Implementation Note KAME Project http://www.kame.net/ $KAME: IMPLEMENTATION,v 1.216 2001/05/25 07:43:01 jinmei Exp $ $FreeBSD: head/share/doc/IPv6/IMPLEMENTATION 228993 2011-12-30 11:11:54Z uqs $ NOTE: The document tries to describe behaviors/implementation choices of the latest KAME/*BSD stack. The description here may not be applicable to KAME-integrated *BSD releases, as we have certain amount of changes between them. Still, some of the content can be useful for KAME-integrated *BSD releases. Table of Contents 1. IPv6 1.1 Conformance 1.2 Neighbor Discovery 1.3 Scope Zone Index 1.3.1 Kernel internal 1.3.2 Interaction with API 1.3.3 Interaction with users (command line) 1.4 Plug and Play 1.4.1 Assignment of link-local, and special addresses 1.4.2 Stateless address autoconfiguration on hosts 1.4.3 DHCPv6 1.5 Generic tunnel interface 1.6 Address Selection 1.6.1 Source Address Selection 1.6.2 Destination Address Ordering 1.7 Jumbo Payload 1.8 Loop prevention in header processing 1.9 ICMPv6 1.10 Applications 1.11 Kernel Internals 1.12 IPv4 mapped address and IPv6 wildcard socket 1.12.1 KAME/BSDI3 and KAME/FreeBSD228 1.12.2 KAME/FreeBSD[34]x 1.12.2.1 KAME/FreeBSD[34]x, listening side 1.12.2.2 KAME/FreeBSD[34]x, initiating side 1.12.3 KAME/NetBSD 1.12.3.1 KAME/NetBSD, listening side 1.12.3.2 KAME/NetBSD, initiating side 1.12.4 KAME/BSDI4 1.12.4.1 KAME/BSDI4, listening side 1.12.4.2 KAME/BSDI4, initiating side 1.12.5 KAME/OpenBSD 1.12.5.1 KAME/OpenBSD, listening side 1.12.5.2 KAME/OpenBSD, initiating side 1.12.6 More issues 1.12.7 Interaction with SIIT translator 1.13 sockaddr_storage 1.14 Invalid addresses on the wire 1.15 Node's required addresses 1.15.1 Host case 1.15.2 Router case 1.16 Advanced API 1.17 DNS resolver 2. Network Drivers 2.1 FreeBSD 2.2.x-RELEASE 2.2 BSD/OS 3.x 2.3 NetBSD 2.4 FreeBSD 3.x-RELEASE 2.5 FreeBSD 4.x-RELEASE 2.6 OpenBSD 2.x 2.7 BSD/OS 4.x 3. Translator 3.1 FAITH TCP relay translator 3.2 IPv6-to-IPv4 header translator 4. IPsec 4.1 Policy Management 4.2 Key Management 4.3 AH and ESP handling 4.4 IPComp handling 4.5 Conformance to RFCs and IDs 4.6 ECN consideration on IPsec tunnels 4.7 Interoperability 4.8 Operations with IPsec tunnel mode 4.8.1 RFC2401 IPsec tunnel mode approach 4.8.2 draft-touch-ipsec-vpn approach 5. ALTQ 6. Mobile IPv6 6.1 KAME node as correspondent node 6.2 KAME node as home agent/mobile node 6.3 Old Mobile IPv6 code 7. Coding style 8. Policy on technology with intellectual property right restriction 1. IPv6 1.1 Conformance The KAME kit conforms, or tries to conform, to the latest set of IPv6 specifications. For future reference we list some of the relevant documents below (NOTE: this is not a complete list - this is too hard to maintain...). For details please refer to specific chapter in the document, RFCs, manpages come with KAME, or comments in the source code. Conformance tests have been performed on past and latest KAME STABLE kit, at TAHI project. Results can be viewed at http://www.tahi.org/report/KAME/. We also attended Univ. of New Hampshire IOL tests (http://www.iol.unh.edu/) in the past, with our past snapshots. RFC1639: FTP Operation Over Big Address Records (FOOBAR) * RFC2428 is preferred over RFC1639. ftp clients will first try RFC2428, then RFC1639 if failed. RFC1886: DNS Extensions to support IPv6 RFC1933: (see RFC2893) RFC1981: Path MTU Discovery for IPv6 RFC2080: RIPng for IPv6 * KAME-supplied route6d, bgpd and hroute6d support this. RFC2283: Multiprotocol Extensions for BGP-4 * so-called "BGP4+". * KAME-supplied bgpd supports this. RFC2292: Advanced Sockets API for IPv6 * see RFC3542 RFC2362: Protocol Independent Multicast-Sparse Mode (PIM-SM) * RFC2362 defines the packet formats and the protcol of PIM-SM. RFC2373: IPv6 Addressing Architecture * KAME supports node required addresses, and conforms to the scope requirement. RFC2374: An IPv6 Aggregatable Global Unicast Address Format * KAME supports 64-bit length of Interface ID. RFC2375: IPv6 Multicast Address Assignments * Userland applications use the well-known addresses assigned in the RFC. RFC2428: FTP Extensions for IPv6 and NATs * RFC2428 is preferred over RFC1639. ftp clients will first try RFC2428, then RFC1639 if failed. RFC2460: IPv6 specification RFC2461: Neighbor discovery for IPv6 * See 1.2 in this document for details. RFC2462: IPv6 Stateless Address Autoconfiguration * See 1.4 in this document for details. RFC2463: ICMPv6 for IPv6 specification * See 1.9 in this document for details. RFC2464: Transmission of IPv6 Packets over Ethernet Networks RFC2465: MIB for IPv6: Textual Conventions and General Group * Necessary statistics are gathered by the kernel. Actual IPv6 MIB support is provided as patchkit for ucd-snmp. RFC2466: MIB for IPv6: ICMPv6 group * Necessary statistics are gathered by the kernel. Actual IPv6 MIB support is provided as patchkit for ucd-snmp. RFC2467: Transmission of IPv6 Packets over FDDI Networks RFC2472: IPv6 over PPP RFC2492: IPv6 over ATM Networks * only PVC is supported. RFC2497: Transmission of IPv6 packet over ARCnet Networks RFC2545: Use of BGP-4 Multiprotocol Extensions for IPv6 Inter-Domain Routing RFC2553: (see RFC3493) RFC2671: Extension Mechanisms for DNS (EDNS0) * see USAGE for how to use it. * not supported on kame/freebsd4 and kame/bsdi4. RFC2673: Binary Labels in the Domain Name System * KAME/bsdi4 supports A6, DNAME and binary label to some extent. * KAME apps/bind8 repository has resolver library with partial A6, DNAME and binary label support. RFC2675: IPv6 Jumbograms * See 1.7 in this document for details. RFC2710: Multicast Listener Discovery for IPv6 RFC2711: IPv6 router alert option RFC2732: Format for Literal IPv6 Addresses in URL's * The spec is implemented in programs that handle URLs (like freebsd ftpio(3) and fetch(1), or netbsd ftp(1)) RFC2874: DNS Extensions to Support IPv6 Address Aggregation and Renumbering * KAME/bsdi4 supports A6, DNAME and binary label to some extent. * KAME apps/bind8 repository has resolver library with partial A6, DNAME and binary label support. RFC2893: Transition Mechanisms for IPv6 Hosts and Routers * IPv4 compatible address is not supported. * automatic tunneling (4.3) is not supported. * "gif" interface implements IPv[46]-over-IPv[46] tunnel in a generic way, and it covers "configured tunnel" described in the spec. See 1.5 in this document for details. RFC2894: Router renumbering for IPv6 RFC3041: Privacy Extensions for Stateless Address Autoconfiguration in IPv6 RFC3056: Connection of IPv6 Domains via IPv4 Clouds * So-called "6to4". * "stf" interface implements it. Be sure to read draft-itojun-ipv6-transition-abuse-01.txt below before configuring it, there can be security issues. RFC3142: An IPv6-to-IPv4 transport relay translator * FAITH tcp relay translator (faithd) implements this. See 3.1 for more details. RFC3152: Delegation of IP6.ARPA * libinet6 resolvers contained in the KAME snaps support to use the ip6.arpa domain (with the nibble format) for IPv6 reverse lookups. RFC3484: Default Address Selection for IPv6 * the selection algorithm for both source and destination addresses is implemented based on the RFC, though some rules are still omitted. RFC3493: Basic Socket Interface Extensions for IPv6 * IPv4 mapped address (3.7) and special behavior of IPv6 wildcard bind socket (3.8) are, - supported and turned on by default on KAME/FreeBSD[34] and KAME/BSDI4, - supported but turned off by default on KAME/NetBSD and KAME/FreeBSD5, - not supported on KAME/FreeBSD228, KAME/OpenBSD and KAME/BSDI3. see 1.12 in this document for details. * The AI_ALL and AI_V4MAPPED flags are not supported. RFC3542: Advanced Sockets API for IPv6 (revised) * For supported library functions/kernel APIs, see sys/netinet6/ADVAPI. * Some of the updates in the draft are not implemented yet. See TODO.2292bis for more details. RFC4007: IPv6 Scoped Address Architecture * some part of the documentation (especially about the routing model) is not supported yet. * zone indices that contain scope types have not been supported yet. draft-ietf-ipngwg-icmp-name-lookups-09: IPv6 Name Lookups Through ICMP draft-ietf-ipv6-router-selection-07.txt: Default Router Preferences and More-Specific Routes * router-side: both router preference and specific routes are supported. * host-side: only router preference is supported. draft-ietf-pim-sm-v2-new-02.txt A revised version of RFC2362, which includes the IPv6 specific packet format and protocol descriptions. draft-ietf-dnsext-mdns-00.txt: Multicast DNS * kame/mdnsd has test implementation, which will not be built in default compilation. The draft will experience a major change in the near future, so don't rely upon it. draft-ietf-ipngwg-icmp-v3-02.txt: ICMPv6 for IPv6 specification (revised) * See 1.9 in this document for details. draft-itojun-ipv6-tcp-to-anycast-01.txt: Disconnecting TCP connection toward IPv6 anycast address draft-ietf-ipv6-rfc2462bis-06.txt: IPv6 Stateless Address Autoconfiguration (revised) draft-itojun-ipv6-transition-abuse-01.txt: Possible abuse against IPv6 transition technologies (expired) * KAME does not implement RFC1933/2893 automatic tunnel. * "stf" interface implements some address filters. Refer to stf(4) for details. Since there's no way to make 6to4 interface 100% secure, we do not include "stf" interface into GENERIC.v6 compilation. * kame/openbsd completely disables IPv4 mapped address support. * kame/netbsd makes IPv4 mapped address support off by default. * See section 1.12.6 and 1.14 for more details. draft-itojun-ipv6-flowlabel-api-01.txt: Socket API for IPv6 flow label field * no consideration is made against the use of routing headers and such. 1.2 Neighbor Discovery Our implementation of Neighbor Discovery is fairly stable. Currently Address Resolution, Duplicated Address Detection, and Neighbor Unreachability Detection are supported. In the near future we will be adding an Unsolicited Neighbor Advertisement transmission command as an administration tool. Duplicated Address Detection (DAD) will be performed when an IPv6 address is assigned to a network interface, or the network interface is enabled (ifconfig up). It is documented in RFC2462 5.4. If DAD fails, the address will be marked "duplicated" and message will be generated to syslog (and usually to console). The "duplicated" mark can be checked with ifconfig. It is administrators' responsibility to check for and recover from DAD failures. We may try to improve failure recovery in future KAME code. A successor version of RFC2462 (called rfc2462bis) clarifies the behavior when DAD fails (i.e., duplicate is detected): if the duplicate address is a link-local address formed from an interface identifier based on the hardware address which is supposed to be uniquely assigned (e.g., EUI-64 for an Ethernet interface), IPv6 operation on the interface should be disabled. The KAME implementation supports this as follows: if this type of duplicate is detected, the kernel marks "disabled" in the ND specific data structure for the interface. Every IPv6 I/O operation in the kernel checks this mark, and the kernel will drop packets received on or being sent to the "disabled" interface. Whether the IPv6 operation is disabled or not can be confirmed by the ndp(8) command. See the man page for more details. DAD procedure may not be effective on certain network interfaces/drivers. If a network driver needs long initialization time (with wireless network interfaces this situation is popular), and the driver mistakingly raises IFF_RUNNING before the driver becomes ready, DAD code will try to transmit DAD probes to not-really-ready network driver and the packet will not go out from the interface. In such cases, network drivers should be corrected. Some of network drivers loop multicast packets back to themselves, even if instructed not to do so (especially in promiscuous mode). In such cases DAD may fail, because the DAD engine sees inbound NS packet (actually from the node itself) and considers it as a sign of duplicate. In this case, drivers should be corrected to honor IFF_SIMPLEX behavior. For example, you may need to check source MAC address on an inbound packet, and reject it if it is from the node itself. Neighbor Discovery specification (RFC2461) does not talk about neighbor cache handling in the following cases: (1) when there was no neighbor cache entry, node received unsolicited RS/NS/NA/redirect packet without link-layer address (2) neighbor cache handling on medium without link-layer address (we need a neighbor cache entry for IsRouter bit) For (1), we implemented workaround based on discussions on IETF ipngwg mailing list. For more details, see the comments in the source code and email thread started from (IPng 7155), dated Feb 6 1999. IPv6 on-link determination rule (RFC2461) is quite different from assumptions in BSD IPv4 network code. To implement the behavior in RFC2461 section 6.3.6 (3), the kernel needs to know the default outgoing interface. To configure the default outgoing interface, use commands like "ndp -I de0" as root. Then the kernel will have a "default" route to the interface with the cloning "C" bit being on. This default route will cause to make a neighbor cache entry for every destination that does not match an explicit route entry. Note that we intentionally disable configuring the default interface by default. This is because we found it sometimes caused inconvenient situation while it was rarely useful in practical usage. For example, consider a destination that has both IPv4 and IPv6 addresses but is only reachable via IPv4. Since our getaddrinfo(3) prefers IPv6 by default, an (TCP) application using the library with PF_UNSPEC first tries to connect to the IPv6 address. If we turn on RFC 2461 6.3.6 (3), we have to wait for quite a long period before the first attempt to make a connection fails. If we turn it off, the first attempt will immediately fail with EHOSTUNREACH, and then the application can try the next, reachable address. The notion of the default interface is also disabled when the node is acting as a router. The reason is that routers tend to control all routes stored in the kernel and the default route automatically installed would rather confuse the routers. Note that the spec misuse the word "host" and "node" in several places in Section 5.2 of RFC 2461. We basically read the word "node" in this section as "host," and thus believe the implementation policy does not break the specification. To avoid possible DoS attacks and infinite loops, KAME stack will accept only 10 options on ND packet. Therefore, if you have 20 prefix options attached to RA, only the first 10 prefixes will be recognized. If this troubles you, please contact the KAME team and/or modify nd6_maxndopt in sys/netinet6/nd6.c. If there are high demands we may provide a sysctl knob for the variable. Proxy Neighbor Advertisement support is implemented in the kernel. For instance, you can configure it by using the following command: # ndp -s fe80::1234%ne0 0:1:2:3:4:5 proxy where ne0 is the interface which attaches to the same link as the proxy target. There are certain limitations, though: - It does not send unsolicited multicast NA on configuration. This is MAY behavior in RFC2461. - It does not add random delay before transmission of solicited NA. This is SHOULD behavior in RFC2461. - We cannot configure proxy NDP for off-link address. The target address for proxying must be link-local address, or must be in prefixes configured to node which does proxy NDP. - RFC2461 is unclear about if it is legal for a host to perform proxy ND. We do not prohibit hosts from doing proxy ND, but there will be very limited use in it. Starting mid March 2000, we support Neighbor Unreachability Detection (NUD) on p2p interfaces, including tunnel interfaces (gif). NUD is turned on by default. Before March 2000 the KAME stack did not perform NUD on p2p interfaces. If the change raises any interoperability issues, you can turn off/on NUD by per-interface basis. Use "ndp -i interface -nud" to turn it off. Consult ndp(8) for details. RFC2461 specifies upper-layer reachability confirmation hint. Whenever upper-layer reachability confirmation hint comes, ND process can use it to optimize neighbor discovery process - ND process can omit real ND exchange and keep the neighbor cache state in REACHABLE. We currently have two sources for hints: (1) setsockopt(IPV6_REACHCONF) defined by the RFC3542 API, and (2) hints from tcp(6)_input. It is questionable if they are really trustworthy. For example, a rogue userland program can use IPV6_REACHCONF to confuse the ND process. Neighbor cache is a system-wide information pool, and it is bad to allow a single process to affect others. Also, tcp(6)_input can be hosed by hijack attempts. It is wrong to allow hijack attempts to affect the ND process. Starting June 2000, the ND code has a protection mechanism against incorrect upper-layer reachability confirmation. The ND code counts subsequent upper-layer hints. If the number of hints reaches the maximum, the ND code will ignore further upper-layer hints and run real ND process to confirm reachability to the peer. sysctl net.inet6.icmp6.nd6_maxnudhint defines the maximum # of subsequent upper-layer hints to be accepted. (from April 2000 to June 2000, we rejected setsockopt(IPV6_REACHCONF) from non-root process - after a local discussion, it looks that hints are not that trustworthy even if they are from privileged processes) If inbound ND packets carry invalid values, the KAME kernel will drop these packet and increment statistics variable. See "netstat -sn", icmp6 section. For detailed debugging session, you can turn on syslog output from the kernel on errors, by turning on sysctl MIB net.inet6.icmp6.nd6_debug. nd6_debug can be turned on at bootstrap time, by defining ND6_DEBUG kernel compilation option (so you can debug behavior during bootstrap). nd6_debug configuration should only be used for test/debug purposes - for a production environment, nd6_debug must be set to 0. If you leave it to 1, malicious parties can inject broken packet and fill up /var/log partition. 1.3 Scope Zone Index IPv6 uses scoped addresses. It is therefore very important to specify the scope zone index (link index for a link-local address, or site index for a site-local address) with an IPv6 address. Without a zone index, a scoped IPv6 address is ambiguous to the kernel, and the kernel would not be able to determine the outbound zone for a packet to the scoped address. KAME code tries to address the issue in several ways. The entire architecture of scoped addresses is documented in RFC4007. One non-trivial point of the architecture is that the link scope is (theoretically) larger than the interface scope. That is, two different interfaces can belong to a same single link. However, in a normal operation, we can assume that there is 1-to-1 relationship between links and interfaces. In other words, we can usually put links and interfaces in the same scope type. The current KAME implementation assumes the 1-to-1 relationship. In particular, we use interface names such as "ne1" as unique link identifiers. This would be much more human-readable and intuitive than numeric identifiers, but please keep your mind on the theoretical difference between links and interfaces. Site-local addresses are very vaguely defined in the specs, and both the specification and the KAME code need tons of improvements to enable its actual use. For example, it is still very unclear how we define a site, or how we resolve host names in a site. There is work underway to define behavior of routers at site border, but, we have almost no code for site boundary node support (neither forwarding nor routing) and we bet almost noone has. We recommend, at this moment, you to use global addresses for experiments - there are way too many pitfalls if you use site-local addresses. 1.3.1 Kernel internal In the kernel, the link index for a link-local scope address is embedded into the 2nd 16bit-word (the 3rd and 4th bytes) in the IPv6 address. For example, you may see something like: fe80:1::200:f8ff:fe01:6317 in the routing table and the interface address structure (struct in6_ifaddr). The address above is a link-local unicast address which belongs to a network link whose link identifier is 1 (note that it eqauls to the interface index by the assumption of our implementation). The embedded index enables us to identify IPv6 link-local addresses over multiple links effectively and with only a little code change. The use of the internal format must be limited inside the kernel. In particular, addresses sent by an application should not contain the embedded index (except via some very special APIs such as routing sockets). Instead, the index should be specified in the sin6_scope_id field of a sockaddr_in6 structure. Obviously, packets sent to or received from must not contain the embedded index either, since the index is meaningful only within the sending/receiving node. In order to deal with the differences, several kernel routines are provided. These are available by including . Typically, the following functions will be most generally used: - int sa6_embedscope(struct sockaddr_in6 *sa6, int defaultok); Embed sa6->sin6_scope_id into sa6->sin6_addr. If sin6_scope_id is 0, defaultok is non-0, and the default zone ID (see RFC4007) is configured, the default ID will be used instead of the value of the sin6_scope_id field. On success, sa6->sin6_scope_id will be reset to 0. This function returns 0 on success, or a non-0 error code otherwise. - int sa6_recoverscope(struct sockaddr_in6 *sa6); Extract embedded zone ID in sa6->sin6_addr and set sa6->sin6_scope_id to that ID. The embedded ID will be cleared with 0. This function returns 0 on success, or a non-0 error code otherwise. - int in6_clearscope(struct in6_addr *in6); Reset the embedded zone ID in 'in6' to 0. This function never fails, and returns 0 if the original address is intact or non 0 if the address is modified. The return value doesn't matter in most cases; currently, the only point where we care about the return value is ip6_input() for checking whether the source or destination addresses of the incoming packet is in the embedded form. - int in6_setscope(struct in6_addr *in6, struct ifnet *ifp, u_int32_t *zoneidp); Embed zone ID determined by the address scope type for 'in6' and the interface 'ifp' into 'in6'. If zoneidp is non NULL, *zoneidp will also have the zone ID. This function returns 0 on success, or a non-0 error code otherwise. The typical usage of these functions is as follows: sa6_embedscope() will be used at the socket or transport layer to convert a sockaddr_in6 structure passed by an application into the kernel-internal form. In this usage, the second argument is often the 'ip6_use_defzone' global variable. sa6_recoverscope() will also be used at the socket or transport layer to convert an in6_addr structure with the embedded zone ID into a sockaddr_in6 structure with the corresponding ID in the sin6_scope_id field (and without the embedded ID in sin6_addr). in6_clearscope() will be used just before sending a packet to the wire to remove the embedded ID. In general, this must be done at the last stage of an output path, since otherwise the address would lose the ID and could be ambiguous with regard to scope. in6_setscope() will be used when the kernel receives a packet from the wire to construct the kernel internal form for each address field in the packet (typical examples are the source and destination addresses of the packet). In the typical usage, the third argument 'zoneidp' will be NULL. A non-NULL value will be used when the validity of the zone ID must be checked, e.g., when forwarding a packet to another link (see ip6_forward() for this usage). An application, when sending a packet, is basically assumed to specify the appropriate scope zone of the destination address by the sin6_scope_id field (this might be done transparently from the application with getaddrinfo() and the extended textual format - see below), or at least the default scope zone(s) must be configured as a last resort. In some cases, however, an application could specify an ambiguous address with regard to scope, expecting it is disambiguated in the kernel by some other means. A typical usage is to specify the outgoing interface through another API, which can disambiguate the unspecified scope zone. Such a usage is not recommended, but the kernel implements some trick to deal with even this case. A rough sketch of the trick can be summarized as the following sequence. sa6_embedscope(dst, ip6_use_defzone); in6_selectsrc(dst, ..., &ifp, ...); in6_setscope(&dst->sin6_addr, ifp, NULL); sa6_embedscope() first tries to convert sin6_scope_id (or the default zone ID) into the kernel-internal form. This can fail with an ambiguous destination, but it still tries to get the outgoing interface (ifp) in the attempt of determining the source address of the outgoing packet using in6_selectsrc(). If the interface is detected, and the scope zone was originally ambiguous, in6_setscope() can finally determine the appropriate ID with the address itself and the interface, and construct the kernel-internal form. See, for example, comments in udp6_output() for more concrete example. In any case, kernel routines except ones in netinet6/scope6.c MUST NOT directly refer to the embedded form. They MUST use the above interface functions. In particular, kernel routines MUST NOT have the following code fragment: /* This is a bad practice. Don't do this */ if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) sin6->sin6_addr.s6_addr16[1] = htons(ifp->if_index); This is bad for several reasons. First, address ambiguity is not specific to link-local addresses (any non-global multicast addresses are inherently ambiguous, and this is particularly true for interface-local addresses). Secondly, this is vulnerable to future changes of the embedded form (the embedded position may change, or the zone ID may not actually be the interface index). Only scope6.c routines should know the details. The above code fragment should thus actually be as follows: /* This is correct. */ in6_setscope(&sin6->sin6_addr, ifp, NULL); (and catch errors if possible and necessary) 1.3.2 Interaction with API There are several candidates of API to deal with scoped addresses without ambiguity. The IPV6_PKTINFO ancillary data type or socket option defined in the advanced API (RFC2292 or RFC3542) can specify the outgoing interface of a packet. Similarly, the IPV6_PKTINFO or IPV6_RECVPKTINFO socket options tell kernel to pass the incoming interface to user applications. These options are enough to disambiguate scoped addresses of an incoming packet, because we can uniquely identify the corresponding zone of the scoped address(es) by the incoming interface. However, they are too strong for outgoing packets. For example, consider a multi-sited node and suppose that more than one interface of the node belongs to a same site. When we want to send a packet to the site, we can only specify one of the interfaces for the outgoing packet with these options; we cannot just say "send the packet to (one of the interfaces of) the site." Another kind of candidates is to use the sin6_scope_id member in the sockaddr_in6 structure, defined in RFC2553. The KAME kernel interprets the sin6_scope_id field properly in order to disambiguate scoped addresses. For example, if an application passes a sockaddr_in6 structure that has a non-zero sin6_scope_id value to the sendto(2) system call, the kernel should send the packet to the appropriate zone according to the sin6_scope_id field. Similarly, when the source or the destination address of an incoming packet is a scoped one, the kernel should detect the correct zone identifier based on the address and the receiving interface, fill the identifier in the sin6_scope_id field of a sockaddr_in6 structure, and then pass the packet to an application via the recvfrom(2) system call, etc. However, the semantics of the sin6_scope_id is still vague and on the way to standardization. Additionally, not so many operating systems support the behavior above at this moment. In summary, - If your target system is limited to KAME based ones (i.e. BSD variants and KAME snaps), use the sin6_scope_id field assuming the kernel behavior described above. - Otherwise, (i.e. if your program should be portable on other systems than BSDs) + Use the advanced API to disambiguate scoped addresses of incoming packets. + To disambiguate scoped addresses of outgoing packets, * if it is okay to just specify the outgoing interface, use the advanced API. This would be the case, for example, when you should only consider link-local addresses and your system assumes 1-to-1 relationship between links and interfaces. * otherwise, sorry but you lose. Please rush the IETF IPv6 community into standardizing the semantics of the sin6_scope_id field. Routing daemons and configuration programs, like route6d and ifconfig, will need to manipulate the "embedded" zone index. These programs use routing sockets and ioctls (like SIOCGIFADDR_IN6) and the kernel API will return IPv6 addresses with the 2nd 16bit-word filled in. The APIs are for manipulating kernel internal structure. Programs that use these APIs have to be prepared about differences in kernels anyway. getaddrinfo(3) and getnameinfo(3) support an extended numeric IPv6 syntax, as documented in RFC4007. You can specify the outgoing link, by using the name of the outgoing interface as the link, like "fe80::1%ne0" (again, note that we assume there is 1-to-1 relationship between links and interfaces.) This way you will be able to specify a link-local scoped address without much trouble. Other APIs like inet_pton(3) and inet_ntop(3) are inherently unfriendly with scoped addresses, since they are unable to annotate addresses with zone identifier. 1.3.3 Interaction with users (command line) Most of user applications now support the extended numeric IPv6 syntax. In this case, you can specify outgoing link, by using the name of the outgoing interface like "fe80::1%ne0" (sorry for the duplicated notice, but please recall again that we assume 1-to-1 relationship between links and interfaces). This is even the case for some management tools such as route(8) or ndp(8). For example, to install the IPv6 default route by hand, you can type like # route add -inet6 default fe80::9876:5432:1234:abcd%ne0 (Although we suggest you to run dynamic routing instead of static routes, in order to avoid configuration mistakes.) Some applications have command line options for specifying an appropriate zone of a scoped address (like "ping6 -I ne0 ff02::1" to specify the outgoing interface). However, you can't always expect such options. Additionally, specifying the outgoing "interface" is in theory an overspecification as a way to specify the outgoing "link" (see above). Thus, we recommend you to use the extended format described above. This should apply to the case where the outgoing interface is specified. In any case, when you specify a scoped address to the command line, NEVER write the embedded form (such as ff02:1::1 or fe80:2::fedc), which should only be used inside the kernel (see Section 1.3.1), and is not supposed to work. 1.4 Plug and Play The KAME kit implements most of the IPv6 stateless address autoconfiguration in the kernel. Neighbor Discovery functions are implemented in the kernel as a whole. Router Advertisement (RA) input for hosts is implemented in the kernel. Router Solicitation (RS) output for endhosts, RS input for routers, and RA output for routers are implemented in the userland. 1.4.1 Assignment of link-local, and special addresses IPv6 link-local address is generated from IEEE802 address (ethernet MAC address). Each of interface is assigned an IPv6 link-local address automatically, when the interface becomes up (IFF_UP). Also, direct route for the link-local address is added to routing table. Here is an output of netstat command: Internet6: Destination Gateway Flags Netif Expire fe80::%ed0/64 link#1 UC ed0 fe80::%ep0/64 link#2 UC ep0 Interfaces that has no IEEE802 address (pseudo interfaces like tunnel interfaces, or ppp interfaces) will borrow IEEE802 address from other interfaces, such as ethernet interfaces, whenever possible. If there is no IEEE802 hardware attached, last-resort pseudorandom value, which is from MD5(hostname), will be used as source of link-local address. If it is not suitable for your usage, you will need to configure the link-local address manually. If an interface is not capable of handling IPv6 (such as lack of multicast support), link-local address will not be assigned to that interface. See section 2 for details. Each interface joins the solicited multicast address and the link-local all-nodes multicast addresses (e.g. fe80::1:ff01:6317 and ff02::1, respectively, on the link the interface is attached). In addition to a link-local address, the loopback address (::1) will be assigned to the loopback interface. Also, ::1/128 and ff01::/32 are automatically added to routing table, and loopback interface joins node-local multicast group ff01::1. 1.4.2 Stateless address autoconfiguration on hosts In IPv6 specification, nodes are separated into two categories: routers and hosts. Routers forward packets addressed to others, hosts does not forward the packets. net.inet6.ip6.forwarding defines whether this node is a router or a host (router if it is 1, host if it is 0). It is NOT recommended to change net.inet6.ip6.forwarding while the node is in operation. IPv6 specification defines behavior for "host" and "router" quite differently, and switching from one to another can cause serious troubles. It is recommended to configure the variable at bootstrap time only. The first step in stateless address configuration is Duplicated Address Detection (DAD). See 1.2 for more detail on DAD. When a host hears Router Advertisement from the router, a host may autoconfigure itself by stateless address autoconfiguration. This behavior can be controlled by the net.inet6.ip6.accept_rtadv sysctl variable and a per-interface flag managed in the kernel. The latter, which we call "if_accept_rtadv" here, can be changed by the ndp(8) command (see the manpage for more details). When the sysctl variable is set to 1, and the flag is set, the host autoconfigures itself. By autoconfiguration, network address prefixes for the receiving interface (usually global address prefix) are added. The default route is also configured. Routers periodically generate Router Advertisement packets. To request an adjacent router to generate RA packet, a host can transmit Router Solicitation. To generate an RS packet at any time, use the "rtsol" command. The "rtsold" daemon is also available. "rtsold" generates Router Solicitation whenever necessary, and it works greatly for nomadic usage (notebooks/laptops). If one wishes to ignore Router Advertisements, use sysctl to set net.inet6.ip6.accept_rtadv to 0. Additionally, ndp(8) command can be used to control the behavior per-interface basis. To generate Router Advertisement from a router, use the "rtadvd" daemon. Note that the IPv6 specification assumes the following items and that nonconforming cases are left unspecified: - Only hosts will listen to router advertisements - Hosts have a single network interface (except loopback) This is therefore unwise to enable net.inet6.ip6.accept_rtadv on routers, or multi-interface hosts. A misconfigured node can behave strange (KAME code allows nonconforming configuration, for those who would like to do some experiments). To summarize the sysctl knob: accept_rtadv forwarding role of the node --- --- --- 0 0 host (to be manually configured) 0 1 router 1 0 autoconfigured host (spec assumes that hosts have a single interface only, autoconfigred hosts with multiple interfaces are out-of-scope) 1 1 invalid, or experimental (out-of-scope of spec) The if_accept_rtadv flag is referred only when accept_rtadv is 1 (the latter two cases). The flag does not have any effects when the sysctl variable is 0. See 1.2 in the document for relationship between DAD and autoconfiguration. 1.4.3 DHCPv6 We supply a tiny DHCPv6 server/client in kame/dhcp6. However, the implementation is premature (for example, this does NOT implement address lease/release), and it is not in default compilation tree on some platforms. If you want to do some experiment, compile it on your own. DHCPv6 and autoconfiguration also needs more work. "Managed" and "Other" bits in RA have no special effect to stateful autoconfiguration procedure in DHCPv6 client program ("Managed" bit actually prevents stateless autoconfiguration, but no special action will be taken for DHCPv6 client). 1.5 Generic tunnel interface GIF (Generic InterFace) is a pseudo interface for configured tunnel. Details are described in gif(4) manpage. Currently v6 in v6 v6 in v4 v4 in v6 v4 in v4 are available. Use "gifconfig" to assign physical (outer) source and destination address to gif interfaces. Configuration that uses same address family for inner and outer IP header (v4 in v4, or v6 in v6) is dangerous. It is very easy to configure interfaces and routing tables to perform infinite level of tunneling. Please be warned. gif can be configured to be ECN-friendly. See 4.5 for ECN-friendliness of tunnels, and gif(4) manpage for how to configure. If you would like to configure an IPv4-in-IPv6 tunnel with gif interface, read gif(4) carefully. You may need to remove IPv6 link-local address automatically assigned to the gif interface. 1.6 Address Selection 1.6.1 Source Address Selection The KAME kernel chooses the source address for an outgoing packet sent from a user application as follows: 1. if the source address is explicitly specified via an IPV6_PKTINFO ancillary data item or the socket option of that name, just use it. Note that this item/option overrides the bound address of the corresponding (datagram) socket. 2. if the corresponding socket is bound, use the bound address. 3. otherwise, the kernel first tries to find the outgoing interface of the packet. If it fails, the source address selection also fails. If the kernel can find an interface, choose the most appropriate address based on the algorithm described in RFC3484. The policy table used in this algorithm is stored in the kernel. To install or view the policy, use the ip6addrctl(8) command. The kernel does not have pre-installed policy. It is expected that the default policy described in the draft should be installed at the bootstrap time using this command. This draft allows an implementation to add implementation-specific rules with higher precedence than the rule "Use longest matching prefix." KAME's implementation has the following additional rules (that apply in the appeared order): - prefer addresses on alive interfaces, that is, interfaces with the UP flag being on. This rule is particularly useful for routers, since some routing daemons stop advertising prefixes (addresses) on interfaces that have become down. - prefer addresses on "preferred" interfaces. "Preferred" interfaces can be specified by the ndp(8) command. By default, no interface is preferred, that is, this rule does not apply. Again, this rule is particularly useful for routers, since there is a convention, among router administrators, of assigning "stable" addresses on a particular interface (typically a loopback interface). In any case, addresses that break the scope zone of the destination, or addresses whose zone do not contain the outgoing interface are never chosen. When the procedure above fails, the kernel usually returns EADDRNOTAVAIL to the application. In some cases, the specification explicitly requires the implementation to choose a particular source address. The source address for a Neighbor Advertisement (NA) message is an example. Under the spec (RFC2461 7.2.2) NA's source should be the target address of the corresponding NS's target. In this case we follow the spec rather than the above rule. If you would like to prohibit the use of deprecated address for some reason, configure net.inet6.ip6.use_deprecated to 0. The issue related to deprecated address is described in RFC2462 5.5.4 (NOTE: there is some debate underway in IETF ipngwg on how to use "deprecated" address). As documented in the source address selection document, temporary addresses for privacy extension are less preferred to public addresses by default. However, for administrators who are particularly aware of the privacy, there is a system-wide sysctl(3) variable "net.inet6.ip6.prefer_tempaddr". When the variable is set to non-zero, the kernel will rather prefer temporary addresses. The default value of this variable is 0. 1.6.2 Destination Address Ordering KAME's getaddrinfo(3) supports the destination address ordering algorithm described in RFC3484. Getaddrinfo(3) needs to know the source address for each destination address and policy entries (described in the previous section) for the source and destination addresses. To get the source address, the library function opens a UDP socket and tries to connect(2) for the destination. To get the policy entry, the function issues sysctl(3). 1.7 Jumbo Payload KAME supports the Jumbo Payload hop-by-hop option used to send IPv6 packets with payloads longer than 65,535 octets. But since currently KAME does not support any physical interface whose MTU is more than 65,535, such payloads can be seen only on the loopback interface(i.e. lo0). If you want to try jumbo payloads, you first have to reconfigure the kernel so that the MTU of the loopback interface is more than 65,535 bytes; add the following to the kernel configuration file: options "LARGE_LOMTU" #To test jumbo payload and recompile the new kernel. Then you can test jumbo payloads by the ping6 command with -b and -s options. The -b option must be specified to enlarge the size of the socket buffer and the -s option specifies the length of the packet, which should be more than 65,535. For example, type as follows; % ping6 -b 70000 -s 68000 ::1 The IPv6 specification requires that the Jumbo Payload option must not be used in a packet that carries a fragment header. If this condition is broken, an ICMPv6 Parameter Problem message must be sent to the sender. KAME kernel follows the specification, but you cannot usually see an ICMPv6 error caused by this requirement. If KAME kernel receives an IPv6 packet, it checks the frame length of the packet and compares it to the length specified in the payload length field of the IPv6 header or in the value of the Jumbo Payload option, if any. If the former is shorter than the latter, KAME kernel discards the packet and increments the statistics. You can see the statistics as output of netstat command with `-s -p ip6' option: % netstat -s -p ip6 ip6: (snip) 1 with data size < data length So, KAME kernel does not send an ICMPv6 error unless the erroneous packet is an actual Jumbo Payload, that is, its packet size is more than 65,535 bytes. As described above, KAME kernel currently does not support physical interface with such a huge MTU, so it rarely returns an ICMPv6 error. TCP/UDP over jumbogram is not supported at this moment. This is because we have no medium (other than loopback) to test this. Contact us if you need this. IPsec does not work on jumbograms. This is due to some specification twists in supporting AH with jumbograms (AH header size influences payload length, and this makes it real hard to authenticate inbound packet with jumbo payload option as well as AH). There are fundamental issues in *BSD support for jumbograms. We would like to address those, but we need more time to finalize the task. To name a few: - mbuf pkthdr.len field is typed as "int" in 4.4BSD, so it cannot hold jumbogram with len > 2G on 32bit architecture CPUs. If we would like to support jumbogram properly, the field must be expanded to hold 4G + IPv6 header + link-layer header. Therefore, it must be expanded to at least int64_t (u_int32_t is NOT enough). - We mistakingly use "int" to hold packet length in many places. We need to convert them into larger numeric type. It needs a great care, as we may experience overflow during packet length computation. - We mistakingly check for ip6_plen field of IPv6 header for packet payload length in various places. We should be checking mbuf pkthdr.len instead. ip6_input() will perform sanity check on jumbo payload option on input, and we can safely use mbuf pkthdr.len afterwards. - TCP code needs careful updates in bunch of places, of course. 1.8 Loop prevention in header processing IPv6 specification allows arbitrary number of extension headers to be placed onto packets. If we implement IPv6 packet processing code in the way BSD IPv4 code is implemented, kernel stack may overflow due to long function call chain. KAME sys/netinet6 code is carefully designed to avoid kernel stack overflow. Because of this, KAME sys/netinet6 code defines its own protocol switch structure, as "struct ip6protosw" (see netinet6/ip6protosw.h). In addition to this, we restrict the number of extension headers (including the IPv6 header) in each incoming packet, in order to prevent a DoS attack that tries to send packets with a massive number of extension headers. The upper limit can be configured by the sysctl value net.inet6.ip6.hdrnestlimit. In particular, if the value is 0, the node will allow an arbitrary number of headers. As of writing this document, the default value is 50. IPv4 part (sys/netinet) remains untouched for compatibility. Because of this, if you receive IPsec-over-IPv4 packet with massive number of IPsec headers, kernel stack may blow up. IPsec-over-IPv6 is okay. 1.9 ICMPv6 After RFC2463 was published, IETF ipngwg has decided to disallow ICMPv6 error packet against ICMPv6 redirect, to prevent ICMPv6 storm on a network medium. KAME already implements this into the kernel. RFC2463 requires rate limitation for ICMPv6 error packets generated by a node, to avoid possible DoS attacks. KAME kernel implements two rate- limitation mechanisms, tunable via sysctl: - Minimum time interval between ICMPv6 error packets KAME kernel will generate no more than one ICMPv6 error packet, during configured time interval. net.inet6.icmp6.errratelimit controls the interval (default: disabled). - Maximum ICMPv6 error packet-per-second KAME kernel will generate no more than the configured number of packets in one second. net.inet6.icmp6.errppslimit controls the maximum packet-per-second value (default: 200pps) Basically, we need to pick values that are suitable against the bandwidth of link layer devices directly attached to the node. In some cases the default values may not fit well. We are still unsure if the default value is sane or not. Comments are welcome. 1.10 Applications For userland programming, we support IPv6 socket API as specified in RFC2553/3493, RFC3542 and upcoming internet drafts. TCP/UDP over IPv6 is available and quite stable. You can enjoy "telnet", "ftp", "rlogin", "rsh", "ssh", etc. These applications are protocol independent. That is, they automatically chooses IPv4 or IPv6 according to DNS. 1.11 Kernel Internals (*) TCP/UDP part is handled differently between operating system platforms. See 1.12 for details. The current KAME has escaped from the IPv4 netinet logic. While ip_forward() calls ip_output(), ip6_forward() directly calls if_output() since routers must not divide IPv6 packets into fragments. ICMPv6 should contain the original packet as long as possible up to 1280. UDP6/IP6 port unreach, for instance, should contain all extension headers and the *unchanged* UDP6 and IP6 headers. So, all IP6 functions except TCP6 never convert network byte order into host byte order, to save the original packet. tcp6_input(), udp6_input() and icmp6_input() can't assume that IP6 header is preceding the transport headers due to extension headers. So, in6_cksum() was implemented to handle packets whose IP6 header and transport header is not continuous. TCP/IP6 nor UDP/IP6 header structure don't exist for checksum calculation. To process IP6 header, extension headers and transport headers easily, KAME requires network drivers to store packets in one internal mbuf or one or more external mbufs. A typical old driver prepares two internal mbufs for 100 - 208 bytes data, however, KAME's reference implementation stores it in one external mbuf. "netstat -s -p ip6" tells you whether or not your driver conforms KAME's requirement. In the following example, "cce0" violates the requirement. (For more information, refer to Section 2.) Mbuf statistics: 317 one mbuf two or more mbuf:: lo0 = 8 cce0 = 10 3282 one ext mbuf 0 two or more ext mbuf Each input function calls IP6_EXTHDR_CHECK in the beginning to check if the region between IP6 and its header is continuous. IP6_EXTHDR_CHECK calls m_pullup() only if the mbuf has M_LOOP flag, that is, the packet comes from the loopback interface. m_pullup() is never called for packets coming from physical network interfaces. TCP6 reassembly makes use of IP6 header to store reassemble information. IP6 is not supposed to be just before TCP6, so ip6tcpreass structure has a pointer to TCP6 header. Of course, it has also a pointer back to mbuf to avoid m_pullup(). Like TCP6, both IP and IP6 reassemble functions never call m_pullup(). xxx_ctlinput() calls in_mrejoin() on PRC_IFNEWADDR. We think this is one of 4.4BSD implementation flaws. Since 4.4BSD keeps ia_multiaddrs in in_ifaddr{}, it can't use multicast feature if the interface has no unicast address. So, if an application joins to an interface and then all unicast addresses are removed from the interface, the application can't send/receive any multicast packets. Moreover, if a new unicast address is assigned to the interface, in_mrejoin() must be called. KAME's interfaces, however, have ALWAYS one link-local unicast address. These extensions have thus not been implemented in KAME. 1.12 IPv4 mapped address and IPv6 wildcard socket RFC2553/3493 describes IPv4 mapped address (3.7) and special behavior of IPv6 wildcard bind socket (3.8). The spec allows you to: - Accept IPv4 connections by AF_INET6 wildcard bind socket. - Transmit IPv4 packet over AF_INET6 socket by using special form of the address like ::ffff:10.1.1.1. but the spec itself is very complicated and does not specify how the socket layer should behave. Here we call the former one "listening side" and the latter one "initiating side", for reference purposes. Almost all KAME implementations treat tcp/udp port number space separately between IPv4 and IPv6. You can perform wildcard bind on both of the address families, on the same port. There are some OS-platform differences in KAME code, as we use tcp/udp code from different origin. The following table summarizes the behavior. listening side initiating side (AF_INET6 wildcard (connection to ::ffff:10.1.1.1) socket gets IPv4 conn.) --- --- KAME/BSDI3 not supported not supported KAME/FreeBSD228 not supported not supported KAME/FreeBSD3x configurable supported default: enabled KAME/FreeBSD4x configurable supported default: enabled KAME/NetBSD configurable supported default: disabled KAME/BSDI4 enabled supported KAME/OpenBSD not supported not supported The following sections will give you more details, and how you can configure the behavior. Comments on listening side: It looks that RFC2553/3493 talks too little on wildcard bind issue, specifically on (1) port space issue, (2) failure mode, (3) relationship between AF_INET/INET6 wildcard bind like ordering constraint, and (4) behavior when conflicting socket is opened/closed. There can be several separate interpretation for this RFC which conform to it but behaves differently. So, to implement portable application you should assume nothing about the behavior in the kernel. Using getaddrinfo() is the safest way. Port number space and wildcard bind issues were discussed in detail on ipv6imp mailing list, in mid March 1999 and it looks that there's no concrete consensus (means, up to implementers). You may want to check the mailing list archives. We supply a tool called "bindtest" that explores the behavior of kernel bind(2). The tool will not be compiled by default. If a server application would like to accept IPv4 and IPv6 connections, it should use AF_INET and AF_INET6 socket (you'll need two sockets). Use getaddrinfo() with AI_PASSIVE into ai_flags, and socket(2) and bind(2) to all the addresses returned. By opening multiple sockets, you can accept connections onto the socket with proper address family. IPv4 connections will be accepted by AF_INET socket, and IPv6 connections will be accepted by AF_INET6 socket (NOTE: KAME/BSDI4 kernel sometimes violate this - we will fix it). If you try to support IPv6 traffic only and would like to reject IPv4 traffic, always check the peer address when a connection is made toward AF_INET6 listening socket. If the address is IPv4 mapped address, you may want to reject the connection. You can check the condition by using IN6_IS_ADDR_V4MAPPED() macro. This is one of the reasons the author of the section (itojun) dislikes special behavior of AF_INET6 wildcard bind. Comments on initiating side: Advise to application implementers: to implement a portable IPv6 application (which works on multiple IPv6 kernels), we believe that the following is the key to the success: - NEVER hardcode AF_INET nor AF_INET6. - Use getaddrinfo() and getnameinfo() throughout the system. Never use gethostby*(), getaddrby*(), inet_*() or getipnodeby*(). - If you would like to connect to destination, use getaddrinfo() and try all the destination returned, like telnet does. - Some of the IPv6 stack is shipped with buggy getaddrinfo(). Ship a minimal working version with your application and use that as last resort. If you would like to use AF_INET6 socket for both IPv4 and IPv6 outgoing connection, you will need tweaked implementation in DNS support libraries, as documented in RFC2553/3493 6.1. KAME libinet6 includes the tweak in getipnodebyname(). Note that getipnodebyname() itself is not recommended as it does not handle scoped IPv6 addresses at all. For IPv6 name resolution getaddrinfo() is the preferred API. getaddrinfo() does not implement the tweak. When writing applications that make outgoing connections, story goes much simpler if you treat AF_INET and AF_INET6 as totally separate address family. {set,get}sockopt issue goes simpler, DNS issue will be made simpler. We do not recommend you to rely upon IPv4 mapped address. 1.12.1 KAME/BSDI3 and KAME/FreeBSD228 The platforms do not support IPv4 mapped address at all (both listening side and initiating side). AF_INET6 and AF_INET sockets are totally separated. Port number space is totally separate between AF_INET and AF_INET6 sockets. It should be noted that KAME/BSDI3 and KAME/FreeBSD228 are not conformant to RFC2553/3493 section 3.7 and 3.8. It is due to code sharing reasons. 1.12.2 KAME/FreeBSD[34]x KAME/FreeBSD3x and KAME/FreeBSD4x use shared tcp4/6 code (from sys/netinet/tcp*) and shared udp4/6 code (from sys/netinet/udp*). They use unified inpcb/in6pcb structure. 1.12.2.1 KAME/FreeBSD[34]x, listening side The platform can be configured to support IPv4 mapped address/special AF_INET6 wildcard bind (enabled by default). There is no kernel compilation option to disable it. You can enable/disable the behavior with sysctl (per-node), or setsockopt (per-socket). Wildcard AF_INET6 socket grabs IPv4 connection if and only if the following conditions are satisfied: - there's no AF_INET socket that matches the IPv4 connection - the AF_INET6 socket is configured to accept IPv4 traffic, i.e. getsockopt(IPV6_V6ONLY) returns 0. (XXX need checking) 1.12.2.2 KAME/FreeBSD[34]x, initiating side KAME/FreeBSD3x supports outgoing connection to IPv4 mapped address (::ffff:10.1.1.1), if the node is configured to accept IPv4 connections by AF_INET6 socket. (XXX need checking) 1.12.3 KAME/NetBSD KAME/NetBSD uses shared tcp4/6 code (from sys/netinet/tcp*) and shared udp4/6 code (from sys/netinet/udp*). The implementation is made differently from KAME/FreeBSD[34]x. KAME/NetBSD uses separate inpcb/in6pcb structures, while KAME/FreeBSD[34]x uses merged inpcb structure. It should be noted that the default configuration of KAME/NetBSD is not conformant to RFC2553/3493 section 3.8. It is intentionally turned off by default for security reasons. The platform can be configured to support IPv4 mapped address/special AF_INET6 wildcard bind (disabled by default). Kernel behavior can be summarized as follows: - default: special support code will be compiled in, but is disabled by default. It can be controlled by sysctl (net.inet6.ip6.v6only), or setsockopt(IPV6_V6ONLY). - add "INET6_BINDV6ONLY": No special support code for AF_INET6 wildcard socket will be compiled in. AF_INET6 sockets and AF_INET sockets are totally separate. The behavior is similar to what described in 1.12.1. sysctl setting will affect per-socket configuration at in6pcb creation time only. In other words, per-socket configuration will be copied from sysctl configuration at in6pcb creation time. To change per-socket behavior, you must perform setsockopt or reopen the socket. Change in sysctl configuration will not change the behavior or sockets that are already opened. 1.12.3.1 KAME/NetBSD, listening side Wildcard AF_INET6 socket grabs IPv4 connection if and only if the following conditions are satisfied: - there's no AF_INET socket that matches the IPv4 connection - the AF_INET6 socket is configured to accept IPv4 traffic, i.e. getsockopt(IPV6_V6ONLY) returns 0. You cannot bind(2) with IPv4 mapped address. This is a workaround for port number duplicate and other twists. 1.12.3.2 KAME/NetBSD, initiating side When getsockopt(IPV6_V6ONLY) is 0 for a socket, you can make an outgoing traffic to IPv4 destination over AF_INET6 socket, using IPv4 mapped address destination (::ffff:10.1.1.1). When getsockopt(IPV6_V6ONLY) is 1 for a socket, you cannot use IPv4 mapped address for outgoing traffic. 1.12.4 KAME/BSDI4 KAME/BSDI4 uses NRL-based TCP/UDP stack and inpcb source code, which was derived from NRL IPv6/IPsec stack. We guess it supports IPv4 mapped address and speical AF_INET6 wildcard bind. The implementation is, again, different from other KAME/*BSDs. 1.12.4.1 KAME/BSDI4, listening side NRL inpcb layer supports special behavior of AF_INET6 wildcard socket. There is no way to disable the behavior. Wildcard AF_INET6 socket grabs IPv4 connection if and only if the following condition is satisfied: - there's no AF_INET socket that matches the IPv4 connection 1.12.4.2 KAME/BSDI4, initiating side KAME/BSDi4 supports connection initiation to IPv4 mapped address (like ::ffff:10.1.1.1). 1.12.5 KAME/OpenBSD KAME/OpenBSD uses NRL-based TCP/UDP stack and inpcb source code, which was derived from NRL IPv6/IPsec stack. It should be noted that KAME/OpenBSD is not conformant to RFC2553/3493 section 3.7 and 3.8. It is intentionally omitted for security reasons. 1.12.5.1 KAME/OpenBSD, listening side KAME/OpenBSD disables special behavior on AF_INET6 wildcard bind for security reasons (if IPv4 traffic toward AF_INET6 wildcard bind is allowed, access control will become much harder). KAME/BSDI4 uses NRL-based TCP/UDP stack as well, however, the behavior is different due to OpenBSD's security policy. As a result the behavior of KAME/OpenBSD is similar to KAME/BSDI3 and KAME/FreeBSD228 (see 1.12.1 for more detail). 1.12.5.2 KAME/OpenBSD, initiating side KAME/OpenBSD does not support connection initiation to IPv4 mapped address (like ::ffff:10.1.1.1). 1.12.6 More issues IPv4 mapped address support adds a big requirement to EVERY userland codebase. Every userland code should check if an AF_INET6 sockaddr contains IPv4 mapped address or not. This adds many twists: - Access controls code becomes harder to write. For example, if you would like to reject packets from 10.0.0.0/8, you need to reject packets to AF_INET socket from 10.0.0.0/8, and to AF_INET6 socket from ::ffff:10.0.0.0/104. - If a protocol on top of IPv4 is defined differently with IPv6, we need to be really careful when we determine which protocol to use. For example, with FTP protocol, we can not simply use sa_family to determine FTP command sets. The following example is incorrect: if (sa_family == AF_INET) use EPSV/EPRT or PASV/PORT; /*IPv4*/ else if (sa_family == AF_INET6) use EPSV/EPRT or LPSV/LPRT; /*IPv6*/ else error; The correct code, with consideration to IPv4 mapped address, would be: if (sa_family == AF_INET) use EPSV/EPRT or PASV/PORT; /*IPv4*/ else if (sa_family == AF_INET6 && IPv4 mapped address) use EPSV/EPRT or PASV/PORT; /*IPv4 command set on AF_INET6*/ else if (sa_family == AF_INET6 && !IPv4 mapped address) use EPSV/EPRT or LPSV/LPRT; /*IPv6*/ else error; It is too much to ask for every body to be careful like this. The problem is, we are not sure if the above code fragment is perfect for all situations. - By enabling kernel support for IPv4 mapped address (outgoing direction), servers on the kernel can be hosed by IPv6 native packet that has IPv4 mapped address in IPv6 header source, and can generate unwanted IPv4 packets. draft-itojun-ipv6-transition-abuse-01.txt, draft-cmetz-v6ops-v4mapped-api- harmful-00.txt, and draft-itojun-v6ops-v4mapped-harmful-01.txt has more on this scenario. Due to the above twists, some of KAME userland programs has restrictions on the use of IPv4 mapped addresses: - rshd/rlogind do not accept connections from IPv4 mapped address. This is to avoid malicious use of IPv4 mapped address in IPv6 native packet, to bypass source-address based authentication. - ftp/ftpd assume that you are on dual stack network. IPv4 mapped address will be decoded in userland, and will be passed to AF_INET sockets (in other words, ftp/ftpd do not support SIIT environment). 1.12.7 Interaction with SIIT translator SIIT translator is specified in RFC2765. KAME node cannot become a SIIT translator box, nor SIIT end node (a node in SIIT cloud). To become a SIIT translator box, we need to put additional code for that. We do not have the code in our tree at this moment. There are multiple reasons that we are unable to become SIIT end node. (1) SIIT translators require end nodes in the SIIT cloud to be IPv6-only. Since we are unable to compile INET-less kernel, we are unable to become SIIT end node. (2) As presented in 1.12.6, some of our userland code assumes dual stack network. (3) KAME stack filters out IPv6 packets with IPv4 mapped address in the header, to secure non-SIIT case (which is much more common). Effectively KAME node will reject any packets via SIIT translator box. See section 1.14 for more detail about the last item. There are documentation issues too - SIIT document requires very strange things. For example, SIIT document asks IPv6-only (meaning no IPv4 code) node to be able to construct IPv4 IPsec headers. If a node knows how to construct IPv4 IPsec headers, that is not an IPv6-only node, it is a dual-stack node. The requirements imposed in SIIT document contradict with the other part of the document itself. 1.13 sockaddr_storage When RFC2553 was about to be finalized, there was discussion on how struct sockaddr_storage members are named. One proposal is to prepend "__" to the members (like "__ss_len") as they should not be touched. The other proposal was that don't prepend it (like "ss_len") as we need to touch those members directly. There was no clear consensus on it. As a result, RFC2553 defines struct sockaddr_storage as follows: struct sockaddr_storage { u_char __ss_len; /* address length */ u_char __ss_family; /* address family */ /* and bunch of padding */ }; On the contrary, XNET draft defines as follows: struct sockaddr_storage { u_char ss_len; /* address length */ u_char ss_family; /* address family */ /* and bunch of padding */ }; In December 1999, it was agreed that RFC2553bis (RFC3493) should pick the latter (XNET) definition. KAME kit prior to December 1999 used RFC2553 definition. KAME kit after December 1999 (including December) will conform to XNET definition, based on RFC3493 discussion. If you look at multiple IPv6 implementations, you will be able to see both definitions. As an userland programmer, the most portable way of dealing with it is to: (1) ensure ss_family and/or ss_len are available on the platform, by using GNU autoconf, (2) have -Dss_family=__ss_family to unify all occurrences (including header file) into __ss_family, or (3) never touch __ss_family. cast to sockaddr * and use sa_family like: struct sockaddr_storage ss; family = ((struct sockaddr *)&ss)->sa_family 1.14 Invalid addresses on the wire Some of IPv6 transition technologies embed IPv4 address into IPv6 address. These specifications themselves are fine, however, there can be certain set of attacks enabled by these specifications. Recent specification documents covers up those issues, however, there are already-published RFCs that does not have protection against those (like using source address of ::ffff:127.0.0.1 to bypass "reject packet from remote" filter). To name a few, these address ranges can be used to hose an IPv6 implementation, or bypass security controls: - IPv4 mapped address that embeds unspecified/multicast/loopback/broadcast IPv4 address (if they are in IPv6 native packet header, they are malicious) ::ffff:0.0.0.0/104 ::ffff:127.0.0.0/104 ::ffff:224.0.0.0/100 ::ffff:255.0.0.0/104 - 6to4 (RFC3056) prefix generated from unspecified/multicast/loopback/ broadcast/private IPv4 address 2002:0000::/24 2002:7f00::/24 2002:e000::/24 2002:ff00::/24 2002:0a00::/24 2002:ac10::/28 2002:c0a8::/32 - IPv4 compatible address that embeds unspecified/multicast/loopback/broadcast IPv4 address (if they are in IPv6 native packet header, they are malicious). Note that, since KAME doe snot support RFC1933/2893 auto tunnels, KAME nodes are not vulnerable to these packets. ::0.0.0.0/104 ::127.0.0.0/104 ::224.0.0.0/100 ::255.0.0.0/104 Also, since KAME does not support RFC1933/2893 auto tunnels, seeing IPv4 compatible is very rare. You should take caution if you see those on the wire. If we see IPv6 packets with IPv4 mapped address (::ffff:0.0.0.0/96) in the header in dual-stack environment (not in SIIT environment), they indicate that someone is trying to impersonate IPv4 peer. The packet should be dropped. IPv6 specifications do not talk very much about IPv6 unspecified address (::) in the IPv6 source address field. Clarification is in progress. Here are couple of comments: - IPv6 unspecified address can be used in IPv6 source address field, if and only if we have no legal source address for the node. The legal situations include, but may not be limited to, (1) MLD while no IPv6 address is assigned to the node and (2) DAD. - If IPv6 TCP packet has IPv6 unspecified address, it is an attack attempt. The form can be used as a trigger for TCP DoS attack. KAME code already filters them out. - The following examples are seemingly illegal. It seems that there's general consensus among ipngwg for those. (1) Mobile IPv6 home address option, (2) offlink packets (so routers should not forward them). KAME implements (2) already. KAME code is carefully written to avoid such incidents. More specifically, KAME kernel will reject packets with certain source/destination address in IPv6 base header, or IPv6 routing header. Also, KAME default configuration file is written carefully, to avoid those attacks. draft-itojun-ipv6-transition-abuse-01.txt, draft-cmetz-v6ops-v4mapped-api- harmful-00.txt and draft-itojun-v6ops-v4mapped-harmful-01.txt has more on this issue. 1.15 Node's required addresses RFC2373 section 2.8 talks about required addresses for an IPv6 node. The section talks about how KAME stack manages those required addresses. 1.15.1 Host case The following items are automatically assigned to the node (or the node will automatically joins the group), at bootstrap time: - Loopback address - All-nodes multicast addresses (ff01::1) The following items will be automatically handled when the interface becomes IFF_UP: - Its link-local address for each interface - Solicited-node multicast address for link-local addresses - Link-local allnodes multicast address (ff02::1) The following items need to be configured manually by ifconfig(8) or prefix(8). Alternatively, these can be autoconfigured by using stateless address autoconfiguration. - Assigned unicast/anycast addresses - Solicited-Node multicast address for assigned unicast address Users can join groups by using appropriate system calls like setsockopt(2). 1.15.2 Router case In addition to the above, routers needs to handle the following items. The following items need to be configured manually by using ifconfig(8). o The subnet-router anycast addresses for the interfaces it is configured to act as a router on (prefix::/64) o All other anycast addresses with which the router has been configured The router will join the following multicast group when rtadvd(8) is available for the interface. o All-Routers Multicast Addresses (ff02::2) Routing daemons will join appropriate multicast groups, as necessary, like ff02::9 for RIPng. Users can join groups by using appropriate system calls like setsockopt(2). 1.16 Advanced API Current KAME kernel implements RFC3542 API. It also implements RFC2292 API, for backward compatibility purposes with *BSD-integrated codebase. KAME tree ships with RFC3542 headers. *BSD-integrated codebase implements either RFC2292, or RFC3542, API. see "COVERAGE" document for detailed implementation status. Here are couple of issues to mention: - *BSD-integrated binaries, compiled for RFC2292, will work on KAME kernel. For example, OpenBSD 2.7 /sbin/rtsol will work on KAME/openbsd kernel. - KAME binaries, compiled using RFC3542, will not work on *BSD-integrated kenrel. For example, KAME /usr/local/v6/sbin/rtsol will not work on OpenBSD 2.7 kernel. - RFC3542 API is not compatible with RFC2292 API. RFC3542 #define symbols conflict with RFC2292 symbols. Therefore, if you compile programs that assume RFC2292 API, the compilation itself goes fine, however, the compiled binary will not work correctly. The problem is not KAME issue, but API issue. For example, Solaris 8 implements RFC3542 API. If you compile RFC2292-based code on Solaris 8, the binary can behave strange. There are few (or couple of) incompatible behavior in RFC2292 binary backward compatibility support in KAME tree. To enumerate: - Type 0 routing header lacks support for strict/loose bitmap. Even if we see packets with "strict" bit set, those bits will not be made visible to the userland. Background: RFC2292 document is based on RFC1883 IPv6, and it uses strict/loose bitmap. RFC3542 document is based on RFC2460 IPv6, and it has no strict/loose bitmap (it was removed from RFC2460). KAME tree obeys RFC2460 IPv6, and lacks support for strict/loose bitmap. The RFC3542 documents leave some particular cases unspecified. The KAME implementation treats them as follows: - The IPV6_DONTFRAG and IPV6_RECVPATHMTU socket options for TCP sockets are ignored. That is, the setsocktopt() call will succeed but the specified value will have no effect. 1.17 DNS resolver KAME ships with modified DNS resolver, in libinet6.a. libinet6.a has a couple of extensions against libc DNS resolver: - Can take "options insecure1" and "options insecure2" in /etc/resolv.conf, which toggles RES_INSECURE[12] option flag bit. - EDNS0 receive buffer size notification support. It can be enabled by "options edns0" in /etc/resolv.conf. See USAGE for details. - IPv6 transport support (queries/responses over IPv6). Most of BSD official releases now has it already. - Partial A6 chain chasing/DNAME/bit string label support (KAME/BSDI4). 2. Network Drivers KAME requires three items to be added into the standard drivers: (1) (freebsd[234] and bsdi[34] only) mbuf clustering requirement. In this stable release, we changed MINCLSIZE into MHLEN+1 for all the operating systems in order to make all the drivers behave as we expect. (2) multicast. If "ifmcstat" yields no multicast group for a interface, that interface has to be patched. To avoid troubles, we suggest you to comment out the device drivers for unsupported/unnecessary cards, from the kernel configuration file. If you accidentally enable unsupported drivers, some of the userland tools may not work correctly (routing daemons are typical example). In the following sections, "official support" means that KAME developers are using that ethernet card/driver frequently. (NOTE: In the past we required all pcmcia drivers to have a call to in6_ifattach(). We have no such requirement any more) 2.1 FreeBSD 2.2.x-RELEASE Here is a list of FreeBSD 2.2.x-RELEASE drivers and its conditions: driver mbuf(1) multicast(2) official support? --- --- --- --- (Ethernet) ar looks ok - - cnw ok ok yes (*) ed ok ok yes ep ok ok yes fe ok ok yes sn looks ok - - (*) vx looks ok - - wlp ok ok - (*) xl ok ok yes zp ok ok - (FDDI) fpa looks ok ? - (ATM) en ok ok yes (Serial) lp ? - not work sl ? - not work sr looks ok ok - (**) You may want to add an invocation of "rtsol" in "/etc/pccard_ether", if you are using notebook computers and PCMCIA ethernet card. (*) These drivers are distributed with PAO (http://www.jp.freebsd.org/PAO/). (**) There was some report says that, if you make sr driver up and down and then up, the kernel may hang up. We have disabled frame-relay support from sr driver and after that this looks to be working fine. If you need frame-relay support to come back, please contact KAME developers. 2.2 BSD/OS 3.x The following lists BSD/OS 3.x device drivers and its conditions: driver mbuf(1) multicast(2) official support? --- --- --- --- (Ethernet) cnw ok ok yes de ok ok - df ok ok - eb ok ok - ef ok ok yes exp ok ok - mz ok ok yes ne ok ok yes we ok ok - (FDDI) fpa ok ok - (ATM) en maybe ok - (Serial) ntwo ok ok yes sl ? - not work appp ? - not work You may want to use "@insert" directive in /etc/pccard.conf to invoke "rtsol" command right after dynamic insertion of PCMCIA ethernet cards. 2.3 NetBSD The following table lists the network drivers we have tried so far. driver mbuf(1) multicast(2) official support? --- --- --- --- (Ethernet) awi pcmcia/i386 ok ok - bah zbus/amiga NG(*) cnw pcmcia/i386 ok ok yes ep pcmcia/i386 ok ok - fxp pci/i386 ok(*2) ok - tlp pci/i386 ok ok - le sbus/sparc ok ok yes ne pci/i386 ok ok yes ne pcmcia/i386 ok ok yes rtk pci/i386 ok ok - wi pcmcia/i386 ok ok yes (ATM) en pci/i386 ok ok - (*) This may need some fix, but I'm not sure what arcnet interfaces assume... 2.4 FreeBSD 3.x-RELEASE Here is a list of FreeBSD 3.x-RELEASE drivers and its conditions: driver mbuf(1) multicast(2) official support? --- --- --- --- (Ethernet) cnw ok ok -(*) ed ? ok - ep ok ok - fe ok ok yes fxp ?(**) lnc ? ok - sn ? ? -(*) wi ok ok yes xl ? ok - (*) These drivers are distributed with PAO as PAO3 (http://www.jp.freebsd.org/PAO/). (**) there were trouble reports with multicast filter initialization. More drivers will just simply work on KAME FreeBSD 3.x-RELEASE but have not been checked yet. 2.5 FreeBSD 4.x-RELEASE Here is a list of FreeBSD 4.x-RELEASE drivers and its conditions: driver multicast --- --- (Ethernet) lnc/vmware ok 2.6 OpenBSD 2.x Here is a list of OpenBSD 2.x drivers and its conditions: driver mbuf(1) multicast(2) official support? --- --- --- --- (Ethernet) de pci/i386 ok ok yes fxp pci/i386 ?(*) le sbus/sparc ok ok yes ne pci/i386 ok ok yes ne pcmcia/i386 ok ok yes wi pcmcia/i386 ok ok yes (*) There seem to be some problem in driver, with multicast filter configuration. This happens with certain revision of chipset on the card. Should be fixed by now by workaround in sys/net/if.c, but still not sure. 2.7 BSD/OS 4.x The following lists BSD/OS 4.x device drivers and its conditions: driver mbuf(1) multicast(2) official support? --- --- --- --- (Ethernet) de ok ok yes exp (*) You may want to use "@insert" directive in /etc/pccard.conf to invoke "rtsol" command right after dynamic insertion of PCMCIA ethernet cards. (*) exp driver has serious conflict with KAME initialization sequence. A workaround is committed into sys/i386/pci/if_exp.c, and should be okay by now. 3. Translator We categorize IPv4/IPv6 translator into 4 types. Translator A --- It is used in the early stage of transition to make it possible to establish a connection from an IPv6 host in an IPv6 island to an IPv4 host in the IPv4 ocean. Translator B --- It is used in the early stage of transition to make it possible to establish a connection from an IPv4 host in the IPv4 ocean to an IPv6 host in an IPv6 island. Translator C --- It is used in the late stage of transition to make it possible to establish a connection from an IPv4 host in an IPv4 island to an IPv6 host in the IPv6 ocean. Translator D --- It is used in the late stage of transition to make it possible to establish a connection from an IPv6 host in the IPv6 ocean to an IPv4 host in an IPv4 island. KAME provides an TCP relay translator for category A. This is called "FAITH". We also provide IP header translator for category A. 3.1 FAITH TCP relay translator FAITH system uses TCP relay daemon called "faithd" helped by the KAME kernel. FAITH will reserve an IPv6 address prefix, and relay TCP connection toward that prefix to IPv4 destination. For example, if the reserved IPv6 prefix is 3ffe:0501:0200:ffff::, and the IPv6 destination for TCP connection is 3ffe:0501:0200:ffff::163.221.202.12, the connection will be relayed toward IPv4 destination 163.221.202.12. destination IPv4 node (163.221.202.12) ^ | IPv4 tcp toward 163.221.202.12 FAITH-relay dual stack node ^ | IPv6 TCP toward 3ffe:0501:0200:ffff::163.221.202.12 source IPv6 node faithd must be invoked on FAITH-relay dual stack node. For more details, consult kame/kame/faithd/README and RFC3142. 3.2 IPv6-to-IPv4 header translator (to be written) 4. IPsec IPsec is implemented as the following three components. (1) Policy Management (2) Key Management (3) AH, ESP and IPComp handling in kernel Note that KAME/OpenBSD does NOT include support for KAME IPsec code, as OpenBSD team has their home-brew IPsec stack and they have no plan to replace it. IPv6 support for IPsec is, therefore, lacking on KAME/OpenBSD. http://www.netbsd.org/Documentation/network/ipsec/ has more information including usage examples. 4.1 Policy Management The kernel implements experimental policy management code. There are two ways to manage security policy. One is to configure per-socket policy using setsockopt(3). In this cases, policy configuration is described in ipsec_set_policy(3). The other is to configure kernel packet filter-based policy using PF_KEY interface, via setkey(8). The policy entry will be matched in order. The order of entries makes difference in behavior. 4.2 Key Management The key management code implemented in this kit (sys/netkey) is a home-brew PFKEY v2 implementation. This conforms to RFC2367. The home-brew IKE daemon, "racoon" is included in the kit (kame/kame/racoon, or usr.sbin/racoon). Basically you'll need to run racoon as daemon, then setup a policy to require keys (like ping -P 'out ipsec esp/transport//use'). The kernel will contact racoon daemon as necessary to exchange keys. In IKE spec, there's ambiguity about interpretation of "tunnel" proposal. For example, if we would like to propose the use of following packet: IP AH ESP IP payload some implementation proposes it as "AH transport and ESP tunnel", since this is more logical from packet construction point of view. Some implementation proposes it as "AH tunnel and ESP tunnel". Racoon follows the latter route (previously it followed the former, and the latter interpretation seems to be popular/consensus). This raises real interoperability issue. We hope this to be resolved quickly. racoon does not implement byte lifetime for both phase 1 and phase 2 (RFC2409 page 35, Life Type = kilobytes). 4.3 AH and ESP handling IPsec module is implemented as "hooks" to the standard IPv4/IPv6 processing. When sending a packet, ip{,6}_output() checks if ESP/AH processing is required by checking if a matching SPD (Security Policy Database) is found. If ESP/AH is needed, {esp,ah}{4,6}_output() will be called and mbuf will be updated accordingly. When a packet is received, {esp,ah}4_input() will be called based on protocol number, i.e. (*inetsw[proto])(). {esp,ah}4_input() will decrypt/check authenticity of the packet, and strips off daisy-chained header and padding for ESP/AH. It is safe to strip off the ESP/AH header on packet reception, since we will never use the received packet in "as is" form. By using ESP/AH, TCP4/6 effective data segment size will be affected by extra daisy-chained headers inserted by ESP/AH. Our code takes care of the case. Basic crypto functions can be found in directory "sys/crypto". ESP/AH transform are listed in {esp,ah}_core.c with wrapper functions. If you wish to add some algorithm, add wrapper function in {esp,ah}_core.c, and add your crypto algorithm code into sys/crypto. Tunnel mode works basically fine, but comes with the following restrictions: - You cannot run routing daemon across IPsec tunnel, since we do not model IPsec tunnel as pseudo interfaces. - Authentication model for AH tunnel must be revisited. We'll need to improve the policy management engine, eventually. - Path MTU discovery does not work across IPv6 IPsec tunnel gateway due to insufficient code. AH specification does not talk much about "multiple AH on a packet" case. We incrementally compute AH checksum, from inside to outside. Also, we treat inner AH to be immutable. For example, if we are to create the following packet: IP AH1 AH2 AH3 payload we do it incrementally. As a result, we get crypto checksums like below: AH3 has checksum against "IP AH3' payload". where AH3' = AH3 with checksum field filled with 0. AH2 has checksum against "IP AH2' AH3 payload". AH1 has checksum against "IP AH1' AH2 AH3 payload", Also note that AH3 has the smallest sequence number, and AH1 has the largest sequence number. To avoid traffic analysis on shorter packets, ESP output logic supports random length padding. By setting net.inet.ipsec.esp_randpad (or net.inet6.ipsec6.esp_randpad) to positive value N, you can ask the kernel to randomly pad packets shorter than N bytes, to random length smaller than or equal to N. Note that N does not include ESP authentication data length. Also note that the random padding is not included in TCP segment size computation. Negative value will turn off the functionality. Recommended value for N is like 128, or 256. If you use a too big number as N, you may experience inefficiency due to fragmented packets. 4.4 IPComp handling IPComp stands for IP payload compression protocol. This is aimed for payload compression, not the header compression like PPP VJ compression. This may be useful when you are using slow serial link (say, cell phone) with powerful CPU (well, recent notebook PCs are really powerful...). The protocol design of IPComp is very similar to IPsec, though it was defined separately from IPsec itself. Here are some points to be noted: - IPComp is treated as part of IPsec protocol suite, and SPI and CPI space is unified. Spec says that there's no relationship between two so they are assumed to be separate in specs. - IPComp association (IPCA) is kept in SAD. - It is possible to use well-known CPI (CPI=2 for DEFLATE for example), for outbound/inbound packet, but for indexing purposes one element from SPI/CPI space will be occupied anyway. - pfkey is modified to support IPComp. However, there's no official SA type number assignment yet. Portability with other IPComp stack is questionable (anyway, who else implement IPComp on UN*X?). - Spec says that IPComp output processing must be performed before AH/ESP output processing, to achieve better compression ratio and "stir" data stream before encryption. The most meaningful processing order is: (1) compress payload by IPComp, (2) encrypt payload by ESP, then (3) attach authentication data by AH. However, with manual SPD setting, you are able to violate the ordering (KAME code is too generic, maybe). Also, it is just okay to use IPComp alone, without AH/ESP. - Though the packet size can be significantly decreased by using IPComp, no special consideration is made about path MTU (spec talks nothing about MTU consideration). IPComp is designed for serial links, not ethernet-like medium, it seems. - You can change compression ratio on outbound packet, by changing deflate_policy in sys/netinet6/ipcomp_core.c. You can also change outbound history buffer size by changing deflate_window_out in the same source code. (should it be sysctl accessible, or per-SAD configurable?) - Tunnel mode IPComp is not working right. KAME box can generate tunnelled IPComp packet, however, cannot accept tunneled IPComp packet. - You can negotiate IPComp association with racoon IKE daemon. - KAME code does not attach Adler32 checksum to compressed data. see ipsec wg mailing list discussion in Jan 2000 for details. 4.5 Conformance to RFCs and IDs The IPsec code in the kernel conforms (or, tries to conform) to the following standards: "old IPsec" specification documented in rfc182[5-9].txt "new IPsec" specification documented in: rfc240[1-6].txt rfc241[01].txt rfc2451.txt rfc3602.txt IPComp: RFC2393: IP Payload Compression Protocol (IPComp) IKE specifications (rfc240[7-9].txt) are implemented in userland as "racoon" IKE daemon. Currently supported algorithms are: old IPsec AH null crypto checksum (no document, just for debugging) keyed MD5 with 128bit crypto checksum (rfc1828.txt) keyed SHA1 with 128bit crypto checksum (no document) HMAC MD5 with 128bit crypto checksum (rfc2085.txt) HMAC SHA1 with 128bit crypto checksum (no document) HMAC RIPEMD160 with 128bit crypto checksum (no document) old IPsec ESP null encryption (no document, similar to rfc2410.txt) DES-CBC mode (rfc1829.txt) new IPsec AH null crypto checksum (no document, just for debugging) keyed MD5 with 96bit crypto checksum (no document) keyed SHA1 with 96bit crypto checksum (no document) HMAC MD5 with 96bit crypto checksum (rfc2403.txt HMAC SHA1 with 96bit crypto checksum (rfc2404.txt) HMAC SHA2-256 with 96bit crypto checksum (draft-ietf-ipsec-ciph-sha-256-00.txt) HMAC SHA2-384 with 96bit crypto checksum (no document) HMAC SHA2-512 with 96bit crypto checksum (no document) HMAC RIPEMD160 with 96bit crypto checksum (RFC2857) AES XCBC MAC with 96bit crypto checksum (RFC3566) new IPsec ESP null encryption (rfc2410.txt) DES-CBC with derived IV (draft-ietf-ipsec-ciph-des-derived-01.txt, draft expired) DES-CBC with explicit IV (rfc2405.txt) 3DES-CBC with explicit IV (rfc2451.txt) BLOWFISH CBC (rfc2451.txt) CAST128 CBC (rfc2451.txt) RIJNDAEL/AES CBC (rfc3602.txt) AES counter mode (rfc3686.txt) each of the above can be combined with new IPsec AH schemes for ESP authentication. IPComp RFC2394: IP Payload Compression Using DEFLATE The following algorithms are NOT supported: old IPsec AH HMAC MD5 with 128bit crypto checksum + 64bit replay prevention (rfc2085.txt) keyed SHA1 with 160bit crypto checksum + 32bit padding (rfc1852.txt) The key/policy management API is based on the following document, with fair amount of extensions: RFC2367: PF_KEY key management API 4.6 ECN consideration on IPsec tunnels KAME IPsec implements ECN-friendly IPsec tunnel, described in draft-ietf-ipsec-ecn-02.txt. Normal IPsec tunnel is described in RFC2401. On encapsulation, IPv4 TOS field (or, IPv6 traffic class field) will be copied from inner IP header to outer IP header. On decapsulation outer IP header will be simply dropped. The decapsulation rule is not compatible with ECN, since ECN bit on the outer IP TOS/traffic class field will be lost. To make IPsec tunnel ECN-friendly, we should modify encapsulation and decapsulation procedure. This is described in draft-ietf-ipsec-ecn-02.txt, chapter 3.3. KAME IPsec tunnel implementation can give you three behaviors, by setting net.inet.ipsec.ecn (or net.inet6.ipsec6.ecn) to some value: - RFC2401: no consideration for ECN (sysctl value -1) - ECN forbidden (sysctl value 0) - ECN allowed (sysctl value 1) Note that the behavior is configurable in per-node manner, not per-SA manner (draft-ietf-ipsec-ecn-02 wants per-SA configuration, but it looks too much for me). The behavior is summarized as follows (see source code for more detail): encapsulate decapsulate --- --- RFC2401 copy all TOS bits drop TOS bits on outer from inner to outer. (use inner TOS bits as is) ECN forbidden copy TOS bits except for ECN drop TOS bits on outer (masked with 0xfc) from inner (use inner TOS bits as is) to outer. set ECN bits to 0. ECN allowed copy TOS bits except for ECN use inner TOS bits with some CE (masked with 0xfe) from change. if outer ECN CE bit inner to outer. is 1, enable ECN CE bit on set ECN CE bit to 0. the inner. General strategy for configuration is as follows: - if both IPsec tunnel endpoint are capable of ECN-friendly behavior, you'd better configure both end to "ECN allowed" (sysctl value 1). - if the other end is very strict about TOS bit, use "RFC2401" (sysctl value -1). - in other cases, use "ECN forbidden" (sysctl value 0). The default behavior is "ECN forbidden" (sysctl value 0). For more information, please refer to: draft-ietf-ipsec-ecn-02.txt RFC2481 (Explicit Congestion Notification) KAME sys/netinet6/{ah,esp}_input.c (Thanks goes to Kenjiro Cho for detailed analysis) 4.7 Interoperability IPsec, IPComp (in kernel) and IKE (in userland as "racoon") has been tested at several interoperability test events, and it is known to interoperate with many other implementations well. Also, KAME IPsec has quite wide coverage for IPsec crypto algorithms documented in RFC (we do not cover algorithms with intellectual property issues, though). Here are (some of) platforms we have tested IPsec/IKE interoperability in the past, no particular order. Note that both ends (KAME and others) may have modified their implementation, so use the following list just for reference purposes. 6WIND, ACC, Allied-telesis, Altiga, Ashley-laurent (vpcom.com), BlueSteel, CISCO IOS, Checkpoint FW-1, Compaq Tru54 UNIX X5.1B-BL4, Cryptek, Data Fellows (F-Secure), Ericsson, F-Secure VPN+ 5.40, Fitec, Fitel, FreeS/WAN, HITACHI, HiFn, IBM AIX 5.1, III, IIJ (fujie stack), Intel Canada, Intel Packet Protect, MEW NetCocoon, MGCS, Microsoft WinNT/2000/XP, NAI PGPnet, NEC IX5000, NIST (linux IPsec + plutoplus), NetLock, Netoctave, Netopia, Netscreen, Nokia EPOC, Nortel GatewayController/CallServer 2000 (not released yet), NxNetworks, OpenBSD isakmpd on OpenBSD, Oullim information technologies SECUREWORKS VPN gateway 3.0, Pivotal, RSA, Radguard, RapidStream, RedCreek, Routerware, SSH, SecGo CryptoIP v3, Secure Computing, Soliton, Sun Solaris 8, TIS/NAI Gauntret, Toshiba, Trilogy AdmitOne 2.6, Trustworks TrustedClient v3.2, USAGI linux, VPNet, Yamaha RT series, ZyXEL Here are (some of) platforms we have tested IPComp/IKE interoperability in the past, in no particular order. Compaq, IRE, SSH, NetLock, FreeS/WAN, F-Secure VPN+ 5.40 VPNC (vpnc.org) provides IPsec conformance tests, using KAME and OpenBSD IPsec/IKE implementations. Their test results are available at http://www.vpnc.org/conformance.html, and it may give you more idea about which implementation interoperates with KAME IPsec/IKE implementation. 4.8 Operations with IPsec tunnel mode First of all, IPsec tunnel is a very hairy thing. It seems to do a neat thing like VPN configuration or secure remote accesses, however, it comes with lots of architectural twists. RFC2401 defines IPsec tunnel mode, within the context of IPsec. RFC2401 defines tunnel mode packet encapsulation/decapsulation on its own, and does not refer other tunnelling specifications. Since RFC2401 advocates filter-based SPD database matches, it would be natural for us to implement IPsec tunnel mode as filters - not as pseudo interfaces. There are some people who are trying to separate IPsec "tunnel mode" from the IPsec itself. They would like to implement IPsec transport mode only, and combine it with tunneling pseudo devices. The prime example is found in draft-touch-ipsec-vpn-01.txt. However, if you really define pseudo interfaces separately from IPsec, IKE daemons would need to negotiate transport mode SAs, instead of tunnel mode SAs. Therefore, we cannot really mix RFC2401-based interpretation and draft-touch-ipsec-vpn-01.txt interpretation. The KAME stack implements can be configured in two ways. You may need to recompile your kernel to switch the behavior. - RFC2401 IPsec tunnel mode approach (4.8.1) - draft-touch-ipsec-vpn approach (4.8.2) Works in all kernel configuration, but racoon(8) may not interoperate. There are pros and cons on these approaches: RFC2401 IPsec tunnel mode (filter-like) approach PRO: SPD lookup fits nicely with packet filters (if you integrate them) CON: cannot run routing daemons across IPsec tunnels CON: it is very hard to control source address selection on originating cases ???: IPv6 scope zone is kept the same draft-touch-ipsec-vpn (transportmode + Pseudo-interface) approach PRO: run routing daemons across IPsec tunnels PRO: source address selection can be done normally, by looking at IPsec tunnel pseudo devices CON: on outbound, possibility of infinite loops if routing setup is wrong CON: due to differences in encap/decap logic from RFC2401, it may not interoperate with very picky RFC2401 implementations (those who check TOS bits, for example) CON: cannot negotiate IKE with other IPsec tunnel-mode devices (the other end has to implement ???: IPv6 scope zone is likely to be different from the real ethernet interface The recommendation is different depending on the situation you have: - use draft-touch-ipsec-vpn if you have the control over the other end. this one is the best in terms of simplicity. - if the other end is normal IPsec device with RFC2401 implementation, you need to use RFC2401, otherwise you won't be able to run IKE. - use RFC2401 approach if you just want to forward packets back and forth and there's no plan to use IPsec gateway itself as an originating device. 4.8.1 RFC2401 IPsec tunnel mode approach To configure your device as RFC2401 IPsec tunnel mode endpoint, you will use "tunnel" keyword in setkey(8) "spdadd" directives. Let us assume the following topology (A and B could be a network, like prefix/length): ((((((((((((The internet)))))))))))) | | |C (global) |D your device peer's device |A (private) |B ==+===== VPN net ==+===== VPN net The policy configuration directive is like this. You will need manual SAs, or IKE daemon, for actual encryption: # setkey -c < B] payload and will not match the policy (= sent in clear). - When you want to run routing protocols on top of the IPsec tunnel, it is not possible. As there is no pseudo device that identifies the IPsec tunnel, you cannot identify where the routing information came from. As a result, you can't run routing daemons. 4.8.2 draft-touch-ipsec-vpn approach With this approach, you will configure gif(4) tunnel interfaces, as well as IPsec transport mode SAs. # gifconfig gif0 C D # ifconfig gif0 A B # setkey -c <