CVS log for kame/kame/sys/netinet6/ip6_output.c

Up to [KAME snap] / kame / kame / sys / netinet6

Request diff between arbitrary revisions - Display revisions graphically

Keyword substitution: kv
Default branch: MAIN

ANSIfy, to reduce diffs with *BSDs.
based on diff from dunceor at gmail.com.
XXX not tested yet, sorry, we need to ressurect buildlab configs.

random thoughts:
- IGMPv3 support needs a serious cleanup.  argument passing is horrible, style
  does not meet style(9) at all.  i'm afraid there could be a lot of bugs.
- DCCP/SCTP/NAT-PT need someone to maintain them, otherwise we should drop
  them at once.
- we should pick a platform (one of *BSD) and try to get KAME tree in sync
  with more recent version of *BSD.  otherwise, KAME tree gets more and more
  out-of-date.

fixed unclosed if clause.

reorder statement to meet the above comment (or should we change comment?)

back out 1.491

knf

remove redundant if statement (see the comment right above it)

remove redundant code setting up ip6_hlim (openbsd)

tab -> space

indent

cleanup SCOPEDROUTING (unifdef -USCOPEDROUTING)

- clear rthdr2 related data structures in ip6_clearpktopts.

fixed a kernel crash by an ASM join from a userland application

reduce mtu value of ro_pmtu when the mtu value is greater than ifmtu.

fixed ip6_getpmtu() logic for freebsd5.

no need to check if a variable is 0 or not in if(variable!=0) sentence.

avoided possible overflow in ip6_insert_jumboopt()
(it's not a real threat with the typical value of MCLBYTES, though).

style

fixed a compilation error

ports some MLDv1 enhancements into MLDv2.
 - use callout timer mechanism for protocol timer
 - impose a random delay (when necessary) according to rfc2462bis

removed a garbage whitespace at the end of line

- corrected operator priority
- avoided the use of an additional argument to in6_selectroute()

allowed in6_selectif() to select the outgoing IF without trying to get a route
to the nexthop.  otherwise, ambiguous scoped destinations may match a rejected
route, and an application that expects specifying the outgoing IF resolves
scope zone ambiguity will fail (even though such expectation is a bad
practice).

disabled a leftover in the RFC3542 work about IPV6_RECVPATHMTU.
(from freebsd, see ftp://ftp.kame.net/pub/mail-list/snap-users/8775)

corrected error handling for IPV6_LEAVE_GROUP

do not specify M_WAITOK during the output process.

MIP6/NEMO: catch up with the recent update of the scope framework.

revised scope-related interfaces, clarifying the relationship between
the kernel-internal form and the public sockaddr_in6 form.

changed the word '2292bis' to 'RFC3542' when appropriate.

remove freebsd[234]-specific code, since they are no longer supported in KAME-SNAP

fixed a kernel crash due to a LOR by a multicast-group join without
specifying any receiving interface.

fixed a freebsd5 kernel crash when WITNESS debug option is enabled.
(M_WAITOK leads to an assertion failure when a non-sleepable
fine-grain-lock exists and WITNESS debug option is on. If this option is off,
M_WAITOK is just regarded as M_NOWAIT)

Unified several similar code enclosed within #ifdef MIP6_MCOA

removed a duplicated code

- sys/netipsec (FAST_IPSEC) has #define for time_second
- use pfil_hook for PF

fixed a path MTU discovery failure on FreeBSD5.

correct validation when ipv6mr_interface == 0.  from djm@openbsd

sync w/ 2.0.
needs checking/fixing: igmp3, pf, altq, dccp, sctp
FAST_IPSEC path is not considered, intentionally.

New Mobile IPv6 code which also supports NEMO.

Make it compilable on FreeBSD 5.3 with IPSEC option

sync with freebsd53 and makes it compilable
(their behavior is not confirmed yet...)

fix openbsd compilation

when IPsec encapsulation is performed, we have no way to tell which
"icmp6 too big" message corresponds to which route.  fragment to 1280.
found by iij seil team

update ro_pmtu on IPsec tunnel mode.  otherwise, ro_pmtu != ro will be
misunderstood as "routing header present".  iij seil team

- revised delay for DAD according to rfc2462bis-02: impose a random delay (when
  necessary) before sending the corresponding MLD report, while still listening
  to the multicast group, rather than before sending an NS.
- improved timer granularity for MLD.  the timer now uses the system's most
  fine-grained timer tick.

typo: missing '_' in __FreeBSD__
(critical!)

remove support for freebsd[23], simplify #ifdef

nuke some bsdi #ifdef.  make pf.c compatible w/ nd6_output() arg.

simplify #ifdef by de-support bsdi[34], freebsd[23].  basically,
- remove contents of "#ifdef __bsdi__"
- change "#if (defined(__FreeBSD__) && __FreeBSD_Version >= [34]) into
  "#ifdef __FreeBSD__"
- and other bunch of changes.   compiled ok on netbsd and freebsd4

make it compilable on freebsd521+kame

ToDo: reflect changes in freebsd521, test, ...

simplied the fix to FreeBSD-SA-04:06.ipv6.  The previous one worried too much
even though we actually validate the parameters.  This code also is more
compatible with other *BSDs, which do copyin within setsockopt().

applies quite the same fix as proposed in FreeBSD-SA-04:06.ipv6

FreeBSD Security Advisory: FreeBSD-SA-04:06.ipv6

indentation

code clean-up; IN6_IS_LOCAL_GROUP() macro changes into !in6_is_mld_target()

use standard address-comparison macro, as it's not necessary to
compare addresses in protocol-independent manner

fixed a MLDv2-related kernel crash owing to a lack of initialization of 'struct ifnet' variable

compares IPv6 group address properly

- rthdr2 should not be processed if rthdr0 exists.
- fixed a indentation style when defining a block inside a function.

fixed overkilling: nd6_output() takes a sockaddr_in6 from the beginning.
some incorrect arguments to rtalloc1 were also fixed.

rewrote rthdr address swapping part.

recovered scope zone boundary check in ip6_output().

use scope6_check_id() for the 1st of a routing header in case the default
scope zone is specified and the 1st hop is a socped address.

removed an unused variable.

disallowed the use of the unspecified address for IPV6_JOIN_GROUP and
IPV6_LEAVE_GROUP.  The special case was not documented in the official
API specification, and for whatever purpose, this should actually have
not been working for a long time.

in the process of IPV6_LEAVE_GROUP, catch explictly the case of
ipv6mr_interface != 0 but ifp == NULL.
the previous code was not dangerous, but was incorrect.

comment wording in ip6_setmoptions(): "no need to keep the mbuf" is not really
accurate since im6o is actually malloced.

recovered an overkilled comment.
used sockaddr_in6{} passed to in6_selectroute for resetting cached destination.
(just to recover the original logic.  there should be no behavior change)

removed space chars in a blank line.

makes it compilable

restructure of header files based on draft-ietf-mip6-mipext-advapi-00.

knf

missing in6_clearscope() for routing header type 1

embed if-index for link-local multicast addresses specified via socket-API

A parameter to specify a destination address for in6_selectsrc() and
in6_selectroute() is gotten back to sockaddr_in6 from in6_addr.
Confirmed on only freebsd4.

makes MLDv2 compilable
(behavior is not confirmed yet...)

refines MLDv2-code for maintenance convenience
 - creates mldv2.c to concentrate almost all the MLDv2-related functions here
 - renamed MLDv2-ready in6_multi manupulation functions

don't lookup route when sending a multicast packet.

KNF

avoid -Wuninitialized

move away from in6_{get,set}pktaddrs framework, go back to embedded
scopeid in s6_addr16[1].  this is to make *BSD merge easier.

TODO: check missing/removed scope border code
TODO: variable name re-sync w/ *BSD platforms
TODO: fix non-netbsd platforms
TODO: fix mip6, sctp, dccp, mldv2

sync with mldv2-20031021
	- prohibit improper MSF delta-API call
	  (BLOCK(S1,G1) when there is an empty source-list for G1,
	   UNBLOCK(S1,G1) when there is an empty block-source-list for G1)
	- express EXCLUDE{NULL} by 'i6ms_blknumsrc != 0', instead of 'i6ms_grpjoin'

fixed typo that collapsed MLDv2

fixed memory leak in MLDv2 MSF list

fixed a bug that group comparison sometimes fails in MLDv2, since multicast address in in6_multi contains port number

correct previous commit; skip ipsec if IPV6_FORWARDING

(non-openbsd) do not lookup security policy if IPV6_FORWARDING.
avoids infinite ipsec encapsulation on
	ip6_input -> ip6_forward -(tunnel mode)-> ip6_output
case.

removed unnecesarry structure members.

NULL is not equivalent to 0

fixed typos

after the introduction of netbsd ifconfig create/delete, if_index can no longer
be used for bound checking due to the wrapping of if_index - we can only use
if_indexlim for bound check, and then check if ifindex2ifnet[x] is NULL or not.

TODO: check other OS (freebsd[45])

draft-chakrabarti-mobileip-mipext-advapi-02 support.
not tested.

changed mobility header protocol no. and MIP6 related ICMPv6 type no.
based on IANA assignment.

arg to suser() have changed in openbsd 3.4

wording and typo fix in comment lines

style

always copy ip6_pktopt.  remove needcopy and needfree argument/structure member
accordingly.

be consistent on naming advanced API functions; use ip6_XXXpktopt(s).

simplified getsockopt(IPV6_PKTOPTIONS):
* RFC3542 (effectively) deprecated the
* semantics of the 2292-style pktoptions.
* Since it was not reliable in nature (i.e.,
* applications had to expect the lack of some
* information after all), it would make sense
* to simplify this part by always returning
* empty data.

- remove a couple of NULL checks (should panic)
- DIAGNOSTIC is not for printf()
- IN6P_HOPOPTS should only be set on privileged in6p (ip6_output.c)
  so panic if mismatch happens
- ip6_savecontrol() argument is redundant

randomize flowlabel/fragment ID.

u_short -> u_int16_t
don't set ip6_plen where unneeded (i.e. before calling ip6_output)

don't use m_cat to mbuf of different types.  KAME PR 495

removed an essentially unused variable.

typo in a panic message.

- sync with Freebsd 5.1-RELEASE
- removed FreeBSD-specific ifdef for M_MOVE_PKTHDR(), since all *BSDs have M_MOVE_PKTHDR() now

typo.

CN or HA doesn't add type2 routing header to a Binding Acknowledgement
packte when a deregistration packet was issued on home link and the
deregistration was failed by some reasons.

comment wording.

simplification.

put *bytag ipsec policy lookup into *byaddr, to keep consistency of lookup
ordering.

do not require root privilege for getsockopt(IPV6_2292HOPOPTS or
IPV6_2292DSTOPTS) since these operations do not retrieve the actual header
values.

comment wording

a clean up; removed an unused macro definition.

missing m_free in the previous change.

correctly caught an error from sooptcopyin() in ip6_ctloutput().

removed an (almost) unused variable.

enable PF for freebsd4

underlying logic for looking up security policy by pf tag.

cleanup for netbsd PF

drop NEW_STRUCT_ROUTE for netbsd

nuke IPV6_REACHCONF

IPV6_REACHCONF is obsoleted.

editorial cleanups wrt adv API;
  - replaced rfc2292bis-xx with RFC 3542 if it was referred to a particular I-D
    name.
  - revised a stale statement

deprecated IPV6_REACHCONF.

style; removed an empty comment line.

- ip6_setmoptions()
  clear the embedded scope id of a multicast link-local address
  before inserting it to the multicast membership list.

removed a meaningless code fragment.

use array index for readability

use 'struct route_in6' if NEW_STRUCTROUTE is disabled

loosened INRIA's MLDv2 copyright

Obtained from: Hitoshi Asaeda

answers to MLD query for link-local multicast group address.
(stopped embedding the ifindex of group address in struct in6_multi)

Reported by: Kentaro Ohara <Kentarou.Oohara@jp.yokogawa.com>
(users@jp.ipv6.org #3055)

- restructured mip6 code.
  divided cn/ha/mn part to separate files.  this makes it easier to
  merge a paticular part (ex. CN) to original BSD or to integrate
  a paticular part (ex. HA) to routers.

makes it compilable on freebsd-4.8

remove "m->m_flags &= ~M_PKTHDR" when m becomes an intermediate mbuf in a
mbuf chain.  it is safer this way.

fixed a panic, useing IKE w/ ipsec tunnel

- introduced M_MOVE_PKTHDR to netbsd.
  TODO: freebsd5

use M_MOVE_PKTHDR instead of M_COPY_PKTHDR when appropriate, in order to
avoid mtag leakage.
(this is an emergent fix.  we may need to change more.)

- update src_sa and dst_sa after calling ipsec6_output_tunnel() which
  may change the src/dst addresses of a mbuf.
  because src_sa and dst_sa are not pointers, they do not point correct
  information automatically.

makes it compilable on freebsd5, too

changed 2nd and 3rd argument of ip6_getpktaddrs() from sockaddr_in6 **
to sockaddr_in6 * in order to avoid sharing a pointer in an mbuf.

changes for *bsd-specific files will follow.

looks okay for freebsd4.  will compilation check for others.

- fixed m_copy() behavior of freebsd as openbsd/netbsd do.
  mtags are copyed back from the copied packet to the original packet.
  it is not necessary to copy ip6aux manually after m_copy() any more.

comment wording

- do not move mtags when copying mbuf in ip6_mloopback().  the aux
  information stored in the tags may be needed in the later process of
  the output routine.  the copied mbuf doesn't need the aux infomation
  because the packet is just passed to looutput() at the end of the
  ip6_mloopback().

transition from m_aux framework to m_tag framework.
XXX freebsd4/bsdi3 won't compile for a while

MIP6: remove mip6_ip6route

MIP6: correct ordering rules for rthdr2

comment wording

MIP6: fix a memory leak and mip6 route cache

uses 'struct sockaddr_storage' to prevent compilation warning and unexpected buffer overrun

add CONSTCOND to while (0)

removed redundant code

- changed mobility header formats to those of id-19.

- plugged memory leakage in some erroneous cases
  (from Sam Leffler)
- removed a noisy log

fixed a kernel crash occurring when specifying multiple source addresses in MLDv2-related API

panic() doesn't need \n

introduced a local variable to refine code

s/igmp/mld/

removed a duplicated include file

make it compile without MIP6

- MIP6: XXX do not check inbound policy if nxthop of rthdr is me.

- always uses a source address of a binding update as a destination address
  of a binding ack (not using existing binding cache entry) so that
  the binding ack can be delivered to the correct location.

sync w/ 1.6.  GENERIC.KAME/i386 compiles, not tested.
- NATPT and IP6FW support were dropped (you can use ipfilter).
- altq, ecn, RTF_CACHE, SCOPEDROUTING need checking

padding length before jumbo payload option need to be "0".

- we can't use an ipsec tunnel mode if an outgoing packet has a home
address destination option.

typo

correct signedness mixup in pointer passing.

KNF - return is not a function.

fixed a kernel-hangup with MLDV2 option, when an IPv6 prefix is removed

remove whitespace in empty line

fixed the getsockopt function.

need for ip6_setpktaddr on frament packets(use ip6_getpktaddr in nd6_ns_output)

MLDv2 host-side implementation for NetBSD and FreeBSD-4,
based on draft-vida-mld-v2-03.txt.

ToDo: porting to OpenBSD & BSDI4, much more test

forgot to increment some mip6 stat values.

Initialize ip6 with mtod() outside of #ifdef IPSEC

MIP6: fix bug a HAO and not work frag. packet. discard mh->pproto != NONE

initialize finaldst_sa

MIP6: duplicated call mip6_destopt_discard()

added a new socket option/cmsg type "IPV6_PREFER_TEMPADDR" to control
source address selection wrt temporary addresses per socket/packet.

- fbsd/430
  set INP_IPV4 flag in a PCB when IPV6_V6ONLY=0 is specified by setsockopt,
  also, reset INP_IPV4 flag in a PCB when IPV6_V6ONLY=1 is specified.
  we should reconsider the logic around here.  This is too complex, isnt' it?

ip6_pktopt controls original IPv6 header, not the outer IPv6 header on IPsec
tunnel case.

MIP6: eps protection of BU to home agent

MIP6: add MAKE_MOBILITYHDR macro, change BA padding

MIP6:
-mip6 tunnel and framgent problem
-bug fix proxy ND(not work unicast NS)

- sync with my working directory for other mip6 developer's convenience.
- add MN state machine (partially), now MN sends BU to CNs.
- no CN code yet.

whitespace cleanup

whitespace

correct IPV6_CHECKSUM (noone is using it?)

KNF for merge

more KNF

KNF for merge

'fall through' is not a valid LINT keyword.

remove obsolete comment

implement RFC2460 section 5, last paragraph (ICMPv6 too big with < 1280).
need testing

do not update rmx_mtu if it is 0 (obey ifmtu)

remove IFF_FRAGMENTABLE for now

KNF

bad indent

- a type 2 routing header output processing.

- the first step for mip6 id17 support.
  this commit is for the KAME mip6 developpers.  do not try to use this.

do not synchronize an ND link MTU with the corresponding physical link MTU.
instead, when detecting the link's "real" MTU always consider those two values
and take the smaller one.
introduced a macro IN6_LINKMTU(ifp) for this purpose.

changed the semantics of IPV6_USE_MIN_MTU based on discussions in the ipv6 wg
ML;
by default, unicast packets will follow path MTU while multicast packets
will be sent at the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
including unicast ones will be sent at the minimum MTU.  Multicast
packets will always be sent at the minimum MTU unless IP6PO_MINMTU_DISABLE
is explicitly specified.

this semantics will be described in forthcoming rfc2292bis-07.

with this change, the corresponding value ip6_pktopts{} is no longer a flag,
and the macro value IP6PO_MINMTU was obsoleted.

use RTUSE whenever defined.

use nd_ifinfo[].linkmtu instead of ifp->if_mtu to handle link MTU issues
wrt IPv6, whenever possible.

when using a route entry, increment statistics members.

revived a part of code that handled the RTV_MTU flag of rtentries so that
use would be able to disable path MTU discovery for particular destinations
(paths) by hand.

revised the case that the reported path MTU is less than the minimum
MTU (1280);
just use the minimum MTU instead of the reported one.
from now on, the RTV_MTU flag will never be used for IPv6 routes.

do not updated the "dst" address, which was the next-hop address,
for multicast destinations.

- applied recent clarifications in rfc2292bis
  + ignore IPV6_DONTFRAG and IPV6_RECVPATHMTU for TCP sockets
  + disallowed to set non-unspecified address by IPV6_PKTINFO for TCP sockets
- corrected the return value from getsocketopt(IPV6_DONTFRAG) and
   getsocketopt(IPV6_USE_MIN_MTU)

use EADDRNOTAVAIL instead of ENXIO in some cases.

fixed kernel panic in the case of IPV6_LEAVE_GROUP without an interface.
the kernel will choose the group based on the address only (however,
scope zone ambiguity will not be allowed by default).

based on: a report from Tomomi Suzuki <stomomi@ebina.hitachi.co.jp>

let ip6_getpmtu() take sockaddr_in6

changed the type of the address part of in6_multi{} to sockaddr_in6
so that it would contain scope zone ID field.
related functions are modified accordingly.

style; avoided to make a separate block just to define local variables.

simplified the trick of "fake loopback interface"; we can actually use
the rt->rt_ifa->ia_ifp as the "original interface".

replaced IN6_IS_ADDR_UNSPECIFIED with SA6_IS_ADDR_UNSPECIFIED
whenever possible.

in ip6_mloopback() made sure to keep mbuf aux for the original packet.

totally revised IPv6 scoped address architecture in the kernel.
- sticked to use sockaddr_in6, not in6_addr, as much as possible.
- did not refer to ip6_src/ip6_dst (which may have ambiguity on scope zones),
  but used full sockaddr_in6 structures attached to the packet.
- replaced special cases for link-local addresses with code that used
  generic functions such as in6_addr2zoneid().
- additional cleanups mainly for scoped address handling.

handled getsockopt(IPV6_RECVRTHDR) correctly
[KAME PR 403]

check the IPV6_MINMTU flag in ip6_output as well as IP6PO_MINMTU in the
packet options.  the former can still be set by icmp6_reflect or gif_output.

less warnings.

[cleanup] removed meaningless functions "ip6_update_recvpcbopt" and
"ip6_reset_rcvopt".

avoided NULL pointer reference.

remove opt_inet6.h,mip6_var.h&etc. includes from if_hif.h&mip6.h

update the tclass API implementation not to erase the ECN bits.
 - do not override if a non-zero value is already set.
 - check the diffserv field and the ecn field separately.
as suggested in draft-itojun-ipv6-tclass-api-03.txt.

made it compile at !freebsd[34]

getsockopt(IPV6_TCLASS) should return the system default (not -1) when
no explicit value was specified.

removed IPV6_CHECKSUM cases from ip6_ctloutput().

allowed getsockopt(IPV6_RECV{HOPOPTS, DSTOPTS, RTHDRDSTOPS}).
we do not need to prohibit this operation for non-privileged users.

removed definitions of in6p_xxx for the NRL-derived systems.

obsoleted IN6P_MINMTU in the pcb flag to specify using the minimum MTU.
instead, we should consistently use IP6PO_MINMTU in the outputopts.

obsoleted the usage of sticky IPV6_HOPLIMIT according to rfc2292bis-03.

separated overriding rule of advanced API options according to rfc2292bis-03.
e.g. an IPV6_HOPOPTS ancillary data object only disables IPV6_HOPOPTS
     sticky option.

clarifications for the IPV6_PKTINFO option:
  - do not allow 0 length for IPV6_PKTINFO (rfc2292bis-03)
  - clear existing pktinfo regardless of the option context (sticky vs
    ancillary)

simplified ip6_setpktoption():
  - uniformed the semantics of the length argument so that we could share
    the same validation code.
  - always call ip6_clearpktopts when len is 0, regardless if the option
    is sticky (this is correct for spec conformance)

disabled IPV6_[O]TCLASS in the RFC 2292 sticky context.

removed the 'needfree' argument of ip6_clearpktopts(), because we can
deal with this issue by a data origin way (by the pktopt->needfree member).

added a prototype in the definition of ip6_setpktoption()

comment wording

added the IPV6_NEXTHOP case in ip6_ctloutput()

typo

rejected odd offset values for IPV6_CHECKSUM.

made it compile on NRL-derived systems.

- added a new function ip6_raw_ctloutput, which only handles IPV6_CHECKSUM.
  this function will be called from *bsd/rip6 ctloutput.
- removed IPV6_CHECKSUM cases from ip6_ctloutput accordingly.

one more cleanup to reduce ifdefs

cleanup in ip6_ctloutput; centralized macro definitions for NRL-derived
systems to share the code.

made it compile on NRL-origin systems

implemented the IPV6_PATHMTU (get)socket option

implemented a new IPv6 socket option IPV6_DONTFRAG as described in
rfc2292bis-03.

s/^ \t/\t/

style; switch is not a function

	- add code for merging the user specified rthdr by ip6_pktopts and
	  the rthdr for the mip6 route optimization.

comment wording

typo (openbsd)

sync with openbsd 3.0.  compiles but not tested.  ALTQ does not work for now.

initialized the last arg to ip6_process_hopopts(), because the recent
code requires it to be 0 when a jumbo payload option is contained.

correct per-ifa stat (freebsd4)
reduce #ifdef (we get a bit of meaningless code on freebsd2/3/openbsd/bsdi3)

increase per-ifa output byte/packet count, presented on netstat -is{,b} (bsdi4)

clean up:
removed a code fragment in ip6_setpktoption() that performed the validation
of the address specified via IPV6_PKTINFO and was temporarily disabled by
'ifdef 0'.
added rationale and issues of the change as comments.

added a new function scope6_check_id() to check sin6_scope_id from an
application.  see netinet6/scope6.c for more details.
(only tested on freebsd4.  will do on bsdi4.)

simplified in6_embedscope():
  - do not take PCB and ifpp as argument.  this function now assumes the caller
    set the correct zone ID in advance.
  - centralized the SCOPEDROUTING condition in this function, not in the
    callers.
(check has only been done on freebsd4.  I'll do it on bsdi4 too)

disabled the check for the address part of IPV6_PKTINFO, and let
in6_selectsrc() do the check later.
(at this moment, I just 'ifdef 0'ed the part.  I'll clean it up later)

clarified cached route manipulation:
- check if the cached route is up in in6_selectsrc()
- removed a corresponding check in ip6_output to avoid redundancy

- NULL-clear m->m_pkthdr.rcvif before calling ip6_mforward()
- avoid NULL (or invalid) pointer access when a printing log message

white space fix

fix a null pointer access to the pktopts when MIP6 is enabled
and a rthdr for mip6 is inserted.

do not touch any extention headers during the mip6 processing
if IPV6_FORWARDING is set.

replace __FUNCTION__ with __FILE__ and __LINE__.

remove pktopt_xxx variables for mip6 related exthdr processing,
instead, introduce mip6_pktopts structure.

there's no possibility that we insert dest1 twice.

insert dest1 exthdr when MIP6 inserts rthdr.

make it impossible to change IPV6_V6ONLY if INET6_BINDV6ONLY is declared.
netbsd.

fix memory leak. I had forgot to free pktopts in ip6_outout.

fix compilation on NRL-based stacks (IPV6_OTCLASS).

pedant

make IPV6_TCLASS to take int, not u_int8_t.  -1 means the default value.
follows 2292bis.  fix a couple of #ifdef bugs while i'm here.
XXX we need to keep backward binary compat for IPV6_OTCLASS as bsdi ships
with IPV6_TCLASS.

freebsd-current introduce ifnet_byindex() instead of ifindex2ifnet[]
for SMPng.

let's say hello to new ifdef storm with smile.

MIP6:
fix using incorrect link mtu when inserting a rthdr for MIP6 route
optimization.

changed the return value type of in6_addr2zoneid(), from u_int32_t to
int64_t, so that the caller can tell an error from valid "4+28" ID values.

added a bit trickter validation on message length in ip6_setpktoptions().
I think the current code is safe enough (i.e. the kernel will not panic),
but this check will make the kernel robester.

printed more logs in the SCOPEDEBUG case (there are still some scope zone
breakage.)

check the IPV6_UNSPECSRC flag first in the srcaddr validation
in ip6_output.

added a new function "ip6_setpktoption" to centralize the routine for
handling outgoing packet options both as socket options and as ancillary
data, with some strickter validation.

clarified a comment; it should be safe to call malloc with M_WAITOK
in ip6_setpktoptions(), because the function is never called under
an interrupt context.

free mbuf in case error in the functions that has an mbuf pointer in its
argument list.

made IPV6_NEXTHOP work correctly for link-local hops

arg to openbsd ipsp_process_packet changed by mistake (1.209)

netbsd don't need opt_hoge.h. remove it from sources.
freebsd3 need opt_hoge.h, add opt_mip6.h to sys/conf/files.

 Modified Files:
 	freebsd3/sys/conf/options kame/sys/netinet6/dest6.c
 	kame/sys/netinet6/icmp6.c kame/sys/netinet6/in6.c
 	kame/sys/netinet6/in6_proto.c kame/sys/netinet6/in6_src.c
 	kame/sys/netinet6/ip6_forward.c kame/sys/netinet6/ip6_input.c
 	kame/sys/netinet6/ip6_output.c kame/sys/netinet6/mip6.c
 	kame/sys/netinet6/mip6_binding.c kame/sys/netinet6/mip6_ha.c
 	kame/sys/netinet6/mip6_icmp6.c kame/sys/netinet6/mip6_prefix.c
 	kame/sys/netinet6/mip6_subnet.c kame/sys/netinet6/nd6_rtr.c

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.

 Modified Files:
	kame/sys/netinet6/dest6.c kame/sys/netinet6/icmp6.c
	kame/sys/netinet6/in6.c kame/sys/netinet6/in6.h
	kame/sys/netinet6/in6_ifattach.c
	kame/sys/netinet6/in6_ifattach.h kame/sys/netinet6/in6_proto.c
	kame/sys/netinet6/in6_src.c kame/sys/netinet6/in6_var.h
	kame/sys/netinet6/ip6_forward.c kame/sys/netinet6/ip6_input.c
	kame/sys/netinet6/ip6_output.c kame/sys/netinet6/mip6.c
	kame/sys/netinet6/mip6.h kame/sys/netinet6/mip6_ha.c
 	kame/sys/netinet6/nd6.h kame/sys/netinet6/nd6_rtr.c
 	freebsd4/sys/net/if_types.h freebsd4/sys/conf/files
 	freebsd4/sys/conf/options
 Added Files:
 	kame/sys/net/if_hif.c kame/sys/net/if_hif.h
 	kame/sys/netinet6/mip6_binding.c
 	kame/sys/netinet6/mip6_icmp6.c kame/sys/netinet6/mip6_prefix.c
 	kame/sys/netinet6/mip6_subnet.c kame/sys/netinet6/mip6_var.h

implemented IPV6_NEXTHOP ancillary data/socket option.
(the next hop should be an AF_INET6 socket,
 socket option case has not been tested yet)

3rd argument 'port' of ip6_fw_chk is unnecessary.

- allowed MLD messages to have the unspecified address as source.
- renamed the macro IPV6_DADOUTPUT to IPV6_UNSPECSRC.  The former is now
  more suitable.

added several improvements on IPv6 output routine
(I should've done some of them by separate commits...sorry):

- added a seprate function in6_selectroute() to centralize the route and
  outgoing interface selection algorithm
- call in6_selectroute from in6_selecsrc() 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')
- renamed in6_addr2scopeid to in6_addr2zoneid to be more intuitive in terms
  of the scope architecture
- made in6_addr2scopeid 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

Although the change is quite large, I just compiled the code on bsdi4.
Please check it on other platforms, thanks.

do not set ro_dst.sin6_scope_id just to check scope zone breakage.
(just for safety; the previous code should not have bad effects)

style: added spaces after /* and before */
(this might be a matter-of-taste issue, though)

interface-local multicast support;
- install ff01::%ifname/32 in in6.c.
- change source address selection for ff01::xxx.
- change outgoing interface selection for ff01::xxxx.
- prohibit packets to ff01::xxx from being sent on the wire.

comment why dest2 is handled specially on header chain concatenation

try using (lower 32bit of src) ^ (lower 32bit of dst) for rtalloc_mpath hash

fill scopeid field on interface local address.  ping6 ff01::1%lo0 now works
(but not ff01::1%ne2... why)

use host-endian value for multipath hashing