Referencing $(srcdir) in the gemspec is not really an option, as "gem build"
includes the full path in the gem, so we need to build in $(srcdir). As there
does not seem to be a way to control the output of "gem build", we manually
move the gem to $(builddir) in OOT builds.
The inbound flag is used to determine if we have to install an update or a new
SA in the kernel. As we do not have allocated SPIs and therefore can't update
an existing SA in the HA plugin, always set the flag to FALSE.
Before 698ed656 we had extra logic for that case, but handling it directly in
the HA plugin is simpler.
While d0d85683 works around a crasher related to the use of libunwind, other
build hangs have been seen in the all test cases. Try to
--disable-unwind-backtraces to see if libunwind is really related to those
and if it fixes these issues.
While it is currently unclear why it happens, canceling threads waiting in the
new_query condvar does not work as expected. The behavior is not fully
reproducible: Either cancel(), join() or destroying the condvar hangs.
The issue has been seen in the http-fetcher unit tests, where the stream service
triggers the use of the resolver for "localhost" hosts. It is reproducible with
any cleanup following a host_create_from_dns() use on a Ubuntu 14.04 x64 system.
Further, the issue is related to the use of libunwind, as only builds with
--enable-unwind-backtraces are affected.
As we broadcast() the new_query condvar before destruction, a hard cancel() of
these threads is actually not required. Instead we let these threads clean up
themselves after receiving the condvar signal.
Ubuntu 12.04 does not seem to provide a sane pkg-config for libiptc or libip4tc.
The monolithic build fails due to missing symbols, so disable it until we have
a newer Ubuntu release.
Unloading libraries calls any library constructor/destructor functions. Some
libraries can't handle that in our excessive unit test use. GnuTLS leaks
a /dev/urandom file descriptor, letting unit tests fail with arbitrary
out-of-resources errors.
Implement a forecast plugin that supports forwarding of multi- and broadcast
messages between a LAN and clients or between connected clients. It uses
IPsec policy marks to send packets over multiple identical multi- and broadcast
policies.
Introduce a connmark plugin that uses Netfilter conntracks mark to select the
correct return-path SAs for client-initiated connections. This can be used
to distinguish transport mode clients behind the same NAT router.
Fixes#365.
In this test two hosts establish a transport mode connection from behind
moon. sun uses the connmark plugin to distinguish the flows.
This is an example that shows how one can terminate L2TP/IPsec connections
from two hosts behind the same NAT. For simplification of the test, we use
an SSH connection instead, but this works for any connection initiated flow
that conntrack can track.
Currently supports transport mode connections using IPv4 only, and requires
a unique mark configured on the connection.
To select the correct outbound SA when multiple connections match (i.e.
multiple peers connected from the same IP address / NAT router) marks must be
configured. This mark should usually be unique, which can be configured in
ipsec.conf using mark=0xffffffff.
The plugin inserts CONNMARK netfilter target rules: Any peer-initiated flow
is tagged with the assigned mark as connmark. On the return path, the mark
gets restored from the conntrack entry to select the correct outbound SA.
As with other configuration backends, XAuth is activated with a two round
client authentication using pubkey and xauth. In load-tester, this is configured
with initiator_auth=pubkey|xauth.
Fixes#835.
As we now reuse the reqid for identical SAs, the behavior changes for
transport connections to multiple peers behind the same NAT. Instead of
rejecting the SA, we now have two valid SAs active. For the reverse path,
however, sun sends traffic always over the newer SA, resembling the behavior
before we introduced explicit SA conflicts for different reqids.
With make-before-break IKEv2 re-authentication, virtual IP addresses must be
assigned overlapping to the same peer. With the remote IKE address, the backend
can detect re-authentication attempts by comparing the remote host address and
port. This allows proper reassignment of the virtual IP if it is re-requested.
This change removes the mem-pool.reassign_online option, as it is obsolete now.
IPs get automatically reassigned if a peer re-requests the same address, and
additionally connects from the same address and port.
Migrates the attribute framework and associated plugins from libhydra back
to libcharon. libcharon is the only user of this framework since pluto is gone.
With these changes, we can pass the full IKE_SA state to attribute providers
and handlers, bringing more flexibility to these plugins.
While it has some tests that we don't directly cover with the new unit tests,
most of them require special infrastructure and therefore have not been used
for a long time.
With these changes, charon dynamically allocates reqids for CHILD_SAs. This
allows the reuse of reqids for identical policies, and basically allows multiple
CHILD_SAs with the same selectors. As reqids do not uniquely define a CHILD_SA,
a new unique identifier for CHILD_SAs is introduced, and the kernel backends
use a proto/dst/SPI tuple to identify CHILD_SAs.
charon-tkm is not yet updated and expires are actually broken with this merge.
As some significant refactorings are required, this is fixed using a separate
merge.
References #422, #431, #463.
To quickly check out IKE_SAs and find associated CHILD_SAs, the
child_sa_manager stores relations between CHILD_SAs and IKE_SAs. It provides
CHILD_SA specific IKE_SA checkout functions wrapping the ike_sa_manager.
As we now use the same reqid for multiple CHILD_SAs with the same selectors,
having marks based on the reqid makes not that much sense anymore. Instead we
use unique marks that use a custom identifier. This identifier is reused during
rekeying, keeping the marks constant for any rule relying on it (for example
installed by updown).
This also simplifies handling of reqid allocation, as we do not have to query
the marks that is not yet assigned for an unknown reqid.
To reassign reqids where appropriate, we explicitly allocate or confirm them
centrally on the kernel-interface.
Currently the state is stored in the kernel-interface wrapper for all
backends, but we may add appropriate methods to each backend to implement
a custom reqid allocation logic, if required.
The kernel backend uses an inbound parameter these days, where it makes
no sense to pass the update flag. The kernel backend decides itself how
it handles SA installation based on the inbound flag.
While we can handle the first selector only in BEET mode in kernel-netlink,
passing the full list gives the backend more flexibility how to handle this
information.
The reqid is not strictly required, as we set the reqid with the update
call when installing the negotiated SA.
If we don't need a reqid at this stage, we can later allocate the reqid in
the kernel backend once the SA parameters have been fully negotaited. This
allows us to assign the same reqid for the same selectors to avoid conflicts
on backends this is necessary.
With the introduction of PT-TLS, we started sending TLS close notifies after
the application layer completes (7bbf7aa9). While this makes sense for TCP based
transports, it is not required in EAP methods. In EAP, handshake completion
can be directly signaled using the outer EAP-SUCCESS message. This also saves
one round-trip in the EAP exchange.
Windows 7/8 does not seem to like TLS close notifies at all in EAP, and either
stalls (EAP-TTLS) or disconnects (PEAP).
Fixes#556.
A type is returned only if upper layers successfully created a record, that is
returns NEED_MORE. If we do not check for the return value, we might check a
previous record or the uninitialized type variable and falsely reset the
sequence number.
While they usually are not included in a normal strongSwan build, the XPC
header indirectly defines these Mach types. To build charon-xpc, which uses
both XPC and strongSwan includes, we have to redefine these types.
On OS X 10.10, when installing a virtual IP on a tun device, there is a chance
that a RTM_IFANNOUNCE is sent before the IP is ready on that link when calling
getifaddrs(). As we don't get an RTM_NEWADDR event either, that race lets us
miss the virtual IP install event, failing the add_ip() call.
While our ASN.1 parser can't handle BER indefinite length encoding, the OpenSSL
backend can. Some PKCS#12 containers get encoded this way, so we should
support loading such files in the pem plugin.
To make accept/recvmsg cancelable, we wrap them with poll. As poll itself
does not honor pending cancellations when entering the function, we use our
variant that checks for pending cancellation requests before entering poll().
In recent ruby versions, extensions get built with -Wformat. As we use custom
printf specifiers, that does not work for us. As there does not seem to be a
reliable way to override -Wformat, we use a variable for the format string,
which prevents gcc from doing the -Wformat check in that particular situation.
While it was always documented as boolean setting, the option is currently
handled as integer value, for which yes/no values do not work. Instead the
default of TRUE is used for a no value.
The option has been moved a lot during the last years, and in some locations
was handled as bool, in some as integer. In the latest codebase it
congruently used integer, which is actually not what is documented and used
in testing.
Fixes#781.
Martin Willi