Skip to content
Fetching contributors…
Cannot retrieve contributors at this time
1090 lines (912 sloc) 48.3 KB
$KAME: CHANGELOG,v 1.1642 2001/04/07 11:46:20 jinmei Exp $
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 specical 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 RTF_LLINFO flag to the
corresponding host route, 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 0 to
this variable 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 all interfaces are NOT up.
Note that the variable effects all interfaces. When you set 0 to
the variable, 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 3014 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.
Jump to Line
Something went wrong with that request. Please try again.