Imported Upstream version 5.1.0

3 files changed, 1132 insertions, 158 deletions

diff --git a/src/libhydra/plugins/kernel_pfroute/Makefile.am b/src/libhydra/plugins/kernel_pfroute/Makefile.am
index df3109eb8..9d1621366 100644
--- a/src/libhydra/plugins/kernel_pfroute/Makefile.am
+++ b/src/libhydra/plugins/kernel_pfroute/Makefile.am
@@ -1,8 +1,10 @@
-
-INCLUDES = -I${linux_headers} -I$(top_srcdir)/src/libstrongswan \
+AM_CPPFLAGS = \
+	-I${linux_headers} \
+	-I$(top_srcdir)/src/libstrongswan \
 	-I$(top_srcdir)/src/libhydra
 
-AM_CFLAGS = -rdynamic
+AM_CFLAGS = \
+	-rdynamic
 
 if MONOLITHIC
 noinst_LTLIBRARIES = libstrongswan-kernel-pfroute.la
diff --git a/src/libhydra/plugins/kernel_pfroute/Makefile.in b/src/libhydra/plugins/kernel_pfroute/Makefile.in
index a4895df2a..b0324ac18 100644
--- a/src/libhydra/plugins/kernel_pfroute/Makefile.in
+++ b/src/libhydra/plugins/kernel_pfroute/Makefile.in
@@ -62,7 +62,7 @@ am__aclocal_m4_deps = $(top_srcdir)/m4/config/libtool.m4 \
 	$(top_srcdir)/m4/macros/with.m4 \
 	$(top_srcdir)/m4/macros/enable-disable.m4 \
 	$(top_srcdir)/m4/macros/add-plugin.m4 \
-	$(top_srcdir)/configure.in
+	$(top_srcdir)/configure.ac
 am__configure_deps = $(am__aclocal_m4_deps) $(CONFIGURE_DEPENDENCIES) \
 	$(ACLOCAL_M4)
 mkinstalldirs = $(install_sh) -d
@@ -103,7 +103,10 @@ am_libstrongswan_kernel_pfroute_la_OBJECTS = kernel_pfroute_plugin.lo \
 	kernel_pfroute_net.lo
 libstrongswan_kernel_pfroute_la_OBJECTS =  \
 	$(am_libstrongswan_kernel_pfroute_la_OBJECTS)
-libstrongswan_kernel_pfroute_la_LINK = $(LIBTOOL) --tag=CC \
+AM_V_lt = $(am__v_lt_@AM_V@)
+am__v_lt_ = $(am__v_lt_@AM_DEFAULT_V@)
+am__v_lt_0 = --silent
+libstrongswan_kernel_pfroute_la_LINK = $(LIBTOOL) $(AM_V_lt) --tag=CC \
 	$(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=link $(CCLD) \
 	$(AM_CFLAGS) $(CFLAGS) \
 	$(libstrongswan_kernel_pfroute_la_LDFLAGS) $(LDFLAGS) -o $@
@@ -116,13 +119,26 @@ am__depfiles_maybe = depfiles
 am__mv = mv -f
 COMPILE = $(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) \
 	$(CPPFLAGS) $(AM_CFLAGS) $(CFLAGS)
-LTCOMPILE = $(LIBTOOL) --tag=CC $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) \
-	--mode=compile $(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) \
-	$(AM_CPPFLAGS) $(CPPFLAGS) $(AM_CFLAGS) $(CFLAGS)
+LTCOMPILE = $(LIBTOOL) $(AM_V_lt) --tag=CC $(AM_LIBTOOLFLAGS) \
+	$(LIBTOOLFLAGS) --mode=compile $(CC) $(DEFS) \
+	$(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) \
+	$(AM_CFLAGS) $(CFLAGS)
+AM_V_CC = $(am__v_CC_@AM_V@)
+am__v_CC_ = $(am__v_CC_@AM_DEFAULT_V@)
+am__v_CC_0 = @echo "  CC    " $@;
+AM_V_at = $(am__v_at_@AM_V@)
+am__v_at_ = $(am__v_at_@AM_DEFAULT_V@)
+am__v_at_0 = @
 CCLD = $(CC)
-LINK = $(LIBTOOL) --tag=CC $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) \
-	--mode=link $(CCLD) $(AM_CFLAGS) $(CFLAGS) $(AM_LDFLAGS) \
-	$(LDFLAGS) -o $@
+LINK = $(LIBTOOL) $(AM_V_lt) --tag=CC $(AM_LIBTOOLFLAGS) \
+	$(LIBTOOLFLAGS) --mode=link $(CCLD) $(AM_CFLAGS) $(CFLAGS) \
+	$(AM_LDFLAGS) $(LDFLAGS) -o $@
+AM_V_CCLD = $(am__v_CCLD_@AM_V@)
+am__v_CCLD_ = $(am__v_CCLD_@AM_DEFAULT_V@)
+am__v_CCLD_0 = @echo "  CCLD  " $@;
+AM_V_GEN = $(am__v_GEN_@AM_V@)
+am__v_GEN_ = $(am__v_GEN_@AM_DEFAULT_V@)
+am__v_GEN_0 = @echo "  GEN   " $@;
 SOURCES = $(libstrongswan_kernel_pfroute_la_SOURCES)
 DIST_SOURCES = $(libstrongswan_kernel_pfroute_la_SOURCES)
 am__can_run_installinfo = \
@@ -136,6 +152,7 @@ DISTFILES = $(DIST_COMMON) $(DIST_SOURCES) $(TEXINFOS) $(EXTRA_DIST)
 ACLOCAL = @ACLOCAL@
 ALLOCA = @ALLOCA@
 AMTAR = @AMTAR@
+AM_DEFAULT_VERBOSITY = @AM_DEFAULT_VERBOSITY@
 AR = @AR@
 AUTOCONF = @AUTOCONF@
 AUTOHEADER = @AUTOHEADER@
@@ -148,6 +165,8 @@ CCDEPMODE = @CCDEPMODE@
 CFLAGS = @CFLAGS@
 CHECK_CFLAGS = @CHECK_CFLAGS@
 CHECK_LIBS = @CHECK_LIBS@
+COVERAGE_CFLAGS = @COVERAGE_CFLAGS@
+COVERAGE_LDFLAGS = @COVERAGE_LDFLAGS@
 CPP = @CPP@
 CPPFLAGS = @CPPFLAGS@
 CYGPATH_W = @CYGPATH_W@
@@ -163,6 +182,7 @@ ECHO_T = @ECHO_T@
 EGREP = @EGREP@
 EXEEXT = @EXEEXT@
 FGREP = @FGREP@
+GENHTML = @GENHTML@
 GPERF = @GPERF@
 GPRBUILD = @GPRBUILD@
 GREP = @GREP@
@@ -171,6 +191,7 @@ INSTALL_DATA = @INSTALL_DATA@
 INSTALL_PROGRAM = @INSTALL_PROGRAM@
 INSTALL_SCRIPT = @INSTALL_SCRIPT@
 INSTALL_STRIP_PROGRAM = @INSTALL_STRIP_PROGRAM@
+LCOV = @LCOV@
 LD = @LD@
 LDFLAGS = @LDFLAGS@
 LEX = @LEX@
@@ -217,6 +238,7 @@ SET_MAKE = @SET_MAKE@
 SHELL = @SHELL@
 SOCKLIB = @SOCKLIB@
 STRIP = @STRIP@
+UNWINDLIB = @UNWINDLIB@
 VERSION = @VERSION@
 YACC = @YACC@
 YFLAGS = @YFLAGS@
@@ -245,6 +267,7 @@ charon_natt_port = @charon_natt_port@
 charon_plugins = @charon_plugins@
 charon_udp_port = @charon_udp_port@
 clearsilver_LIBS = @clearsilver_LIBS@
+cmd_plugins = @cmd_plugins@
 datadir = @datadir@
 datarootdir = @datarootdir@
 dbusservicedir = @dbusservicedir@
@@ -322,10 +345,14 @@ top_srcdir = @top_srcdir@
 urandom_device = @urandom_device@
 xml_CFLAGS = @xml_CFLAGS@
 xml_LIBS = @xml_LIBS@
-INCLUDES = -I${linux_headers} -I$(top_srcdir)/src/libstrongswan \
+AM_CPPFLAGS = \
+	-I${linux_headers} \
+	-I$(top_srcdir)/src/libstrongswan \
 	-I$(top_srcdir)/src/libhydra
 
-AM_CFLAGS = -rdynamic
+AM_CFLAGS = \
+	-rdynamic
+
 @MONOLITHIC_TRUE@noinst_LTLIBRARIES = libstrongswan-kernel-pfroute.la
 @MONOLITHIC_FALSE@plugin_LTLIBRARIES = libstrongswan-kernel-pfroute.la
 libstrongswan_kernel_pfroute_la_SOURCES = \
@@ -409,7 +436,7 @@ clean-pluginLTLIBRARIES:
 	  rm -f "$${dir}/so_locations"; \
 	done
 libstrongswan-kernel-pfroute.la: $(libstrongswan_kernel_pfroute_la_OBJECTS) $(libstrongswan_kernel_pfroute_la_DEPENDENCIES) $(EXTRA_libstrongswan_kernel_pfroute_la_DEPENDENCIES) 
-	$(libstrongswan_kernel_pfroute_la_LINK) $(am_libstrongswan_kernel_pfroute_la_rpath) $(libstrongswan_kernel_pfroute_la_OBJECTS) $(libstrongswan_kernel_pfroute_la_LIBADD) $(LIBS)
+	$(AM_V_CCLD)$(libstrongswan_kernel_pfroute_la_LINK) $(am_libstrongswan_kernel_pfroute_la_rpath) $(libstrongswan_kernel_pfroute_la_OBJECTS) $(libstrongswan_kernel_pfroute_la_LIBADD) $(LIBS)
 
 mostlyclean-compile:
 	-rm -f *.$(OBJEXT)
@@ -421,25 +448,25 @@ distclean-compile:
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/kernel_pfroute_plugin.Plo@am__quote@
 
 .c.o:
-@am__fastdepCC_TRUE@	$(COMPILE) -MT $@ -MD -MP -MF $(DEPDIR)/$*.Tpo -c -o $@ $<
-@am__fastdepCC_TRUE@	$(am__mv) $(DEPDIR)/$*.Tpo $(DEPDIR)/$*.Po
-@AMDEP_TRUE@@am__fastdepCC_FALSE@	source='$<' object='$@' libtool=no @AMDEPBACKSLASH@
+@am__fastdepCC_TRUE@	$(AM_V_CC)$(COMPILE) -MT $@ -MD -MP -MF $(DEPDIR)/$*.Tpo -c -o $@ $<
+@am__fastdepCC_TRUE@	$(AM_V_at)$(am__mv) $(DEPDIR)/$*.Tpo $(DEPDIR)/$*.Po
+@AMDEP_TRUE@@am__fastdepCC_FALSE@	$(AM_V_CC)source='$<' object='$@' libtool=no @AMDEPBACKSLASH@
 @AMDEP_TRUE@@am__fastdepCC_FALSE@	DEPDIR=$(DEPDIR) $(CCDEPMODE) $(depcomp) @AMDEPBACKSLASH@
-@am__fastdepCC_FALSE@	$(COMPILE) -c $<
+@am__fastdepCC_FALSE@	$(AM_V_CC@am__nodep@)$(COMPILE) -c $<
 
 .c.obj:
-@am__fastdepCC_TRUE@	$(COMPILE) -MT $@ -MD -MP -MF $(DEPDIR)/$*.Tpo -c -o $@ `$(CYGPATH_W) '$<'`
-@am__fastdepCC_TRUE@	$(am__mv) $(DEPDIR)/$*.Tpo $(DEPDIR)/$*.Po
-@AMDEP_TRUE@@am__fastdepCC_FALSE@	source='$<' object='$@' libtool=no @AMDEPBACKSLASH@
+@am__fastdepCC_TRUE@	$(AM_V_CC)$(COMPILE) -MT $@ -MD -MP -MF $(DEPDIR)/$*.Tpo -c -o $@ `$(CYGPATH_W) '$<'`
+@am__fastdepCC_TRUE@	$(AM_V_at)$(am__mv) $(DEPDIR)/$*.Tpo $(DEPDIR)/$*.Po
+@AMDEP_TRUE@@am__fastdepCC_FALSE@	$(AM_V_CC)source='$<' object='$@' libtool=no @AMDEPBACKSLASH@
 @AMDEP_TRUE@@am__fastdepCC_FALSE@	DEPDIR=$(DEPDIR) $(CCDEPMODE) $(depcomp) @AMDEPBACKSLASH@
-@am__fastdepCC_FALSE@	$(COMPILE) -c `$(CYGPATH_W) '$<'`
+@am__fastdepCC_FALSE@	$(AM_V_CC@am__nodep@)$(COMPILE) -c `$(CYGPATH_W) '$<'`
 
 .c.lo:
-@am__fastdepCC_TRUE@	$(LTCOMPILE) -MT $@ -MD -MP -MF $(DEPDIR)/$*.Tpo -c -o $@ $<
-@am__fastdepCC_TRUE@	$(am__mv) $(DEPDIR)/$*.Tpo $(DEPDIR)/$*.Plo
-@AMDEP_TRUE@@am__fastdepCC_FALSE@	source='$<' object='$@' libtool=yes @AMDEPBACKSLASH@
+@am__fastdepCC_TRUE@	$(AM_V_CC)$(LTCOMPILE) -MT $@ -MD -MP -MF $(DEPDIR)/$*.Tpo -c -o $@ $<
+@am__fastdepCC_TRUE@	$(AM_V_at)$(am__mv) $(DEPDIR)/$*.Tpo $(DEPDIR)/$*.Plo
+@AMDEP_TRUE@@am__fastdepCC_FALSE@	$(AM_V_CC)source='$<' object='$@' libtool=yes @AMDEPBACKSLASH@
 @AMDEP_TRUE@@am__fastdepCC_FALSE@	DEPDIR=$(DEPDIR) $(CCDEPMODE) $(depcomp) @AMDEPBACKSLASH@
-@am__fastdepCC_FALSE@	$(LTCOMPILE) -c -o $@ $<
+@am__fastdepCC_FALSE@	$(AM_V_CC@am__nodep@)$(LTCOMPILE) -c -o $@ $<
 
 mostlyclean-libtool:
 	-rm -f *.lo
diff --git a/src/libhydra/plugins/kernel_pfroute/kernel_pfroute_net.c b/src/libhydra/plugins/kernel_pfroute/kernel_pfroute_net.c
index 7ac3e8a3c..976170c57 100644
--- a/src/libhydra/plugins/kernel_pfroute/kernel_pfroute_net.c
+++ b/src/libhydra/plugins/kernel_pfroute/kernel_pfroute_net.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (C) 2009-2012 Tobias Brunner
+ * Copyright (C) 2009-2013 Tobias Brunner
  * Hochschule fuer Technik Rapperswil
  *
  * This program is free software; you can redistribute it and/or modify it
@@ -16,6 +16,7 @@
 #include <sys/types.h>
 #include <sys/socket.h>
 #include <net/if.h>
+#include <net/if_dl.h>
 #include <ifaddrs.h>
 #include <net/route.h>
 #include <unistd.h>
@@ -26,8 +27,10 @@
 #include <hydra.h>
 #include <utils/debug.h>
 #include <networking/host.h>
+#include <networking/tun_device.h>
 #include <threading/thread.h>
 #include <threading/mutex.h>
+#include <threading/condvar.h>
 #include <threading/rwlock.h>
 #include <collections/hashtable.h>
 #include <collections/linked_list.h>
@@ -37,11 +40,21 @@
 #error Cannot compile this plugin on systems where 'struct sockaddr' has no sa_len member.
 #endif
 
+/** properly align sockaddrs */
+#ifdef __APPLE__
+/* Apple always uses 4 bytes */
+#define SA_ALIGN 4
+#else
+/* while on other platforms like FreeBSD it depends on the architecture */
+#define SA_ALIGN sizeof(long)
+#endif
+#define SA_LEN(len) ((len) > 0 ? (((len)+SA_ALIGN-1) & ~(SA_ALIGN-1)) : SA_ALIGN)
+
 /** delay before firing roam events (ms) */
 #define ROAM_DELAY 100
 
-/** buffer size for PF_ROUTE messages */
-#define PFROUTE_BUFFER_SIZE 4096
+/** delay before reinstalling routes (ms) */
+#define ROUTE_DELAY 100
 
 typedef struct addr_entry_t addr_entry_t;
 
@@ -55,9 +68,6 @@ struct addr_entry_t {
 
 	/** virtual IP managed by us */
 	bool virtual;
-
-	/** Number of times this IP is used, if virtual */
-	u_int refcount;
 };
 
 /**
@@ -126,6 +136,9 @@ struct addr_map_entry_t {
 	/** The IP address */
 	host_t *ip;
 
+	/** The address entry for this IP address */
+	addr_entry_t *addr;
+
 	/** The interface this address is installed on */
 	iface_entry_t *iface;
 };
@@ -156,8 +169,17 @@ static bool addr_map_entry_equals(addr_map_entry_t *a, addr_map_entry_t *b)
 static bool addr_map_entry_match_up_and_usable(addr_map_entry_t *a,
 											   addr_map_entry_t *b)
 {
-	return iface_entry_up_and_usable(b->iface) &&
-		   a->ip->ip_equals(a->ip, b->ip);
+	return !b->addr->virtual && iface_entry_up_and_usable(b->iface) &&
+			a->ip->ip_equals(a->ip, b->ip);
+}
+
+/**
+ * Used with get_match this finds an address entry if it is installed as virtual
+ * IP address
+ */
+static bool addr_map_entry_match_virtual(addr_map_entry_t *a, addr_map_entry_t *b)
+{
+	return b->addr->virtual && a->ip->ip_equals(a->ip, b->ip);
 }
 
 /**
@@ -166,7 +188,112 @@ static bool addr_map_entry_match_up_and_usable(addr_map_entry_t *a,
  */
 static bool addr_map_entry_match_up(addr_map_entry_t *a, addr_map_entry_t *b)
 {
-	return iface_entry_up(b->iface) && a->ip->ip_equals(a->ip, b->ip);
+	return !b->addr->virtual && iface_entry_up(b->iface) &&
+			a->ip->ip_equals(a->ip, b->ip);
+}
+
+typedef struct route_entry_t route_entry_t;
+
+/**
+ * Installed routing entry
+ */
+struct route_entry_t {
+	/** Name of the interface the route is bound to */
+	char *if_name;
+
+	/** Gateway for this route */
+	host_t *gateway;
+
+	/** Destination net */
+	chunk_t dst_net;
+
+	/** Destination net prefixlen */
+	u_int8_t prefixlen;
+};
+
+/**
+ * Clone a route_entry_t object.
+ */
+static route_entry_t *route_entry_clone(route_entry_t *this)
+{
+	route_entry_t *route;
+
+	INIT(route,
+		.if_name = strdup(this->if_name),
+		.gateway = this->gateway ? this->gateway->clone(this->gateway) : NULL,
+		.dst_net = chunk_clone(this->dst_net),
+		.prefixlen = this->prefixlen,
+	);
+	return route;
+}
+
+/**
+ * Destroy a route_entry_t object
+ */
+static void route_entry_destroy(route_entry_t *this)
+{
+	free(this->if_name);
+	DESTROY_IF(this->gateway);
+	chunk_free(&this->dst_net);
+	free(this);
+}
+
+/**
+ * Hash a route_entry_t object
+ */
+static u_int route_entry_hash(route_entry_t *this)
+{
+	return chunk_hash_inc(chunk_from_thing(this->prefixlen),
+						  chunk_hash(this->dst_net));
+}
+
+/**
+ * Compare two route_entry_t objects
+ */
+static bool route_entry_equals(route_entry_t *a, route_entry_t *b)
+{
+	if (a->if_name && b->if_name && streq(a->if_name, b->if_name) &&
+		chunk_equals(a->dst_net, b->dst_net) && a->prefixlen == b->prefixlen)
+	{
+		return (!a->gateway && !b->gateway) || (a->gateway && b->gateway &&
+					a->gateway->ip_equals(a->gateway, b->gateway));
+	}
+	return FALSE;
+}
+
+typedef struct net_change_t net_change_t;
+
+/**
+ * Queued network changes
+ */
+struct net_change_t {
+	/** Name of the interface that got activated (or an IP appeared on) */
+	char *if_name;
+};
+
+/**
+ * Destroy a net_change_t object
+ */
+static void net_change_destroy(net_change_t *this)
+{
+	free(this->if_name);
+	free(this);
+}
+
+/**
+ * Hash a net_change_t object
+ */
+static u_int net_change_hash(net_change_t *this)
+{
+	return chunk_hash(chunk_create(this->if_name, strlen(this->if_name)));
+}
+
+/**
+ * Compare two net_change_t objects
+ */
+static bool net_change_equals(net_change_t *a, net_change_t *b)
+{
+	return streq(a->if_name, b->if_name);
 }
 
 typedef struct private_kernel_pfroute_net_t private_kernel_pfroute_net_t;
@@ -197,19 +324,54 @@ struct private_kernel_pfroute_net_t
 	hashtable_t *addrs;
 
 	/**
-	 * mutex to lock access to the PF_ROUTE socket
+	 * List of tun devices we installed for virtual IPs
 	 */
-	mutex_t *mutex_pfroute;
+	linked_list_t *tuns;
 
 	/**
-	 * PF_ROUTE socket to communicate with the kernel
+	 * mutex to communicate exclusively with PF_KEY
 	 */
-	int socket;
+	mutex_t *mutex;
+
+	/**
+	 * condvar to signal if PF_KEY query got a response
+	 */
+	condvar_t *condvar;
+
+	/**
+	 * installed routes
+	 */
+	hashtable_t *routes;
+
+	/**
+	 * mutex for routes
+	 */
+	mutex_t *routes_lock;
+
+	/**
+	 * interface changes which may trigger route reinstallation
+	 */
+	hashtable_t *net_changes;
+
+	/**
+	 * mutex for route reinstallation triggers
+	 */
+	mutex_t *net_changes_lock;
+
+	/**
+	 * time of last route reinstallation
+	 */
+	timeval_t last_route_reinstall;
+
+	/**
+	 * pid to send PF_ROUTE messages with
+	 */
+	pid_t pid;
 
 	/**
-	 * PF_ROUTE socket to receive events
+	 * PF_ROUTE socket to communicate with the kernel
 	 */
-	int socket_events;
+	int socket;
 
 	/**
 	 * sequence number for messages sent to the kernel
@@ -217,11 +379,121 @@ struct private_kernel_pfroute_net_t
 	int seq;
 
 	/**
+	 * Sequence number a query is waiting for
+	 */
+	int waiting_seq;
+
+	/**
+	 * Allocated reply message from kernel
+	 */
+	struct rt_msghdr *reply;
+
+	/**
 	 * time of last roam event
 	 */
 	timeval_t last_roam;
+
+	/**
+	 * Time in ms to wait for IP addresses to appear/disappear
+	 */
+	int vip_wait;
 };
 
+
+/**
+ * Forward declaration
+ */
+static status_t manage_route(private_kernel_pfroute_net_t *this, int op,
+							 chunk_t dst_net, u_int8_t prefixlen,
+							 host_t *gateway, char *if_name);
+
+/**
+ * Clear the queued network changes.
+ */
+static void net_changes_clear(private_kernel_pfroute_net_t *this)
+{
+	enumerator_t *enumerator;
+	net_change_t *change;
+
+	enumerator = this->net_changes->create_enumerator(this->net_changes);
+	while (enumerator->enumerate(enumerator, NULL, (void**)&change))
+	{
+		this->net_changes->remove_at(this->net_changes, enumerator);
+		net_change_destroy(change);
+	}
+	enumerator->destroy(enumerator);
+}
+
+/**
+ * Act upon queued network changes.
+ */
+static job_requeue_t reinstall_routes(private_kernel_pfroute_net_t *this)
+{
+	enumerator_t *enumerator;
+	route_entry_t *route;
+
+	this->net_changes_lock->lock(this->net_changes_lock);
+	this->routes_lock->lock(this->routes_lock);
+
+	enumerator = this->routes->create_enumerator(this->routes);
+	while (enumerator->enumerate(enumerator, NULL, (void**)&route))
+	{
+		net_change_t *change, lookup = {
+			.if_name = route->if_name,
+		};
+		/* check if a change for the outgoing interface is queued */
+		change = this->net_changes->get(this->net_changes, &lookup);
+		if (change)
+		{
+			manage_route(this, RTM_ADD, route->dst_net, route->prefixlen,
+						 route->gateway, route->if_name);
+		}
+	}
+	enumerator->destroy(enumerator);
+	this->routes_lock->unlock(this->routes_lock);
+
+	net_changes_clear(this);
+	this->net_changes_lock->unlock(this->net_changes_lock);
+	return JOB_REQUEUE_NONE;
+}
+
+/**
+ * Queue route reinstallation caused by network changes for a given interface.
+ *
+ * The route reinstallation is delayed for a while and only done once for
+ * several calls during this delay, in order to avoid doing it too often.
+ * The interface name is freed.
+ */
+static void queue_route_reinstall(private_kernel_pfroute_net_t *this,
+								  char *if_name)
+{
+	net_change_t *update, *found;
+	timeval_t now;
+	job_t *job;
+
+	INIT(update,
+		.if_name = if_name
+	);
+
+	this->net_changes_lock->lock(this->net_changes_lock);
+	found = this->net_changes->put(this->net_changes, update, update);
+	if (found)
+	{
+		net_change_destroy(found);
+	}
+	time_monotonic(&now);
+	if (timercmp(&now, &this->last_route_reinstall, >))
+	{
+		timeval_add_ms(&now, ROUTE_DELAY);
+		this->last_route_reinstall = now;
+
+		job = (job_t*)callback_job_create((callback_job_cb_t)reinstall_routes,
+										  this, NULL, NULL);
+		lib->scheduler->schedule_job_ms(lib->scheduler, job, ROUTE_DELAY);
+	}
+	this->net_changes_lock->unlock(this->net_changes_lock);
+}
+
 /**
  * Add an address map entry
  */
@@ -230,13 +502,9 @@ static void addr_map_entry_add(private_kernel_pfroute_net_t *this,
 {
 	addr_map_entry_t *entry;
 
-	if (addr->virtual)
-	{	/* don't map virtual IPs */
-		return;
-	}
-
 	INIT(entry,
 		.ip = addr->ip,
+		.addr = addr,
 		.iface = iface,
 	);
 	entry = this->addrs->put(this->addrs, entry, entry);
@@ -252,14 +520,10 @@ static void addr_map_entry_remove(addr_entry_t *addr, iface_entry_t *iface,
 {
 	addr_map_entry_t *entry, lookup = {
 		.ip = addr->ip,
+		.addr = addr,
 		.iface = iface,
 	};
 
-	if (addr->virtual)
-	{	/* these are never mapped, but this check avoid problems if a virtual IP
-		 * equals a regular one */
-		return;
-	}
 	entry = this->addrs->remove(this->addrs, &lookup);
 	free(entry);
 }
@@ -296,35 +560,114 @@ static void fire_roam_event(private_kernel_pfroute_net_t *this, bool address)
 }
 
 /**
+ * Data for enumerator over rtmsg sockaddrs
+ */
+typedef struct {
+	/** implements enumerator */
+	enumerator_t public;
+	/** copy of attribute bitfield */
+	int types;
+	/** bytes remaining in buffer */
+	int remaining;
+	/** next sockaddr to enumerate */
+	struct sockaddr *addr;
+} rt_enumerator_t;
+
+METHOD(enumerator_t, rt_enumerate, bool,
+	rt_enumerator_t *this, int *xtype, struct sockaddr **addr)
+{
+	int i, type;
+
+	if (this->remaining < sizeof(this->addr->sa_len) ||
+		this->remaining < this->addr->sa_len)
+	{
+		return FALSE;
+	}
+	for (i = 0; i < RTAX_MAX; i++)
+	{
+		type = (1 << i);
+		if (this->types & type)
+		{
+			this->types &= ~type;
+			*addr = this->addr;
+			*xtype = i;
+			this->remaining -= SA_LEN(this->addr->sa_len);
+			this->addr = (struct sockaddr*)((char*)this->addr +
+											SA_LEN(this->addr->sa_len));
+			return TRUE;
+		}
+	}
+	return FALSE;
+}
+
+/**
+ * Create an enumerator over sockaddrs in rt/if messages
+ */
+static enumerator_t *create_rt_enumerator(int types, int remaining,
+										  struct sockaddr *addr)
+{
+	rt_enumerator_t *this;
+
+	INIT(this,
+		.public = {
+			.enumerate = (void*)_rt_enumerate,
+			.destroy = (void*)free,
+		},
+		.types = types,
+		.remaining = remaining,
+		.addr = addr,
+	);
+	return &this->public;
+}
+
+/**
+ * Create a safe enumerator over sockaddrs in rt_msghdr
+ */
+static enumerator_t *create_rtmsg_enumerator(struct rt_msghdr *hdr)
+{
+	return create_rt_enumerator(hdr->rtm_addrs, hdr->rtm_msglen - sizeof(*hdr),
+								(struct sockaddr *)(hdr + 1));
+}
+
+/**
+ * Create a safe enumerator over sockaddrs in ifa_msghdr
+ */
+static enumerator_t *create_ifamsg_enumerator(struct ifa_msghdr *hdr)
+{
+	return create_rt_enumerator(hdr->ifam_addrs, hdr->ifam_msglen - sizeof(*hdr),
+								(struct sockaddr *)(hdr + 1));
+}
+
+/**
  * Process an RTM_*ADDR message from the kernel
  */
 static void process_addr(private_kernel_pfroute_net_t *this,
-						 struct rt_msghdr *msg)
+						 struct ifa_msghdr *ifa)
 {
-	struct ifa_msghdr *ifa = (struct ifa_msghdr*)msg;
-	sockaddr_t *sockaddr = (sockaddr_t*)(ifa + 1);
+	struct sockaddr *sockaddr;
 	host_t *host = NULL;
 	enumerator_t *ifaces, *addrs;
 	iface_entry_t *iface;
 	addr_entry_t *addr;
 	bool found = FALSE, changed = FALSE, roam = FALSE;
-	int i;
+	enumerator_t *enumerator;
+	char *ifname = NULL;
+	int type;
 
-	for (i = 1; i < (1 << RTAX_MAX); i <<= 1)
+	enumerator = create_ifamsg_enumerator(ifa);
+	while (enumerator->enumerate(enumerator, &type, &sockaddr))
 	{
-		if (ifa->ifam_addrs & i)
+		if (type == RTAX_IFA)
 		{
-			if (RTA_IFA & i)
-			{
-				host = host_create_from_sockaddr(sockaddr);
-				break;
-			}
-			sockaddr = (sockaddr_t*)((char*)sockaddr + sockaddr->sa_len);
+			host = host_create_from_sockaddr(sockaddr);
+			break;
 		}
 	}
+	enumerator->destroy(enumerator);
 
-	if (!host)
+	if (!host || host->is_anyaddr(host))
 	{
+		DESTROY_IF(host);
 		return;
 	}
 
@@ -352,21 +695,17 @@ static void process_addr(private_kernel_pfroute_net_t *this,
 						addr_map_entry_remove(addr, iface, this);
 						addr_entry_destroy(addr);
 					}
-					else if (ifa->ifam_type == RTM_NEWADDR && addr->virtual)
-					{
-						addr->refcount = 1;
-					}
 				}
 			}
 			addrs->destroy(addrs);
 
 			if (!found && ifa->ifam_type == RTM_NEWADDR)
 			{
+				INIT(addr,
+					.ip = host->clone(host),
+				);
 				changed = TRUE;
-				addr = malloc_thing(addr_entry_t);
-				addr->ip = host->clone(host);
-				addr->virtual = FALSE;
-				addr->refcount = 1;
+				ifname = strdup(iface->ifname);
 				iface->addrs->insert_last(iface->addrs, addr);
 				addr_map_entry_add(this, addr, iface);
 				if (iface->usable)
@@ -386,6 +725,15 @@ static void process_addr(private_kernel_pfroute_net_t *this,
 	this->lock->unlock(this->lock);
 	host->destroy(host);
 
+	if (roam && ifname)
+	{
+		queue_route_reinstall(this, ifname);
+	}
+	else
+	{
+		free(ifname);
+	}
+
 	if (roam)
 	{
 		fire_roam_event(this, TRUE);
@@ -393,15 +741,54 @@ static void process_addr(private_kernel_pfroute_net_t *this,
 }
 
 /**
+ * Re-initialize address list of an interface if it changes state
+ */
+static void repopulate_iface(private_kernel_pfroute_net_t *this,
+							 iface_entry_t *iface)
+{
+	struct ifaddrs *ifap, *ifa;
+	addr_entry_t *addr;
+
+	while (iface->addrs->remove_last(iface->addrs, (void**)&addr) == SUCCESS)
+	{
+		addr_map_entry_remove(addr, iface, this);
+		addr_entry_destroy(addr);
+	}
+
+	if (getifaddrs(&ifap) == 0)
+	{
+		for (ifa = ifap; ifa != NULL; ifa = ifa->ifa_next)
+		{
+			if (ifa->ifa_addr && streq(ifa->ifa_name, iface->ifname))
+			{
+				switch (ifa->ifa_addr->sa_family)
+				{
+					case AF_INET:
+					case AF_INET6:
+						INIT(addr,
+							.ip = host_create_from_sockaddr(ifa->ifa_addr),
+						);
+						iface->addrs->insert_last(iface->addrs, addr);
+						addr_map_entry_add(this, addr, iface);
+						break;
+					default:
+						break;
+				}
+			}
+		}
+		freeifaddrs(ifap);
+	}
+}
+
+/**
  * Process an RTM_IFINFO message from the kernel
  */
 static void process_link(private_kernel_pfroute_net_t *this,
-						 struct rt_msghdr *hdr)
+						 struct if_msghdr *msg)
 {
-	struct if_msghdr *msg = (struct if_msghdr*)hdr;
 	enumerator_t *enumerator;
 	iface_entry_t *iface;
-	bool roam = FALSE;
+	bool roam = FALSE, found = FALSE, update_routes = FALSE;
 
 	this->lock->write_lock(this->lock);
 	enumerator = this->ifaces->create_enumerator(this->ifaces);
@@ -413,7 +800,7 @@ static void process_link(private_kernel_pfroute_net_t *this,
 			{
 				if (!(iface->flags & IFF_UP) && (msg->ifm_flags & IFF_UP))
 				{
-					roam = TRUE;
+					roam = update_routes = TRUE;
 					DBG1(DBG_KNL, "interface %s activated", iface->ifname);
 				}
 				else if ((iface->flags & IFF_UP) && !(msg->ifm_flags & IFF_UP))
@@ -423,12 +810,44 @@ static void process_link(private_kernel_pfroute_net_t *this,
 				}
 			}
 			iface->flags = msg->ifm_flags;
+			repopulate_iface(this, iface);
+			found = TRUE;
 			break;
 		}
 	}
 	enumerator->destroy(enumerator);
+
+	if (!found)
+	{
+		INIT(iface,
+			.ifindex = msg->ifm_index,
+			.flags = msg->ifm_flags,
+			.addrs = linked_list_create(),
+		);
+		if (if_indextoname(iface->ifindex, iface->ifname))
+		{
+			DBG1(DBG_KNL, "interface %s appeared", iface->ifname);
+			iface->usable = hydra->kernel_interface->is_interface_usable(
+										hydra->kernel_interface, iface->ifname);
+			repopulate_iface(this, iface);
+			this->ifaces->insert_last(this->ifaces, iface);
+			if (iface->usable)
+			{
+				roam = update_routes = TRUE;
+			}
+		}
+		else
+		{
+			free(iface);
+		}
+	}
 	this->lock->unlock(this->lock);
 
+	if (update_routes)
+	{
+		queue_route_reinstall(this, strdup(iface->ifname));
+	}
+
 	if (roam)
 	{
 		fire_roam_event(this, TRUE);
@@ -445,61 +864,98 @@ static void process_route(private_kernel_pfroute_net_t *this,
 }
 
 /**
- * Receives events from kernel
+ * Receives PF_ROUTE messages from kernel
  */
-static job_requeue_t receive_events(private_kernel_pfroute_net_t *this)
+static bool receive_events(private_kernel_pfroute_net_t *this, int fd,
+						   watcher_event_t event)
 {
-	unsigned char buf[PFROUTE_BUFFER_SIZE];
-	struct rt_msghdr *msg = (struct rt_msghdr*)buf;
-	int len;
-	bool oldstate;
-
-	oldstate = thread_cancelability(TRUE);
-	len = recvfrom(this->socket_events, buf, sizeof(buf), 0, NULL, 0);
-	thread_cancelability(oldstate);
-
+	struct {
+		union {
+			struct rt_msghdr rtm;
+			struct if_msghdr ifm;
+			struct ifa_msghdr ifam;
+		};
+		char buf[sizeof(struct sockaddr_storage) * RTAX_MAX];
+	} msg;
+	int len, hdrlen;
+
+	len = recv(this->socket, &msg, sizeof(msg), MSG_DONTWAIT);
 	if (len < 0)
 	{
 		switch (errno)
 		{
 			case EINTR:
-				/* interrupted, try again */
-				return JOB_REQUEUE_DIRECT;
 			case EAGAIN:
-				/* no data ready, select again */
-				return JOB_REQUEUE_DIRECT;
+				return TRUE;
 			default:
 				DBG1(DBG_KNL, "unable to receive from PF_ROUTE event socket");
 				sleep(1);
-				return JOB_REQUEUE_FAIR;
+				return TRUE;
 		}
 	}
 
-	if (len < sizeof(msg->rtm_msglen) || len < msg->rtm_msglen ||
-		msg->rtm_version != RTM_VERSION)
+	if (len < offsetof(struct rt_msghdr, rtm_flags) || len < msg.rtm.rtm_msglen)
 	{
-		DBG2(DBG_KNL, "received corrupted PF_ROUTE message");
-		return JOB_REQUEUE_DIRECT;
+		DBG1(DBG_KNL, "received invalid PF_ROUTE message");
+		return TRUE;
 	}
-
-	switch (msg->rtm_type)
+	if (msg.rtm.rtm_version != RTM_VERSION)
+	{
+		DBG1(DBG_KNL, "received PF_ROUTE message with unsupported version: %d",
+			 msg.rtm.rtm_version);
+		return TRUE;
+	}
+	switch (msg.rtm.rtm_type)
 	{
 		case RTM_NEWADDR:
 		case RTM_DELADDR:
-			process_addr(this, msg);
+			hdrlen = sizeof(msg.ifam);
 			break;
 		case RTM_IFINFO:
-		/*case RTM_IFANNOUNCE <- what about this*/
-			process_link(this, msg);
+			hdrlen = sizeof(msg.ifm);
 			break;
 		case RTM_ADD:
 		case RTM_DELETE:
-			process_route(this, msg);
+		case RTM_GET:
+			hdrlen = sizeof(msg.rtm);
+			break;
 		default:
+			return TRUE;
+	}
+	if (msg.rtm.rtm_msglen < hdrlen)
+	{
+		DBG1(DBG_KNL, "ignoring short PF_ROUTE message");
+		return TRUE;
+	}
+	switch (msg.rtm.rtm_type)
+	{
+		case RTM_NEWADDR:
+		case RTM_DELADDR:
+			process_addr(this, &msg.ifam);
 			break;
+		case RTM_IFINFO:
+			process_link(this, &msg.ifm);
+			break;
+		case RTM_ADD:
+		case RTM_DELETE:
+			process_route(this, &msg.rtm);
+			break;
+		default:
+			break;
+	}
+
+	this->mutex->lock(this->mutex);
+	if (msg.rtm.rtm_pid == this->pid && msg.rtm.rtm_seq == this->waiting_seq)
+	{
+		/* seems like the message someone is waiting for, deliver */
+		this->reply = realloc(this->reply, msg.rtm.rtm_msglen);
+		memcpy(this->reply, &msg, msg.rtm.rtm_msglen);
 	}
+	/* signal on any event, add_ip()/del_ip() might wait for it */
+	this->condvar->broadcast(this->condvar);
+	this->mutex->unlock(this->mutex);
 
-	return JOB_REQUEUE_DIRECT;
+	return TRUE;
 }
 
 
@@ -530,6 +986,10 @@ static bool filter_addresses(address_enumerator_t *data,
 	{   /* skip virtual interfaces added by us */
 		return FALSE;
 	}
+	if (!(data->which & ADDR_TYPE_REGULAR) && !(*in)->virtual)
+	{	/* address is regular, but not requested */
+		return FALSE;
+	}
 	ip = (*in)->ip;
 	if (ip->get_family(ip) == AF_INET6)
 	{
@@ -578,9 +1038,12 @@ static bool filter_interfaces(address_enumerator_t *data, iface_entry_t** in,
 METHOD(kernel_net_t, create_address_enumerator, enumerator_t*,
 	private_kernel_pfroute_net_t *this, kernel_address_type_t which)
 {
-	address_enumerator_t *data = malloc_thing(address_enumerator_t);
-	data->this = this;
-	data->which = which;
+	address_enumerator_t *data;
+
+	INIT(data,
+		.this = this,
+		.which = which,
+	);
 
 	this->lock->read_lock(this->lock);
 	return enumerator_create_nested(
@@ -591,6 +1054,12 @@ METHOD(kernel_net_t, create_address_enumerator, enumerator_t*,
 				(void*)address_enumerator_destroy);
 }
 
+METHOD(kernel_net_t, get_features, kernel_feature_t,
+	private_kernel_pfroute_net_t *this)
+{
+	return KERNEL_REQUIRE_EXCLUDE_ROUTE;
+}
+
 METHOD(kernel_net_t, get_interface_name, bool,
 	private_kernel_pfroute_net_t *this, host_t* ip, char **name)
 {
@@ -616,6 +1085,19 @@ METHOD(kernel_net_t, get_interface_name, bool,
 		this->lock->unlock(this->lock);
 		return TRUE;
 	}
+	/* check if it is a virtual IP */
+	entry = this->addrs->get_match(this->addrs, &lookup,
+								  (void*)addr_map_entry_match_virtual);
+	if (entry)
+	{
+		if (name)
+		{
+			*name = strdup(entry->iface->ifname);
+			DBG2(DBG_KNL, "virtual IP %H is on interface %s", ip, *name);
+		}
+		this->lock->unlock(this->lock);
+		return TRUE;
+	}
 	/* maybe it is installed on an ignored interface */
 	entry = this->addrs->get_match(this->addrs, &lookup,
 								  (void*)addr_map_entry_match_up);
@@ -627,44 +1109,484 @@ METHOD(kernel_net_t, get_interface_name, bool,
 	return FALSE;
 }
 
-METHOD(kernel_net_t, get_source_addr, host_t*,
-	private_kernel_pfroute_net_t *this, host_t *dest, host_t *src)
+METHOD(kernel_net_t, add_ip, status_t,
+	private_kernel_pfroute_net_t *this, host_t *vip, int prefix,
+	char *ifname)
+{
+	enumerator_t *ifaces, *addrs;
+	iface_entry_t *iface;
+	addr_entry_t *addr;
+	tun_device_t *tun;
+	bool timeout = FALSE;
+
+	tun = tun_device_create(NULL);
+	if (!tun)
+	{
+		return FAILED;
+	}
+	if (prefix == -1)
+	{
+		prefix = vip->get_address(vip).len * 8;
+	}
+	if (!tun->up(tun) || !tun->set_address(tun, vip, prefix))
+	{
+		tun->destroy(tun);
+		return FAILED;
+	}
+
+	/* wait until address appears */
+	this->mutex->lock(this->mutex);
+	while (!timeout && !get_interface_name(this, vip, NULL))
+	{
+		timeout = this->condvar->timed_wait(this->condvar, this->mutex,
+											this->vip_wait);
+	}
+	this->mutex->unlock(this->mutex);
+	if (timeout)
+	{
+		DBG1(DBG_KNL, "virtual IP %H did not appear on %s",
+			 vip, tun->get_name(tun));
+		tun->destroy(tun);
+		return FAILED;
+	}
+
+	this->lock->write_lock(this->lock);
+	this->tuns->insert_last(this->tuns, tun);
+
+	ifaces = this->ifaces->create_enumerator(this->ifaces);
+	while (ifaces->enumerate(ifaces, &iface))
+	{
+		if (streq(iface->ifname, tun->get_name(tun)))
+		{
+			addrs = iface->addrs->create_enumerator(iface->addrs);
+			while (addrs->enumerate(addrs, &addr))
+			{
+				if (addr->ip->ip_equals(addr->ip, vip))
+				{
+					addr->virtual = TRUE;
+				}
+			}
+			addrs->destroy(addrs);
+			/* during IKEv1 reauthentication, children get moved from
+			 * old the new SA before the virtual IP is available. This
+			 * kills the route for our virtual IP, reinstall. */
+			queue_route_reinstall(this, strdup(iface->ifname));
+			break;
+		}
+	}
+	ifaces->destroy(ifaces);
+	/* lets do this while holding the lock, thus preventing another thread
+	 * from deleting the TUN device concurrently, hopefully listeners are quick
+	 * and cause no deadlocks */
+	hydra->kernel_interface->tun(hydra->kernel_interface, tun, TRUE);
+	this->lock->unlock(this->lock);
+
+	return SUCCESS;
+}
+
+METHOD(kernel_net_t, del_ip, status_t,
+	private_kernel_pfroute_net_t *this, host_t *vip, int prefix,
+	bool wait)
 {
-	return NULL;
+	enumerator_t *enumerator;
+	tun_device_t *tun;
+	host_t *addr;
+	bool timeout = FALSE, found = FALSE;
+
+	this->lock->write_lock(this->lock);
+	enumerator = this->tuns->create_enumerator(this->tuns);
+	while (enumerator->enumerate(enumerator, &tun))
+	{
+		addr = tun->get_address(tun, NULL);
+		if (addr && addr->ip_equals(addr, vip))
+		{
+			this->tuns->remove_at(this->tuns, enumerator);
+			hydra->kernel_interface->tun(hydra->kernel_interface, tun,
+										 FALSE);
+			tun->destroy(tun);
+			found = TRUE;
+			break;
+		}
+	}
+	enumerator->destroy(enumerator);
+	this->lock->unlock(this->lock);
+
+	if (!found)
+	{
+		return NOT_FOUND;
+	}
+	/* wait until address disappears */
+	if (wait)
+	{
+		this->mutex->lock(this->mutex);
+		while (!timeout && get_interface_name(this, vip, NULL))
+		{
+			timeout = this->condvar->timed_wait(this->condvar, this->mutex,
+												this->vip_wait);
+		}
+		this->mutex->unlock(this->mutex);
+		if (timeout)
+		{
+			DBG1(DBG_KNL, "virtual IP %H did not disappear from tun", vip);
+			return FAILED;
+		}
+	}
+	return SUCCESS;
 }
 
-METHOD(kernel_net_t, get_nexthop, host_t*,
-	private_kernel_pfroute_net_t *this, host_t *dest, host_t *src)
+/**
+ * Append a sockaddr_in/in6 of given type to routing message
+ */
+static void add_rt_addr(struct rt_msghdr *hdr, int type, host_t *addr)
 {
-	return NULL;
+	if (addr)
+	{
+		int len;
+
+		len = *addr->get_sockaddr_len(addr);
+		memcpy((char*)hdr + hdr->rtm_msglen, addr->get_sockaddr(addr), len);
+		hdr->rtm_msglen += SA_LEN(len);
+		hdr->rtm_addrs |= type;
+	}
 }
 
-METHOD(kernel_net_t, add_ip, status_t,
-	private_kernel_pfroute_net_t *this, host_t *virtual_ip, int prefix,
-	char *iface)
+/**
+ * Append a subnet mask sockaddr using the given prefix to routing message
+ */
+static void add_rt_mask(struct rt_msghdr *hdr, int type, int family, int prefix)
 {
-	return FAILED;
+	host_t *mask;
+
+	mask = host_create_netmask(family, prefix);
+	if (mask)
+	{
+		add_rt_addr(hdr, type, mask);
+		mask->destroy(mask);
+	}
 }
 
-METHOD(kernel_net_t, del_ip, status_t,
-	private_kernel_pfroute_net_t *this, host_t *virtual_ip, int prefix,
-	bool wait)
+/**
+ * Append an interface name sockaddr_dl to routing message
+ */
+static void add_rt_ifname(struct rt_msghdr *hdr, int type, char *name)
 {
-	return FAILED;
+	struct sockaddr_dl sdl = {
+		.sdl_len = sizeof(struct sockaddr_dl),
+		.sdl_family = AF_LINK,
+		.sdl_nlen = strlen(name),
+	};
+
+	if (strlen(name) <= sizeof(sdl.sdl_data))
+	{
+		memcpy(sdl.sdl_data, name, sdl.sdl_nlen);
+		memcpy((char*)hdr + hdr->rtm_msglen, &sdl, sdl.sdl_len);
+		hdr->rtm_msglen += SA_LEN(sdl.sdl_len);
+		hdr->rtm_addrs |= type;
+	}
+}
+
+/**
+ * Add or remove a route
+ */
+static status_t manage_route(private_kernel_pfroute_net_t *this, int op,
+							 chunk_t dst_net, u_int8_t prefixlen,
+							 host_t *gateway, char *if_name)
+{
+	struct {
+		struct rt_msghdr hdr;
+		char buf[sizeof(struct sockaddr_storage) * RTAX_MAX];
+	} msg = {
+		.hdr = {
+			.rtm_version = RTM_VERSION,
+			.rtm_type = op,
+			.rtm_flags = RTF_UP | RTF_STATIC,
+			.rtm_pid = this->pid,
+			.rtm_seq = ref_get(&this->seq),
+		},
+	};
+	host_t *dst;
+	int type;
+
+	if (prefixlen == 0 && dst_net.len)
+	{
+		status_t status;
+		chunk_t half;
+
+		half = chunk_clonea(dst_net);
+		half.ptr[0] |= 0x80;
+		prefixlen = 1;
+		status = manage_route(this, op, half, prefixlen, gateway, if_name);
+		if (status != SUCCESS)
+		{
+			return status;
+		}
+	}
+
+	dst = host_create_from_chunk(AF_UNSPEC, dst_net, 0);
+	if (!dst)
+	{
+		return FAILED;
+	}
+
+	if ((dst->get_family(dst) == AF_INET && prefixlen == 32) ||
+		(dst->get_family(dst) == AF_INET6 && prefixlen == 128))
+	{
+		msg.hdr.rtm_flags |= RTF_HOST | RTF_GATEWAY;
+	}
+
+	msg.hdr.rtm_msglen = sizeof(struct rt_msghdr);
+	for (type = 0; type < RTAX_MAX; type++)
+	{
+		switch (type)
+		{
+			case RTAX_DST:
+				add_rt_addr(&msg.hdr, RTA_DST, dst);
+				break;
+			case RTAX_NETMASK:
+				if (!(msg.hdr.rtm_flags & RTF_HOST))
+				{
+					add_rt_mask(&msg.hdr, RTA_NETMASK,
+								dst->get_family(dst), prefixlen);
+				}
+				break;
+			case RTAX_IFP:
+				if (if_name)
+				{
+					add_rt_ifname(&msg.hdr, RTA_IFP, if_name);
+				}
+				break;
+			case RTAX_GATEWAY:
+				if (gateway)
+				{
+					add_rt_addr(&msg.hdr, RTA_GATEWAY, gateway);
+				}
+				break;
+			default:
+				break;
+		}
+	}
+	dst->destroy(dst);
+
+	if (send(this->socket, &msg, msg.hdr.rtm_msglen, 0) != msg.hdr.rtm_msglen)
+	{
+		if (errno == EEXIST)
+		{
+			return ALREADY_DONE;
+		}
+		DBG1(DBG_KNL, "%s PF_ROUTE route failed: %s",
+			 op == RTM_ADD ? "adding" : "deleting", strerror(errno));
+		return FAILED;
+	}
+	return SUCCESS;
 }
 
 METHOD(kernel_net_t, add_route, status_t,
 	private_kernel_pfroute_net_t *this, chunk_t dst_net, u_int8_t prefixlen,
 	host_t *gateway, host_t *src_ip, char *if_name)
 {
-	return FAILED;
+	status_t status;
+	route_entry_t *found, route = {
+		.dst_net = dst_net,
+		.prefixlen = prefixlen,
+		.gateway = gateway,
+		.if_name = if_name,
+	};
+
+	this->routes_lock->lock(this->routes_lock);
+	found = this->routes->get(this->routes, &route);
+	if (found)
+	{
+		this->routes_lock->unlock(this->routes_lock);
+		return ALREADY_DONE;
+	}
+	found = route_entry_clone(&route);
+	this->routes->put(this->routes, found, found);
+	status = manage_route(this, RTM_ADD, dst_net, prefixlen, gateway, if_name);
+	this->routes_lock->unlock(this->routes_lock);
+	return status;
 }
 
 METHOD(kernel_net_t, del_route, status_t,
 	private_kernel_pfroute_net_t *this, chunk_t dst_net, u_int8_t prefixlen,
 	host_t *gateway, host_t *src_ip, char *if_name)
 {
-	return FAILED;
+	status_t status;
+	route_entry_t *found, route = {
+		.dst_net = dst_net,
+		.prefixlen = prefixlen,
+		.gateway = gateway,
+		.if_name = if_name,
+	};
+
+	this->routes_lock->lock(this->routes_lock);
+	found = this->routes->get(this->routes, &route);
+	if (!found)
+	{
+		this->routes_lock->unlock(this->routes_lock);
+		return NOT_FOUND;
+	}
+	this->routes->remove(this->routes, found);
+	route_entry_destroy(found);
+	status = manage_route(this, RTM_DELETE, dst_net, prefixlen, gateway,
+						  if_name);
+	this->routes_lock->unlock(this->routes_lock);
+	return status;
+}
+
+/**
+ * Do a route lookup for dest and return either the nexthop or the source
+ * address.
+ */
+static host_t *get_route(private_kernel_pfroute_net_t *this, bool nexthop,
+						 host_t *dest, host_t *src)
+{
+	struct {
+		struct rt_msghdr hdr;
+		char buf[sizeof(struct sockaddr_storage) * RTAX_MAX];
+	} msg = {
+		.hdr = {
+			.rtm_version = RTM_VERSION,
+			.rtm_type = RTM_GET,
+			.rtm_pid = this->pid,
+			.rtm_seq = ref_get(&this->seq),
+		},
+	};
+	host_t *host = NULL;
+	enumerator_t *enumerator;
+	struct sockaddr *addr;
+	bool failed = FALSE;
+	int type;
+
+retry:
+	msg.hdr.rtm_msglen = sizeof(struct rt_msghdr);
+	for (type = 0; type < RTAX_MAX; type++)
+	{
+		switch (type)
+		{
+			case RTAX_DST:
+				add_rt_addr(&msg.hdr, RTA_DST, dest);
+				break;
+			case RTAX_IFA:
+				add_rt_addr(&msg.hdr, RTA_IFA, src);
+				break;
+			case RTAX_IFP:
+				if (!nexthop)
+				{	/* add an empty IFP to ensure we get a source address */
+					add_rt_ifname(&msg.hdr, RTA_IFP, "");
+				}
+				break;
+			default:
+				break;
+		}
+	}
+	this->mutex->lock(this->mutex);
+
+	while (this->waiting_seq)
+	{
+		this->condvar->wait(this->condvar, this->mutex);
+	}
+	this->waiting_seq = msg.hdr.rtm_seq;
+	if (send(this->socket, &msg, msg.hdr.rtm_msglen, 0) == msg.hdr.rtm_msglen)
+	{
+		while (TRUE)
+		{
+			if (this->condvar->timed_wait(this->condvar, this->mutex, 1000))
+			{	/* timed out? */
+				break;
+			}
+			if (this->reply->rtm_msglen < sizeof(*this->reply) ||
+				msg.hdr.rtm_seq != this->reply->rtm_seq)
+			{
+				continue;
+			}
+			enumerator = create_rtmsg_enumerator(this->reply);
+			while (enumerator->enumerate(enumerator, &type, &addr))
+			{
+				if (nexthop)
+				{
+					if (type == RTAX_DST && this->reply->rtm_flags & RTF_HOST)
+					{	/* probably a cloned/cached direct route, only use that
+						 * as fallback if no gateway is found */
+						host = host ?: host_create_from_sockaddr(addr);
+					}
+					if (type == RTAX_GATEWAY)
+					{	/* could actually be a MAC address */
+						host_t *gtw = host_create_from_sockaddr(addr);
+						if (gtw)
+						{
+							DESTROY_IF(host);
+							host = gtw;
+						}
+					}
+				}
+				else
+				{
+					if (type == RTAX_IFA)
+					{
+						host = host_create_from_sockaddr(addr);
+					}
+				}
+			}
+			enumerator->destroy(enumerator);
+			break;
+		}
+	}
+	else
+	{
+		failed = TRUE;
+	}
+	/* signal completion of query to a waiting thread */
+	this->waiting_seq = 0;
+	this->condvar->signal(this->condvar);
+	this->mutex->unlock(this->mutex);
+
+	if (failed)
+	{
+		if (src)
+		{	/* the given source address might be gone, try again without */
+			src = NULL;
+			msg.hdr.rtm_seq = ref_get(&this->seq);
+			msg.hdr.rtm_addrs = 0;
+			memset(msg.buf, sizeof(msg.buf), 0);
+			goto retry;
+		}
+		DBG1(DBG_KNL, "PF_ROUTE lookup failed: %s", strerror(errno));
+	}
+	if (!host)
+	{
+		return NULL;
+	}
+	if (!nexthop)
+	{	/* make sure the source address is not virtual and usable */
+		addr_entry_t *entry, lookup = {
+			.ip = host,
+		};
+
+		this->lock->read_lock(this->lock);
+		entry = this->addrs->get_match(this->addrs, &lookup,
+									(void*)addr_map_entry_match_up_and_usable);
+		this->lock->unlock(this->lock);
+		if (!entry)
+		{
+			host->destroy(host);
+			return NULL;
+		}
+	}
+	DBG2(DBG_KNL, "using %H as %s to reach %H", host,
+		 nexthop ? "nexthop" : "address", dest);
+	return host;
+}
+
+METHOD(kernel_net_t, get_source_addr, host_t*,
+	private_kernel_pfroute_net_t *this, host_t *dest, host_t *src)
+{
+	return get_route(this, FALSE, dest, src);
+}
+
+METHOD(kernel_net_t, get_nexthop, host_t*,
+	private_kernel_pfroute_net_t *this, host_t *dest, host_t *src)
+{
+	return get_route(this, TRUE, dest, src);
 }
 
 /**
@@ -711,22 +1633,22 @@ static status_t init_address_list(private_kernel_pfroute_net_t *this)
 
 				if (!iface)
 				{
-					iface = malloc_thing(iface_entry_t);
+					INIT(iface,
+						.ifindex = if_nametoindex(ifa->ifa_name),
+						.flags = ifa->ifa_flags,
+						.addrs = linked_list_create(),
+						.usable = hydra->kernel_interface->is_interface_usable(
+										hydra->kernel_interface, ifa->ifa_name),
+					);
 					memcpy(iface->ifname, ifa->ifa_name, IFNAMSIZ);
-					iface->ifindex = if_nametoindex(ifa->ifa_name);
-					iface->flags = ifa->ifa_flags;
-					iface->addrs = linked_list_create();
-					iface->usable = hydra->kernel_interface->is_interface_usable(
-										hydra->kernel_interface, ifa->ifa_name);
 					this->ifaces->insert_last(this->ifaces, iface);
 				}
 
 				if (ifa->ifa_addr->sa_family != AF_LINK)
 				{
-					addr = malloc_thing(addr_entry_t);
-					addr->ip = host_create_from_sockaddr(ifa->ifa_addr);
-					addr->virtual = FALSE;
-					addr->refcount = 1;
+					INIT(addr,
+						.ip = host_create_from_sockaddr(ifa->ifa_addr),
+					);
 					iface->addrs->insert_last(iface->addrs, addr);
 					addr_map_entry_add(this, addr, iface);
 				}
@@ -758,16 +1680,30 @@ METHOD(kernel_net_t, destroy, void,
 	private_kernel_pfroute_net_t *this)
 {
 	enumerator_t *enumerator;
+	route_entry_t *route;
 	addr_entry_t *addr;
 
-	if (this->socket > 0)
+	enumerator = this->routes->create_enumerator(this->routes);
+	while (enumerator->enumerate(enumerator, NULL, (void**)&route))
 	{
-		close(this->socket);
+		manage_route(this, RTM_DELETE, route->dst_net, route->prefixlen,
+					 route->gateway, route->if_name);
+		route_entry_destroy(route);
 	}
-	if (this->socket_events)
+	enumerator->destroy(enumerator);
+	this->routes->destroy(this->routes);
+	this->routes_lock->destroy(this->routes_lock);
+
+	if (this->socket != -1)
 	{
-		close(this->socket_events);
+		lib->watcher->remove(lib->watcher, this->socket);
+		close(this->socket);
 	}
+
+	net_changes_clear(this);
+	this->net_changes->destroy(this->net_changes);
+	this->net_changes_lock->destroy(this->net_changes_lock);
+
 	enumerator = this->addrs->create_enumerator(this->addrs);
 	while (enumerator->enumerate(enumerator, NULL, (void**)&addr))
 	{
@@ -776,8 +1712,11 @@ METHOD(kernel_net_t, destroy, void,
 	enumerator->destroy(enumerator);
 	this->addrs->destroy(this->addrs);
 	this->ifaces->destroy_function(this->ifaces, (void*)iface_entry_destroy);
+	this->tuns->destroy(this->tuns);
 	this->lock->destroy(this->lock);
-	this->mutex_pfroute->destroy(this->mutex_pfroute);
+	this->mutex->destroy(this->mutex);
+	this->condvar->destroy(this->condvar);
+	free(this->reply);
 	free(this);
 }
 
@@ -787,11 +1726,11 @@ METHOD(kernel_net_t, destroy, void,
 kernel_pfroute_net_t *kernel_pfroute_net_create()
 {
 	private_kernel_pfroute_net_t *this;
-	bool register_for_events = TRUE;
 
 	INIT(this,
 		.public = {
 			.interface = {
+				.get_features = _get_features,
 				.get_interface = _get_interface_name,
 				.create_address_enumerator = _create_address_enumerator,
 				.get_source_addr = _get_source_addr,
@@ -803,45 +1742,51 @@ kernel_pfroute_net_t *kernel_pfroute_net_create()
 				.destroy = _destroy,
 			},
 		},
+		.pid = getpid(),
 		.ifaces = linked_list_create(),
 		.addrs = hashtable_create(
 								(hashtable_hash_t)addr_map_entry_hash,
 								(hashtable_equals_t)addr_map_entry_equals, 16),
+		.routes = hashtable_create((hashtable_hash_t)route_entry_hash,
+								   (hashtable_equals_t)route_entry_equals, 16),
+		.net_changes = hashtable_create(
+								   (hashtable_hash_t)net_change_hash,
+								   (hashtable_equals_t)net_change_equals, 16),
+		.tuns = linked_list_create(),
 		.lock = rwlock_create(RWLOCK_TYPE_DEFAULT),
-		.mutex_pfroute = mutex_create(MUTEX_TYPE_DEFAULT),
+		.mutex = mutex_create(MUTEX_TYPE_DEFAULT),
+		.condvar = condvar_create(CONDVAR_TYPE_DEFAULT),
+		.routes_lock = mutex_create(MUTEX_TYPE_DEFAULT),
+		.net_changes_lock = mutex_create(MUTEX_TYPE_DEFAULT),
+		.vip_wait = lib->settings->get_int(lib->settings,
+					"%s.plugins.kernel-pfroute.vip_wait", 1000, hydra->daemon),
 	);
-
-	if (streq(hydra->daemon, "starter"))
-	{   /* starter has no threads, so we do not register for kernel events */
-		register_for_events = FALSE;
-	}
+	timerclear(&this->last_route_reinstall);
+	timerclear(&this->last_roam);
 
 	/* create a PF_ROUTE socket to communicate with the kernel */
 	this->socket = socket(PF_ROUTE, SOCK_RAW, AF_UNSPEC);
-	if (this->socket < 0)
+	if (this->socket == -1)
 	{
 		DBG1(DBG_KNL, "unable to create PF_ROUTE socket");
 		destroy(this);
 		return NULL;
 	}
 
-	if (register_for_events)
+	if (streq(hydra->daemon, "starter"))
 	{
-		/* create a PF_ROUTE socket to receive events */
-		this->socket_events = socket(PF_ROUTE, SOCK_RAW, AF_UNSPEC);
-		if (this->socket_events < 0)
+		/* starter has no threads, so we do not register for kernel events */
+		if (shutdown(this->socket, SHUT_RD) != 0)
 		{
-			DBG1(DBG_KNL, "unable to create PF_ROUTE event socket");
-			destroy(this);
-			return NULL;
+			DBG1(DBG_KNL, "closing read end of PF_ROUTE socket failed: %s",
+				 strerror(errno));
 		}
-
-		lib->processor->queue_job(lib->processor,
-			(job_t*)callback_job_create_with_prio(
-					(callback_job_cb_t)receive_events, this, NULL,
-					(callback_job_cancel_t)return_false, JOB_PRIO_CRITICAL));
 	}
-
+	else
+	{
+		lib->watcher->add(lib->watcher, this->socket, WATCHER_READ,
+						  (watcher_cb_t)receive_events, this);
+	}
 	if (init_address_list(this) != SUCCESS)
 	{
 		DBG1(DBG_KNL, "unable to get interface list");