XP fails, because the tun/tap driver does not forward IPv6 frames kernel->userland if in "tun" mode * options.c: set IPv6 version to 20100228-1 * release as patch 20100228-1 Sun Mar 7 19:17:33 CET 2010 * options.c: set IPv6 version to 20100307-1 * TODO.IPv6: add note about OpenBSD TODO (#16) * route.c: set (and remove) "magic next hop" fe80::8 for IPv6 routes on Win32 * install-win32/settings.in: bump TAP driver version from 9.6 to 9.7 and TAP_RELDATE to "07/03/2010" * tap-win32/proto.h: add data types and definitions needed for IPv6 * tap-win32/types.h: add m_UserToTap_IPv6 ethernet header for IPv6 packets * tap-win32/tapdrvr.c: implement support for IPv6 in TUN mode: - IPv6 packets User->OS need correct ether type - IPv6 packets OS->User get correctly forwarded - IPv6 neighbour discovery packets for "fe80::8" (magic address installed as route-nexthop by OpenVPN.exe) get answered locally * TEST SUCCESS: WindowsXP/32bit: client-tun/net30, v4+v6 * tun.c: if IPv6 requested in TUN mode, and TUN/TAP driver version is older than 9.7, log warning and disable IPv6 (won't work anyway). * release as patch 20100307-1 Sat Jul 10 14:37:52 CEST 2010 * TEST SUCCESS: point-to-point tun mode with --ifconfig-ipv6 between Solaris10/sparc and Linux (Michal Ludvig) (using the whiteboard tun driver on Solaris, otherwise "no IPv6") Sun Aug 8 12:30:44 CEST 2010 * route.c: split NetBSD and OpenBSD parts of add_route_ipv6() and delete_route_ipv6(), implement OpenBSD variant (needs "-prefixlen nn" while NetBSD uses "/nn") * tun.c: implement IPv6 ifconfig for OpenBSD * tun.c: destroy tunX interface at tun_close() on OpenBSD (cleanup) * TEST SUCCESS: OpenBSD 4.7: client-tun/net30, v4+v6 Thu Sep 2 21:18:32 CEST 2010 * tun.c: the TAP binary in 2.2-beta3 has the IPv6 related changes, but the version number is 9.8 now -> check for 9.8, not 9.7 Wed Sep 22 22:20:37 CEST 2010 * tun.c: bugfix for Linux/iproute2/"topology subnet". Works :-) * TEST SUCCESS: Linux/ifconfig: client-tun/net30+subnet, v4+v6 * TEST SUCCESS: Linux/iproute2: client-tun/net30+subnet, v4+v6 * options.c: tag as 20100922-1 so "allmerged" users can see IPv6 change