4 years agotesting: Fix URL for kernel sources
Tobias Brunner [Mon, 20 Mar 2017 09:13:33 +0000 (10:13 +0100)]
testing: Fix URL for kernel sources

4 years agoike-sa-manager: Remove superfluous assignment
Thomas Egerer [Wed, 15 Mar 2017 15:24:49 +0000 (16:24 +0100)]
ike-sa-manager: Remove superfluous assignment

Memory is allocated with calloc, hence set to zero, thus assigning the
numerical value 0 is not required.

Signed-off-by: Thomas Egerer <>
4 years agoike: Log remote IP when deleting half-open IKE_SAs
Tobias Brunner [Wed, 15 Mar 2017 14:56:39 +0000 (15:56 +0100)]
ike: Log remote IP when deleting half-open IKE_SAs

4 years agocoverage: Exclude test suites and /usr from coverage report
Tobias Brunner [Wed, 15 Mar 2017 14:54:38 +0000 (15:54 +0100)]
coverage: Exclude test suites and /usr from coverage report

4 years agotravis: Create coverage report via
Tobias Brunner [Wed, 15 Mar 2017 09:19:02 +0000 (10:19 +0100)]
travis: Create coverage report via

4 years agoVersion bump to 5.5.2dr7 5.5.2dr7
Andreas Steffen [Mon, 6 Mar 2017 19:21:40 +0000 (20:21 +0100)]
Version bump to 5.5.2dr7

4 years agoaikpub2: Removed aikpub2 tool
Andreas Steffen [Mon, 6 Mar 2017 05:16:10 +0000 (06:16 +0100)]
aikpub2: Removed aikpub2 tool

The aikpub2 tool has been replaced by pki --pub|--req --keyid hex ..
where keyid indicates the TPM 2.0 private key object handle. Thus
either the public key in PKCS#1 format can be extracted or a PKCS#10
certificate request signed by the TPM private key can be generated.

4 years agopki: Add key object handle of smartcard or TPM private key as an argument to pki...
Andreas Steffen [Mon, 6 Mar 2017 11:53:48 +0000 (12:53 +0100)]
pki: Add key object handle of smartcard or TPM private key as an argument to pki --keyid

4 years agoutils: chunk_from_hex() skips optional 0x prefix
Andreas Steffen [Mon, 6 Mar 2017 10:51:10 +0000 (11:51 +0100)]
utils: chunk_from_hex() skips optional 0x prefix

4 years agopki: Edited keyid parameter use in various pki man pages and usage outputs
Andreas Steffen [Sun, 5 Mar 2017 10:19:39 +0000 (11:19 +0100)]
pki: Edited keyid parameter use in various pki man pages and usage outputs

4 years agoquick-mode: Correctly prepare NAT-OA payloads as responder
Tobias Brunner [Fri, 3 Mar 2017 16:11:04 +0000 (17:11 +0100)]
quick-mode: Correctly prepare NAT-OA payloads as responder

The initiator's address was sent back twice previously.

Fixes #2268.

4 years agoVersion bump to 5.5.2dr6 5.5.2dr6
Andreas Steffen [Fri, 3 Mar 2017 08:34:50 +0000 (09:34 +0100)]
Version bump to 5.5.2dr6

4 years agoAdd keyid of smartcard or TPM private key as an argument to pki --req
Andreas Steffen [Thu, 2 Mar 2017 19:30:24 +0000 (20:30 +0100)]
Add keyid of smartcard or TPM private key as an argument to pki --req

4 years agotesting: load-testconfig script loads config from source dir
Tobias Brunner [Thu, 2 Mar 2017 08:50:34 +0000 (09:50 +0100)]
testing: load-testconfig script loads config from source dir

It now does replace the IPs too. This way it's easier to play around
with a config (otherwise a do-tests run was required to build the
config files in the build dir).

4 years agolibipsec: Enforce a minimum of 256 for SPIs
Tobias Brunner [Thu, 2 Mar 2017 10:51:27 +0000 (11:51 +0100)]
libipsec: Enforce a minimum of 256 for SPIs

RFC 4303 reserves the SPIs between 1 and 255 for future use.  This also
avoids an overflow and a division by zero if spi_min is 0 and spi_max is

4 years agolibipsec: Fix min/max SPI
Tobias Brunner [Thu, 2 Mar 2017 09:11:32 +0000 (10:11 +0100)]
libipsec: Fix min/max SPI

4 years agocontroller: Don't listen for CHILD_SA state changes when terminating IKE_SAs
Tobias Brunner [Thu, 2 Mar 2017 07:58:07 +0000 (08:58 +0100)]
controller: Don't listen for CHILD_SA state changes when terminating IKE_SAs

We actually want to wait until the IKE_SA is destroyed, not any of the
CHILD_SAs (even though there might not be that much of a difference
depending on the number of CHILD_SAs).

Fixes #2261.

4 years agokernel: Make range of SPIs for IPsec SAs configurable
Tobias Brunner [Tue, 21 Feb 2017 18:21:01 +0000 (19:21 +0100)]
kernel: Make range of SPIs for IPsec SAs configurable

4 years agosettings: Add support for hex integers (0x prefix) via get_int()
Tobias Brunner [Tue, 21 Feb 2017 18:12:32 +0000 (19:12 +0100)]
settings: Add support for hex integers (0x prefix) via get_int()

4 years agolibipsec: Log a packet's ports and protocol in case of a policy mismatch
Tobias Brunner [Fri, 17 Feb 2017 11:40:36 +0000 (12:40 +0100)]
libipsec: Log a packet's ports and protocol in case of a policy mismatch

4 years agohost: Don't log port if it is zero
Tobias Brunner [Fri, 17 Feb 2017 11:32:04 +0000 (12:32 +0100)]
host: Don't log port if it is zero

4 years agolibipsec: Match IPsec policies against ports of processed packets
Tobias Brunner [Wed, 15 Feb 2017 14:33:38 +0000 (15:33 +0100)]
libipsec: Match IPsec policies against ports of processed packets

Fixes #2252.

4 years agoNEWS: Mention the new addrblock features
Martin Willi [Thu, 2 Mar 2017 07:23:01 +0000 (08:23 +0100)]
NEWS: Mention the new addrblock features

4 years agoaddrblock: Use dynamic TS narrowing instead of rejecting the whole CHILD_SA
Martin Willi [Wed, 22 Feb 2017 09:01:19 +0000 (10:01 +0100)]
addrblock: Use dynamic TS narrowing instead of rejecting the whole CHILD_SA

Previously, the client had to propose no wider selectors than the certificate
permits, otherwise the complete CHILD_SA was rejected. However, with IKEv2
we can dynamically narrow the selectors to what the certificate allows. This
makes client and gateway configurations very simple by just proposing,
narrowed to selectors the client is permitted to route into the network.

4 years agoaddrblock: Support an optional non-strict mode accepting certs without addrblock
Martin Willi [Wed, 22 Feb 2017 08:43:31 +0000 (09:43 +0100)]
addrblock: Support an optional non-strict mode accepting certs without addrblock

This allows a gateway to enforce the addrblock policy on certificates that
actually have the extension only. For (legacy) certificates not having the
extension, traffic selectors are validated/narrowed by other means, most
likely by the configuration.

4 years agochild-cfg: Always apply hosts to traffic selectors if proposing transport mode
Tobias Brunner [Mon, 20 Feb 2017 10:36:30 +0000 (11:36 +0100)]
child-cfg: Always apply hosts to traffic selectors if proposing transport mode

Usually, %dynamic is used as traffic selector for transport mode SAs,
however, if wildcard traps are used then the remote TS will be a subnet.
With strongSwan at the remote end that usually works fine as the local
%dynamic TS narrows the proposed TS appropriately.  But some
implementations reject non-host TS for transport mode SAs.
Another problem could be if several distinct subnets are configured for a
wildcard trap, as we'd then propose unrelated subnets on that transport
mode SA, which might be problematic even for strongSwan (switch to tunnel
mode and duplicate policies).

Closes strongswan/strongswan#61.

4 years agotraffic-selector: Allow calling set_address() for any traffic selector
Tobias Brunner [Mon, 20 Feb 2017 10:28:17 +0000 (11:28 +0100)]
traffic-selector: Allow calling set_address() for any traffic selector

Users may check is_host(), is_dynamic() or includes() before calling this
if restrictions are required (most actually already do).

4 years agoMerge branch 'pki-addrblock'
Martin Willi [Mon, 27 Feb 2017 08:37:51 +0000 (09:37 +0100)]
Merge branch 'pki-addrblock'

Add support to the x509 plugin and pki to generate certificates with the
RFC 3779 addrblock extension.

4 years agopki: Add a note about constructing RFC 3779 compliant certificates to manpage
Martin Willi [Wed, 22 Feb 2017 07:52:00 +0000 (08:52 +0100)]
pki: Add a note about constructing RFC 3779 compliant certificates to manpage

4 years agopki: Support an --addrblock option for issued certificates
Martin Willi [Tue, 21 Feb 2017 14:07:42 +0000 (15:07 +0100)]
pki: Support an --addrblock option for issued certificates

4 years agopki: Support an --addrblock option for self-signed certificates
Martin Willi [Tue, 21 Feb 2017 14:07:03 +0000 (15:07 +0100)]
pki: Support an --addrblock option for self-signed certificates

4 years agopki: Add a helper function parse traffic selectors from CIDR subnets or ranges
Martin Willi [Tue, 21 Feb 2017 14:06:15 +0000 (15:06 +0100)]
pki: Add a helper function parse traffic selectors from CIDR subnets or ranges

4 years agox509: Do not mark generated addrblock extension as critical
Martin Willi [Tue, 21 Feb 2017 15:54:57 +0000 (16:54 +0100)]
x509: Do not mark generated addrblock extension as critical

While RFC 3779 says we SHOULD mark it is critical, this has severe side effects
in practice. The addrblock extension is not widely used nor implemented, and
only a few applications can handle this extension. By marking it critical,
none of these applications can make use of such certificates where included
addrblocks do not matter, such as TLS/HTTPS.

If an application wants to make use of addrblocks, that is usually an explicit
decision. Then the very same application obviously can handle addrblocks, and
there is no need for the extension to be critical. In other words, for local
policy checks it is a local matter to handle the extension, hence making it
critical is usually not of much help.

4 years agox509: Support encoding the RFC 3779 addrblock extension
Martin Willi [Tue, 21 Feb 2017 14:05:17 +0000 (15:05 +0100)]
x509: Support encoding the RFC 3779 addrblock extension

4 years agobuilder: Define a builder part for X.509 RFC 3779 address blocks
Martin Willi [Tue, 21 Feb 2017 14:04:18 +0000 (15:04 +0100)]
builder: Define a builder part for X.509 RFC 3779 address blocks

4 years agoplugin-loader: Fix hashing of registered plugin features
Tobias Brunner [Fri, 24 Feb 2017 17:22:23 +0000 (18:22 +0100)]
plugin-loader: Fix hashing of registered plugin features

This strangely never caused any noticeable issues, but was the reason for
build failures in certain test cases (mostly BLISS) due to missing plugin
features when built with specific options on Travis (was not reproducible

4 years agoVersion bump to 5.5.2dr5 5.5.2dr5
Andreas Steffen [Thu, 23 Feb 2017 16:31:11 +0000 (17:31 +0100)]
Version bump to 5.5.2dr5

4 years agoUse of TPM 2.0 private keys for signatures via tpm plugin
Andreas Steffen [Thu, 16 Feb 2017 07:39:47 +0000 (08:39 +0100)]
Use of TPM 2.0 private keys for signatures via tpm plugin

4 years agoImplement signatures with private keys bound to TPM 2.0
Andreas Steffen [Mon, 13 Feb 2017 18:06:18 +0000 (19:06 +0100)]
Implement signatures with private keys bound to TPM 2.0

4 years agoandroid: New release after fixing potential ANR issue
Tobias Brunner [Mon, 20 Feb 2017 15:34:43 +0000 (16:34 +0100)]
android: New release after fixing potential ANR issue

4 years agoandroid: Send network change events from a separate thread via JNI
Tobias Brunner [Wed, 15 Feb 2017 15:08:35 +0000 (16:08 +0100)]
android: Send network change events from a separate thread via JNI

Doing this from the main UI thread (which delivers the broadcast) might
cause an ANR if there is a delay (e.g. while acquiring a mutex in the
native parts). There might also have been a race condition during
termination previously because Unregister() was not synchronized so there
might have been dangling events that got delivered while or after the mutex
in the native parts was destroyed.

4 years agoikev1: Respond to DPDs for rekeyed IKE_SAs
Tobias Brunner [Mon, 13 Feb 2017 10:54:53 +0000 (11:54 +0100)]
ikev1: Respond to DPDs for rekeyed IKE_SAs

Some devices always use the oldest IKE_SA to send DPDs and will delete
all IKE_SAs when there is no response. If uniqueness is not enforced
rekeyed IKE_SAs might not get deleted until they expire so we should
respond to DPDs.

References #2090.

4 years agoike-sa: Optionally try to migrate to the best path on routing priority changes
Martin Willi [Mon, 10 Oct 2016 13:59:52 +0000 (15:59 +0200)]
ike-sa: Optionally try to migrate to the best path on routing priority changes

When multihomed, a setup might prefer to dynamically stay on the cheapest
available path by using MOBIKE migrations. If the cheapest path goes away and
comes back, we currently stay on the more expensive path to reduce noise and
prevent potential migration issues. This is usually just fine for links not
generating real cost.

If we have more expensive links in the setup, it can be desirable to always
migrate to the cheapest link available. By setting charon.prefer_best_path,
charon tries to migrate to the path using the highest priority link, allowing
an external application to update routes to indirectly control MOBIKE behavior.
This option has no effect if MOBIKE is unavailable.

4 years agoikev2: Ignore roam events without MOBIKE but static local address
Tobias Brunner [Tue, 15 Nov 2016 14:42:33 +0000 (15:42 +0100)]
ikev2: Ignore roam events without MOBIKE but static local address

Disabling MOBIKE and statically configuring a local address should be
enough indication that the user doesn't want to roam to a different
address.  There might not be any routes that indicate we can use the
current address but it might still work (e.g. if the address is on an
interface that is not referenced in any routes and the address itself
is neither).  This way we avoid switching to another address for routes
that might be available on the system.

We currently don't make much use of COND_STALE anyway when MOBIKE is not
enabled, e.g. to avoid sending DPDs if the connection is seemingly down.
With MOBIKE enabled we don't exactly check that state but we do don't
send DPDs if there is no route/source address available.

4 years agoike-cfg: Add helper function to determine if a given IP address was configured
Tobias Brunner [Wed, 1 Feb 2017 17:00:57 +0000 (18:00 +0100)]
ike-cfg: Add helper function to determine if a given IP address was configured

4 years agoMerge branch 'vici-updates'
Tobias Brunner [Thu, 16 Feb 2017 18:24:17 +0000 (19:24 +0100)]
Merge branch 'vici-updates'

Adds several new features for the VICI interface and swanctl.

4 years agoNEWS: VICI updates
Tobias Brunner [Wed, 15 Feb 2017 16:49:06 +0000 (17:49 +0100)]
NEWS: VICI updates

4 years agovici: Only log messages if there actually is a listener
Tobias Brunner [Mon, 13 Feb 2017 16:38:49 +0000 (17:38 +0100)]
vici: Only log messages if there actually is a listener

4 years agovici: Let has_event_listeners() actually check if clients are registered
Tobias Brunner [Wed, 8 Feb 2017 14:20:58 +0000 (15:20 +0100)]
vici: Let has_event_listeners() actually check if clients are registered

Fixes: 8d96f90a7983 ("vici: Add function to test if an event should be

4 years agovici: Add support for mediation extension
Tobias Brunner [Tue, 7 Feb 2017 11:04:30 +0000 (12:04 +0100)]
vici: Add support for mediation extension

4 years agopeer-cfg: Store mediated_by as name and not peer-cfg reference
Tobias Brunner [Tue, 7 Feb 2017 10:30:49 +0000 (11:30 +0100)]
peer-cfg: Store mediated_by as name and not peer-cfg reference

This way updates to the mediation config are respected and the order in
which configs are configured/loaded does not matter.

The SQL plugin currently maintains the strong relationship between
mediated and mediation connection (we could theoretically change that to a
string too).

4 years agovici: Include uniqueness policy in list-conns
Tobias Brunner [Wed, 1 Feb 2017 10:02:22 +0000 (11:02 +0100)]
vici: Include uniqueness policy in list-conns

4 years agoswanctl: Add --rekey command
Tobias Brunner [Tue, 24 Jan 2017 15:34:32 +0000 (16:34 +0100)]
swanctl: Add --rekey command

4 years agovici: Add command to initiate SA rekeying
Tobias Brunner [Tue, 24 Jan 2017 15:26:48 +0000 (16:26 +0100)]
vici: Add command to initiate SA rekeying

4 years agovici: Use unique names for CHILD_SAs in the list-sas command
Tobias Brunner [Fri, 9 Dec 2016 13:45:41 +0000 (14:45 +0100)]
vici: Use unique names for CHILD_SAs in the list-sas command

The original name is returned in the new "name" attribute.

This fixes an issue with bindings that map VICI messages to
dictionaries.  For instance, in roadwarrior scenarios where every
CHILD_SA has the same name only the information of the last CHILD_SA
would end up in the dictionary for that name.

4 years agoswanctl: Allow specifying pubkeys directly via 0x/0s prefix
Tobias Brunner [Wed, 18 Jan 2017 16:46:27 +0000 (17:46 +0100)]
swanctl: Allow specifying pubkeys directly via 0x/0s prefix

4 years agovici: Add support to load CA certificates from tokens and paths in authority sections
Tobias Brunner [Wed, 30 Nov 2016 14:41:18 +0000 (15:41 +0100)]
vici: Add support to load CA certificates from tokens and paths in authority sections

4 years agovici: Add support to load certificates from file paths
Tobias Brunner [Wed, 30 Nov 2016 14:09:04 +0000 (15:09 +0100)]
vici: Add support to load certificates from file paths

Probably not that useful via swanctl.conf but could be when used via VICI.

4 years agovici: Add support to load certificates from tokens
Tobias Brunner [Wed, 30 Nov 2016 11:44:51 +0000 (12:44 +0100)]
vici: Add support to load certificates from tokens

4 years agoswanctl: Add `token` secrets for keys on tokens/smartcards
Tobias Brunner [Fri, 18 Nov 2016 15:40:34 +0000 (16:40 +0100)]
swanctl: Add `token` secrets for keys on tokens/smartcards

4 years agovici: Add command to load a private key from a token
Tobias Brunner [Fri, 18 Nov 2016 14:01:18 +0000 (15:01 +0100)]
vici: Add command to load a private key from a token

PINs are stored in a "hidden" credential set, so that its shared
secrets are not exposed via VICI.  Since they are not explicitly loaded as
shared secrets via VICI a client might consider them as removed secrets and
remove them.

4 years agovici: List namespace/peer-cfg name with policies and allow filtering
Tobias Brunner [Mon, 13 Feb 2017 17:18:58 +0000 (18:18 +0100)]
vici: List namespace/peer-cfg name with policies and allow filtering

The two names are also transmitted in separate keys.

4 years agoswanctl: Pass optional connection name to --initiate/install/uninstall
Tobias Brunner [Tue, 11 Oct 2016 16:27:29 +0000 (18:27 +0200)]
swanctl: Pass optional connection name to --initiate/install/uninstall

4 years agovici: Explicitly use peer name when uninstalling trap and shunt policies
Tobias Brunner [Wed, 16 Nov 2016 17:13:59 +0000 (18:13 +0100)]
vici: Explicitly use peer name when uninstalling trap and shunt policies

Also adds an `ike` parameter to the `uninstall` command.

4 years agostroke: Use peer name as namespace for shunt policies
Tobias Brunner [Wed, 8 Feb 2017 15:13:32 +0000 (16:13 +0100)]
stroke: Use peer name as namespace for shunt policies

The same goes for the start-action-job.  When unrouting, we search for
the first policy with a matching child-cfg.

4 years agoshunt-manager: Add an optional namespace for each shunt
Tobias Brunner [Wed, 16 Nov 2016 16:59:22 +0000 (17:59 +0100)]
shunt-manager: Add an optional namespace for each shunt

This will allow us to reuse the names of child configs e.g. when they
are defined in different connections.

4 years agovici: Add support for NT Hash secrets
Tobias Brunner [Wed, 16 Nov 2016 16:12:33 +0000 (17:12 +0100)]
vici: Add support for NT Hash secrets

Fixes #1002.

4 years agovici: Add support for IPv6 Transport Proxy Mode
Tobias Brunner [Wed, 16 Nov 2016 14:58:34 +0000 (15:58 +0100)]
vici: Add support for IPv6 Transport Proxy Mode

4 years agovici: Add support for certificate policies
Tobias Brunner [Wed, 16 Nov 2016 14:37:23 +0000 (15:37 +0100)]
vici: Add support for certificate policies

4 years agovici: Add missing dscp setting for IKE_SAs
Tobias Brunner [Fri, 11 Nov 2016 09:40:53 +0000 (10:40 +0100)]
vici: Add missing dscp setting for IKE_SAs

Fixes #2170.

4 years agoswanctl: Automatically unload removed shared keys
Tobias Brunner [Wed, 9 Nov 2016 15:49:35 +0000 (16:49 +0100)]
swanctl: Automatically unload removed shared keys

4 years agovici: Add possibility to remove shared keys by a unique identifier
Tobias Brunner [Wed, 9 Nov 2016 15:27:01 +0000 (16:27 +0100)]
vici: Add possibility to remove shared keys by a unique identifier

This identifier can be set when adding/replacing a secret.  The unique
identifiers of all secrets may be enumerated.

4 years agomem-cred: Add methods to add/remove shared keys with unique identifiers
Tobias Brunner [Wed, 9 Nov 2016 15:20:03 +0000 (16:20 +0100)]
mem-cred: Add methods to add/remove shared keys with unique identifiers

Also added is a method to enumerate the unique identifiers.

4 years agoswanctl: Automatically unload removed private keys
Tobias Brunner [Wed, 9 Nov 2016 11:25:00 +0000 (12:25 +0100)]
swanctl: Automatically unload removed private keys

4 years agovici: Add commands to enumerate and remove private keys
Tobias Brunner [Wed, 9 Nov 2016 10:49:32 +0000 (11:49 +0100)]
vici: Add commands to enumerate and remove private keys

They are identified by their SHA-1 key identifier.

4 years agomem-cred: Add method to remove a private key with a specific fingerprint
Tobias Brunner [Wed, 9 Nov 2016 10:22:11 +0000 (11:22 +0100)]
mem-cred: Add method to remove a private key with a specific fingerprint

4 years agoswanctl: Add possibility to query a specific pool by name
Tobias Brunner [Wed, 9 Nov 2016 09:37:56 +0000 (10:37 +0100)]
swanctl: Add possibility to query a specific pool by name

4 years agovici: Update get_pools() in Python and Ruby bindings
Tobias Brunner [Thu, 8 Dec 2016 17:14:40 +0000 (18:14 +0100)]
vici: Update get_pools() in Python and Ruby bindings

4 years agovici: Add option to query a specific pool
Tobias Brunner [Wed, 9 Nov 2016 09:18:01 +0000 (10:18 +0100)]
vici: Add option to query a specific pool

4 years agobypass-lan: Don't use interfaces in policies
Tobias Brunner [Mon, 13 Feb 2017 18:06:24 +0000 (19:06 +0100)]
bypass-lan: Don't use interfaces in policies

After an interface disappeared we can't remove the policies correctly as
the name doesn't resolve to the previous index anymore.
And making the policies so specific might not provide that much benefit.

To handle the interfaces on the policies correctly would require some
changes to the child-cfg, kernel-interface etc. so they'd take interface
indices directly so we could target the policies correctly even if an
interface disappeared (or reappeared and got a new index).

4 years agotesting: Fix ALLOWED_HOSTS in strongTNC settings.ini
Tobias Brunner [Thu, 16 Feb 2017 17:24:25 +0000 (18:24 +0100)]
testing: Fix ALLOWED_HOSTS in strongTNC settings.ini

4 years agotesting: Fix swanctl/ocsp-disabled scenario after changing the log messages
Tobias Brunner [Thu, 16 Feb 2017 16:51:16 +0000 (17:51 +0100)]
testing: Fix swanctl/ocsp-disabled scenario after changing the log messages

4 years agorevocation: More accurately describe the flags to disable OCSP/CRL validation
Tobias Brunner [Wed, 25 Jan 2017 15:17:38 +0000 (16:17 +0100)]
revocation: More accurately describe the flags to disable OCSP/CRL validation

These options disable validation as such, e.g. even from cached CRLs, not
only the fetching.  Also made the plugin's validate() implementation a
no-op if both options are disabled.

4 years agochild-sa: Do not install mark on inbound kernel SA
Eyal Birger [Wed, 25 Jan 2017 10:26:42 +0000 (12:26 +0200)]
child-sa: Do not install mark on inbound kernel SA

The SA ID (src, dst, proto, spi) is unique on ingress.

As such, explicit inbound marking is not needed to match an SA.

On the other hand, requiring inbound SAs to use marks forces the
installation of a mechanism for marking traffic (e.g. iptables) based
on some criteria.

Defining the criteria becomes complicated, for example when required to
support multiple SAs from the same src, especially when traffic is UDP

This commit removes the assignment of the child_sa mark_in to the inbound SA.

Policies can be arbitrated by existing means - e.g, via netfilter policy
matching or using VTI interfaces - without the need to classify the flows prior
to state matching.

Since the reqid allocator regards the mark value, there is no risk of matching
the wrong policy.

And as explicit marking was required for route-based VPN to work before this
change, it should not cause regressions in existing setups.

Closes strongswan/strongswan#59.

4 years agounit-tests: Allow default test timeout to be configured via compile option
Thomas Egerer [Fri, 10 Feb 2017 14:27:11 +0000 (15:27 +0100)]
unit-tests: Allow default test timeout to be configured via compile option

Signed-off-by: Thomas Egerer <>
4 years agotkm: Fix get_auth_octets() signature
Tobias Brunner [Mon, 13 Feb 2017 17:36:01 +0000 (18:36 +0100)]
tkm: Fix get_auth_octets() signature

Fixes: 267c1f7083d4 ("keymat: Allow keymat to modify signature scheme(s)")

4 years agokernel-netlink: Use RTA_SRC to specify route source in kernel-based lookups
Martin Willi [Thu, 19 Jan 2017 10:23:45 +0000 (11:23 +0100)]
kernel-netlink: Use RTA_SRC to specify route source in kernel-based lookups

For table dumps the kernel accepts RTA_PREFSRC to filter the routes, which is
what we do when doing userspace route calculations. For kernel-based route
lookups, however, the RTA_PREFSRC attribute is ignored and we must specify
RTA_SRC for policy based route lookups.

4 years agokernel-netlink: Use kernel-based route lookup if we do not install routes
Martin Willi [Thu, 19 Jan 2017 10:03:55 +0000 (11:03 +0100)]
kernel-netlink: Use kernel-based route lookup if we do not install routes

For gateways with many connections, installing routes is often disabled,
as we can use a static route configuration to achieve proper routing with
a single rule. If this is the case, there is no need to dump all routes and
do userspace route lookups, as there is no need to exclude routes we installed

Doing kernel-based route lookups is not only faster with may routes, but also
can use the full power of Linux policy based routing; something we can hardly
rebuild in userspace when calculating routes.

4 years agoswanctl: List CHILD_SA marks, if set
Martin Willi [Fri, 6 Jan 2017 12:01:34 +0000 (13:01 +0100)]
swanctl: List CHILD_SA marks, if set

4 years agovici: Include the Netfilter marks in listed CHILD_SAs
Martin Willi [Fri, 6 Jan 2017 11:42:04 +0000 (12:42 +0100)]
vici: Include the Netfilter marks in listed CHILD_SAs

4 years agovici: Explicitly set the Python encoding type
Martin Willi [Tue, 8 Dec 2015 16:13:59 +0000 (17:13 +0100)]
vici: Explicitly set the Python encoding type

When using vici over RPyC and its (awesome) splitbrain, encoding and decoding
strings fails in vici, most likely because of the Monkey-Patch magic splitbrain

When specifying the implicit UTF-8 as encoding scheme explicitly, Python uses
the correct method to encode/decode the string, making vici useable in
splitbrain contexts.

4 years agoMerge branch 'mid-sync'
Tobias Brunner [Wed, 8 Feb 2017 14:11:20 +0000 (15:11 +0100)]
Merge branch 'mid-sync'

Adds support for handling IKEV2_MESSAGE_ID_SYNC notifies as responder
(usually the original initiator) as defined in RFC 6311.  Some HA solutions
use these notifies to set the new IKEv2 message IDs after a failover event.

4 years agounit-tests: Add test cases for MID sync exchanges
Tobias Brunner [Tue, 4 Oct 2016 17:35:51 +0000 (19:35 +0200)]
unit-tests: Add test cases for MID sync exchanges

4 years agoikev2: Ignore IKEV2_MESSAGE_ID_SYNC notifies if extension is disabled
Tobias Brunner [Tue, 4 Oct 2016 15:07:30 +0000 (17:07 +0200)]
ikev2: Ignore IKEV2_MESSAGE_ID_SYNC notifies if extension is disabled

If this is the first message by the peer, i.e. we expect MID 0, the
message is not pre-processed in the task manager so we ignore it in the

We also make sure to ignore such messages if the extension is disabled
and the peer already sent us one INFORMATIONAL, e.g. a DPD (we'd otherwise
consider the message with MID 0 as a retransmit).

4 years agoikev2: Don't increase expected MID after handling MID sync message
Tobias Brunner [Tue, 4 Oct 2016 13:15:36 +0000 (15:15 +0200)]
ikev2: Don't increase expected MID after handling MID sync message

If the responder never sent a message the expected MID is 0.  While
the sent MID (M1) SHOULD be increased beyond the known value, it's
not necessarily the case.
Since M2 - 1 would then equal UINT_MAX setting that MID would get ignored
and while we'd return 0 in the notify we'd actually expect 1 afterwards.

4 years agoikev2: Don't cache response to MID sync request
Tobias Brunner [Mon, 19 Sep 2016 09:16:06 +0000 (11:16 +0200)]
ikev2: Don't cache response to MID sync request

4 years agoikev2: Accept INFORMATIONAL messages with MID 0 if used to sync MIDs
Tobias Brunner [Fri, 16 Sep 2016 15:44:39 +0000 (17:44 +0200)]
ikev2: Accept INFORMATIONAL messages with MID 0 if used to sync MIDs

We are very picky to only allow MID 0 for these messages (while we
currently don't support IPSEC_REPLAY_COUNTER_SYNC notifies we accept

4 years agoikev2: Negotiate support for IKE message ID synchronisation during IKE_AUTH
Tobias Brunner [Fri, 16 Sep 2016 15:37:59 +0000 (17:37 +0200)]
ikev2: Negotiate support for IKE message ID synchronisation during IKE_AUTH

4 years agoikev2: Add task to handle IKEV2_MESSAGE_ID_SYNC notifies as responder
Tobias Brunner [Fri, 16 Sep 2016 15:26:41 +0000 (17:26 +0200)]
ikev2: Add task to handle IKEV2_MESSAGE_ID_SYNC notifies as responder

4 years agoike: Publish getter for the current message ID on IKE_SA
Tobias Brunner [Fri, 16 Sep 2016 14:19:25 +0000 (16:19 +0200)]
ike: Publish getter for the current message ID on IKE_SA