[Announce] Openswan 2.6.47 released

[Samir Hussain]

Xelerance has released Openswan 2.6.47

https://download.openswan.org/openswan/openswan-2.6.47.tar.gz
https://download.openswan.org/openswan/openswan-2.6.47.tar.gz.asc


v2.6.47 (March 28, 2016)

Added feature to allow DNS query for external IP address of a gateway.

* Make certificate directories in correct place [MCR]
* Order of addconn and pluto is non-deterministic, so stick
  addconn output elsewhere [MCR]
* As a result of change to orient with family=0, this test case now
  binds an interface [MCR]
* With revision to permit left=%any, right=%defaultroute, and orient by
  private key it is permissible to have a conn that does not specify a
  host IP for "our side" [MCR]
* When looking for a matching interface, and conn family is 0, and
  both ends are zero, just pick the first interface that matches
  on ports [MCR]
* Initial test case for loading connection with right=%defaultroute [MCR]
* Exit if connection not found, rather than core dump later [MCR]
* Clean out .o files [MCR]
* Test case on why defaultroute is not a valid IP address: discovered in
  DrTaylorPlumage with ikev1-double-nat [MCR]
* Created functional test case for loading a mixed v6 in v4 conn [MCR]
* Created new test case for loading and orientating a mixed v6 in v4
  conn [MCR]
* Added new keyword: endaddrfamily.  renamed connaddrfamily
  to clientaddrfamily [MCR]
* Always look for v6 and v4 addresses in left/right=, updating
  the end-family only if it was not set [MCR]
* Refactor lp07 so that it can be used by lp41 [MCR]
* Adjust build-every-rev to exit smarter [MCR]
* Permit unit test cases to cause returned addresses to be sorted [MCR]
* alg_info_ike leak is unstable [MCR]
* Some missing leaks [MCR]
* With port numbers in play, the desired_port may not be set, in which
  case, look for plutos port [MCR]
* Change log to indicate a match on IP, going along with match on
  private key [MCR]
* The NO_KERNEL interface type was originally intended for pluto
  functional testing, and so it has some cruft related to not
  matching port numbers against 500. This presents problems when
  unit testing with (fake) ports other than 500, and so really the
  ignoring of port numbers should be a seperately enabled feature,
  as the unit testing really needs to use the actual NO_KERNEL
  interface, since it has no kernel [MCR]
* Added additional interface for port 4500 [MCR]
* Try to pick matching port when picking an interface [MCR]
* When picking an appropriate interface, make sure that the
  port numbers match [MCR]
* Updated test cases for logging of IP address in orient test [MCR]
* Turn off IPv6 since mock output code does not speak IPv6. [MCR]
* Bit a bit more flexible in where the address family comes from, and
  log the resulting interface better [MCR]
* Fixed order of htons() and init_iface_port, and added include
  for inet_pton [MCR]
* Provide for optional INIT_LOADED to be called. Used in lp18
  for assert [MCR]
* More extensive debug of orient() --- log which private key
  was found, and also if the pick_interface was able to find an
  interface. Also change the family searched for to be the (derived)
  value for the connection, rather than the end [MCR]
* Provide example of how to translate enum to string for keyword_host [MCR]
* Updates to lp18 for revised debugging of orient [MCR]
* Use preformatted interface address [MCR]
* Pick an interface that matches the right family [MCR]
* Reformat comments [MCR]
* Added ip_oriented flag to indicate if orientation was bound
  to IP [MCR]
* Added init_iface_port to include setting of q->socktypename [MCR]
* lp31 discovered that addrtypeof, addrbytesptr and samaddr
  should be tolerant of receiving a NULL [MCR]
* Changes to fmt_connection means that the nexthop for right=%any is
  no longer guess/assumed or logged [MCR]
* The changes to the conn load verification for IKEv1 PSK
  right=%any should use a left=%any, rather than an explicit
  address. [MCR]
* Change update_host_pair to return indicate of whether orient
  worked; as if it did not then a different address might be in
  order. [MCR]
* Updates to unit test cases for socktypename addition [MCR]
* Added socknametype to iface_port structure so that socket
  family can easily be logged [MCR]
* Make it clear that IPHOSTNAME types are not v4 or v6, and should not
  initialize the nexthop in any specific way, and should fit into the
  right=%any checks for IKEv1 as well [MCR]
* Try to guess what kind of family the conn is, if the conn has
  a this or that with a family set the host for the side that does
  not have a family to  that family [MCR]
* Try to set the address family from left or from right, if set. [MCR]
* Only diagnose an address-type mis-match if both sides are
  specified by a literal address [MCR]
* Check for core dumps. Write test case 2b into gdbinit file [MCR]
* Arbitrarily decide to use IPv6 ANY address when right is
  default route [MCR]
* Set the address type based upon which kind of address was parsed [MCR]
* Also validate that the conn is properly loaded into pluto [MCR]
* Correctly parse a site local (e.g. fec0::1) address [MCR]
* New test case to validate IPv6 site local addresses in left/right= [MCR]
* Log the string value involved in the debug of the looseenum [MCR]
* IPv6 address of cassidy.sandelman.ca actually did change [MCR]
* lp40 updated for DNS delayed rebase [MCR]
* find_ID_host_pair debug now includes dump of exact parameter [MCR]
* Check the orientations after the secrets are loaded, as
  possession-of-private-key test needs private keys [MCR]
* Removed confusing comment [MCR]
* Changes to conn to be really h2h [MCR]
* Orient test which loads keys after conns [MCR]
* h2h should use host to host items [MCR]
* Added h2h and brokenspace as possible test cases for
  readwriteconf crash [MCR]
* Change connection list as per DNS changes to show IP address
  discovered [MCR]
* gcc 5.0 fixes [MCR]
* Reconciled leak of ID to fact that IDhost_pair is never freed [MCR]
* Update lp08 with proper CHILDSA_DEL name for state, after
  state_names added [MCR]
* Updated description to explain three unit subtests [MCR]
* There was a IDhost_pair leak, which was located, as one list was
  never properly emptied as the clear_host_pair routine was incorrectly
  calling  the host_pair free routine when it meant to remove a
  connection from a list [MCR]
* Updated Makefile and explanation of how to get updated pcap file [MCR]
* Unit tests do not speak 3des-md5, modp1024 [MCR]
* Make sure the installed_time for a public key is set from regression
controlled time [MCR]
* Updated parameters for test case to match files named after tests [MCR]
* A half-open, prospective_parent_sa that is in progress only
  gets priority over new DNS answers, if the DNS query had an
  error [MCR]
* If no addresses are available from DNS yet, but there is a hint, then
  the hint should be attempted [MCR]
* Protect connection_check_ddns1 against corrupt IPhp_next loops [MCR]
* Try to be smarter about when a connection is stuck: consider
  connections which have never come up as well [MCR]
* Blacklist a bunch of replies for DNS lookups [MCR]
* In order to avoid DNS errors causing more DNS lookups, only do new DNS
  lookups when there is a timeout --- other attempts will use additional
  addresses only [MCR]
* Have adns return getaddrinfo()-style EAI errors, even for old
  nquery work [MCR]
* Be careful not to remove connections which were not yet added
  to host_pair [MCR]
* Output sanifier now removes kernel state numbers from output [MCR]
* Updated to reflect changes to debugging [MCR]
* Slight tweak to comment [MCR]
* Make sure to set the DNS list pointer upon receiving new answers [MCR]
* Guard against no connections in search routine [MCR]
* Added additional debugging to delayed DNS lookup continuation [MCR]
* Init generic CR before filling in DNS name so that qtid gets
  logged sanely [MCR]
* Keep track of states that are created to potentially bring up
  a parent SA. This is needed to tell if there is an ongoing initiation
  for a delayed-DNS effort, or if one should be made. Do not make an
  attempt to bring up the conn unless the policy is set to UP. Use
  returned state number from the initiate process to always get correct
  state in test harness [MCR]
* Make test validate that handle_adns_answer() does not cause conn
  to be set to UP [MCR]
* Make clean would clean up whackfile, so on reffile use, cp it
  to OUTPUT [MCR]
* Use return serial number to pull up correct serial number [MCR]
* Return state number for newly created states, as there is no way to
  track them until they are authenticated [MCR]
* Fix lp33,lp34,lp35 to include seam_initiate, and show DNS name
  in status [MCR]
* Added missing test cases [MCR]
* Show ccache statistics [MCR]
* Try to do straight build first [MCR]
* Try to use ccache when building [MCR]
* Add make clean target [MCR]
* Slight adjustment to list of leaks [MCR]
* When doing DNS lookups, use the connaddr family as the hint as
  to what kind of records to lookup (A vs AAAA) [MCR]
* Added seam_initiate and seam_adns appropriate to fix up tests [MCR]
* Do DNS lookup and then initiate connection [MCR]
* Whitespace changes [MCR]
* Split up sendI1 so continuation part can be called again [MCR]
* Move kick_adns_connection from dnskey to initiate [MCR]
* Reworked lp33 to include actual initiate and dns continuation code [MCR]
* Create kick_adns_connection routine so that DNS replies kick new
  connections immediately [MCR]
* Copy parentI1 main code into lp33 and add aDNS steps [MCR]
* Added A and AAAA records to rr_typename [MCR]
* Remember if an end has a valid address when DNS lookups are delayed so
  that we do not initiate until DNS lookups have had a chance [MCR]
* Rearranged a bunch of seam so that lp33 can import real adns
  code properly [MCR]
* Removed include of connections.c, use connections.o: add set
  of includes [MCR]
* Include seam_dnskey explicitly, as test case 33 will use real code [MCR]
* Updates after rebase [MCR]
* Add test case lp33 for dns delayed, when there is no hint [MCR]
* For unclear reasons the lookup of cassidy.sandelman.ca/KEY RR fails.
  Could be due to obsolete RR? [MCR]
* Rename lp28-parentR2anychoice to lp32 to keep sequence [MCR]
* Rename lp27-IDhostpair to lp31 to keep sequence [MCR]
* Tweak lp30-dnskick [MCR]
* Tweak seam_log [MCR]
* Test pcap output now uses TESTNAME, so set it correctly [MCR]
* Clean up PID file, and create .gdbinit with arguments [MCR]
* Deal with some leaks; use stop_adns() properly to clear up children.
  Make sure that ipanswers list, after sorting, is restored so that all
  items get freed (affects regression testing) [MCR]
* Process each dns request before making a new one to keep order
  the same [MCR]
* Use sort_addr_info to canonicalize the output to deal
  with differences in gai.conf [MCR]
* Added make explicitly to package list [MCR]
* Use DBG_log to get consistent output [MCR]
* As structure is used as temporary, and copied, make sure to zero
  it first [MCR]
* Output results if failure [MCR]
* Adjust unit test cases for update_host_pair() seam. Rename lp28-dns to
  lp30-dnskick, add needed canonicalization. Adjust output from moving
  dump_addr_info() into pluto as it uses DBG_log() rather than printf(),
  and outputs to stderr [MCR]
* Removed last vestiges of DYNAMICDNS and processing converted
  to IPHOSTNAME [MCR]
* Tweak adnstest [MCR]
* Process DNS getaddrinfo() replies, and attach them to continuation [MCR]
* Refactor dump_addr_info debug into seperate file [MCR]
* Cleanup leaks of addrinfo structures [MCR]
* Added test case for serialization/deserialization of addrinfo [MCR]
* Basic test case for looking up KEY RR; one success, one failure [MCR]
* Use standard openswan_log() for messages rather than syslog() [MCR]
* Make it easy to generate cpp processed files for inspection [MCR]
* Added new utility strtochunk() [MCR]
* Created adnstest case to validate operation of dnskey.c and adns.c [MCR]
* Small refactor of start_adns_query so that it can accept things
  other than struct id [MCR]
* Initial test case for dnslookups [MCR]
* Adjust comments on functions [MCR]
* Updated trace with IP address in hint [MCR]
* Do not include EF unless defined [MCR]
* Added seam for kick_adns_connection_lookup [MCR]
* Include gdb instructions for testing pluto [MCR]
* _pluto_adns is no longer seperate executable [MCR]
* Change definition of progname to const [MCR]
* Moved init_adns() call earlier, and make sure it exits properly [MCR]
* Make certificate directories; removed --adns path argument from help [MCR]
* Move test for SAref and SAbind into kernel.c [MCR]
* Added setproctitle() [from BSD licensed sendmail via pppd] and use it
  rather than global_argv hack. Use setproctitle() in adns
  sub-process [MCR]
* Create dummy kick_adns_connection [MCR]
* _pluto_adns is no longer a seperate program, but is part of the pluto
  executable, forked out for use. This makes it much easier for embedded
  systems to have a sane (if simple) async DNS resolver. Future work
  will switch to c-ares This patch also includes changing progname to
  a const globally [MCR]
* Move whack out of pluto directory [MCR]
* Removed DYNAMIC DNS from whack client [MCR]
* Added lp27 to test list [MCR]
* Do not show hostname string if the host_type is IP address [MCR]
* Comment out I9 [MCR]
* Switch update order to make it run update1 first [MCR]
* Fix up lp06 test case to work [MCR]
* Remove LWRES support --- it broke awhile ago [MCR]
* Force ikev2 [MCR]
* If nexthop is invalid, then do not show it [MCR]
* Ipsecconf already included the hostname into a string, but
  now it needs to include the hint as well. This code plus test
  cases probably produces a whack file with the correct hint [MCR]
* Fix emacs variables [MCR]
* Gdb init for test case [MCR]
* Refactor lp02-parentI1, so it can be reused by lp27 [MCR]
* Obsolete is a qualifier for a keyword, not a type of keyword [MCR]
* Add functional/06 test case [MCR]
* Removed redundant kw_list->string member [MCR]
* Added processing of new loose_enum_arg type, added %dns and test
  it out in a functional test [MCR]
* Figure out left/right-ness of keyword so that it can be logged
  better in errors [MCR]
* Added loose_enumarg processing [MCR]
* When setup properly, the h2hR2 test case works fine: just needs
  to have actual keys [MCR]
* Use parker end-point and parker secrets [MCR]
* h2h R2 packet processing - broken [MCR]
* Make output file parameterized by testname [MCR]
* h2h I2 packet processing [MCR]
* Make lp10 a template test case [MCR]
* Make output file parameterized by testname [MCR]
* h2h R1 packet processing [MCR]
* Process arguments more carefully [MCR]
* Added test case for h2h I1 [MCR]