Fetching contributors…
Cannot retrieve contributors at this time
3314 lines (2775 sloc) 144 KB
$KAME: CHANGELOG.2001,v 1.3 2002/06/12 11:58:32 jinmei Exp $
2001-12-28 JINMEI, Tatuya <>
* kame/sys/netinet6/ip6_input.c (ip6_savecontrol): do not separate
the case for IPV6_RTHDRDSTOPTS and the case for IPV6_DSTOPTS,
according to rfc2292bis-03.
* kame/sys/netinet6/ip6_var.h: accordingly replaced the dest1 and
dest2 members of ip6_recvpktopts{} with a single "dest" member.
2001-12-27 JINMEI, Tatuya <>
* *bsd/usr.sbin/ip6addrctl: a dedicated directory was digged in
order for the tool to be compiled as a standard package.
* kame/kame/ip6addrctl: renamed from kame/kame/addrselect
according to the change of the command name.
Thu Dec 27 17:09:51 JST 2001
* dhcp6: use draft 22 Inform/Reply exchange as suggested in
draft-droms-dnsconfig-dhcpv6-00.txt. no backward compatibility
Thu Dec 27 11:42:38 JST 2001
* kame/sys/netinet6/mip6*
- implemented authdata sub-optoin. this implementation uses
KAME/IPSEC spd/sa database. because of this reason, OpenBSD can't
use authdata sub-option. the spd and sa can be set using
setkey program. the protocol number is currently ipv6-opts(60).
it needs some discussion how to set spd/sa for authdata/mip6.
currently, i do not provide a switch to disable authdata
feature. the users must specify spd and sa between the mn and
ha/cn. otherwise, the home registration/bu operation will fail.
take a look into to understand how to set up
spd/sa for authdata sub-option.
2001-12-25 JINMEI, Tatuya <>
* kame/kame/libinet6/ip6opt.c (inet6_opt_next, inet6_opt_find):
- corrected the calculation of the "previous length" returned by
these functions. The returned length should contain the length of
the returned option.
- corrected the option length value returned by these functions so
that the length would only contain the data length (i.e. without
the type and length fields)
2001-12-22 JINMEI, Tatuya <>
* kame/sys/netinet6/in6_src.c (in6_selectroute): corrected a
condition to get a route to the next hop specified by the
IPV6_NEXTHOP option. Also, a statistics counter was incremented
in this function.
2001-12-21 JINMEI, Tatuya <>
* kame/kame/addrselect: renamed the program to ip6addrctl so that
the name would describe its function well.
2001-12-21 JINMEI, Tatuya <>
* kame/sys/netinet6/ip6_output.c (ip6_raw_ctloutput): added as a new
function ip6_raw_ctloutput, which only handles IPV6_CHECKSUM. The
purpose for the separate function is to share the logic by all
A new validation rule to reject odd offset values was added,
according to the latest specification of the advanced API.
2001-12-21 JINMEI, Tatuya <>
* kame/sys/netinet6/ip6_output.c (ip6_ctloutput): supported a new
IPv6 socket option IPV6_PATHMTU defined in rfc2292bis-03.
Fri Dec 21 14:38:15 JST 2001
* sys/netinet6/in6_gif.c: correct default hoplimit value for
IPvX-over-IPv6 tunnel (has been set to 0 due to #include mistake).
2001-12-20 JINMEI, Tatuya <>
* kame/sys/netinet6/ip6_output.c: supported a new IPv6 socket
option IPV6_DONTFRAG as described in rfc2292bis-03. This option
is used to disable IPv6-layer fragmentation.
Thu Dec 20 06:48:57 JST 2001
* kame/racoon:
fixed that the negotiation always failed when "verify_identifier"
was defined with aggressive mode.
2001-12-19 JINMEI, Tatuya <>
* kame/kame/libinet6/getaddrinfo.c (get_addrselectpolicy): enabled
for FreeBSD.
2001-12-19 JINMEI, Tatuya <>
* kame/kame/addrselect/addrselect.c: used ioctl()s just added below
to set/delete policy entries.
2001-12-19 JINMEI, Tatuya <>
* kame/sys/netinet6/{in6.c, in6_src.c}:
- implemented two new ioctls to set/delete address selection
policy entry
- disabled write operation of net.inet6.ip6.addrselpolicy
with these changes, address selection policy can now be installed
on all supported platforms.
2001-12-18 JINMEI, Tatuya <>
* kame/{kame,sys}/(many files): replaced mld6_xxx and MLD6_XXX
with mld_xxx and MLD_XXX according to the change below.
2001-12-18 JINMEI, Tatuya <>
* kame/sys/netinet/icmp6.h: revised MLD-related definitions:
- used mld_xxx and MLD_xxx instead of mld6_xxx and MLD6_xxx
according to the official defintions in rfc2292bis
- changed the first member of mld_hdr{} from mld_hdr to
mld_icmp6_hdr to avoid name space conflict in C++
Sat Dec 15 01:45:38 JST 2001
* kame/racoon:
Suported the 6144-bit DH MODP group. Fixed some memory leaks.
2001-12-14 JINMEI, Tatuya <>
* kame/sys/netinet6/in6.c (in6_control): always called
pfxlist_onlink_check() after the process of SIOCAIFADDR_IN6. The
function call was mistakenly put in a wrong block with another
Based on: a report from <>
Fri Dec 14 17:19:56 JST 2001
* kame/sys/netinet6/ip6_mroute.c
- s/nexpire/n6expire/, since this global variable is duplicated.
(ip_mroute.c also has a global variable "nexpire" on *BSD
except FreeBSD)
2001-12-13 Shin'ichi Fujisawa <>
* kame/sys/natpt_{defs,dispatch}.[ch]:
- Add a code to send ICMP "packet too big" message to IPv4
host when IPv4 packet is too big (at most 1232 octets,
excluding the IPv4 header) to translate into IPv6. 1232
comes from 1280 (IPv6 minimum MTU) minus 40 for the IPv6
header and 8 for the Fragment header.
A problem of the host who does not listen to ICMP is still
Thu Dec 13 00:31:47 JST 2001
* kame/racoon:
the retransmission logic has been improved. racoon stores packets
that have been sent. when racoon receives a packet that has been
processed already, racoon will reply the saved packet simply.
the default timer for waiting each negotiation has also been changed.
2001-12-12 Shin'ichi Fujisawa <>
* kame/sys/netinet6/natpt_{dispatch,tslot,var}.[ch]:
- Support translation from ICMPv6 that occurred with IPv6 host
into ICMPv4 in "IPv4->IPv6" translation.
This translation is done even if there is no "IPv6->IPv4"
translation rule, but only this ICMPv6 packet is a target of
translation. Because translator examines whether this is
ICMP of return more first than translation rule, eventually
"IPv6->IPv4" translation rule is not adapted to.
2001-12-12 JINMEI, Tatuya <>
* kame/kame/libinet6/getaddrinfo.c: implemented more rules for
destination address ordering; matching label, higher precedence,
and longest prefix matching. Due to differences on sysctl, the
first two rules are not available for FreeBSD at this moment.
2001-12-11 Shin'ichi Fujisawa <>
* kame/sys/netinet6/natpt_{defs,dispatch,trans,tslot,var}.[ch]:
- Support fragmented IPv4 packet translation.
Fragmented IPv4 packet is translated into fragmented one or
two IPv6 packet. IPv4 packet larger than 1280 bytes
(monolithic or fragmented) is translated into fragmented
two IPv6 pakcets. Sorry, DF flag is not cared yet.
- Reassembly is not supported.
The following limitation occurs from this restriction.
o Fragmented ICMPv4 packet is not translated into ICMPv6
packet. Because length of ICMPv6 packet is necessary
when translator calculates ICMPv6 checksum, but the
length is not provided without reassembly.
o IPv4 packet which is fragmented and needs conversion
of payload is not translated into IPv6 packet.
i.e. ftp command.
I think there is hardly a case that FTP command or ICMP
packet is fragmented except ICMP ECHO/ECHOREPLY message.
Therefor, I think this restriction is not so serious.
- Mind DF flag.
- Send back ICMP "packet too big" message to sender when
DF flag is set and IPv4 packet size is larger than 1280
There is a host that sends a IPv4 packet which DF bit is
set but does not listen to an ICMP message. Or an ICMP
"packet too big" message may be dropped at a firewall
prior to this host.
This problem is still under consideration now.
Tue Dec 4 20:58:54 JST 2001
* kame/sys/net/if_hif.[hc]
* kame/sys/netinet6/{mip6.[hc],mip6_binding.c}
* kame/sys/netinet6/nd6_rtr.c
- do movement detection in pfxlist_onlink_check() instead of
nd6_ra_input(). this change makes it possible to utilize
the change of the status of (detached) prefixes to detect
2001-12-04 JINMEI, Tatuya <>
* bsdi4/contrib/bind/src/lib/irs/dns_ho.c (ho_byaddr):
* bsdi4/contrib/bind/src/lib/resolv/res_init.c (res_setoptions):
use then as the upper domain for IPv6 reverse
loookups (with the nibble format), unless the "no-ip6-arpa"
resolver option is specified.
Tue Dec 4 10:14:02 JST 2001
* libinet6/name6.c: on IPv6 reverse name lookup, query
then (both using nibble format).
2001-12-03 JINMEI, Tatuya <>
* netbsd/sys/netinet/tcp_subr.c (ipsec4_hdrsiz_tcp): corrected
IP+TCP headers to calculate the header size.
This fix is very important, because without the fix the kernel
would bypass the path acquiring SAs even if the policy requires
All KAME netbsd users who use IPsec for TCP/IPv4 should apply this
Mon Dec 3 21:24:25 JST 2001
* kame/sys/net/if_hif.c, kame/sys/netinet6/mip6*
- change the movement detection algorithm from prefix infomation
based to coa based. this is more stateless addrconf friendly.
- fix incorrect ifa_rtrequest setting in hif_ioctl.
2001/11/30 20:51:35 JST
* ALTQ works again for openbsd-3.0.
- the ALTQ base is enabled in "sys/conf/GENERIC" but other ALTQ
options are enabled in "openbsd/sys/arch/i386/conf/GENERIC.KAME".
Thu Nov 29 18:32:39 JST 2001
* openbsd: switched to openbsd 3.0. still needs a torture-testing.
ALTQ is unavailable for the moment.
2001-11-28 Shin'ichi Fujisawa <>
* kame/sys/netinet6/natpt_{defs,dispatch,trans,tslot,var}.[ch]:
- Support translation from fragmented IPv6 packet to IPv4.
IPv4 packet is also fragmented.
Restriction: The first fragmented packet must have all of
header of ICMP/TCP/UDP, because this translation does not
support assembly of fragmented packets.
Tue Nov 27 19:03:30 JST 2001
* kame/sys/crypto/cast128,kame/sys/netinet6/esp_core.c:
fixed the cast128 calculation with a short cipher key length.
the memory was overridden when the key length was less than 16 bytes.
Tue Nov 27 16:03:20 JST 2001
* kame/kame/pim6sd/:
- SSM prefix definition is now based on draft-ietf-ipngwg-uni-
- joins all MLDv2 routers group on all interfaces listening to MLDv2
- updated MLDv2-related code (from Mickael Hoerdt)
- never uses timer with id == 0, since it means "unspecified"
- frees/discards/resets everything with regard to the stopping VIF.
Tue Nov 27 01:35:26 JST 2001
* kame/kame/racoon:
- when racoon receives a IKE packet, it checks the value of the
responder's cookie strictly.
- fixed to recover the state when a error happens during the phase1
negotiation. the return value was overwritten a illegal value.
- added a debugging code to find memory leaks. two memory leaks
have been fixed.
2001-11-26 JINMEI, Tatuya <>
* kame/sys/netinet6/ip6_output.c(ip6_output): initialized the last
argument to ip6_process_hopopts(), because the callee requires it
to be 0 when a jumbo payload option is contained.
(We may have to reconsider the requirement, though.)
2001-11-25 JINMEI, Tatuya <>
* freebsd4/{altq*, hroute6d, ppp, pvc*, ifconfig}/Makefile:
linked KAME's libinet6 in order to make sure to use the latest
getaddrinfo() and getnameinfo(). Those applications may not work
well with libc due to latest changes of handling IPv6 scoped
addresses, particularly the lack of NI_WITHSCOPEID.
Wed Nov 21 14:17:10 JST 2001
* sys/netinet/ip_output.c: update outgoing interface only if ipsec
tunnel mode encapsulation is applied. the change is necessary to
use IPsec over multicast (like IP_MULTICAST_IF).
Mon Nov 19 14:34:06 JST 2001
* sys/netinet6/udp6_output.c: initialize IPv4 source address on packet
correctly, when IPv4 mapped address is used on AF_INET6 socket.
affects bsdi4 only.
Thu Nov 15 20:09:26 JST 2001
* kame/sys/netinet6/mip6*
- fix a binding request bug. now CN sends a binding request for
the mobile node of a expiring binding cache.
- add auth sub-option handling.
currently zero-length auth sub-option (means no auth) is
sent/received with bu.
- stop using in6_control() for address handling. instead,
use in6_update_ifa/in6_purgeaddr.
2001-11-15 Shin'ichi Fujisawa <>
* kame/kame/natptconfig/showsubs.c:
- Invert the source and destination address of the remote side
of "natptconfig show xlate" command output.
Prior to this change, "natptconfig show xlate" command
indicates it in order having with kernel internal table.
However this output is against intuition (and does not fit
header line either).
2001-11-15 Shin'ichi Fujisawa <>
* kame/sys/netinet6/natpt_rule.c:
- Adapt a translation rule when packet is TCP and only SYN
flag is set. Until now, this NAT-PT starts translation
even if TCP packet with no SYN flag is coming. However,
this behavior seems no good, action was changed in this way.
2001-11-14 JINMEI, Tatuya <>
* kame/sys/netinet6/raw_ip6.c (rip6_usrreq):
* freebsd4/sys/netinet6/raw_ip6.c (rip6_send):
allow (with warning messages) an unspecified address family for
the destination address to provide backward compatibility to
buggy(!) applications.
We may change our mind later on this.
2001-11-13 JINMEI, Tatuya <>
* kame/kame/(many files): removed all references to
2001-11-13 JINMEI, Tatuya <>
* *bsd/lib/libinet6/Makefile: added scope.c to SRCS in order to
deal with scope zone IDs. The library functions defined in the
file are KAME specific, and not standard ones.
2001-11-13 JINMEI, Tatuya <>
* kame/kame/rtsold/rtsold.c (ifconfig): got a proper link ID for
each interface.
* kame/kame/rtsold/rtsol.c (sendpacket):
* kame/kame/rtsold/probe.c (sendprobe):
set the link ID for the outgoing link-local packet.
2001-11-13 JINMEI, Tatuya <>
* kame/kame/rtsold/rtsol.c (sockopen): initialize the family and
the length members of a sockaddr_in6 structure correctly.
2001-11-13 JINMEI, Tatuya <>
* kame/sys/netinet6/raw_ip6.c (rip6_usrreq):
* freebsd4/sys/netinet6/raw_ip6.c (rip6_send):
validated the length and the address family of the destination for
the send operation. This fix is correct per se, but breaks
backward compatibility to old rtsol(d) due to a bug of the
2001-11-13 JINMEI, Tatuya <>
* kame/sys/netinet6/(many files): changed the way of qualifying
the zone of address scope. The current code basically assumes the
application sets the scope zone correctly, and only uses the
default zone value as the last resort (if configured so).
However, a workaround was provided for backward compatibility; the
kernel also tries to disambiguate the scope zone after selecting
the source address and the outgoing interface.
2001-11-10 JINMEI, Tatuya <>
* kame/kame/ping6/ping6.c (main): used a pair of the
instead of the IPV6_HOPLIMIT sticky option, because the usage
of IPV6_HOPLIMIT might be disabled in a future version of
2001-11-10 JINMEI, Tatuya <>
* {kame, *BSD}/sys/netinet6/various files: introduced a new sysctl
variable "net.inet[6].ip6.use_defaultzone" to enable the notion of
default scope zone. When the variable is non-0, the kernel will
try to set an appropriate zone ID for an address given from an
application if its scope zone is ambiguous.
2001-11-10 JINMEI, Tatuya <>
* kame/sys/netinet6/ip6_output.c (ip6_setpktoption): disabled the
check for the address part of IPV6_PKTINFO, and let
in6_selectsrc() do the check.
* kame/sys/netinet6/in6_src.c (in6_selectsrc): validate the
address specified by IPV6_PKTINFO. This change will introduce
additional cost when the option is specified as a sticky option,
but this type of usage should be rare and can be ignored.
2001-11-10 JINMEI, Tatuya <>
* freebsd4/sys/netinet6/in6_pcb.c (in6_pcbladdr, in6_pcbbind):
set the default scope zone ID when unspecified.
* freebsd4/sys/netinet6/in6_pcb.c (in6_setsockaddr,
in6_setpeeraddr): use in6_recoverscope(), which is more generic.
Thu Nov 8 14:59:31 JST 2001
* bsdi4/sbin/ifconfig/ifconfig.c: support "tunnel" and "deletetunnel"
keyword for setting/removing gif physical address.
2001-11-08 Shin'ichi Fujisawa <>
* kame/kame/natptconfig/{cfparse,yylex}.[cy]:
- Keyword "proto" is necessary before protocol when you write
protocol dependant rule. It is written that "proto" is
necessary in man page, and "natptconfig show rules" command
also indicates "proto".
2001-11-07 JINMEI, Tatuya <>
* kame/sys/netinet6/in6.c (in6_update_ifa): when a user (perhaps
mistakenly) tried to install an IPv6 address with 0 valid
lifetime, just returned without any operation that had side
effects. There should be no difference on the behaviour from the
user's side, but the code should be safer than before because of
the lack of the effects.
Wed Nov 7 12:30:49 JST 2001
* kame/sys/netinet6/mip6_var.h,mip6_binding.c
- add ifnet info to the binding cache structure
to distinguish one bc on the home link from the other bcs
which has a same interface identifier on the other link.
(need to recompile mip6control)
Wed Nov 7 10:48:51 JST 2001
* kame/kame/racoon:
racoon will compare between the list of "sainfo" and the initiator's
proposal even when "generate_policy" is defined.
2001-11-06 JINMEI, Tatuya <>
* kame/sys/netinet6/nd6.c (nd6_purge): corrected purging the
default router list in nd6_purge(), so that both two loops would
surely start from the head of the list. Otherwise, we'd see some
garbage routers after purging the interface.
2001-11-06 JINMEI, Tatuya <>
* kame/sys/netinet6/in6.c (in6_update_ifa): set address lifetimes
before calling in6_ifinit(). The previous code should be correct,
but at least this one should not be wrong. Additionally, it
surely solves the situation that configuring an IPv6 address on
some NetBSD PCMCIA cards causes kernel panic.
Suggested by {karino, sugyo}@kame.
Tue Nov 6 17:03:39 JST 2001
* kame/sys/netinet6/ipsec.c:
fixed the behavior when there is no inbound policy for the ipsec
tunneled packet.
when there is no suitable inbound policy for the packet of the ipsec
tunnel mode, the kernel never decapsulate the tunneled packet
as the ipsec tunnel mode even when the system wide policy is "none".
then the kernel leaves the generic tunnel module to process this
packet. if there is no rule of the generic tunnel, the packet
is rejected and the statistics will be counted up.
Tue Nov 6 16:47:45 JST 2001
* kame/sys/netinet6/udp6_usrreq.c, netbsd/sys/netinet/udp_usrreq.c:
remove in6_mcmatch(). now sockets after bind(2) won't grab packets
to joined multicast groups (behavior now looks like 4.4BSD IPv4
2001-11-04 JINMEI, Tatuya <>
* kame/sys/netinet6/nd6_rtr.c (defrouter_select): do not override
a working router unless the new one is really better. "better"
means (1) the old router is not known to be reachable, or (2) the
new router has a really higher preference value than the old one.
2001-11-04 JINMEI, Tatuya <>
clarified cached route manipulation:
* kame/sys/netinet6/in6_src.c (in6_selectroute): check if the
cached route is up before using the cache. With this check, we
can select more appropriate source addresses.
* kame/sys/netinet6/ip6_output.c (ip6_output): removed the check
for the cached route to avoid redundancy.
Sun Nov 4 15:29:28 JST 2001
* kame/kame/v6test/cksum.c: calculates PIM checksum without
encapsulated packet if it is a Register message.
2001-11-02 JINMEI, Tatuya <>
* bsdi4/sys/netinet/in.c (in_control): fixed a bug that the kernel
can panic in process of the SIOCSIFADDR ioctl unless the
COMPAT_IFIOCTL kernel compilation option is specified.
2001-10-31 JINMEI, Tatuya <>
* kame/sys/netinet6/nd6_rtr.c (defrtrlist_update): corrected the
default router selection algorithm so that round-robin will surely
be performed when there's no reachable router.
Also, do not try to install more than one default router,
regardless of the availability of the multipath routing. This is
because I'm not sure about the benefits for stub hosts comparing
to the risk of making the code complicated and the possibility of
introducing bugs. (the previous code was actually buggy.)
The bug was found by recent TAHI tests, and was reported by Yukiyo
Akisada <>
2001-10-31 Shin'ichi Fujisawa <>
* kame/sys/netinet6/natpt_{rule,tslot}.c:
- Support IPv4 -> IPv6 translation.
Translation is restricted, one ipv4 host can connect to ipv6
host. This connection is one-on-one. So, if you want to
connect more than one IPv6 host, same number of IPv4 addresses
are necessary.
For example, assume you add following rule.
map from daddr
to daddr 2001:240:10a:5555:260:8ff:feb0:96f4
(I bend this line because line is long, but describe it in
single line)
IPv4 packet destined to this IPv4 address described just after
first "daddr" is a target of translation of IPv4->IPv6, and is
redirected to 2001:240:10a:5555:260:8ff:feb0:96f4, also
described just after second "daddr".
This IPv4 address ( in above example) had better
not to be NAT-PT box address. It will be easy to use ifconfig
command. For example,
ifconfig fxp0 inet netmask alias
Of course, change interface name, ip address and netmask in
accordance with your network.
Tue Oct 30 00:01:35 JST 2001
* kame/sys/netkey/key.c:
the system wide policy should be returned when no policy found
in the SPD. the packet was rejected in ipsec[46]_tunnel_validate().
Fri Oct 26 22:29:15 JST 2001
* kame/sys/netinet6/mip6.c
- add a binding request sending code.
2001-10-26 JINMEI, Tatuya <>
* freebsd4/usr.sbin/ppp/ipv6cp.c (ipcp_SetIPv6address): embedded the
link ID to the link-local addresses of both ends.
* freebsd4/usr.sbin/ppp/ncpaddr.c (adjust_linklocal): do not move
the embedded value to sin6_scope_id, since the kernel routing
engine does not understand this form.
* freebsd4/usr.sbin/ppp/ipv6cp.c (ipv6cp_Layer{Up,Down}): read
ppp.linkup and ppp.linkdown for IPv6. I'm not 100% sure if this
can be done without other fixes.
Fri Oct 26 18:27:29 JST 2001
* mip6
- changed the mip6 ioctl request structure. (this REQUIRES UPDATING
the mip6control command you are using.)
- protected the mip6 ioctl path by splnet() during retreiving
binding cache entries.
Fri Oct 26 10:06:23 JST 2001
* kame/kame/racoon:
fixed that the negotiation of the tunnel mode always failed.
2001-10-25 Shin'ichi Fujisawa <>
* kame/sys/netinet6/natpt_{defs,rule,transtslot}.[ch]:
- Support bi-directional IPv4 NAT.
This NAT binds global address to private address
one-on-one statically. Sessions can be initiated from
hosts in the public network as well as the private
network. So, if you can set nat session as bidirectional,
you can connect to the host located behind this NAT from
public IPv4 network.
You need several IPv4 address, an equal number of host to
be connected to. Sorry, manual page is not ready yet.
2001-10-25 JINMEI, Tatuya <>
* kame/kame/ndp/ndp.c (dump): skipped non-host routes when
printing neighbor cache entries, since they were just confusing.
2001-10-25 JINMEI, Tatuya <>
* kame/sys/netinet6/nd6.c (nd6_rtrequest): ignored a route when it
is created by cloning and is not a neighbor. Otherwise, it could
mistakenly regard a cloned route to an off-link destination as
an on-link neighbor, and create a neighbor cache. Then we'd see
some strange neighbor solicitations to the destination.
Some additional clarifications on this file were made with this
2001-10-25 JINMEI, Tatuya <>
* kame/sys/netinet6/in6_gif.c (in6_gif_output): let ip6_output()
(and a subroutine) to validate the cache and to make a new one, in
order to avoid missing logic and code duplication.
* kame/sys/netinet/in_gif.c (in_gif_output):
added a check if the cached route (per gif) is up before using it.
This should be verified for all type of cached routes, but we've
forgotten the check. With this change, FreeBSD will be able to
reflect a change of route for the other end point of the tunnel
as soon as possible.
(OpenBSD IPv6 only)merged recent changes into the openbsd specific
code path.
2001-10-25 JINMEI, Tatuya <>
* kame/sys/netinet6/nd6.h: removed the ndpr_addr member from the
nd_prefix structure, which was effectively not used anywhere. The
structure is kernel-internal, so there should be no compatibility
issue to existing applications.
2001-10-25 Shin'ichi Fujisawa <>
* kame/kame/natptconfig/{cfparse.y,defs.h,yylex.c}:
- Add "daddr" keyword to specify that translator converts
destination address, and add element of struct ruletab{}
to hold this address.
- Add translation rule to convert IPv4->IPv6. I am sorry,
many of them do not work yet.
2001-10-25 JINMEI, Tatuya <>
* kame/sys/netinet6/nd6.c (nd6_is_addr_neighbor): corrected an
expression to detect whether a given address matches an on-link
prefix. The previous one mistakenly detected every address as
2001-10-25 Shin'ichi Fujisawa <>
* kame/sys/netinet6/natpt_{rule,tslot}.c:
- Change dport handling (destination port mapping) to use a
new member of cSlot structure.
Wed Oct 24 13:43:34 JST 2001
* kame/sys/netinet/ip6.h, kame/sys/netinet6/mip6*
change the packet format of binding update/ack
according to the latest draft.
Wed Oct 24 11:15:40 2001 SUMIKAWA Munechika <>
* freebsd4: fix the bug IPv6 over vlan do not work when updating
2001-10-23 JINMEI, Tatuya <>
added a per-gif timer to refresh a (possibly) stale cached route
towards the tunnel end. Based on a comment from
* kame/sys/net/if_gif.h: added a new member "rtcache_expire" to
introduce per-gif timer (see below).
* kame/sys/netinet/in6_gif.c (in6_gif_output):
* kame/sys/netinet/in_gif.c (in_gif_output):
purged the cached route after in[6]_gif_rtcachetime seconds.
2001-10-23 JINMEI, Tatuya <>
* kame/sys/netinet6/in6_gif.c (in6_gif_output):
* kame/sys/netinet/in_gif.c (in_gif_output):
Corrected the loop prevention algorithm. The current one is not
effective, because it tries to detect the loop condition only
when generating a new route, and it does not remove the cached
route even when a loop is detected.
Luckily, the bug does not necessarily mean we'll see infinite
loop, since we have another mechanism for loop prevention in
2001-10-23 JINMEI, Tatuya <>
* kame/sys/netinet6/ip6_output.c: NULL-clear m->m_pkthdr.rcvif
before calling ip6_mforward(). [KAME PR sys/387]
2001-10-23 JINMEI, Tatuya <>
* kame/kame/v6test: removed the -c option introduced on
2001-09-16, just because we do not need the function.
Tue Oct 23 10:11:07 JST 2001
* kame/kame/racoon:
racoon gives up the negotiation when racoon cannot find any pre-shared
key in the case of the aggressive mode AND verify_identifier is set on.
Mon Oct 22 18:28:07 JST 2001
* kame/sys/netinet6/mip6_binding.c
- fix a bug of the updating routine of the BC entries
of CNs when the unregistering has finished.
Mon Oct 22 14:54:26 JST 2001
* nodeinfod: support IPv4 node addresses.
* sys/netinet6/icmp6.c: do not respond to IPv4 node addresses query,
with a bogus packet.
Sat Oct 20 03:09:33 JST 2001
* freebsd4/sys/netinet/udp_usrreq.c:
the data in the ip header should be restored when the extended udp
header and data checksum is calculated. this caused some trouble
in the code which the ip header is not modified. for example,
inbound policy lookup failed.
Fri Oct 19 16:57:58 2001 SUMIKAWA Munechika <>
* freebsd4/ports/{ct,v6eval}: upgrade to 1.3.
Fri Oct 19 15:41:09 JST 2001
* sys/netinet6/nd6.c: repair binary backward compatibility breakage
for SIOCGPRLST_IN6 ioctl.
Fri Oct 19 09:54:19 JST 2001
* kame/sys/key.c:
the value of the prefixlen in the sadb_address structure is fixed.
when pfkey message relative to SA is sent, the prefixlen was incorrect.
reported by <>.
Thu Oct 18 17:16:19 JST 2001
* kame/sys/netinet6/mip6*
- fix a bug that a HA sends wrong seq number when issuing a
'too small seqno' binding ack.
Thu Oct 18 14:25:36 JST 2001
* kame/sys/netinet6/ip6_output.c
do not touch any extention headers during the mip6 processing
if IPV6_FORWARDING is set.
2001-10-17 Shin'ichi Fujisawa <>
* kame/sys/netinet6/natpt_trans.c:
- Fix a bug to miscalculate when compute a checksum in IPv4 ->
IPv4 translation.
- Remove natpt_recalculateTCP4Checksum() and related include
file. This function became useless by modification mentioned above.
Wed Oct 17 00:03:54 JST 2001
* kame/kame/racoon:
fixed the internal value of the wild card of the upper layer protocol.
the value of the wile card is differnt between the kernel and IKE
protocol. reported by <>.
Tue Oct 16 23:56:34 JST 2001
* kame/kame/racoon:
to compare the policy between own policy and peer's policy in the
responder side, it should be loose. for example, the initiator
proposes "any", and the responder requires "tcp", this case is
admitted. also, the initiator proposes "tcp", and the responder
accepts "any", this case is admitted too. however this behavior
should be considered.
Tue Oct 16 19:22:52 JST 2001
* kame/sys/netinet6/mip6_binding.c
fix the bug of the validation routine of an incoming tunneled
packet from a homeagent to a mobile node.
Tue Oct 16 16:09:22 JST 2001
* bsdi4/sys/netinet/ip_icmp.c:
do not panic even if last arg to icmp_error() is NULL.
this situation can happen when ICMPv4 too big is generated.
from markus@openbsd. bsdi4 only.
Mon Oct 15 21:25:29 JST 2001
* icmp6.c
swap the ip6_src and the address in the homeaddress destopt
(if exists) in icmp6_error() routine.
2001-10-15 Shin'ichi Fujisawa <>
* kame/sys/netinet6/natpt_{defs.h,trans.c}
- Fix a bug to miscalculate when compute a checksum in
IPv6 -> IPv4 translation.
Thu Oct 11 21:57:50 JST 2001
* fix BU list management function.
* introduce MIP6_ALLOW_COA_FALLBACK kernel option.
this allows you to use a coa as a src address if the peer
doesn't recognize a home address destination option.
may arouse a mip6 believe's anger, but very useful.
Wed Oct 10 17:38:05 JST 2001
* kame/kame/racoon:
Fixed racoon crash when uni-directional policy is defined.
racoon negoticate two SAs even in the case of the uni-directional
Wed Oct 10 04:52:54 JST 2001
* bsdi4: use PULLDOWN_TEST (m_pulldown) codepath. improves conformance
when ipv6 extension headers are present.
Wed Oct 10 10:27:24 JST 2001
* netbsd/sys/netinet/raw_ip.c: fix a typo which could lead to kernel
panic when ICMPv4 is returned against raw ip socket.
reported by kato@wide
2001-10-10 JINMEI, Tatuya <>
* kame/kame/rtadvd/rtadvd.c (find_prefix, prefix_match): corrected
prefix calculation:
- avoid invalid pointer access when bytelen is a multiple of 8
- avoid using the right shift operator to make the code look safer
2001-10-09 JINMEI, Tatuya <>
* kame/kame/rtadvd/config.c (getconfig): made configuration parser
compatible with FreeBSD 4.4-RELEASE, in terms of the route
information option:
- allowed rtrXXX instead of rtXXX.
- made route lifetime optional.
In any case, some warning messages are printed as well, so that the
user can notice the change and fix the configuration.
Tue Oct 9 08:43:04 JST 2001
* kame/kame/racoon/schedule.c:
the entry of the schedule is marked with dead before the function in
the entry will be called. some schedules, check_rtsock(),
check_flushsa_stub, woulbe be remained in the scheduler even after they
were excuted. sched_scrub_param() is unnecessary probably anymore.
2001-10-04 JINMEI, Tatuya <>
* bsdi4/contrib/bind/src/lib/irs/dns_ho.c (add_hostent):
calculated alignment correctly. Without this fix, gethostbyname()
would not work correctly when the function handles more than one
This is a back merge of a fix from bsdi to BSD/OS 4.3 (beta).
* freebsd4/lib/libinet6/Makefile
- "options insecure1" is available on FreeBSD4-KAME, too.
(you have to rebuild applications to enjoy this feature)
2001-10-01 Shin'ichi Fujisawa <>
* kame/kame/{natptconfig,natptlog}:
- Add contents of manual page.
- Separate contents of configuration file as natpt.conf.5.
2001-09-26 JINMEI, Tatuya <>
* kame/sys/netinet6/in6_ifattach.c (in6_ifattach): calls
nd6_ifattach() before creating addresses. This is necessary
because an MLD packet may be sent during the creation procedure,
in which a valid nd_ifinfo entry can be referred to.
Wed Sep 26 14:30:46 JST 2001
* kame/kame/racoon:
new directive "verify_identifier" has been added. it can be strict
to check the identifier in the ID payload transmitted by the peer.
the default is off.
Wed Sep 26 00:01:08 JST 2001
* kame/sys/netkey/key.c:
* kame/kame/setkey:
the syntax how to define a policy of a ICMPv6 type and/or a code
has been changed. the previous modification at Fri Sep 21 broke
a backward compatibility, and had no sense. when the policy doesn't
require IPsec for an inbound Neighbor Solicitation with any source
/destination address, the specification is the following;
spdadd ::/0 ::/0 icmp6 135,0 -P out none;
2001-09-25 JINMEI, Tatuya <>
* kame/kame/addrselect/: a tool to configure the policy table (see
below). This program is also an experimental stuff.
2001-09-25 JINMEI, Tatuya <>
* kame/sys/netinet6/in6_src.c: implemented the policy table for
source address selection, according to
The policy table can be configured via the sysctl(3) interface
(except for FreeBSD, at this moment). This implementation is
still experimental, and might be changed in the near future.
2001-09-21 JINMEI, Tatuya <>
* kame/sys/netinet6/nd6.c (fill_[dp]rlist): added as shared
subroutines for nd6_sysctl_[dp]rlist (for FreeBSD) and nd6_sysctl
(for other *BSDs), in order to centralize the complicated logic.
2001-09-21 JINMEI, Tatuya <>
* kame/sys/netinet6/nd6_rtr.c (defrouter_select): removed the
default route if there was neither a default router nor the
default interface.
2001-09-21 JINMEI, Tatuya <>
* kame/sys/netinet6/in6_src.c (in6_selectroute): made sure to fill
in retifp and retrt, regardless of the return value. Without this
change, the kernel could panic, since ip6_output() refers to the
returned ifp even in error cases.
All versions of the kernel after the 20010730 snap should be
upgraded to fix this problem.
2001-09-21 JINMEI, Tatuya <>
* kame/sys/netinet6/(various files): fixed integer overflow for
valid and preferred lifetimes;
- introduced new members in nd_prefix{} and in6_ifaddr{} to record
the timestamp of the latest update
- check expiration based on the difference between the current
time and the timestamp, not on the explicit expiration times
Fri Sep 21 14:19:02 JST 2001
* kame/sys/netinet6/ipsec.c:
When the value of the upper layer of the security policy index (spidx)
structure is ICMPv6, the port field in "src" of the spidx means ICMPv6
type, and the port field in "dst" of the spidx specifies ICMPv6 code.
For example, the following means the policy doesn't require IPsec for
an inbound Neighbor Solicitation.
spdadd ::/0[135] ::/0 icmp6 -P in none;
2001-09-20 JINMEI, Tatuya <>
* kame/sys/netinet6/nd6_rtr.c (prelist_update): corrected the
definition of "storedlifetime" (used in the two-hour rule) for an
address that has an infinite lifetime. Without this special
case, the lifetime of such an address would unintentionally be
2001-09-19 JINMEI, Tatuya <>
* kame/kame/v6test/: added some tiny improvements/corrections:
- added missing NTOHS
- corrected checksum calculation for packets with routing headers
- made the checksum calculation routine against invalid packets
The fist two were based on comments from Yutaka Shimizu
2001-09-19 Shin'ichi Fujisawa <>
* freebsd4/sys/netinet/udp_usrreq.c:
- Remove 'static' attribute from variable "udpcksum".
* kame/kame/sys/netinet6/natpt_trans.c:
- Examine net.inet.udp.checksum when making UDP packet.
* kame/sys/netinet6/natpt_{defs.h,{dispatch,trans,tslot}.c}:
- traceroute6 works. I forgot to review it.
Wed Sep 19 19:06:06 JST 2001
* sys/netinet6/ip6_{in,out}put.c: make IPV6_TCLASS socket option to
take int, not u_int8_t. follows the latest 2292bis draft.
(backward binary compatibility is provided for bsdi43)
Thu Sep 20 01:00:08 JST 2001
* kame/kame/rtsold:
improved the -a option. it can probe a interface automatically when
the interface wake up. it can be started anytime even when there is no
network interface on the list of intarfaces in the kernel.
2001-09-18 JINMEI, Tatuya <>
* kame/kame/libinet6/resolv/res_send.c (res_send):
* bsdi4/contrib/bind/src/lib/resolv/res_send.c (res_nsend):
if a UDP response from a "wrong" server is truncated (and if we
allow to accept such responses), fall back to TCP with the "wrong"
address, in order to avoid connecting to an anycast address.
* kame/kame/v6test/testcap.[ch](tgetnum), getconfig.c: fixed a
bug that you sometimes cannot specify a value if its MSB is on.
* kame/kame/v6test/cksum.c(cksum6), v6test.c(main), v6test.1:
added an option not to generate checksum automatically.
2001-09-15 JINMEI, Tatuya <>
* kame/kame/v6test/getconfig.c (make_rthdr): fixed routing header
2001-09-13 JINMEI, Tatuya <>
* bsdi4/usr.bin/telnet/commands.c (tn): made it sure to terminate
the loop of connection attempt correctly, based on a comment from
2001-09-13 JINMEI, Tatuya <>
* kame/sys/netinet6/nd6.c (nd6_timer): before calling
icmp6_error(), embedded scope zone ID (if necessary) for the
erroneous packet.
Thu Sep 13 08:15:21 JST 2001
* kame/sys/netkey/key.c:
newer SA is prefered for a out-bound packet than old one
when net.key.prefered_oldsa is set to zero.
Thu Sep 13 08:12:46 JST 2001
* kame/sys/netinet6/ipsec.c:
fixed to process a IPv6 packet when ah transport after esp tunnel
should be applied. the SA of AH transport could not be selected
from the SAD because of this bug.
Wed Sep 12 16:19:42 JST 2001
* kame/racoon/proposal.c:
fixed to compare pfs values in two proposals in the case of
"claim" mode. reported by <>
2001-09-13 JINMEI, Tatuya <>
* kame/sys/netinet6/scope6.c (in6_addr2zoneid): changed the return
value type from u_int32_t to int64_t, so that the caller can tell
an error from valid "4+28" ID values. All the callers of this
function were also modified accordingly, with stricter validation
2001-09-03 Shin'ichi Fujisawa <>
* kame/sys/netinet6/natpt_{rule.h,trans.c,var.h}:
- Support tftp translation. Tftp6 client can connect to tftp4
2001-09-10 JINMEI, Tatuya <>
* kame/sys/netinet6/udp6_output.c (udp6_output): (bsdi4 and
netbsd) when sending IPv4 packets represented as IPv4-mapped IPv6
address, passed socket option to ip_output() so that the function
would handle broadcasted packet correctly. For bsdi4, also merged
all-ones broadcast cases from udp_usrreq.c.
2001-09-07 JINMEI, Tatuya <>
* kame/sys/netinet6/ip6_output.c (ip6_setpktoptions): checked if
each cmsghdr pointer had enough size to store the structure.
This could be a security fix, but I think the current code is
practically safe enough. That is, we do not have to be in a harry
to merge this fix to *BSDs.
2001-09-06 JINMEI, Tatuya <>
* *bsd/sys/sys/socket.h (CMSG_FIRSTHDR): checked msg_controllen
in CMSG_FIRSTHDR as described in RFC2292, particularly in case
that the kernel returns an empty list for some reasons.
(based on a note from David Borman <>)
2001-09-05 JINMEI, Tatuya <>
* bsdi4/contrib/bind/src/lib/resolv/res_send.c (res_nsend):
* kame/kame/libinet6/resolv/res_send.c (res_send):
when "insecure1" is specified, do not connect datagram sockets,
so that the kernel can accept responses from an "unknown" server.
2001-09-05 JINMEI, Tatuya <>
* kame/sys/netinet6/ip6_input.c (ip6_sysctl): (bsdi4 only) made
net.inet[6].ip6.v6only modifiable.
Wed Sep 5 12:05:14 JST 2001
* faithd: change the default directory for daemons and configuration
files. was: /usr/local/v6/{libexec,etc}, now: /usr/libexec and /etc.
2001-09-05 JINMEI, Tatuya <>
* bsdi4/sys/netinet/in_pcb.c (in_pcballoc): when
net.inet[6].ip6.v6only is 1, set IN6P_IPV6_V6ONLY (which
corresponds to the IPV6_V6ONLY socket option) for all sockets at
Tue Sep 4 01:32:49 2001 SUMIKAWA Munechika <>
* freebsd4/usr.sbin/ppp: import IPv6 supporting PPP from
2001-09-03 Shin'ichi Fujisawa <>
* kame/sys/netinet6/natpt_*.[ch]:
* freebsd4/usr.sbin/{natptconfig,natptd,natptlog}:
KAME NAT-PT (kernel and user command) code was completely renewed.
Major changes are as follows.
- Changed a coding style to "bsd" (described in IMPLEMENTATION,
chapter 8).
- Does not depend on interface.
'interface' directive was removed in configuration rule, and no
inbound/outbound distinction in each rule.
- Changed a syntax of configuration command (natptconfig) related to
the item mentioned above. The syntax does not change greatly, it
became simpler.
- Use TAILQ (defined in /usr/include/sys/queue.h) instead of my
original list handle routines.
- Removed natptd user command.
Because the facility that natptd aimed is the same as totd.
2001-09-02 JINMEI, Tatuya <>
* bsdi[34]/sys/sys/mbuf.h (MFREE): corrected the point of calling
splmem_fast(), so that the lock would not be freed
unintentionally. A busy kernel, such as the one processing IPsec,
could panic without this fix, so all bsdi users are recommended to
apply this fix.
The fix is a back port from BSD/OS 4.3 beta.
2001-08-31 SUZUKI Shinsuke <>
* pim6sd/pim6_proto.c (parse_pim6_hello): strengthens length check for
Hello packet. Without this fix, pim6sd may crash if it receives a
Hello packet with a too short Hello option.
2001-08-31 JINMEI, Tatuya <>
* kame/sys/netinet6/{in6.c in6_ifattach.c, nd6_rtr.c}: revised the
support of the privacy extension of stateless address
configuration, based on draft-ietf-ipngwg-temp-addresses-v2-00.
See Section 8. of the document to see feature changes.
* kame/sys/netinet6/in6_var.h: added a new member "ia6_createtime"
to the in6_ifaddr structure with the change above, in order to
record the time of creation of an address. This member is set for
all addresses, but is only referred for temporary addresses.
2001-08-30 JINMEI, Tatuya <>
* bsdi4/sys/netinet/tcp_input.c (tcp_rtlookup): always used
rtcalloc() so that both active and passive connections would make
cloned host routes. The previous code, which called rtalloc(),
should not have severe bad effects, but using rtcalloc() is more
suitable to bsdi's logic.
2001-08-28 JINMEI, Tatuya <>
* kame/kame/ndp/ndp.c (ifinfo):
- printed the values of ND6_IFF_ACCEPT_RTADV and
- allowed a user to specify the flag bits by
'ndp -i IF [-]accept_rtadv' or
'ndp -i IF [-] prefer_source', respectively.
2001-08-28 JINMEI, Tatuya <>
* kame/sys/netinet6/nd6.h:
* kame/sys/netinet6/in6_src.c (in6_selectsrc):
introduced a new flag bit "ND6_IFF_PREFER_SOURCE" in the nd_ifinfo
structure, in order to specify "preferred" interfaces for source
address selection.
2001-08-28 JINMEI, Tatuya <>
* kame/sys/netinet6/nd6.[ch]:
* kame/sys/netinet6/nd6_rtr.c (nd6_ra_input):
introduced a new flag bit "ND6_IFF_ACCEPT_RTADV" in the nd_ifinfo
structure, in order to control whether to accept RAs per-interface
basis. The kernel now accepts RAs only when
net.inet6.ip6.accept_rtadv is 1 and the flag is on the receiving
interface. The new stuff does not change the old default
2001-08-28 JINMEI, Tatuya <>
* kame/sys/netinet6/in6_src.c (in6_selectsrc): corrected a
condition on temporary addresses.
* bsdi4/sys/net/if.h:
add ALTQ-compat queue macros to bsdi4 in order to
reduce "#ifdef ALTQ". (in paticular, sys/net/if_gif.c)
(the commit mail didn't go out due to a permission problem in
the repository.)
2001-08-18 JINMEI, Tatuya <>
* kame/sys/netinet6/in6_src.c (in6_selectsrc): changed the number
of longest matching rule in source address selection from 8 to 14,
so that it is easy to assign smaller numbers to more preferred
rules. Additionally, a new rule to prefer addresses on alive
(i.e. IFF_UP) interfaces, based on a suggestion from
*bsd/netstat were also changed accordingly.
Fri Aug 17 05:24:45 JST 2001
* kame/setkey/parse.y: handle FQDN address in SPD/SAD configurations.
when an FQDN resolves into multiple addresses, setkey(8) will
install SPD/SAD entries for all possible combinations.
inspired by Solaris IPsec stack.
2001-08-16 JINMEI, Tatuya <>
* bsdi4/usr.sbin/netstat/stats_inet6.h:
* {freebsd4,netbsd}/usr.bin/netstat/inet6.c (ip6_stats):
printed statistics about the source address selection rules.
(experimental. we may have to reconsider the way to print the
2001-08-16 JINMEI, Tatuya <>
* kame/sys/netinet6/in6_src.c (in6_selectsrc): implemented the
source address selection algorithm based on
draft-ietf-ipngwg-default-addr-select-05. With this change, the
older backend function, in6_ifawithscope(), to choose a source
address was removed. Also,
- added statistics array to ip6stat to see which rule is applied
in the algorithm.
- added a new sysctl variable "net.inet6.ip6.prefer_tempaddr" to
reverse the rule about temporary addresses.
2001-08-15 JINMEI, Tatuya <>
* kame/sys/netinet6/scope6.c (in6_addrscope): treat the loopback
address as global, because there is no ambiguity about scoping.
Otherwise, we'd see ::1%ID (ID != 0) via scope6_addr2default(),
which would rather has bad effects.
Note: I'm still not sure if this change is really safe. We should
take care of the new behavior.
2001-08-14 Keiichi SHIMA <>
* add a home agents list timeout routine.
* add some init functions.
Tue Aug 14 15:01:57 JST 2001
* kame/pim6sd: explicitly assigns NULL to pointers in newly-
malloced structures if there's no specific pointer to be assigned.
Tue Aug 14 05:20:23 JST 2001
* sys/netkey/key.c: drop support for byte lifetime. byte lifetime
is define in a very vague manner, and it can lead to unsynchronized
SAs (= dangling SA left behind) with packet losses, byte count
mismatches and other causes. kernel returns EINVAL if non-zero
byte lifetime is specified on ADD/UPDATE.
Mon Aug 13 22:03:28 JST 2001
* kame/racoon:
supported MODP 2048, 3072, 4096 and 8192-bit.
these are described in draft-ietf-ipsec-ike-modp-groups-01.txt
2001-08-12 JINMEI, Tatuya <>
* kame/sys/netinet6/ip6_output.c (ip6_setpktoption): newly added,
in order to centralize the routine for handling outgoing packet
options both as socket options and as ancillary data.
It also introduced additional data validation. For example,
IPV6_USE_MIN_MTU is not allowed in the context of RFC2292 (i.e. by
2001-08-09, Keiichi SHIMA <>
* add a relay routine for an icmp against an encaplulated packet.
not tested.
* free mbuf in case error in the functions that has an mbuf pointer
in its argument list.
Thu Aug 9 09:37:44 BST 2001
* kame/kame/pim6sd: s/vifi_t/mifi_t/ to sync with kernel implementation
of IPv6 multicast interface.
2001-08-07 JINMEI, Tatuya <>
* kame/sys/netinet6/ip6_output.c (ip6_pcbopt, ip6_setpktoptions):
for the IPV6_NEXTHOP ancillary data/socket option, set
sin6_scope_id to 0 in non-SCOPEDROUTING cases, so that the kernel
can make a neighbor cache for the specified nexthop with a scoped
Tue Aug 7 23:30:59 JST 2001
* kame/racoon:
RFC2407 says that INITIAL-CONTACT is the notify message that
announces the peer the sender of the message was rebooted.
previous interpretation in racoon was to delete all SAs which
source address is the sender of the message.
with the change, racoon only deletes SA which matches BOTH the
source address and the destination accress in the notify message.
Tue Aug 7 09:42:23 JST 2001
* libinet6/getrrsetbyname.c: libc implementation of getrrsetbyname(3).
useful for racoon(8) cert-on-DNS handling. from openbsd.
Mon Aug 6 21:18:58 BST 2001
* kame/kame/v6test/getconfig.c: you can build up a packet from
multiple raw-data entries.
Sun Aug 5 14:00:04 JST 2001
* sys/netinet6/ipsec.c: improve cached ipsec policy lookup on connected
2001-08-03 Keiichi SHIMA <>
add mobileip source code.
currently, many parts of spec are not implemented yet. even
uncompilable under other than freebsd4 now. also very
unstable. of course, without MIP6 kernel option, there is no
problem. all mip6 related code is separated by ifdef MIP6.
this integration is just for KAME mip6 developper's
convinience. the purpose of this early integration is to
decrease the maintenance cost to synchronize KAME main tree
and developping mip6 tree. after this merge, mip6 code
modifications are reflected directly to KAME code.
i will make these codes compilable on the other OSes as soon
as possible.
2001-08-02 JINMEI, Tatuya <>
* kame/kame/advapitest/sender.c (main): added a new option '-n
nexthop' to test IPV6_NEXTHOP.
2001-08-02 JINMEI, Tatuya <>
* kame/sys/netinet6/{ip6_output.c, in6_src.c}: supported
the IPV6_NEXTHOP ancillary data / socket option.
At this moment, the next hop should be an AF_INET6 socket address.
2001-08-02 JINMEI, Tatuya <>
* kame/sys/netinet6/nd6.c (nd6_is_addr_neighbor): check on-link
prefixes (not addresses) to see if a given address is a neighbor
in a given link.
Tue Jul 31 22:54:57 JST 2001
* kame/kame/pim6sd/cfparse.y: command-line debug-option works now.
Tue Jul 31
* kame/racoon:
the phase1 deletion should be postponed until there is no phase2.
this was probably made a consensus at vpn bakeoff on Oct, 1998.
2001-07-29 JINMEI, Tatuya <>
* kame/sys/netinet6/(several files): several improvements on the
IPv6 output routine:
- added a seprate function in6_selectroute() to centralize the
route and outgoing interface selection algorithm
- call in6_selectroute() from in6_selectsrc() and ip6_output(), and
use same codebase for both unicast and multicast packets as much
as possible
- added stricter (and thorough) scope checks for outgoing packets
(e.g. this check would prevent a packet generated by
'ping6 -S ::1 fe80::1%ne0' from being sent)
- renamed in6_addr2scopeid() to in6_addr2zoneid() to be more intuitive
in terms of the scope architecture
- made in6_addr2zoneid take care of the loopback address
- set the tentative flag before joinging multicast addresses, so
that the corresponding MLD packet would not have a tentative
source address
2001-07-29 JINMEI, Tatuya <>
* kame/sys/netinet6/nd6_nbr.c (nd6_ns_output):
* kame/sys/netinet6/nd6_nbr.c (nd6_na_output):
* kame/sys/netinet6/icmp6.c (icmp6_redirect_output):
do not pass a pointer to a pointer to ifnet to ip6_output() for
counting statistics, use the interface on which the ND process is
involved instead.
2001-07-29 JINMEI, Tatuya <>
* kame/sys/netinet6/mld6.c (mld6_input): allowed MLD messages to
have the unspecified address as source.
* kame/sys/netinet6/mld6.c (mld6_sendpkt): use the unspecified
address when there is no valid link-local address available.
Fri Jul 27 17:39:24 JST 2001
* sys/net/if_sec.c: "sec*" pseudo device for decapsulating IPsec
tunnel packets. if we have the device, we depart from RFC2401
tunnelling model (where IPsec tunnels are defined within IPsec
document) to draft-touch-ipsec-vpn-01.txt model on inbound.
all IPsec tunnel-mode SPD entries will automagically be accompanied
with a sec* device. refer to IMPLEMENTATION for more details.
Fri Jul 27 12:49:47 JST 2001
* sys/netinet6/ipsec.c: cache ipsec policy on pcbs, try to avoid
per-packet SPD lookup. idea from thorpej
2001-07-27 JINMEI, Tatuya <>
* bsdi4/sys/net/if_loop.c (looutput): set M_LOOP for IPv6 packets.
Thu Jul 26 23:13:06 JST 2001
* mdnsd: IPv4 support. PR 370 from
Thu Jul 26 02:12:04 JST 2001
* *bsd*/sys/netinet*/in*_pcb.c: call ipsec_init_policy() from
within in*_pcballoc(), not from PRU_ATTACH logic. it is to make
the allocation of ipsec policy struct prior to making inpcb available
to the world (don't show incomplete inpcb to others).
can lead to panic(), as reported on stable@freebsd.
Wed Jul 25 18:15:37 JST 2001
* kame/sys/netinet6/udp6_output.c, freebsd4/sys/netinet6/udp_usrreq.c:
rejects IPv6 packet toward IPv4-mapped address if its source
address is not an IPv4-mapped IPv6 address, since the
converted IPv4 packets would have an unexpected IPv4 source
(TCP6 might have the same bug)
Wed Jul 25 15:33:06 JST 2001
* netbsd/sys/net/route.c: do not fill rmx_mtu too much on RTM_ADD/
RTM_RESOLVE. this was merged by mistake during past kame work.
2001-07-25 JINMEI, Tatuya <>
* kame/sys/netinet6/(various files): correctly supported
interface-local multicast:
- install ff01::%ifname/32 when first configuring an interface.
- changed the source address selection for ff01::xxxx.
- changed the outgoing interface selection for ff01::xxxx.
- prohibit packets towards ff01::xxx from being sent on the wire.
- reject packets towards ff01::xxxx from the wire.
Wed Jul 25 11:52:34 JST 2001
* sys/netinet/if_gif.c: use encap_attach(), instead of
encap_attach_func(). enables us to use radix address match in
Wed Jul 25 03:51:34 JST 2001
* sys/netinet/ip_encap.c: use radix table lookup for entries registered
through encap_attach() API (address pairs). still not sure if it
gives us any performance improvement, but should be much better
when you have thoudsands of gif interfaces.
2001-07-24 JINMEI, Tatuya <>
* kame/sys/netinet6/in6.c (in6_control): in the SIOCDIFADDR_IN6
case, if the corresponding prefix can be removed, just call
prelist_remove(), not changing the expire value.
Tue Jul 24 17:54:10 JST 2001
* sys/netinet6/in6_prefix.c: RIP.
2001-07-24 JINMEI, Tatuya <>
* kame/kame/ping6/ping6.c: use bind(2) to specify the source
address with the -S option. (Unfortunately) IPV6_PKTINFO is not a
good idea, because it cannot handle scoped addresses very well.
2001-07-24 JINMEI, Tatuya <>
* kame/sys/netinet6/raw_ip6.c (rip6_usrreq):
* freebsd4/sys/netinet6/raw_ip6.c (rip6_bind):
* freebsd4/sys/netinet6/raw_ip6.c (rip6_connect):
supported the scoped address format in bind(2) and connect(2) for
raw socket. The former case is essential; bind(2) for a scoped
address would not success without this fix.
Tue Jul 24 03:54:24 JST 2001
* sys/netinet6/nd6.c: repair ndp -R.
* ndp: do not print uninitialized interface on ndp -i.
Mon Jul 23 15:10:08 JST 2001
* sys/netinet6/in6.c: now ff01::/16 prefix means "interface local
multicast", not "node local multicast". therefore, we now
join ff01::1 in per-interface basis. buggy.
* *bsd*/usr.*/netstat/route.c: decode embedded scope identifier for
ff01::/16. NOTE: we still are using IN6_IS_ADDR_MC_NODELOCAL macro,
as there's no "interface local" macro defined in 2553bis-03.
Mon Jul 23 00:28:05 JST 2001
* sys/net/radix_mpath.c: (netbsd, RADIX_MPATH) reject RTM_ADD of
conflicting entries, where all key/mask/gw are the same.
Sun Jul 22 11:07:17 JST 2001
* netbsd/sys/netinet/in.c: improve IFA_ROUTE management, to make sure
multiple interface addresses with the same prefix ( and works right. i bothered as it is mandatory with
RADIX_MPATH environment, and netbsd-current is moving toward this
direction. netbsd only.
- on interface address addition, check if there's the same prefix
already installed, and if so, don't install another one
- on removal, check if there's other interface address which has the
same prefix, and if so, move IFA_ROUTE to him
Sat Jul 21 13:30:21 JST 2001
* sys/netinet6/nd6_rtr.c: simplify default router list manipulation.
try to install multiple default routes into routing table, if
there are multiple RA sources (routers).
Sat Jul 21 03:35:20 JST 2001
* *bsd*/sys/net/route.c: make equal() more pickier about sa_len,
to avoid possible data buffer overrun.
* netbsd/sys/net/radix.c: multipath support in radix routing table.
experimental, netbsd only. see 7.2 of IMPLEMENTATION.
Fri Jul 20 14:20:51 JST 2001
* nodeinfod: a daemon that responds to node information queries.
we are wondering if it is a good idea to migrate to a userland
implementation, for complicated query type handling (like node
addresses, FQDN response with DNS encoding, NI group address,
and such). we still find ping6 -w responses quite useful and
would like to keep it in the kernel (so that every node responds to).
or should we implement CDP? :-) experimental.
2001-07-18 JINMEI, Tatuya <>
* kame/kame/libinet6/getaddrinfo.c (getaddrinfo): enabled
destination address reordering (that I introduced on July 3) by
default, and removed the environment variable knob.
2001-07-18 JINMEI, Tatuya <>
* kame/sys/netinet6/nd6.c (nd6_free): if the reason for the
deletion is just garbage collection, and the neighbor is an active
default router, do not delete it. Instead, reset the GC timer
using the router's lifetime. Simply deleting the entry would
affect default router selection, which is not necessarily a good
thing, especially when we're using router preference values.
2001-07-18 JINMEI, Tatuya <>
* kame/sys/netinet6/nd6_rtr.c (pfxlist_onlink_check): mark
(autoconfigured) prefixes that do not have a reachable router when
we have at least one default router (as well as the former
This change would help the case that we've moved to a new link where
we have a router that does not provide prefixes and we configure
an address by hand.
2001-07-18 JINMEI, Tatuya <>
* bsdi4/contrib/bind/src/lib/resolv/res_init.c (res_setoptions):
applied the fix below.
Wed Jul 18 18:06:25 JST 2001
* libinet6/resolv/res_init.c: by putting "insecure1" or "insecure2"
into /etc/resolv.conf "options" line, you can control the
RES_INSECURE[12] settings in _res.options.
2001-07-18 JINMEI, Tatuya <>
* kame/sys/netinet6/in6.c (in6_createmkludge): was newly added
to allocate space for the kludge at interface initialization time.
Formerly, we dynamically allocated the space in in6_savemkludge()
with malloc(M_WAITOK). However, it was wrong since the function
could be called under an interrupt context (software timer on
address lifetime expiration).
Although this function is a global one, it is expected to be
called only from in6_ifattach().
Tue Jul 17 10:58:21 2001 SUMIKAWA Munechika <>
* freebsd4/sys/kern/kern_exec.c: applied the patch of 'FreeBSD
Security Advisory FreeBSD-SA-01:42.signal'
2001-07-16 Shin'ichi Fujisawa <>
* kame/sys/netinet6/natpt_dispatch.c (configCv4, configCv6):
- configCv4() and configCv6() returns real payload number or
IPPROTO_IP. Packet passes to the next process when value of
IPPROTO_IP was returned.
2001-07-15 Shin'ichi Fujisawa <>
* kame/sys/netinet6/natpt_{dispatch.c,var.h}
- Separate natpt_pyldaddr() from foundFinalPayload(). This
routine finds ICMP/TCP/UDP payload from IPv6 header chain
and returns it's address.
* kame/sys/netinet6/natpt_trans.c
- translate ICMPv6 Error Messages into ICMPv4.
2001-07-15 JINMEI, Tatuya <>
* kame/sys/netinet6/in6.c (in6_update_ifa): do not call malloc
with M_WAITOK, since in6_update_ifa() can be called under an
interrupt context.
Based on a FreeBSD PR 28927 from Garrett Wollman
<> via ume.
Sun Jul 15 00:01:44 JST 2001
* kame/racoon:
the time of execution of several cipher function and the taking time
of each phase can be logged when ENABLE_STATS is defined.
these messages are passed to syslogd as LOG_NOTICE.
2001-07-12 Shin'ichi Fujisawa <>
* kame/sys/netinet6/natpt_{rule,tslot}.c
- You can connect from outside (IPv4 world) to inside (IPv6
world), The following connection is possible.
o inbound from
to 3ffe:501:41c:6000:2a0:24ff:fe95:a4a6
o inbound from any4 port 65303
to 3ffe:501:41c:6000:2a0:24ff:fe95:a4a6 port 23
This facility did not work for a bug until now.
Wed Jul 11 17:55:33 JST 2001
* kame/kame/pim6sd: added SSM (Source-Specific-Multicast) mode
for SSM prefix (ff20::/12 and ff30::/12).
from Mickael Hoerdt <>.
+ supports MLDv2 as well as MLDv1 (cfparse.y, cftoken.l,
mld6*.[ch], timer.c, vif.[ch])
+ uses SPT for multicast address in SSM range
(inet6.[ch], mrt.c, pim6*.[ch], route.c)
* kame/sys/netinet/icmp6.h: added ICMPv6 code for MLDv2 Report
* kame/kame/pim6sd/BUGS.V6
- write down ToDos suggested in the above patch.
Tue Jul 10 15:34:09 JST 2001
* netbsd: switch to NetBSD-1.5.1 from NetBSD-1.5.
Tue Jul 10 15:34:09 JST 2001
* freebsd4/sys/netinet in.c in_pcb.c in_pcb.h
* freebsd4/sys/netinet6 in6_pcb.c in6_pcb.h
* kame/sys/netinet6 in6_ifattach.c
- "When remove a network card, freebsd4 can crash" bug fixed.
When remove interface, release multicast group on interface.
if_detach() -> in_control() -> in_pcbpurgeif0()
if_detach() -> in6_ifdetach() -> in6_pcbpurgeif0()
Mon Jul 9 23:01:10 JST 2001
* kame/racoon:
- print all of rejected attributes in phase 1 proposal when no suitable
proposal are found. suggested by <>
- fixed not to increment the transform number in phase 1 proposal.
it was degraded.
Fri Jul 6 19:52:59 2001 SUMIKAWA Munechika <>
* kame/sys/netinet6/ip6_fw.c: fragmented packets processing logic
was wrong. reviewd by: ume
Fri Jul 6 15:17:25 JST 2001
* netbsd/sys/sys/systm.h: declare ovbcopy() for better code sharing.
Fri Jul 6 14:37:30 JST 2001
* freebsd4/usr.bin/v6test: added v6test to FreeBSD4-KAME
Fri Jul 6 08:33:30 JST 2001
* sys/netinet/ip6.h: IP6_EXTHDR_GET0() did not check the mbuf length,
when off == 0. noone was using this macro.
2001-07-05 JINMEI, Tatuya <>
* bsdi4/sys/net/if_tun.c (tunifioctl): returned EAFNOSUPPORT if
the SIOCSIFADDR command is issued for an unsupported address
family (i.e. all AFs except AF_INET). This helps the IPv6 layer
detect the unavailability of the interface before going further on
initialization, and, as a result, suppress unexpected warning
Wed Jul 4 16:02:26 JST 2001
* openbsd/usr.sbin/inetd/inetd.c: correct UDP source address checks.
2.9 code did not check it for IPv6 traffic. sync with openbsd.
2001-07-03 JINMEI, Tatuya <>
* kame/kame/libinet6/getaddrinfo.c: when the GAI_USE_ORDERING
environment variable is set, reorder the chain that getaddrinfo(3)
would return, based on the logic described in
draft-ietf-ipngwg-default-addr-select. With this extension,
things would be happier when
+ the underlying kernel supports IPv6,
+ "the default interface" is specified by the "ndp -I" command,
+ there is no router around the node,
+ the destination node has both AAAA (or A6) and A resource
records, and
+ the application just tries to connect to all the entries that
getaddrinfo(3) returns.
XXX: items to be considered:
- some of the logic is not implemented.
- this routine opens a socket to get the corresponding source
address for each destination candidate. This might cause
performance effect.
2001-07-03 JINMEI, Tatuya <>
* bsdi4/sys/netinet/in_proto.c (inetsw[]): corrected the sysctl
callback function for net.inet.ip6.XXX names (from icmp6_sysctl to
Mon Jul 2 20:06:28 JST 2001
* sys/netinet6/in6.c: record multicast groups joined from within the
kernel, into struct in6_ifaddr. leave from these groups accordingly
on removal of interface addresses.
* (netbsd) sys/netinet6/in6_pcb.c: remove multicast group information
from pcb, in the early stage of interface removal processing, in
in6_pcbpurgeif0(). without this change kernel may panic on pcmcia
card removal. notified by jinmei.
2001-06-29 JINMEI, Tatuya <>
* kame/sys/netinet6/ip6_forward.c (ip6_forward): do not drop the
packet (nor send an icmp6 error) on a "p2p redirect" case, unless
the packet's destination address is regarded as on-link. With
this check, we can distinguish a routing loop from a packet sent to
a nonexistent address. For the former case, we'd rather let the
packet go to the loop, and detect the loop by traceroute.
Clarification based on a recent discussion about the p2p-pingpong
Thu Jun 28 21:56:12 JST 2001
* kame/libipsec/pfkey.c:
Fixed to calculate the length of the sadb extension in the function
pfkey_send_x5(). Calling pfkey_send_spddelete2() and
pfkey_send_spdget() had a problem. reported by <>
Thu Jun 28 15:12:09 JST 2001
* sys/netkey/key.c:
the behavior of SPDUPDATE has been changed. the kernel always add
a new policy in the case of SPDUPDATE. when there is a policy to be
updated, the kernel will move the state of the policy to be dead, and
then will add new policy. hence, SPDUPDATE doesn't depend on whether
there is a SP or not.
Thu Jun 28 10:23:01 JST 2001
* netbsd/sys/netinet/udp_usrreq.c: correct UDP over IPv6 reception
when the packet is destined to a linklocal address.
Thu Jun 28 02:33:07 JST 2001
* sys/netinet6/nd6.c: refresh default router list on nd6_detach(),
only if we are an autoconfigured host. bug was that, we will lose
default route on "ifconfig gif0 destroy" even if default is not
pointing to gif0. reported by
Thu Jun 28 02:35:18 JST 2001
* kame/racoon:
- fixed to set the inbound policy in the case of "generate_policy".
- supported sadb_x_spdexpire().
above two things are from <>.
Thu Jun 28 02:33:32 JST 2001
* kame/setkey:
enabled to use a service name as a port number.
but these operation should use getaddrinfo().
Wed Jun 27 22:10:43 JST 2001
* sys/netkey/key.c:
* kame/libipsec/pfkey_dump.c:
the lifetime information of the SP entry will send to the userland
from the kernel through pfkey when of spddump. And they can be
displayed by setkey -DP.
Wed Jun 27 19:47:19 JST 2001
* sys/net/pfkeyv2.h:
* sys/netkey/key.c,key_debug.c:
* kame/libipsec/pfkey_dump.c:
* kame/setkey/
printed current sequence number of the SA. accordingly, changed
into sadb_x_sa2_sequence from sadb_x_sa2_reserved3 in the sadb_x_sa2
structure. Also the output of setkey is changed. sequence number
of the sadb is replaced to the end of the output.
Wed Jun 27 14:35:00 JST 2001
* openbsd/sys/netinet/tcp_input.c: make faithd work on openbsd.
OpenBSD 2.9 has been working okay, this is KAME/openbsd29 issue.
Mon Jun 25 16:15:06 2001 SUMIKAWA Munechika <>
* kame/sys/netinet6/ip6_fw.c: use syslog(3) interface for logging
Sat Jun 23 10:55:46 JST 2001
* kame/mdnsd: with -N flag, mdnsd will lookup name-to-address mapping
using ICMPv6 node information query. experimental.
Sat Jun 23 03:10:50 JST 2001
* sys/netinet6/ip6_output.c: disallow setsockopt(IPV6_V6ONLY)
for sockets that are already bound. per discussions on ipngwg
mailing list.
2001-06-21 Shin'ichi Fujisawa <>
* kame/sys/netinet6/natpt_dispatch.c (foundFinalPayload):
- Fix a bug that offset of the payload is calculated
incorrectly. Apply a patch from <>.
2001-06-21 JINMEI, Tatuya <>
* kame/sys/netinet6/icmp6.c (icmp6_reflect): removed ifdef'ed
blocks to keep an older rule about the size of icmp6 echo replies
specified in rfc 1885, in order to make the code simpler.
The history about the behavior was described as a comment just
before the function definition.
2001-06-20 JINMEI, Tatuya <>
* kame/kame/advapitest/sender.c:
* kame/kame/ping6/ping6.c:
removed IPV6_USE_MTU related parts.
2001-06-20 JINMEI, Tatuya <>
* kame/sys/netinet6/(several files): removed IPV6_USE_MTU related
parts. It was introduced as an experimental workaround on
2000-11-28 (see CHANGELOG.2000), but we've found we do not need
this stuff through further discussion (and implementation changes
on the path MTU discovery procedure).
2001-06-20 JINMEI, Tatuya <>
* kame/sys/netinet6/icmp6.c (icmp6_reflect): set IPV6_MINMTU to
avoid path MTU discovery for reflected packets.
This might be controversial, but I believe this makes sense.
2001-06-20 JINMEI, Tatuya <>
* freebsd4/sys/netinet/tcp_subr.c (tcp_rtlookup):
* freebsd4/sys/netinet/tcp_subr.c (tcp_rtlookup6):
made sure to use the correct sa_len for rtalloc(). sizeof(ro_dst)
is not necessarily the correct one, especially in NEW_STRUCT_ROUTE
cases for IPv6.
The previous code could turn the path MTU discovery off as a bad
effect. If you define NEW_STRUCT_ROUTE in a KAME snap
(note that GENERIC.KAME defines this option), be sure to update
the kernel.
Wed Jun 20 14:33:28 2001 SUMIKAWA Munechika <>
* kame/sys/netinet6/{in6_var.h,in6.c,natpt_rule.c,nd6_rtr.c}:
remove in6_len2mask(). this fucnction is duplicated with
Tue Jun 19 16:33:40 JST 2001
* bsdi4/sys/netinet/tcp_input.c: make faithd work on bsdi4.
reported by jinmei
2001-06-18 Shin'ichi Fujisawa <>
* kame/sys/netinet6/natpt_trans.c (translatingICMPv4To6):
- Hop limit of translated IPv6 packet uses ttl of original
IPv4 packet, so that a hop limit is decremented in
* kame/sys/netinet6/natpt_trans.c (tr_icmp4MimicPayload):
- Correct UDP port number stored in ICMP_UNREACH packet
returned from IPv4 network. It was not enough in the change
that I put in 09 Jun 2001.
Now, KAME NAT-PT can process traceroute6 from IPv6 host to
IPv4 host.
Mon Jun 18 16:59:35 JST 2001
* sys/netinet6/icmp6.c: on icmp6 node information query (FQDN),
do not respond with hostnames with two dots (like "").
0-length labels are not distinguishable with multiple name replies.
2001-06-18 JINMEI, Tatuya <>
* kame/sys/netinet6/nd6_rtr.c (in6_tmpifadd): called
pfxlist_onlink_check() at the end of this function, to make sure
a temporary address generated from a detached public one also
detached. This is redundant when the temporary address is
generated when creating a new public address, but is essential
when the address is generated due to deprecation of an old
temporary address.
2001-06-17 JINMEI, Tatuya <>
* kame/sys/netinet6/ip6_output.c: re-enabled clarification on the
dependency between the 1st dst opt header and the routing header,
based on rfc2292bis-02.
This part was disabled when merging a recent mip6 patch from Ericsson,
without much consideration about the rationale. Actually,
disabling this part just for MIP6 convenience is a bad idea (or at
least not a good idea). We need to think carefully about a way to
make the advanced API coexist with MIP6 options.
Fri Jun 15 13:14:31 JST 2001
* openbsd/sys/net/if.[ch]: change meaning of ifnet.if_lastchange
to meet with RFC1573 ifLastChange. sync with openbsd-current.
Thu Jun 14 18:12:57 2001 SUMIKAWA Munechika <>
* kame/sys/netinet/icmp6.h:
- add new types and codes defined in RFC3122(Inverse NDP). they
will be renamed when defined in 2292bis.
- duplicated KAME local types and codes are renumberd.
note that, for the following features, we have lost interoperability
between past KAME releases and future KAME releases (because they
use unofficial numbers). if you play with these features, you must
upgrade all boxes you have.
- mtrace6(8)
- mobile-ip6 home agent discovery
- specific route information on RA, as defined in
Thu Jun 14 17:05:06 JST 2001
* netbsd/sys/net/if.[ch]: change meaning of ifnet.if_lastchange
to meet with RFC1573 ifLastChange. sync with netbsd-current.
Thu Jun 14 13:42:55 JST 2001
* bsdi3/usr.sbin/netstat/mroute6.c
+ fixed a bug that bsdi3's "netstat -gn" does not display multicast
I/F list due to a lack of NEW_STRUCT_ROUTE macro.
Wed Jun 13 22:31:49 JST 2001
* *bsd*/sys/net/if_ethersubr.c: in ether_input(), drop multicast packet
from myself, if the interface is !IFF_SIMPLEX. multicast packets
are explicitly copied via loopback interface in ip_output() and
ip6_output(), so the old codepath caused a duplicate. also the
old codepath affected IPv6 DAD. KAME PR 360
Wed Jun 13 17:29:56 2001 SUMIKAWA Munechika <>
* kame/sys/net/if_gif.c: suppressed update of if_lastchange when
processing packets on BSD/OS and FreeBSD for SNMP requirements.
2001-06-13 Shin'ichi Fujisawa <>
* sys/i386/conf/GENERIC.KAME
- Add NATPT-NAT and add some notes.
* kame/sys/netinet6/natpt_{defs.h,trans.c,tslot.c}
- Support FTP4 non-passive mode.
You can connect IPv4 FTP client to IPv4 FTP server with
passive mode or non passive mode. This is a part of
facility of NAT-PT.
If you want to use this facility (IPv4 NAT), it is necessary
to do uncomment of both following kernel options in
sys/i386/conf/GENERIC.KAME and compile/link kernel.
#options NATPT
#options NATPT_NAT
2001-06-10 Keiichi SHIMA <>
Remove MIP6 code from KAME source tree. MIP6 code is now
under reconstruction. The last KAME snap that includes MIP6 is
20010604 snap.
2001-06-09 Shin'ichi Fujisawa <>
* kame/sys/netinet6/natpt_trans.c (tr_icmp4MimicPayload):
- Correct a packet header stored in a data part of returned
ICMP packet when ICMP_UNREACH returned from IPv4 network.
* kame/sys/netinet6/natpt_{defs.h,trans.c,tslot.c}:
- Remove NATPT_TRACEROUTE flag in struct _cv{}.
Does not use NATPT_TRACEROUTE by this modification.
Fri Jun 8 08:28:56 JST 2001
* sys/net/if_stf.c: inject outgoing packet to bpf. KAME PR 358
2001-06-07 Shin'ichi Fujisawa <>
* kame/sys/netinet6/natpt_tslot.c (checkTracerouteReturn):
- Translate incoming IPv4 ICMP_UNREACH packet when this
ICMP_UNREACH returns from IPv4 network, and it is a error
return for IPv6 host. Only return of traceroute was a
target until now.
* kame/sys/netinet6/natpt_{dispatch.c,tslot.c,var.h}
- Changed routine name as follows due to modification
mentioned above.
checkTracerouteReturn() -> checkIncomingICMP()
2001-06-07 JINMEI, Tatuya <>
* kame/kame/ping6/ping6.c: stop sending echo packets whenever the
upper limit is specified by the -c option, regardless of the -f
option. Based on a comment from Tomohide Nagashima
Thu Jun 7 11:30:38 JST 2001
* kame/racoon:
fixed a segmentation fault when when racoon checks whether there is a
phase 1 sa for phase 2 sa negotiation.
Mon Jun 4 22:45:23 JST 2001
* openbsd: switch base version to 2.9.
Mon Jun 4 JST 2001
* *bsd*/sys/net/rtsock.c: adjust routing socket message length
on route_output(). previous code may send up garbage at the end of
the message.
* netbsd/sys/net/rtsock.c: check mbuf allocation failure in rt_msg1().
2001-06-03 JINMEI, Tatuya <>
* kame/kame/rtadvd/config.c (getconfig): forced users to specify
router lifetimes explicitly. "rtltimeN" must now explicitly be
2001-06-02 JINMEI, Tatuya <>
* kame/kame/ndp/ndp.c (dump): skip routes with the LINK flag and
a non AF_LINK gateway to suppress bark in getnbrinfo().
XXX: such routes should have the GATEWAY flag, not the LINK flag.
However, there is rotten routing software that advertises all
routes that have the GATEWAY flag without careful examination.
Thus, KAME kernel intentionally does not set the LINK flag. What
is to be fixed is not ndp, but such routing software (and the
kernel workaround).
2001-06-01 JINMEI, Tatuya <>
* kame/sys/netinet6/icmp6.c (icmp6_reflect): used the default
hoplimit value when rcvif is NULL. Without the change, icmp6
error packets would be sent with 0 hoplimit.
All KAME snaps after the following change should be updated:
Wed Apr 4 19:49:39 JST 2001
2001-06-01 JINMEI, Tatuya <>
* kame/sys/netinet6/nd6.c (nd6_cache_lladdr): when the link-layer
address of a router changes, select the best router again. This
is important especially when the neighbor entry is newly created,
since it might make a new default router that is probably
reachable, which affects the selection policy.
Fri Jun 1 00:44:00 JST 2001
* sys/netkey/key.c:
Fixed to make a response in key_spdadd().
reported by <>
2001-05-31 Shin'ichi Fujisawa <>
* kame/sys/netinet6/natpt_trans.c:
- translate LPRT -> PORT.
- translate PORT result(*1) to LPRT result(*2).
*1: 200 PORT command successful.
*2: 200 LPRT command successful.
You can connect IPv6 FTP client to IPv4 FTP server with
passive mode or non passive mode.
o When you use non passive mode, and IPv4 FTP server does
not use port Number 20 as a destination port, your ftp
session will fail.
* sys/netinet6/{nd6.c, nd6.h, nd6_rtr.c, mip6_md.c}
implements router-preference on host side. (not enabled
yet. plese include "option RTPREF" in your kernel to make
use of this feature.) [will commit kernel config for other OS,
when I confirm its compilability]
2001-05-31 JINMEI, Tatuya <>
* kame/kame/libinet6/getaddrinfo.c (explore_numeric): when a
numeric address is given to getaddrinfo() with the AI_CANONNAME
flag, set the name itself to each ai_canonname field of the
returned addrinfo chain.
Without the change, getaddrinfo would leave the field NULL, which
would annoy some applications.
(the change follows 2553bis-03)
2001-05-29 Shin'ichi Fujisawa <>
* kame/sys/netinet6/natpt_trans.c
- Translate LPSV to PASV.
- Translate PASV result(*1) to LPSV result(*2).
*1: 227 Entering Passive Mode (...)
*2: 228 Entering Long Passive Mode (...)
Todo: LPRT -> PORT, PORT result -> LPRT result.
2001-05-29 Shin'ichi Fujisawa <>
* kame/sys/netinet6/natpt_{defs.h,trans.c,tslot.c,var.h}
- convert FTP EPRT command to PORT.
Now, you can connect IPv6 host to IPv4 ftp server when passive
mode is off.
Mon May 28 00:24:45 JST 2001
* sys/netinet6/raw_ip6.c: declare struct rip6stat. gather statistics
on SOCK_RAW sockets (except for icmp6 sockets) for debugging aid.
req'ed by yasu.
TODO: netstat(1) support for bsdi4
Sat May 26 23:40:50 JST 2001
* freebsd4/sys/netinet6/raw_ip6.c:
Fixed IPV6_USE_MIN_MTU behavior in rip6_output().
Thu May 24 18:18:24 JST 2001
* kame/racoon:
Enabled passive mode. Racoon never initiate IKE session
if passive mode is defined in the configuration file.
Thu May 24 17:59:51 JST 2001
* sys/netkey/key.c:
Fixed to get a SA to be used when of SADB_ADD, SADB_UPDATE
and SADB_GETSPI. In the case of add, update and getspi, "reqid"
must to be take care when of looking for a SA.
Thu May 24 17:21:05 JST 2001
* sys/netinet6/in6_rmx.c (freebsd[34]): do not configure tcp
send/receive buffer size onto routing table. honor socket buffer
size. not tested.
2001-05-23 Shin'ichi Fujisawa <>
* kame/sys/netinet6/natpt_{defs.h,trans.c}
- Fix a bug in calculation of TCP ack.
- Remove unused function(s). These functions were unnecessary.
decrementSeq() and incrementAck().
Tue May 22 20:01:05 2001 SUMIKAWA Munechika <>
* freebsd4/usr.bin/telnet: make buildable with 2292bis API.
Tue May 22 17:00:41 JST 2001
* sys/netinet6/ip6_input.c: repair mbuf alignment code in
!PULLDOWN_TEST case (netbsd/openbsd are not affected).
M_COPY_PKTHDR should have been used prior to MCLGET.
2001-05-22 JINMEI, Tatuya <>
* kame/kame/rtsold/rtsold.c (rtsol_timer_update): changed the
timeout period after sending MAX_RTR_SOLICITATIONS solicitation
according to the last paragraph of RFC 2461 Section 6.3.7.
In response to: a comment from Pekka Savola <>
2001-05-21 JINMEI, Tatuya <>
* kame/sys/netinet6/udp6_output.c (udp6_output): if the
destination address is an IPv4-mapped IPv6 address which is
supposed to be sent as an IPv4 packet, call in_selectsrc to
determine the source IPv4 address. Also, some parameters to
caclulate the IPv4 checksum are corrected.
2001-05-21 JINMEI, Tatuya <>
* {kame,*bsd}/sys/netinet{,6}/many files: added complete support
of the IPV6_V6ONLY socket option;
- it now prohibits the outbound direction as well as inbound.
- it considers the cases where the option is specified for an already
bound, connected, or listening socket.
- it works for bsdi4 as well as for other BSDs (except OpenBSD).
2001-05-21 JINMEI, Tatuya <>
* kame/sys/netinet6/ip6_input.c (ip6_input): in the 'goto ours'
check, do not accept packet if the corresponding route has the
cloned bit. This fix should solve the problem that some OSes
mistakenly accept a packet to fe80::2%lo0.
XXX: the added check is not intuitive, and I'm not 100% sure this
change will never introduce another kind of problem. We might
reconsidr to solve the problem in a more explicit manner.
The fix was based on a report from Hajimu UMEMOTO
Fri May 18 17:10:10 JST 2001
* kame/racoon:
Fixed to copy "reqid" into a proposal table from policy entries.
The exchange failed when policy level "unique" was specified.
Fri May 18 14:50:56 JST 2001
* kame/setkey:
Enabled to get a protocol number from /etc/protocol as upper layer
protocol. "icmp6" and "ip4" still can be used.
2001-05-17 JINMEI, Tatuya <>
* kame/sys/netinet6/frag6.c (frag6_input): plugged memory leak in
overlapping fragments cases.
Suggested by hitachi guys via
Thu May 17 12:50:40 JST 2001
* sys/netinet6: remove OLDIP6OUTPUT codepath. notified from sumikawa
(i remember i said i would do this, a long time ago)
Wed May 16 12:04:57 JST 2001
* sys/netinet6/ah_input.c:
don't flip ip_id *back* in freebsd4.x case.
reported by <>.
Mon May 14 23:01:30 JST 2001
* sys/netinet/in_gif.c, sys/netinet6/in6_gif.c: drop IFF_LINK0
(multi destination mode) support.
Mon May 14 00:42:52 JST 2001
* kame/kame/mld6query/mld6.c: fixed a bug in argument handling.
from Mickael Hoerdt <>
2001-05-13 JINMEI, Tatuya <>
* bsdi4/INSTALL: added a note about ppp.diff
based on a comment from <>.
2001-05-10 JINMEI, Tatuya <>
* bsdi4/usr.bin/telnet/commands.c (tn): corrected a loop for
connect(2) after getaddrinfo(3).
The fix was from <>
Wed May 9 20:15:22 JST 2001
* ping6/ping6.c: correct signal handling with
"ping6 -f <nonexisting peer>". from
2001-05-07 JINMEI, Tatuya <>
* kame/kame/bindtest: added supplement tests to see which sockets
receive which packets.
Two new option -1 and -2 were added to support TCP cases of the
above test.
Another new option -6 was also added to support the IPV6_V6ONLY
option if available.
* kame/kame/bindtest/ added new tests based on the above
2001-05-07 Shin'ichi Fujisawa <>
* sys/netinet6/natpt_{defs.h,trans.c}
Support EPSV command/response pair to/from PASV command/respons
translation. You can use ftp from IPv6 client to IPv4 ftp
server, if it is passive mode.
2001-05-05 Shin'ichi Fujisawa <>
* sys/netinet6/natpt_{log,trans}.c
Remove warning in compilation.
2001-05-05 Shin'ichi Fujisawa <>
* sys/netinet6/natpt_*.[ch]
* kame/natptconfig/*.[chly]
- Remove the code which considered FAITH.
This facility is not thought about well. If there is need, we had
better reconsider this.
Wed May 2 16:30:34 JST 2001
* kame/racoon:
- logged the openssl version when racoon starts.
- fixed to check whether to process the pfkey message received.
moved the routine into each pfkey handler.
2001-05-01 JINMEI, Tatuya <>
* freebsd4/sys/net/if.c (if_detach): reversed the order of
in6_ifdetach and rnh_walktree(if_rtdel). Since IPv6 interface direct
routes are expected to be removed by the IPv6-specific kernel API,
the old order would cause some inconsistency between the routing
entries and IPv6 specific data structures.
Wed May 2 02:01:25 JST 2001
* sys/netkey/key.c: correct varaible initialization in inbound
tunnel policy checking. From: Gunther Schadow
Tue May 1 16:49:24 JST 2001
* sys/netinet6/ip6_forward.c: turn on PROHIBIT_P2PREDIRECT by default,
based on discussion on ipngwg mailing list. see changelog on
sep 12 2000 by
Sun Apr 29 12:25:48 JST 2001
* sys/netinet6/raw_ip6.c (except freebsd[34]): plug mbuf leak on
Sat Apr 28 00:09:10 JST 2001
* sys/netinet6/nd6.c (freebsd): repair backward binary compatibility
breakage for SIOCGIFINFO_IN6.
see changelog on Sat Feb 17 01:51:43 JST 2001.
if you are on freebsd systems please make sure to recompile
/usr/local/v6/sbin/ndp with the new header files.
/usr/sbin/ndp should work fine with the latest kernel.
Fri Apr 27 17:53:54 JST 2001
* sys/netinet6/in6.c:in6_ifinit(): make it work with stf interface.
initialize ia_addr before if_ioctl(SIOCSIFADDR), to meet historical
practice in in_ifinit().
* sys/net/if_stf.c: pickup a correct outer IPv4 destination address,
even if the gateway portion on the routing table is set to non-6to4
address, like:
# route add -inet6 2002:: -prefixlen 16 ::1 -ifp stf0
* netbsd/sbin/ifconfig/ifconfig.c: allow "deprecated" flag to be
set by user. it is useful to configure outgoing-only stf interface,
# ifconfig stf0 inet6 2002:d2a0:5f68::1 prefixlen 16 alias deprecated
* sys/net/if_stf.c: IFF_LINK0 disables the input path.
2001-04-27 JINMEI, Tatuya <>
* kame/sys/netinet6/nd6.c (nd6_rtrequest): stop process if
nd6_need_cache() is 0 and the route is not a host route (which
should probably be an interface direct route on a link that does
not need neighbor cahces). Otherwise, the function would set the
RTF_LLINFO flag, which would annoy the ndp(8) command.
* kame/sys/netinet6/nd6_rtr.c (nd6_prefix_onlink): did not set
the RTF_GATEWAY flag even for !nd6_need_cache() cases. The
intention was to prevent the process of nd6_rtrequest(), but we do
not need the flag, since we now explicitly disable the process.
Setting the RTF_GATEWAY flag is not the best way, because routes
with the flag can mistakenly be deleted by user applications.
Tue Apr 24 15:29:51 2001 SUMIKAWA Munechika <>
* freebsd4: sync with 4.3-RELEASE
Fri Apr 20 23:59:21 JST 2001
* sys/netinet6/in6_pcb.c: (netbsd): fix #if clause with NOIPPRIVPORTS.
the old code mistakenly allowed bind(2) from non-privileged user
onto privilege ports (netbsd integrated trees, like plain NetBSD 1.5
are not affected). from k-sugyou.
Thu Apr 19 18:26:49 JST 2001
* sys/netinet6/ah_input.c:
Fixed to mismatch AH checksum on FreeBSD4. ip->ip_id was reversed
before calculating IPv4 ah checksum, but this operation is not
necessary for FreeBSD4. Only FreeBSD[23], openbsd and bsdi3 need.
The problem doesn't appear on the FreeBSD original tree.
2001-04-18 JINMEI, Tatuya <>
* kame/kame/pim6[ds]d/mld6_proto.c: clarifications and cleanups;
- removed an incorrect check in accept_listener_query(), which was
rather harmful.
- ignore queries if the receiving node is the querier.
- removed a redundant check in accept_listener_query().
- removed the al_old member of the listaddr structure, and all
references to the member. This should be safe, because there
is effectively no use of the member. This change should also
make the code more conformant to RFC 2710.
All the changes were based on comments from Mickael Hoerdt
Sun Apr 15 14:38:58 JST 2001
* sys/netinet6/ipsec.c: on IPsec tunnel mode encapsulation, do not
copy TTL (or hop limit) value from inner to outer IP header.
Wed Apr 11 18:52:26 JST 2001
* racoon:
Supported to get a certificate from DNS CERT RR.
Also getcertsbyname() is implemented In order to get CERT RRs.
This function can use lwres.a if HAVE_LWRES is defined when racoon
is compiled.
XXX need more local test and interoperability test.
XXX should be arranged too many certificate stuff in racoon.conf.
2001-04-11 JINMEI, Tatuya <>
* kame/sys/netinet6/ip6_output.c (ip6_pcbopt):
* kame/sys/netinet6/ip6_output.c (ip6_setpktoptions):
prevented invalid (anycast or unready) addresses from being
specified as the packet's source address using the IPV6_PKTINFO
socket option or ancillary data.
2001-04-10 Jason R. Thorpe <>
* racoon/pfkey.c: pk_recvacquire(): Make sure the phase1
and phase2 handlers are unbound before the phase 2 handler
is deleted.
* racoon/isakmp.c: ph1_main(), quick_main(): Add the message
to the received-list before processing to ensure the packet
isn't processed twice in case of an error.
isakmp_post_acquire(): Don't unbind the phase1/phase2 handlers;
let the caller do it.
isakmp_newcookie(): Plug memory leaks.
From George Yang <>.
* racoon/ipsec_doi.c: get_ph2approvalx(): When we find a
matching saprop, make sure to flushsaprop(pr0), as the returned
saprop is a copy. Fixes a memory leak.
From George Yang <>.
* racoon/isakmp_quick.c: quick_r2send(): Make sure to vfree(data)
if we fail to allocate a new body. Fixes a memory leak.
From George Yang <>.
Tue Apr 10 22:51:26 JST 2001
* rtadvd/config.c, rtadvd/rtadvd.h, rtadvd/rtadvd.conf.5
you can advertise route information option as stated in
* sys/netinet/icmp6.h (nd_opt_route_info, ND_OPT_ROUTE_INFO)
* kame/sys/netinet/icmp6.h added a macro and a structure
for route information option.
Note that these are still non-standard.
Note: before compiling the latest rtadvd, you have to install the
header file, and (probably) perform "make clean".
2001-04-07 JINMEI, Tatuya <>
* bsdi4/sys/netinet/in_proto.c (inetsw[]): set ipsec_sysctl
correctly. Without this, "netstat -p ipsec -s" does not work.
2001-04-07 JINMEI, Tatuya <>
* bsdi4/sys/netinet/tcp_input.c (tcp_peer_mss): avoided
IPv6 fragmentation when IPV6_USE_MIN_MTU is required.
Fri Apr 6 23:25:19 JST 2001
* racoon:
implemented to generate the policy in the responder side automatically.
If the responder does not have any policy in SPD during phase 2
negotiation, and the directive is set on, then racoon will choice
the first proposal in the SA payload from the initiator, and generate
policy entries from the proposal. This function is for the responder,
and ignored in the initiator case.
XXX should be checked tunnel mode case.
2001-04-04 Jason R. Thorpe <>
* racoon: Add support for the Dmalloc debugging malloc
library. This library gives very nice memory usage
statistics and leak information.
Thu Apr 5 03:42:24 JST 2001
* kame/v6test/getconfig.c: v6test supports arbitrary raw packet
advertisement, such as OSPFv3, Tunnelled packet etc.
Wed Apr 4 22:47:27 JST 2001
* racoon:
support scopeid. base code was from <>.
it should be considered more.
Wed Apr 4 19:49:39 JST 2001
* sys/netinet6/icmp6.c: make sure we do not pass mbuf with
bogus m->m_pkthdr.rcvif, to icmp6_reflect(), on icmp6 error
Wed Apr 4 13:17:00 JST 2001
* kame/v6test/getconfig.c
fixed a bug that you cannot specify an optional value if its
MSB is on.
2001-04-03 Jason R. Thorpe <>
* racoon: Better integration of debugging malloc libraries.
Use wrapper macros (racoon_{malloc,calloc,free,realloc}())
so that debugging malloc implementations can get file/line
info, and also put traditional malloc/calloc/free/realloc
stubs in the main program so that libraries linked with
racoon get the debugging allocators, as well.
Mon Apr 3 JST 2001
* sys/netinet6/tcp6_output.c (freebsd2/bsdi3),
sys/netinet/tcp_output.c (netbsd/openbsd/bsdi4):
support IPV6_USE_MIN_MTU setsockopt on IPv6 tcp.
- pass IPV6_MINMTU down to ip6_output,
- on outgoing segment packing use 1280 (IPV6_MMTU) as packet size,
instead of MSS heard from the peer (except bsdi4/openbsd).
design memo:
do we need to play with MSS advertisement to the peer? I don't
think so, since:
- IPV6_MIN_MTU controls outbound traffic only
- outgoing MSS advertisement will control the inbound segment size.
though i agree that it is be useful to decrase MSS advertisement
to trick inbound PMTUD behavior, i believe it is a separate topic
from IPV6_MIN_MTU.
TODO: generate packets fit into 1280 bytes for bsdi4/openbsd
2001-04-02 JINMEI, Tatuya <>
* bsdi3/sys/net/if.c (ifa_ifwithnet): disabled the check for
IFF_POINTOPOINT|IFF_LOOPBACK in ifa_ifwithnet(). The check for
IFF_POINTOPOINT was already disabled, but the check for
IFF_LOOPBACK should be disabled as well, in order to install
fe80::%lo0/64 correctly.
2001-04-01 JINMEI, Tatuya <>
* kame/sys/netinet6/ip6_input.c (ip6_input): clarified goto-ours
1. separated checks against spoofed ::1 src/dst from the goto-ours
check. This also fixed a bug that the kernel accepted a packet with
src=::1, dst=invalid (not assigned), rcvif=lo0
(you can test it by 'ping6 -S ::1 fe80::xxxx%lo0", where xxxx is
not an interface ID of lo0)
2. omitted a special case for link-local destinations on a
loopback interface, because
- we now have a host route for fe80::1%lo0, so we can accept a
packet to the address using the generic logic.
- we can reject packets to fe80::xxxx%lo0 (xxxx != 1) by the check
for the RTF_GATEWAY bit for rt_flags (ip6_input.c line 872).
Fri Mar 30 10:46:00 JST 2001
* sys/netinet6/nd6_rtr.c: repair inbound RA processing. broken
yesterday (with mobile-ip6 commit).
2001-03-29 JINMEI, Tatuya <>
* kame/sys/netinet6/in6.h (IN6_IS_ADDR_xxx): made the macros safe
to the gcc's -Wcast-qual option.
Based on a comment from Brian Wellington
Thu Mar 29 16:31:34 JST 2001
* sys/netinet6/mip6*: bring in latest ericsson mobile-ip6 code.
based on revision 13 of the mobile-ip6 draft. see
for details.
2001-03-26 Jason R. Thorpe <>
* racoon/isakmp_ident.c: ident_ir2sendmx(): plug memory
leak -- gsstoken wasn't being freed at function exit.
2001-03-26 Jason R. Thorpe <>
* racoon: Changes to Vendor ID payload handling. Determine
which VID we will send on a per-proposal basis; we may need
to send a different one for each proposal depending on the
proposal contents (e.g. GSSAPI auth method). We no longer
set the Vendor ID in the localconf.
When matching the Vendor ID in check_vendorid(), use a table
of known Vendor IDs, and return the index, and maintain a list
of extensions that vendors implement (e.g. GSSAPI auth method).
XXX We have a slight hack to recognize the Windows 2000 Vendor
ID. Need to clarify with the Microsoft IPsec guys.
In Aggressive Mode, as responder, when sending first
response, make sure to include a Vendor ID payload.
In Main Mode, as responder, when sending first response,
make sure to include a Vendor ID payload.
XXX Still more Vendor ID processing fixes to go. And
GSSAPI auth doesn't interoperate with Windows 2000 yet.
Sun Mar 25 18:11:24 JST 2001
* sys/netinet6/{ip6_mroute,in6_prefix}.c: add missing splx.
From: (Constantine Sapuntzakis)
* sys/netinet6/ip6_output.c: correct dangling pointer in jumbogram
output logic.
From: (Constantine Sapuntzakis)
2001-03-23 Shin'ichi Fujisawa <>
* kame/sys/netinet6/natpt_*.[ch]
- Change MALLOC type M_TEMP to M_NATPT.
This causes serious memory leak in FreeBSD4.2 when leave M_TEMP.
2001-03-23 JINMEI, Tatuya <>
* kame/sys/netinet6/icmp6.c (ni6_addrs, ni6_store_addrs):
If the 3rd bit of the icmp6_nodeinfo sysctl variable is clear,
- do not respond to node info FQDN to an RFC3041 temporary address.
- do not include temporary addresses in a node info Node Addresses
This bit is clear by default based on privacy consideration.
Thu Mar 22 08:06:30 JST 2001
* racoon:
fixed to parse modp1536 of DH group. reported by <>
2001-03-22 JINMEI, Tatuya <>
* kame/sys/netinet6/nd6.c (nd6_cache_lladdr): set nd6_gctimer to
ln_expire just after the state transition to STALE. This change
fixed a bug that a longer timer value was mistakenly set, after
transition to the delay ND state. The bug could delay NUD, but it
does not happen in a normal operation. Thus, you do not
necessarily have to update the kernel in a hurry.
The bug was introduced around Jan 20th, 2001, and was found by a
recent TAHI conformance check.
Thu Mar 22 04:56:57 JST 2001
* racoon/policy.c:
fixed to compare between policies when the responder decides to
accept the proposal or not. the upper layer protocol is represented
by 0 in ID payload.
Thu Mar 22 04:15:43 JST 2001
* sys/netinet6/ip6_input.c (netbsd): inject packet to ipfilter
only if it is wire format (not if it went through ipsec tunnel).
2001-03-22 JINMEI, Tatuya <>
* kame/kame/rtadvd/config.c (getconfig): allow router preference
values to be specified in the "raflags" directive.
* kame/sys/netinet/icmp6.h (ND_RA_RTPREF_xxx): added definitions
for router preferences. Note that these are non-standard.
Note: before compiling the latest rtadvd, you have to install the
header file, and (probably) perform "make clean".
Thu Mar 22 01:45:32 JST 2001
* racoon:
fixed potencial of a buffer overrun when adding a ID payload to
the ISAKMP payload. It happened when policy is both to use IPSec
transport mode and not to specify a transport protocol.
reported by <>.
Thu Mar 22 00:26:51 JST 2001
* sys/netinet6/icmp6.c: update MTU on path MTU timeout.
noted by
* ndp/ndp.c: do not dereference null pointer. from tomomi suzuki
2001-03-20 Shin'ichi Fujisawa <>
* sys/netinet6/natpt_trans.c: Put IPPROTO_UDP into IPv6 header.
Because conversion of IP header part is shared with TCP, default
protocol was set as TCP.
* sys/netinet6/natpt_trans.c: Calculate UDP checksum which was
converted to IPv6. I forgot to re-calculate it.
Tue Mar 20 11:56:08 JST 2001
* sys/netinet6/icmp6.c: change interpretation of
net.inet6.icmp6.nodeinfo from true/false to bitmap.
2^0 (= 1) bit: respond/ignore FQDN query (ping6 -w)
2^1 (= 2) bit: respond/ignore NODEINFO query (ping6 -a)
2001-03-19 Shin'ichi Fujisawa <>
* Add initial version of NAT-PT plan.
2001-03-18 Shin'ichi Fujisawa <>
* freebsd4/sys/netinet/ip_input.c
- Move NATPT hook to just before check to see if the packet is
for us. Because NATPT need one more extra IPv4 address to
translate IPv6 to IPv4 when this NATPT hook was put after this
- Change indent to an 8 character tab.
* kame/sys/netinet6/natpt_dispatch.c
- Add IPv6 outbound packet counter.
- Fix bug of comparison that size of packet exceeds mtu.
* kame/sys/netinet6/natpt_trans.c
- Fix a bug to clear UDP header when calculate UDP checksum.
* kame/sys/netinet6/natpt_{defs.h,dispatch.c,tslot.c}
- Remove code evading a bug of SuMiRe NAT.
SuMiRe NAT is obsolete.
Thu Mar 15 20:39:03 JST 2001
* racoon:
- fixed a phase 2 handler deletion. racoon will delete a phase2
handler immediately when hard lifetime expires.
- check a unit of the timer in the configuration file.
Thu Mar 15 17:48:54 JST 2001
* sys/netinet/ip_mroute.c: use sys/netinet/ip_encap.c framework
for inbound packet dispatch.
2001-03-15 JINMEI, Tatuya <>
* kame/sys/netinet6/in6.c (in6_ifinit): always set nd6_rtrequest
to ifa_rtrequest.
* kame/sys/netinet6/nd6_rtr.c (nd6_prefix_onlink): because of the
above change, the function does not set the ifa_rtrequest
* kame/sys/netinet6/in6.c (in6_ifloop_request): set the address
itself as gateway, and set the corresponding host route to the
RTF_LLINFO, so that the route would have the flag, and thus
applications (e.g. routing daemons) that assume traditional kernel
behavior would be happy. Older versions made the route to the
node's own address like this:
2001:200::3ca2:ffef:eff5:f9fd ::1 UH lo0
However, since some routing daemons try to install kernel internal
routes that do not have the RTF_LLINFO flag, this kind of entry
could cause unintentional host routes propagated. The new kernel,
instead, installs
2001:200::c049:d099:ab4b:b637 0:a0:12:34:ab:cd UHL lo0
just like far older versions of the kernel (except for the
existance of the cloned bit), which installed the host route as a
cloned route from the corresponding interface direct route.
2001-03-13 JINMEI, Tatuya <>
* freebsd2/ports/wu-ftpd: upgraded to wu-ftpd 2.6.1. Since there
are security holes in older versions, upgrade is recommended.
Mon Mar 12 20:17:43 JST 2001
* sys/crypto/sha2/sha2.c: hmac-sha2-{256,384,512} support. attaches
96 bits of crypto checksum (not sure if this is right - there's no
draft on this).
TODO: interop tests
2001-03-11 JINMEI, Tatuya <>
* kame/sys/netinet6/ip6_output.c (ip6_output): added a couple of
missing splx() for OpenBSD IPsec.
2001-03-08 Atsushi Onoe <>
* kame/route6d.c: correct deleting host route, based on
report from
2001-03-06 Jason R. Thorpe <>
* kame/racoon/schedule.c: Implement sched_scrub_param(),
which kills all scheduler work queue entries which a
specified parameter.
* kame/racoon/handler.c: Use sched_scrub_param() to make
sure no references to a handler exist when it is freed.
2001-03-06 JINMEI, Tatuya <>
* kame/sys/netinet6/nd6_rtr.c (nd6_prefix_onlink): set RTF_GATEWAY
to an interface direct route when nd6_need_cache() is false, in
order to prevent nd6_rtrequest() from setting RTF_LLINFO (which
annoys the ndp(8) command).
This change is based on a report from nobumichi ozoe
Tue Mar 6 09:22:56 JST 2001
* sys/netinet6/raw_ip6.c: permit IPV6_CHECKSUM socket option for
the following cases only (previously it was allowed for any AF_INET6
socket): raw ip6 socket, and protocol != IPPROTO_ICMPV6.
RFC2292 section 3.1. commented by yoshfuji.
2001-03-05 Jason R. Thorpe <>
* kame/racoon/gssapi.c: Use GSS_C_MECH_CODE when reporting
GSSAPI errors.
2001-03-05 Jason R. Thorpe <>
* kame/racoon/handler.c: Implement deleteallph2(), which
deletes all Phase 2 handlers for a given src/dst/proto.
* kame/racoon/isakmp_inf.c: When processing INITIAL-CONTACT,
try to use the SADB_DELETE `delete all' extension and
deleteallph2() before doing it The Hard Way. For both The
Easy Way and The Hard Way, make sure we only delete SAD entries
for SATYPEs that we manage.
* kame/racoon/pfkey.c: Use a table of SATYPEs that we manage,
and use that table to initialize our PF_KEY state.
2001-03-05 Jason R. Thorpe <>
* kame/libipsec/pfkey.c: Add pfkey_send_delete_all(), which
sends an SADB_DELETE with satype/mode/src/dst.
Mon Mar 5 JST 2001
* sys/netinet6/raw_ip6.c: avoid 1-byte mbuf buffer overrun, and/or
unaligned memory access, in rip6_output(). commented by yoshfuji.
it will be tickled only if a evil/buggy root process runs.
2001-03-04 JINMEI, Tatuya <>
* kame/sys/netinet6/nd6.c (nd6_free): moved a line calling
pfxlist_onlink_check() before the default router selection part,
based on a request from Mattias Pettersson
2001-03-03 Shin'ichi Fujisawa <>
* kame/pma/*: removed.
This function is obsolete.
2001-03-03 Shin'ichi Fujisawa <>
* sys/netinet6/natpt_usrreq.c (natpt_uattach) call natpt_attach if
privileged process.
* sys/netinet6/natpt_rule.c (natpt_in4_len2mask): convert host
byte order of calculated mask.
* sys/conf/options (freebsd4): add NATPT_NAT
* kame/natptconfig/show.c: Fix struct nlist symbol name. Change
string compare method, check string length in the first place.
TODO: It is not so good to read '/dev/kmem', I need to think about
another method to read these variables from kernel.
2001-03-02 JINMEI, Tatuya <>
* kame/sys/netinet6/nd6_rtr.c (prelist_update): clarified the
update routine for an existing prefix. It now tries to update
some parameters even if the L bit is zero. This would be
necessary for the case where a router first advertises the prefix
without the L bit being set, and next the same prefix with the bit
being set.
Fri Mar 2 20:05:23 2001 SUMIKAWA Munechika <>
* netbsd/usr.bin/tftp/main.c: port checking fix.
Fri Mar 2 16:11:17 JST 2001
* sys/crypto/{twofish,rijndael}: change key initialization API to
pass binary keys instead of hexadecimal string keys.
Fri Mar 2 01:48:24 JST 2001
* sys/netinet/ip_input.c (all *bsd): drop packets from outside
if these contain 127/8 in ip_src or ip_dst.
Thu Mar 1 20:01:14 JST 2001
* bsdi3/sys/netinet/raw_ip.c
fixed a bug that bsdi3 KAME with IPsec crashes when it receives
an ICMP packet
Thu Mar 1 18:28:48 JST 2001
* sys/netinet/ip_input.c, sys/netinet6/ip6_input.c (except openbsd):
enforce inbound policy match on all final protocol headers.
previously, inbound policy is enforced only for udp/tcp/icmp.
basically the fix is for transport mode case.
XXX ipsec policy engine still needs a great amount of updates.
* freebsd4/sys/netinet/in_proto.c: add ipcomp input support.
Wed Feb 28 23:26:26 JST 2001
* sys/netinet6/esp_output.c: support random length padding in ESP,
to avoid traffic analysis on short packets. 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.
recommeded value for N is like 128, or 256 (if you use a too big
number as N, you may experience inefficiency due to fragmented
Mon Feb 26 22:53:02 2001 SUMIKAWA Munechika <>
* sys/netinet6/nd6.c : do not overwrite llcache if static ndp is
Mon Feb 26 18:21:08 JST 2001
* sys/netinet6/ip6_{in,out}put.c: support 2553bis-03 IPV6_V6ONLY
socket option (not finished yet).
(1) supply net.inet6.ip6.v6only on all platforms. it shows the
kernel default behavior for IPV6_V6ONLY socket option. mutable
on netbsd/freebsd3/freebsd4, immutable on other platforms.
(2) support IPV6_V6ONLY on all platforms. on netbsd/freebsd3/
freebsd4, the socket option is mutable and has per-socket state.
on other platforms, we can only read it (you can set it if the
value is same as the default - non-changeable - behavior).
TODO: synchronize AF_INET6 socket behavior with 2553bis-03, from
IPV6_V6ONLY POV (at least netbsd does not meet 2553bis-03, as the
socket option controls inbound only, should control both inbound
and outbound). *bsd have their own preference on kernel default
behavior due to security risks, so will change that.
NOTE: netbsd IPV6_BINDV6ONLY is now obsoleted. old binaries should
just work (IPV6_V6ONLY uses same socket option # as IPV6_BINDV6ONLY).
NOTE: freebsd net.inet6.ip6.mapped_addr is obsoleted.
Mon Feb 26 15:36:51 JST 2001
* sys/netinet*/raw*.c: validate inbound IPsec policy on raw socket
processing. from:
2001-02-25 JINMEI, Tatuya <>
* kame/sys/netinet6/nd6.c (nd6_need_cache): separated from
nd6_output() to decide whether to make a neighbor cache entry for
the link of a given interface. This function is (currently)
called from nd6_output() and nd6_prefix_onlink(), and make the
decision policy consistent.
Based on the current policy, we do not make a neighbor cache for a
loopback link, in response to a comment from Miyata-san@YDC.
Thu Feb 22 10:08:27 JST 2001
* racoon:
fixed to check the outbound policy when the responder received the
1st packet in phase 2. the tunnel mode and the transport specified
the pair of IP addresses of the end of the SA had failed.
Thu Feb 22 01:59:01 JST 2001
* sys/netinet6/{dest6,nd6}.c: make variable length header parsing
pickier (= safer).
* sys/netinet6/frag6.c: fix behavior on memory shortage/too many
reass queues.
* openbsd/include/netdb.h: change ai_addrlen from int to socklen_t,
to conform to 2553bis-03.
Wed Feb 21 14:25:55 JST 2001
* racoon:
changed the proposal order of the protocol in the phase 2.
If we want to make a packet "IP2 AH ESP IP1 ULP", the SPD in KAME
expresses AH transport + ESP tunnel. racoon sent the proposal
contained such the order. But lots of implementation interprets
AH tunnel + ESP tunnel in this case. racoon changes the order,
and usually uses this format. If the option, 'complex_bundle'
is enable, racoon uses old format.
Wed Feb 21 13:04:17 JST 2001
* openbsd/usr.sbin/pim6{sd,dd}: compile these on openbsd.
not really tested.
Wed Feb 21 13:00:00 JST 2001
* sys/netinet6/ah_{core,output}.c: one more correction to IPv4 option
chasing in AH processing.
Mon Feb 19 20:51:41 JST 2001
* sys/netinet6/ipsec.c (netbsd only): correct severe memory leak
in ipsec operation. the bug was introduced on 2/14/2001.
Mon Feb 19 15:03:24 JST 2001
* sys/netinet6/ah_{core,output}.c: correct IPv4 option chasing in
AH processing. the change is important if you use IPsec.
Mon Feb 19 11:49:29 JST 2001
* netbsd/usr.sbin/pppd: sync with 1.5 tree.
2001-02-19 Shin'ichi Fujisawa <>
* freebsd4/sys/netinet/ip_input.c: Add a NATPT hook into ip_input().
* freebsd4/sys/netinet/ip_proto.c: Add a NATPT entry into struct
ipprotosw inetsw[].
* sys/netinet6/ip6_input.c: include "opt_natpt.h" when FreeBSD
version is more than 3.
Sorry, there is a problem a little, and NATPT does not work yet.
2001-02-16 Jason R. Thorpe <>
* sys/netkey/key.c: When processing an SADB_DELETE message,
allow SADB_EXT_SA to be blank. In this case, we delete
all non-LARVAL SAs that match the src/dst/protocol. This
is particularly useful in IKE INITIAL-CONTACT processing.
Idea from Bill Sommerfeld <>, who
implemented it in post-Solaris8.
* kame/setkey/parse.y, kame/setkey/token.l,
kame/setkey/setkey.8: Add a "deleteall" command that takes
a src/dst/protocol.
Sat Feb 17 01:51:43 JST 2001
* sys/netinet6/nd6.c: recover backward binary compatibility for
* openbsd/sys/netinet/udp_usrreq.c: remove IPv6 cases from udp_output()
to simplify it and make it easier to audit it (sync with openbsd-
Wed Feb 14 17:37:34 JST 2001
* sys/kern/uipc_mbuf2.c: make sure to return non-shared mbuf on
calls to m_pulldown().
Wed Feb 14 16:27:49 JST 2001
* sys/netinet6/ip6_input.c: merge mbuf data into single mbuf, if
mbuf chain is passed to ip6_input. it makes IPv6 packet processed
okay when L2 code passes non-continuous mbuf data (like L2 bridge
code). affects freebsd[234] and bsdi[34] only.
Tue Feb 13 18:42:17 2001 SUMIKAWA Munechika <>
* freebsd4/sys/{if_sppp.h,if_spppsubr.c}: supported PPPv6 for
kernel level PPP.
2001-02-11 JINMEI, Tatuya <>
* kame/sys/netinet6/in6.c (in6if_do_dad): was added to centralize
the decision of whether DAD should be performed on a particular
interface. This function is currently called from
in6_update_ifa() and in6_ifattach_linklocal().
2001-02-09 JINMEI, Tatuya <>
* kame/sys/netinet6/in6_ifattach.c (in6_ifattach): skipped
assignment of link-local addresses when interfaces become up,
if ip6_auto_linklocal is 0 (its default is 1). By setting this
variable to 0 and manually configuring link-local addresses, you
can use well-known interface identifiers for servers/router/etc.
This variable can be configured via sysctl
net.inet6.ip6.auto_linklocal or by the IP6_AUTO_LINKLOCAL kernel
compilation option. If you configure the variable via sysctl, you
should be sure that no interface has been up.
Note that the variable effects all interfaces. When you set the
variable to 0, you should configure link-local addresses for all
interfaces that you might want to use.
Fri Feb 9 00:51:01 JST 2001
* sys/netinet6/in6.h: correct SA6_ARE_ADDR_EQUAL definition.
kernel responds to ping6 -w again.
Fri Feb 9 00:18:46 JST 2001
* sys/netinet6/icmp6.c: negative {mtudisc,redirect}_{hi,lo}wat
value will turn off the checks. See changelog on 2000/12/09.
2001-02-08 JINMEI, Tatuya <>
* kame/sys/netinet6/nd6.c (nd6_free): returned the pointer to the
next entry of the freed one. Since pfxlist_onlink_check() called
from this function could change the link structure, the caller
would need the "real" successor of the list to avoid to refer to
invalid memory. This fix actually fixed a problem that the kernel
hunged in some (rare) cases.
Wed Feb 7 21:35:26 JST 2001
* sys/kern/uipc_socket.c (netbsd/openbsd/freebsd4/bsdi4):
return ECONNABORTED, if the socket (tcp connection for example)
is disconnected by RST right before accept(2). fixes PR 10698/12027.
checked with SUSv2, XNET 5.2, and Stevens (unix network programming
vol 1 2nd ed) section 5.11.
bsdi3/freebsd[23] needs some other fix here, so they are left behind.
NOTE: unix domain socket behavior needs checking.
2001-02-07 JINMEI, Tatuya <>
* bsdi4/usr.sbin/netstat/route.c (p_sockaddr): removed special
cases of ::/plen (plen > 0). I believe it is better than the
original behavior, which printed those routes "v6-in-v4 default".
Wed Feb 7 16:19:19 JST 2001
* sys/netinet6/raw_ip6.c (netbsd/openbsd): validate/notify path MTU
discovery correctly for raw ip6 sockets.
note: since ping6 does not use connect(2) for the socket for
outgoing traffic, ping6 will have issues with route pointer kept
in in6pcb - need to restart ping6 to reflect new pmtu.
Wed Feb 7 01:08:46 JST 2001
* netbsd/sys/net/route.c: ignore redirect attempt, if we are to
create/update routing entry with the same value in rt_key and
rt_gateway. response to NetBSD PR 4827. experimental.
2001-02-06 JINMEI, Tatuya <>
* kame/sys/netinet6/nd6.c (nd6_p2p_rtrequest): removed. It was
introduced long long ago, and is not necessary any more,
especially after the recent clarification on the address/prefix
Tue Feb 6 13:08:24 JST 2001
* sys/netinet6/nd6*.c: minimize number of log() or printf() on inbound
packet processing path, to avoid /var from get filled up with with
bogus packet storms.
* sys/netinet6/icmp6.c: supply new sysctl net.inet6.icmp6.nd6_debug,
to turn on/off error/warning messages on inbonud ND/ICMPv6 packets.
disabled by default, can be enabled by default if you have
"options ND6_DEBUG".
* sys/netinet6: OLD_LOOPBACK_IF is no longer supported (see changelog
on 2000/7/30).
2001-02-06 JINMEI, Tatuya <>
* kame/sys/netinet6/in6.c (in6_update_ifa): always set the
destination address when specified regardless of the type of the
interface. This would solve a problem that bsdi3 could not add a
route that has ::1 as the gateway (bsdi3 specific, maybe).
This fix was in response to a report from Tomomi Suzuki
Tue Feb 6 10:49:36 JST 2001
* sys/netinet6/ip6_output.c: correct m->m_pkthdr.rcvif setting for
own linklocal address cases. broken around Feb 2.
2001-02-05 JINMEI, Tatuya <>
* *BSD/sbin/ifconfig/ifconfig.c: marked "temporary" for temporary
IPv6 addresses (see the next log).
2001-02-05 JINMEI, Tatuya <>
* kame/sys/netinet6/: implemented RFC 3041 Privacy Extensions for
Stateless Address Autoconfiguration. You can enable this by
# sysctl -w net.inet6.ip6.use_tempaddr=1
before accepting an RA. Once enabled, the kernel will configure
temporary addresses as well as public autoconfigured ones, as
described in the RFC. Also, the kernel will prefer temporary
addresses to public autoconfigured ones as the source address for
a new communication. For more details about the selection
algorithm, see 1.6 of the KAME IMPLEMENTATION file.
2001-02-04 JINMEI, Tatuya <>
* kame/sys/netinet6 (bunch of files): completely revisited address
and prefix manipulation:
- centralize addition or removal of addresses to make code
maintainance easier.
- separate address manipulation and prefix manipulation, including
+ completely separate address lifetimes and prefix lifetimes.
+ separate on-link determination from address autoconfiguration.
+ separate route for a node's own address (to a loopback
interface) from an interface's direct route.
+ separate the notion of detached addresses and the notion of
detached prefixes. The latter can now be designated by the
"D" bit of the output from "ndp -p".
- more conformance to RFC 2462 address autoconfiguration.
- clarify a notion of autoconfigured addresses.
Note: the change is so drastic, so you will probably have to clean
all your source tree up, and then rebuild the entire kernel and
2001-02-04 JINMEI, Tatuya <>
* kame/sys/netinet6/in6.h (IFA6_IS_{DEPRECATED, INVALID}): added
to hide implementation of address lifetimes. As a middle term
solution, we should clarify the relationship among ia6t_xxx
members and the IN6_IFF_DEPRECATED flag. When you add code about
address lifetimes, do not assume its implementation, and stick to
use these macros.
2001-02-04 JINMEI, Tatuya <>
* kame/kame/rtadvd/rtadvd.c (main): the -R (router renumbering)
option is currently ignored (with a warning message). We will
re-enable the option after clarifying the address/prefix
manipulation in the kernel, and implementing missing stuff within
rtadvd (hopefully soon).
2001-02-04 JINMEI, Tatuya <>
* kame/kame/prefix/ (prefix): added as a tiny replacement
of original prefix(8), based on recent clarification on the kernel
prefix/address manipulation engine. "Tiny" means that it does NOT
provide complete backward compatibility with the original one. It
can just do simple set/delete operations using ifconfig(8) as a
backend. Thus, users are rather recommended not to use the prefix
command, but should use ifconfig (8) explicitly.
Sun Feb 4 11:21:15 JST 2001
* freebsd*: do not remove prefix list on net.inet6.ip6.forwarding
change from 0 to 1. the behavior is confusing, and not friendly
with recent changes to on-link determination logic.
2001-02-01 JINMEI, Tatuya <>
* kame/kame/rtadvd/config.c (getconfig):
* kame/kame/rtadvd/advcap.c (agetnum):
use a long long integer to parse numeric values to unitentional
overflow for 32 bits integers. With this change, you can specify
very large value (i.e. values larger than 0x80000000) for valid
and preferred lifetimes.
XXX: we assume the type "long long" is more than 32 bits in length.
2001-02-01 JINMEI, Tatuya <>
* {bsdi[34],freebsd4,netbsd,openbsd}/sbin/ifconfig/ifconfig.c:
added a new modifier "autoconf" for IPv6 addresses.
For example, when you configure an IPv6 address like this
# ifconfig ne0 inet6 3ffe:ffff::1 prefixlen 64 autoconf
and a router advertises RA with the prefix information for
3ffe:ffff::, then the lifetimes of the address will be adjusted
according to the RA.
Wed Jan 31 17:43:23 JST 2001 sakane@ydc.cojp
* racoon:
If the file for saving SA information is specified in the
configuration file, racoon will save a SA information to the file
when negotiation will succeed. If you launch racoon with tye -B
option and if the file is specified, she will install SAs into the
kernel during intialization. racoon cleans the file when she
shutdowns normally. The function is useful when an abnormal
system shutdown happened, because SAs will remain at the peer. In
this case, you can use the -B option then SAs will revive.
XXX racoon simply adds SAs to the end of the file specified.
Tue Jan 30 05:03:01 JST 2001
* route6d: corrected RTA_NETMASK handling (do not look at sa_family).
bug report from IIJ SEIL team.
Mon Jan 29 JST 2001
* {freebsd*,openbsd}/ports/bind: disable it for security issues.
these were based on 8.1.2...
* freebsd4/include/netdb.h: do not pollute namespace by inclusion of
sys/types.h. declare _BSD_SOCKLEN_T_ in machine/ansi.h, avoid
duplicated declarations.
based on SUSv2, netdb.h should not pull in sys/types.h.
if your code become not compilable due to the change, it is because
of bug (or SUSv2 non-conformance) in your code.
(you can't assume that netdb.h would pull in sys/types.h)
Mon Jan 29 00:24:00 JST 2001
* sys/netinet/icmp6.h: synchronize some of declarations, like RA packet
header structure and value #defines with 2292bis-02.
Sun Jan 28 JST 2001
* netbsd: synchronize with netbsd-current change to cloning routes.
(1) cloning routes will be marked with RTF_CLONED. (2) RTF_CLONED
routes go away when RTF_CLONING routes go away (= ARP cache goes away
when interface address goes away) (3) permit overwrite of RTF_CLONED
by static route settings.
2001-01-27 JINMEI, Tatuya <>
* kame/kame/libinet6/getaddrinfo.c: obsoleted EAI_ADDRFAMILY and
EAI_NODATA according to the rfc2553bis-02 draft.
* {bsdi3, freebsd4, netbsd, openbsd}/include/netdb.h:
* bsdi4/contrib/bind/src/include/netdb.h:
removed the definition of the obsoleted error types. In order to
minimize the possibility of binary compatibily breakage, we did
not reorder the error number sequence.
Fri Jan 26 00:12:48 JST 2001
* bsdi3: support ART routing table lookup algorithm.
* bsdi4: support ART routing table lookup algorithm. inet6 only
due to the use of custom host lookup logic in inet.
Thu Jan 25 12:26:32 JST 2001
* racoon:
Fixed to handle variable-length DH shared secret correctly.
from <>
Wed Jan 24 10:31:29 JST 2001
* sys/netinet6/ip6_fw.[ch]: pull security patch (fixes "established"
tcp filter) from FreeBSD-SA-01:08.ipfw.
NOTE: IPv4 ip_fw.[ch] on freebsd[234] are NOT patched (yet).
Wed Jan 24 09:32:16 JST 2001
* sys/net/radix_art.[ch]: ART routing table lookup algorithm.
(memory eater, but should be faster). currently netbsd/openbsd only.
Wed Jan 24 02:46:42 JST 2001
* netbsd: make MIP6 kernel compile again.
2001-01-23 JINMEI, Tatuya <>
* kame/kame/rtadvd/config.c (getconfig): made sure to zero-clear
manually configured prefix information structures. Without this,
rtadvd would not work when you configure advertised prefixes
manually. Please be sure to update. This bug seemed to be
introduced on around 2000-11-11.
Tue Jan 23 22:14:08 JST 2001
* freebsd4: share sys/netinet6/ipsec.h. sys/netinet6/ipsec6.h is gone.
NOTE: "make TARGET=freebsd4 clean" is necessary on upgrade.
2001-01-23 JINMEI, Tatuya <>
* kame/sys/netinet6/in6.c (in6_update_ifa, in6_ifscrub):
considered cases where multiple addresses with a same single
prefix are assigned on multiple interfaces. For example, you can
# ifconfig ne0 3ffe:ffff::1111 alias
# ifconfig ne1 3ffe:ffff::aaaa alias
then both two addresses can be assigned on ne0 and ne1, resp., and
the corresponding interface route 3ffe:ffff::/64 goes to the
interface ne0 (i.e. 1st cofirgured one wins). If you then delete
the address on ne0, the interface route switches the associated
interface to ne1.
(we do not recommend users such a configuration, though).
Tue Jan 23 18:09:08 JST 2001
* sys/netinet6/ip6_input.c: declare ip6aux structure, which is used
to pass around information about the packet, across header processing
routines. we plan to put header chain parsing history into here.
* sys/netinet6/dest6.c: support home agent destination option.
it is made mandatory by mobile-ip6 spec.
XXX may have broke ericsson mobile-ip6
2001-01-23 JINMEI, Tatuya <>
* kame/sys/net/if_gif.c (gif_ioctl): set IFF_RUNNING upon sucess
of the SIOCSIFPHYADDR ioctl (and its variations). This is
necessary to perform DAD on a gif interface.
2001-01-23 JINMEI, Tatuya <>
* kame/sys/netinet6/in6.c (in6_ifinit): clarified on installation
of interface direct routes;
- made an interface route only when
+ prefixlen is smaller than 128, or
+ the destination address is specified.
- set the cloning bit of an interface route only when the prefix
length is smaller than 128. And, in this case, the bit is
always set regardless of the interface type. In particular, an
interface route with the cloning bit would be configured on a
p2p interface.
This mean you can now invoke the following command:
# ifconfig gif0 inet6 3ffe:ffff::1
Note that there is no destination address specified, and the
prefix length is implicitly set 64. Also note that the
corresponding interface route of the address would have the
cloning bit, and neighbor caches would be created when you try to
communicate with a destination that matches the prefix
In summary, by the two consecutive changes, the only possible
commands you can type are:
# ifconfig gif0 inet6 IPv6address prefixlen plen
(where IPv6address is an IPv6 address, plen < 128) and
# ifconfig gif0 inet6 IPv6address IPv6dstaddress prefixlen 128
(where IPv6address and IPv6dstaddress are IPv6 addresses)
2001-01-23 JINMEI, Tatuya <>
* kame/sys/netinet6/in6.c (in6_update_ifa): made the argument
validation stricter; if a destination address is specified for the
SIOCAIFADDR_IN6 ioctl, its prefix MUST be 128. As an effect, the
kernel now refuses the following command:
# ifconfig gif0 inet6 3ffe:ffff::1 3ffe:ffff::2 alias
(note that the default prefix length is 64).
You should explicitly specify 128 as the prefix length:
# ifconfig gif0 inet6 3ffe:ffff::1 3ffe:ffff::2 prefixlen 128 alias
You may think this is too restrictive, but I believe it's better
than before. We've seen many users confused about configration of
p2p interfaces.
Tue Jan 23 13:51:34 JST 2001
* sys/netinet6/ipsec.c: record IPsec decapsulation history information,
so that we can use it for validating inbound packets at L4.
* netbsd: run ipfilter only for wire format packets, not the IPsec-
decapsulated ones.
Tue Jan 23 03:44:03 JST 2001
* netbsd/openbsd: make sure do not return garbage, when accept(2) is
called after socket is disconnected (like RST right after TCP
Mon Jan 22 JST 2001
* route6d: advertise aggregated route, only to interfaces specified
after -A option. found by IIJ IPv6 router torture testers.
2001-01-21 JINMEI, Tatuya <>
* kame/sys/netinet6/nd6_rtr.c (in6_{ifadd, ifdel}): ifa_refcnt
- do not IFAREF in in6_ifadd() (except on nbsd)
- do not duplicate IFAFREE in in6_ifdel() (except on nbsd)
without these changes, a manually configured ifaddr would be freed
two times when its lifetime expired. Although such a situation
would be rare, all *BSD (except NetBSD) users (who use versions
after Jan 2, 2001) are recommended to apply this fix.
Sun Jan 21 JST 2001
* libinet6/getaddrinfo.c: disallow invalid scope identifier
(like fe80::1%junk). it was broken around Jan 5 2001.
2001-01-21 JINMEI, Tatuya <>
* bsdi4/sbin/ifconfig/ifconfig.c (main): set IPv6 address
lifetimes correctly (i.e. set infinite). Without this, ifconfig
would not be able to assign IPv6 addresses.
Sun Jan 21 00:26:47 PST 2001
* sys/netinet6/ip6_input.c: do not accept packet to link-local address
on loopback interface, if the destination address is not assigned
to the node itself. fixes KAME PR 250.
Sun Jan 21 01:37:23 2001 SUMIKAWA Munechika <>
* sys/netinet6/{nd6.c,nd6_nbr.c,nd6.h}: implement garbage
collection for a stale NDP entry, as described in RFC2461
5.3. Each entry will be removed in 1day by default.
2001-01-21 JINMEI, Tatuya <>
* bsdi4/bin/ping/ping.c: corrected the output format when the
command is build without "INET6" being defined.
2001-01-21 JINMEI, Tatuya <>
* bsdi4/sys/netinet/udp_usrreq.c (udp_input):
* bsdi4/sys/netinet/tcp_input.c (tcp_input):
made a local copy of the IPv6 header and called in6_clearscope()
(which was newly implemented with this change) to remove
(possibly) embedded scope identifiers.
XXX: this change would reduce efficiency, but is necessary to
compute checksum correctly for a packet with scoped source or
destination addresses.
All BSD/OS 4.2 users are recommended to apply this fix.
Sat Jan 20 JST 2001
* sys/netinet6/ipsec.c: candidate fix for KAME PR 233 (tunnelled packet
may cause bad ARP).
Fri Jan 19 17:09:17 JST 2001
* sys/netinet/icmp6.h: synchronize RR flag bit definitions to
RFC2894. you need to compile every system you have, to make RR work.
KAME PR 300.
2001-01-19 JINMEI, Tatuya <>
* bsdi4/sys/i386/isa/if_ef.c (efintr): do not use the sentinel
pointer stored in ef_mb, and force the EF_NEWM macro to create a
fresh mbuf chain to store incoming ether frames (with mbuf cluster
if necessary). This fix is essential to the KAME IPv6 stack
(without the m_pulldown stuff), so please be sure to apply this
fix if you use IPv6 on an interface using the ef driver.
2001-01-18 JINMEI, Tatuya <>
* bsdi4/contrib/bind/src/lib/irs/{lcl,nis}_ho.c: set ho_addrinfo
function(s) properly. Without this fix, getaddrinfo() could cause
SEGV when "local" or "nis" is specified to look hosts up. Be sure
to update.
Thu Jan 18 02:00:32 JST 2001
* netbsd, openbsd, bsdi4: fix unsafe typecast in rtrequest1().
clarify 3rd arg handling in eonrtrequest().
* netbsd, openbsd: integrate post-4.4BSD rtrequest1(), and argument
type change on ifa_rtrequest (3rd arg).
2001-01-16 JINMEI, Tatuya <>
* kame/kame/dtcp/dtcps.rb (tunnelcleanup): deleted the pair of
physical addresses by "gifconfig delete" after the session. This
would prevent a garbage pair.
Tue Jan 16 00:43:19 JST 2001
* sys/netinet6/mld6.c: emit MLD6 packet correctly (MTU setting was
wrong and the packet did not go out of the node). the bug was
introduced around November 27, 2000.
Mon Jan 15 12:53:37 JST 2001
* route6d:
when an interface goes down, do not advertise routes associated with
the interface, more like vendor routers. by doing so, we can
avoid blackhole route (advertised even though the router lost
reachability), and we can switch to alternate route if one exists.
NOTE: it may cause reachability problem to the router itself, if
the router has interfaces without global addresses.
Sat Jan 13 12:45:56 JST 2001
* sys/netinet/ip_output.c: allow interface index to be specified with
multicast set/getsockopts, by using (= pass it as network
endian value, 24bit in in_addr). follows RFC1724 section 3.3.
suggested by Dave Thaler.
Fri Jan 12 18:54:52 JST 2001
* bsdi4: use BSD/OS 4.2 as the base version.
Thu Jan 11 02:47:37 JST 2001
* kame/sys/netkey/key.c:
- key_acquire() does not require a secpolicy structure.
There was a possibility of kernel panic.
reported by <dwang@iPolicyNet.COM>.
Thu Jan 11 02:46:09 JST 2001
* racoon:
- saved the type of SA in PF_KEY acquire message from the kernel
in order to reply a error.
- removed from the scheduler immediately if error happened when sending
phase 1 initiation message.
Wed Jan 10 11:41:30 JST 2001
* racoon:
- Fixed to configure the logging level. racoon saves some parameters
before parsing configuration file in order to prefer the parameters
by command line.
Wed Jan 10 00:19:59 JST 2001
* sys/netinet6/icmp6.c: fix SEGV on icmp6 redirect input.
openbsd/netbsd only. was introduced early Dec2000.
Tue Jan 9 16:50:07 JST 2001
* kame/libinet6/getaddrinfo.c: share getaddrinfo source code (again)
across all platforms we have. though we have a jumbo #ifdef at the
bottom for DNS lookup portion, it is good to share the core logic.
while we are at it, simplify lots of things.
freebsd4 behavior change: $GAI support is dropped. classful notation
(127.1) is dropped.
2001-01-07 JINMEI, Tatuya <>
* kame/sys/netinet6/in6_pcb.c (in6_pcbnotify):
* {bsdi4,freebsd[34],openbsd}/sys/netinet6/in6_pcb.c (in6_pcbbind):
allowed an application to bind a socket to a deprecated address.
For openbsd, a new check if the specified address is a node's own
address was added as well.
2001-01-07 JINMEI, Tatuya <>
* freebsd3/ports/libbind8: newly added based on BIND 8.2.2p7. This
libbind supported A6 and DNAME.
2001-01-06 JINMEI, Tatuya <>
* kame/kame/prefix/prefix.c: changed the default values of
prefixes to infinity. Finite default values were bad, because they
would be installed as the lifetimes of routers' addresses, which
would never increment and be invalidated in the future.
NOTE: if you use the kernel that has the change made on 2000-12-03
by Koji Kawano (see CHANGELOG.2000) on a router box, and use the
prefix command to assign addresses for the router, please be sure
to update the prefix command and reboot the router (or at least
reassign the addresses by the new prefix command). The lifetimes
might now be decresing, and will expire in the near future.
We'd even recommend you not to use the prefix command. It has
recently caused many problematic situations, and almost no one
understands the kernel prefix management routines well. You can do
the same thing by the ordinally ifconfig command. See the latest
rc.net6.sample file.
2001-01-06 JINMEI, Tatuya <>
* kame/kame/etc/rc.net6: prohibit the prefix command by default.
Since the command has recently caused many problems, we should
reconsider the prefix management mechanism.
Fri Jan 5 22:51:16 JST 2001
* libinet6/name6.c: nuke getnodeby{name,addr}, which was mentioned
only in draft-ietf-ipngwg-bsd-api-new-01.txt.
Fri Jan 5 13:24:21 JST 2001
* getaddrinfo.c (shared/netbsd/openbsd/freebsd4)
query DNS only once per an AF.
2001-01-05 JINMEI, Tatuya <>
* bsdi4/contrib/bind/src/lib/irs/gethostent.c: allowed gethostbyname
to accept abbreviated IPv4 text addresses (like 127.1).
(XXX: this change is against a bind8 policy, but just for backward
2001-01-02 JINMEI, Tatuya <>
* freebsd3/lib/libinet6/Makefile: do not link resolver files when
USE_LIBBIND6 is YES, so that enhanced libbind could be tested.
* freebsd3/*/ when USE_LIBBIND6 is YES, link libbind
(as well as libinet6).
Tue Jan 2 17:15:44 JST 2001
* sys/netinet6/in6{,_ifattach}.c: correct use of IFAREF/IFFREE on
non-netbsd operating systems. old code had memory leak possibility
on SIOCDIFADDR_IN6. reported by hitachi guys.
2001-01-01 JINMEI, Tatuya <>
* bsdi4/lib/libinet6:
* bsdi4/contrib/bind/src/lib:
massive improvements on DNS resolver including
- IPv6 transport of DNS resolver. An IPv6 textual address can now
be in the /etc/resolv.conf file as a DNS server.
- support of A6 and DNAME with bit string labels. They are still
very experimental, and thus disabled by default. "options a6"
and "options dname" in the /etc/resolv.conf file would enable A6
and DNAME, respectively.
- EDNS0 OPT RR support with UDP buffer size negotiation, which is
disabled by default, and would enabled by "options edns0" in the
/etc/resolv.conf file.
- clarification on the relationship between address family and irs
functions. For example, if you call getaddrinfo with AF_UNSPEC,
and irs.conf specifies to consider the local hosts file first,
then getaddrinfo would try to resolve an address in the
following order:
+ IPv6 addresses in the hosts file.
+ IPv4 addresses in the hosts file.
+ IPv6 addresses via DNS.
+ IPv4 addresses via DNS.
And, if one of the first two attempts succeeded, the remaining
two methods would be skipped. This change of the behavior would
be useful if you put both IPv6 and IPv4 addresses in the hosts
file, and do not want to see redundant DNS packets.