Merge commit '98bb80a243b58c43453e9be69d19d0350286549c' into int-new

.gitignore

@@ -11,3 +11,4 @@
 /config.status
 /configure
 /sysdep/autoconf.h.in
+/sysdep/autoconf.h.in~

.gitlab-ci.yml

@@ -0,0 +1,349 @@
+variables:
+  DEBIAN_FRONTEND: noninteractive
+  LC_ALL: C
+  GIT_STRATEGY: fetch
+  DOCKER_CMD: docker --config="$HOME/.docker/$CI_JOB_ID/"
+  IMG_BASE: registry.labs.nic.cz/labs/bird
+
+stages:
+  - image
+  - build
+
+.docker: &docker_build
+  stage: image
+  allow_failure: true
+  script:
+  - $DOCKER_CMD login -u gitlab-ci-token -p $CI_BUILD_TOKEN registry.labs.nic.cz
+  # Make sure we refresh the base image if it updates (eg. security updates, etc)
+  # If we do just the build, cache is always reused and the freshness of the
+  # base image is never checked. However, pull always asks and updates the
+  # image only if it changed ‒ therefore, the cache is used unless there's a
+  # change.
+  - $DOCKER_CMD pull `sed -ne 's/^FROM //p' "misc/docker/$IMG_NAME/Dockerfile"`
+  - $DOCKER_CMD build -t "bird:$IMG_NAME" "misc/docker/$IMG_NAME"
+  - $DOCKER_CMD tag "bird:$IMG_NAME" "$IMG_BASE:$IMG_NAME"
+  - $DOCKER_CMD push "$IMG_BASE:$IMG_NAME"
+  after_script:
+  - rm -f "$HOME/.docker/$CI_JOB_ID/" # cleanup the credentials
+  tags:
+  # That's Docker in Docker
+  - dind
+
+docker_debian-7-amd64:
+  variables:
+    IMG_NAME: "debian-7-amd64"
+  <<: *docker_build
+
+docker_debian-8-amd64:
+  variables:
+    IMG_NAME: "debian-8-amd64"
+  <<: *docker_build
+
+docker_debian-9-amd64:
+  variables:
+    IMG_NAME: "debian-9-amd64"
+  <<: *docker_build
+
+docker_debian-testing-amd64:
+  variables:
+    IMG_NAME: "debian-testing-amd64"
+  <<: *docker_build
+
+docker_debian-7-i386:
+  variables:
+    IMG_NAME: "debian-7-i386"
+  <<: *docker_build
+
+docker_debian-8-i386:
+  variables:
+    IMG_NAME: "debian-8-i386"
+  <<: *docker_build
+
+docker_debian-9-i386:
+  variables:
+    IMG_NAME: "debian-9-i386"
+  <<: *docker_build
+
+docker_debian-testing-i386:
+  variables:
+    IMG_NAME: "debian-testing-i386"
+  <<: *docker_build
+
+docker_fedora-25-amd64:
+  variables:
+    IMG_NAME: "fedora-25-amd64"
+  <<: *docker_build
+
+docker_fedora-26-amd64:
+  variables:
+    IMG_NAME: "fedora-26-amd64"
+  <<: *docker_build
+
+docker_centos-6-amd64:
+  variables:
+    IMG_NAME: "centos-6-amd64"
+  <<: *docker_build
+
+docker_centos-7-amd64:
+  variables:
+    IMG_NAME: "centos-7-amd64"
+  <<: *docker_build
+
+docker_opensuse-42_3-amd64:
+  variables:
+    IMG_NAME: "opensuse-42.3-amd64"
+  <<: *docker_build
+
+docker_ubuntu-14_04-amd64:
+  variables:
+    IMG_NAME: "ubuntu-14.04-amd64"
+  <<: *docker_build
+
+docker_ubuntu-16_04-amd64:
+  variables:
+    IMG_NAME: "ubuntu-16.04-amd64"
+  <<: *docker_build
+
+.debian-7-i386: &debian-7-i386_env
+  image: registry.labs.nic.cz/labs/bird:debian-7-i386
+  tags:
+  - docker
+  - linux
+  - amd64
+
+.debian-8-i386: &debian-8-i386_env
+  image: registry.labs.nic.cz/labs/bird:debian-8-i386
+  tags:
+  - docker
+  - linux
+  - amd64
+
+.debian-9-i386: &debian-9-i386_env
+  image: registry.labs.nic.cz/labs/bird:debian-9-i386
+  tags:
+  - docker
+  - linux
+  - amd64
+
+.debian-testing-i386: &debian-testing-i386_env
+  image: registry.labs.nic.cz/labs/bird:debian-testing-i386
+  tags:
+  - docker
+  - linux
+  - amd64
+
+.debian-7-amd64: &debian-7-amd64_env
+  image: registry.labs.nic.cz/labs/bird:debian-7-amd64
+  tags:
+  - docker
+  - linux
+  - amd64
+
+.debian-8-amd64: &debian-8-amd64_env
+  image: registry.labs.nic.cz/labs/bird:debian-8-amd64
+  tags:
+  - docker
+  - linux
+  - amd64
+
+.debian-9-amd64: &debian-9-amd64_env
+  image: registry.labs.nic.cz/labs/bird:debian-9-amd64
+  tags:
+  - docker
+  - linux
+  - amd64
+
+.debian-testing-amd64: &debian-testing-amd64_env
+  image: registry.labs.nic.cz/labs/bird:debian-testing-amd64
+  tags:
+  - docker
+  - linux
+  - amd64
+
+.fedora-25-amd64: &fedora-25-amd64_env
+  image: registry.labs.nic.cz/labs/bird:fedora-25-amd64
+  tags:
+  - docker
+  - linux
+  - amd64
+
+.fedora-26-amd64: &fedora-26-amd64_env
+  image: registry.labs.nic.cz/labs/bird:fedora-26-amd64
+  tags:
+  - docker
+  - linux
+  - amd64
+
+.centos-6-amd64: &centos-6-amd64_env
+  image: registry.labs.nic.cz/labs/bird:centos-6-amd64
+  tags:
+  - docker
+  - linux
+  - amd64
+
+.centos-7-amd64: &centos-7-amd64_env
+  image: registry.labs.nic.cz/labs/bird:centos-7-amd64
+  tags:
+  - docker
+  - linux
+  - amd64
+
+.opensuse-42_3-amd64: &opensuse-42_3-amd64_env
+  image: registry.labs.nic.cz/labs/bird:opensuse-42.3-amd64
+  tags:
+  - docker
+  - linux
+  - amd64
+
+.ubuntu-14_04-amd64: &ubuntu-14_04-amd64_env
+  image: registry.labs.nic.cz/labs/bird:ubuntu-14.04-amd64
+  tags:
+  - docker
+  - linux
+  - amd64
+
+.ubuntu-16_04-amd64: &ubuntu-16_04-amd64_env
+  image: registry.labs.nic.cz/labs/bird:ubuntu-16.04-amd64
+  tags:
+  - docker
+  - linux
+  - amd64
+
+# TODO We want to copy these BSDs to our own virtual machines, to make sure someone doesn't update them by accident.
+.freebsd-11-i386: &freebsd-11-i386_env
+  variables:
+    CPPFLAGS: "-I/usr/local/include"
+    LDFLAGS: "-L/usr/local/lib"
+  tags:
+  - freebsd
+  - i386
+  #only:
+  #- master
+  #- triggers
+  #- tags
+
+.freebsd-11-amd64: &freebsd-11-amd64_env
+  variables:
+    CPPFLAGS: "-I/usr/local/include"
+    LDFLAGS: "-L/usr/local/lib"
+  tags:
+  - freebsd
+  - amd64
+  #only:
+  #- master
+  #- triggers
+  #- tags
+
+.build: &build_job
+  stage: build
+  script:
+  - autoreconf
+  - ./configure CPPFLAGS="$CPPFLAGS" LDFLAGS="$LDFLAGS"
+  # Detect which make is available
+  - MAKE=make
+  - which gmake 2>/dev/null >/dev/null && MAKE=gmake
+  - $MAKE
+  # Run tests if they are available (eg. don't fail if "check" isn't a valid make target)
+  - $MAKE check || [ "$?" = 2 ]
+
+build-debian-7-amd64:
+  variables:
+    IPV6: "no"
+  <<: *debian-7-amd64_env
+  <<: *build_job
+
+build-debian-8-amd64:
+  variables:
+    IPV6: "no"
+  <<: *debian-8-amd64_env
+  <<: *build_job
+
+build-debian-9-amd64:
+  variables:
+    IPV6: "no"
+  <<: *debian-9-amd64_env
+  <<: *build_job
+
+build-debian-testing-amd64:
+  variables:
+    IPV6: "no"
+  <<: *debian-testing-amd64_env
+  <<: *build_job
+
+build-fedora-25-amd64:
+  variables:
+    IPV6: "no"
+  <<: *fedora-25-amd64_env
+  <<: *build_job
+
+build-fedora-26-amd64:
+  variables:
+    IPV6: "no"
+  <<: *fedora-26-amd64_env
+  <<: *build_job
+
+build-centos-6-amd64:
+  variables:
+    IPV6: "no"
+  <<: *centos-6-amd64_env
+  <<: *build_job
+
+build-centos-7-amd64:
+  variables:
+    IPV6: "no"
+  <<: *centos-7-amd64_env
+  <<: *build_job
+
+build-opensuse-42_3-amd64:
+  variables:
+    IPV6: "no"
+  <<: *opensuse-42_3-amd64_env
+  <<: *build_job
+
+build-ubuntu-14_04-amd64:
+  variables:
+    IPV6: "no"
+  <<: *ubuntu-14_04-amd64_env
+  <<: *build_job
+
+build-ubuntu-16_04-amd64:
+  variables:
+    IPV6: "no"
+  <<: *ubuntu-16_04-amd64_env
+  <<: *build_job
+
+build-debian-7-i386:
+  variables:
+    IPV6: "no"
+  <<: *debian-7-i386_env
+  <<: *build_job
+
+build-debian-8-i386:
+  variables:
+    IPV6: "no"
+  <<: *debian-8-i386_env
+  <<: *build_job
+
+build-debian-9-i386:
+  variables:
+    IPV6: "no"
+  <<: *debian-9-i386_env
+  <<: *build_job
+
+build-debian-testing-i386:
+  variables:
+    IPV6: "no"
+  <<: *debian-testing-i386_env
+  <<: *build_job
+
+build-freebsd-11-amd64:
+  variables:
+    IPV6: "no"
+  <<: *freebsd-11-amd64_env
+  <<: *build_job
+
+build-freebsd-11-i386:
+  variables:
+    IPV6: "no"
+  <<: *freebsd-11-i386_env
+  <<: *build_job

TODO

@@ -1,45 +0,0 @@
-Core
-~~~~
-- socket open failure should not be fatal
-- &&,||: priorities
-- static: allow specifying a per-route filter program for setting route attributes?
-
-Globals
-~~~~~~~
-- right usage of DBG vs. debug
-- logging and tracing; use appropriate log levels
-- check incoming packets and log errors!!
-- check log calls for trailing newlines and log levels followed by comma
-- check if all protocols set proper packet priorities and TTL's.
-- try compiling with -Wunused
-- does everybody test return value of sk_open?
-- protocols: implement CLI hooks and per-procotol CLI commands
-- protocols: implement reconfigure hook
-- protocols: use locking
-- check use of system includes and sprintf()
-
-Various ideas
-~~~~~~~~~~~~~
-- client: Ctrl-R eats one more enter
-- bgp: timing of updates?
-- netlink: import Linux route attributes to our rta's, so that they can be filtered?
-- config: executable config files
-- filters: user defined attributes?
-- io: use poll if available
-- route recalculation timing and flap dampening [see RFC2439 for algorithms]
-- aggregate engine: standard route aggregation and summarization [RFC2519]
-- aggregate engine: injection of manually configured pseudo-static routes
-- generate default route if any working BGP connection exists (aggregate engine again?)
-- generate default route to IGP's (aggregate engine yet another time?)
-- look at RFC 2386 (QoS-based routing)
-- cli: show tables?
-
-OSPF
-~~~~
-	- check incoming packets using neighbor cache
-	- RFC2328 appendix E: Use a better algorithm
-	- automatic generation of external route tags (RFC1403)
-	- RFC2370 opaque LSA's
-	- Limit export rate of external LSAs (like Gated does)
-	- Bugfix in link state retransmission list (aging)
-	- Graceful OSPF restart - RFC3623

client/client.c

@@ -25,9 +25,10 @@
 #include <fcntl.h>
 #include <unistd.h>
 #include <errno.h>
+#include <sys/select.h>
 #include <sys/socket.h>
-#include <sys/un.h>
 #include <sys/types.h>
+#include <sys/un.h>
 
 #include "nest/bird.h"
 #include "lib/resource.h"

doc/bird.sgml

@@ -3660,6 +3660,12 @@ dsc-iface
 	as a default router. For <cf/sensitive/ option, see <ref id="radv-trigger" name="trigger">.
 	Default: 3 * <cf/max ra	interval/, <cf/sensitive/ yes.
 
+	<tag><label id="radv-iface-linger-time">linger time <m/expr/</tag>
+	When a prefix disappears, it is advertised for some time with 0
+	lifetime, to inform clients the prefix is no longer usable. This option
+	sets the time for how long it is advertised (in seconds). Maximum is
+	3600, 0 means disabled. Default: 300.
+
 	<tag><label id="radv-iface-default-preference-low">default preference low|medium|high</tag>
 	This option specifies the Default Router Preference value to advertise
 	to hosts. Default: medium.

lib/net.h

@@ -526,6 +526,18 @@ int net_classify(const net_addr *N);
 int net_format(const net_addr *N, char *buf, int buflen);
 int rd_format(const u64 rd, char *buf, int buflen);
 
+static inline int ipa_in_net_ip4(ip4_addr a, const net_addr_ip4 *n)
+{ return ip4_zero(ip4_and(ip4_xor(a, n->prefix), ip4_mkmask(n->pxlen))); }
+
+static inline int net_in_net_ip4(const net_addr_ip4 *a, const net_addr_ip4 *b)
+{ return (a->pxlen >= b->pxlen) && ipa_in_net_ip4(a->prefix, b); }
+
+static inline int ipa_in_net_ip6(ip6_addr a, const net_addr_ip6 *n)
+{ return ip6_zero(ip6_and(ip6_xor(a, n->prefix), ip6_mkmask(n->pxlen))); }
+
+static inline int net_in_net_ip6(const net_addr_ip6 *a, const net_addr_ip6 *b)
+{ return (a->pxlen >= b->pxlen) && ipa_in_net_ip6(a->prefix, b); }
+
 int ipa_in_netX(const ip_addr A, const net_addr *N);
 int net_in_netX(const net_addr *A, const net_addr *N);
 

misc/docker/centos-6-amd64/Dockerfile

@@ -0,0 +1,11 @@
+FROM centos:6
+RUN yum -y upgrade
+RUN yum -y install \
+	autoconf \
+	flex \
+	bison \
+	pkgconfig \
+	'readline-devel' \
+	'pkgconfig(ncurses)' \
+	gcc \
+	make

misc/docker/centos-7-amd64/Dockerfile

@@ -0,0 +1,11 @@
+FROM centos:7
+RUN yum -y upgrade
+RUN yum -y install \
+	autoconf \
+	flex \
+	bison \
+	pkgconfig \
+	'readline-devel' \
+	'pkgconfig(ncurses)' \
+	gcc \
+	make

misc/docker/debian-7-amd64/Dockerfile

@@ -0,0 +1,12 @@
+FROM debian:wheezy-slim
+ENV DEBIAN_FRONTEND noninteractive
+RUN sed -i 's/deb.debian.org/ftp.cz.debian.org/' /etc/apt/sources.list
+RUN apt-get -y update
+RUN apt-get -y upgrade
+RUN apt-get -y install \
+	autoconf \
+	build-essential \
+	flex \
+	bison \
+	ncurses-dev \
+	libreadline-dev

misc/docker/debian-7-i386/Dockerfile

@@ -0,0 +1,12 @@
+FROM i386/debian:wheezy-slim
+ENV DEBIAN_FRONTEND noninteractive
+RUN sed -i 's/deb.debian.org/ftp.cz.debian.org/' /etc/apt/sources.list
+RUN apt-get -y update
+RUN apt-get -y upgrade
+RUN apt-get -y install \
+	autoconf \
+	build-essential \
+	flex \
+	bison \
+	ncurses-dev \
+	libreadline-dev

misc/docker/debian-8-amd64/Dockerfile

@@ -0,0 +1,12 @@
+FROM debian:jessie-slim
+ENV DEBIAN_FRONTEND noninteractive
+RUN sed -i 's/deb.debian.org/ftp.cz.debian.org/' /etc/apt/sources.list
+RUN apt-get -y update
+RUN apt-get -y upgrade
+RUN apt-get -y install \
+	autoconf \
+	build-essential \
+	flex \
+	bison \
+	ncurses-dev \
+	libreadline-dev

misc/docker/debian-8-i386/Dockerfile

@@ -0,0 +1,12 @@
+FROM i386/debian:jessie-slim
+ENV DEBIAN_FRONTEND noninteractive
+RUN sed -i 's/deb.debian.org/ftp.cz.debian.org/' /etc/apt/sources.list
+RUN apt-get -y update
+RUN apt-get -y upgrade
+RUN apt-get -y install \
+	autoconf \
+	build-essential \
+	flex \
+	bison \
+	ncurses-dev \
+	libreadline-dev

misc/docker/debian-9-amd64/Dockerfile

@@ -0,0 +1,12 @@
+FROM debian:stretch-slim
+ENV DEBIAN_FRONTEND noninteractive
+RUN sed -i 's/deb.debian.org/ftp.cz.debian.org/' /etc/apt/sources.list
+RUN apt-get -y update
+RUN apt-get -y upgrade
+RUN apt-get -y install \
+	autoconf \
+	build-essential \
+	flex \
+	bison \
+	ncurses-dev \
+	libreadline-dev

misc/docker/debian-9-i386/Dockerfile

@@ -0,0 +1,12 @@
+FROM i386/debian:stretch-slim
+ENV DEBIAN_FRONTEND noninteractive
+RUN sed -i 's/deb.debian.org/ftp.cz.debian.org/' /etc/apt/sources.list
+RUN apt-get -y update
+RUN apt-get -y upgrade
+RUN apt-get -y install \
+	autoconf \
+	build-essential \
+	flex \
+	bison \
+	ncurses-dev \
+	libreadline-dev

misc/docker/debian-testing-amd64/Dockerfile

@@ -0,0 +1,12 @@
+FROM debian:testing-slim
+ENV DEBIAN_FRONTEND noninteractive
+RUN sed -i 's/deb.debian.org/ftp.cz.debian.org/' /etc/apt/sources.list
+RUN apt-get -y update
+RUN apt-get -y upgrade
+RUN apt-get -y install \
+	autoconf \
+	build-essential \
+	flex \
+	bison \
+	ncurses-dev \
+	libreadline-dev

misc/docker/debian-testing-i386/Dockerfile

@@ -0,0 +1,12 @@
+FROM i386/debian:testing-slim
+ENV DEBIAN_FRONTEND noninteractive
+RUN sed -i 's/deb.debian.org/ftp.cz.debian.org/' /etc/apt/sources.list
+RUN apt-get -y update
+RUN apt-get -y upgrade
+RUN apt-get -y install \
+	autoconf \
+	build-essential \
+	flex \
+	bison \
+	ncurses-dev \
+	libreadline-dev

misc/docker/fedora-25-amd64/Dockerfile

@@ -0,0 +1,10 @@
+FROM fedora:25
+RUN dnf -y upgrade
+RUN dnf -y install \
+	autoconf \
+	flex \
+	bison \
+	pkgconfig \
+	'readline-devel' \
+	'pkgconfig(ncurses)' \
+	gcc

misc/docker/fedora-26-amd64/Dockerfile

@@ -0,0 +1,10 @@
+FROM fedora:26
+RUN dnf -y upgrade
+RUN dnf -y install \
+	autoconf \
+	flex \
+	bison \
+	pkgconfig \
+	'readline-devel' \
+	'pkgconfig(ncurses)' \
+	gcc

misc/docker/opensuse-42.3-amd64/Dockerfile

@@ -0,0 +1,11 @@
+FROM opensuse:42.3
+RUN zypper -n up
+RUN zypper -n install \
+	autoconf \
+	flex \
+	bison \
+	pkgconfig \
+	readline-devel \
+	ncurses-devel \
+	gcc \
+	gmake

misc/docker/ubuntu-14.04-amd64/Dockerfile

@@ -0,0 +1,12 @@
+FROM ubuntu:14.04
+ENV DEBIAN_FRONTEND noninteractive
+RUN sed -i 's/deb.debian.org/ftp.cz.debian.org/' /etc/apt/sources.list
+RUN apt-get -y update
+RUN apt-get -y upgrade
+RUN apt-get -y install \
+	autoconf \
+	build-essential \
+	flex \
+	bison \
+	ncurses-dev \
+	libreadline-dev

misc/docker/ubuntu-16.04-amd64/Dockerfile

@@ -0,0 +1,12 @@
+FROM ubuntu:16.04
+ENV DEBIAN_FRONTEND noninteractive
+RUN sed -i 's/deb.debian.org/ftp.cz.debian.org/' /etc/apt/sources.list
+RUN apt-get -y update
+RUN apt-get -y upgrade
+RUN apt-get -y install \
+	autoconf \
+	build-essential \
+	flex \
+	bison \
+	ncurses-dev \
+	libreadline-dev

nest/iface.c

@@ -145,9 +145,9 @@ ifa_send_notify(struct proto *p, unsigned c, struct ifa *a)
   if (p->ifa_notify && (p->proto_state != PS_DOWN))
     {
       if (p->debug & D_IFACES)
-	log(L_TRACE "%s < %s address %N on interface %s %s",
+	log(L_TRACE "%s < address %N on interface %s %s",
-	    p->name, (a->flags & IA_PRIMARY) ? "primary" : "secondary",
+	    p->name, &a->prefix, a->iface->name,
-	    &a->prefix, a->iface->name, (c & IF_CHANGE_UP) ? "added" : "removed");
+	    (c & IF_CHANGE_UP) ? "added" : "removed");
       p->ifa_notify(p, c, a);
     }
 }

proto/radv/config.Y

@@ -27,7 +27,7 @@ static u8 radv_mult_val;	/* Used by radv_mult for second return value */
 CF_DECLS
 
 CF_KEYWORDS(RADV, PREFIX, INTERFACE, MIN, MAX, RA, DELAY, INTERVAL,
-	MANAGED, OTHER, CONFIG, LINK, MTU, REACHABLE, TIME, RETRANS,
+	MANAGED, OTHER, CONFIG, LINGER, LINK, MTU, REACHABLE, TIME, RETRANS,
 	TIMER, CURRENT, HOP, LIMIT, DEFAULT, VALID, PREFERRED, MULT,
 	LIFETIME, SKIP, ONLINK, AUTONOMOUS, RDNSS, DNSSL, NS, DOMAIN,
 	LOCAL, TRIGGER, SENSITIVE, PREFERENCE, LOW, MEDIUM, HIGH)
@@ -80,6 +80,7 @@ radv_iface_start:
   RADV_IFACE->max_ra_int = DEFAULT_MAX_RA_INT;
   RADV_IFACE->min_delay = DEFAULT_MIN_DELAY;
   RADV_IFACE->current_hop_limit = DEFAULT_CURRENT_HOP_LIMIT;
+  RADV_IFACE->linger_time = DEFAULT_LINGER_TIME;
   RADV_IFACE->default_lifetime = -1;
   RADV_IFACE->default_lifetime_sensitive = 1;
   RADV_IFACE->default_preference = RA_PREF_MEDIUM;
@@ -94,6 +95,7 @@ radv_iface_item:
  | LINK MTU expr { RADV_IFACE->link_mtu = $3; }
  | REACHABLE TIME expr { RADV_IFACE->reachable_time = $3; if ($3 > 3600000) cf_error("Reachable time must be in range 0-3600000"); }
  | RETRANS TIMER expr { RADV_IFACE->retrans_timer = $3; }
+ | LINGER TIME expr { RADV_IFACE->linger_time = $3; if ($3 > 3600) cf_error("Linger time must be in range 0-3600"); }
  | CURRENT HOP LIMIT expr { RADV_IFACE->current_hop_limit = $4; if ($4 > 255) cf_error("Current hop limit must be in range 0-255"); }
  | DEFAULT LIFETIME expr radv_sensitive {
      RADV_IFACE->default_lifetime = $3;

proto/radv/packets.c

@@ -70,36 +70,6 @@ struct radv_opt_dnssl
   char domain[];
 };
 
-
-static struct radv_prefix_config default_prefix = {
-  .onlink = 1,
-  .autonomous = 1,
-  .valid_lifetime = DEFAULT_VALID_LIFETIME,
-  .preferred_lifetime = DEFAULT_PREFERRED_LIFETIME
-};
-
-
-static struct radv_prefix_config *
-radv_prefix_match(struct radv_iface *ifa, struct ifa *a)
-{
-  struct radv_proto *p = ifa->ra;
-  struct radv_config *cf = (struct radv_config *) (p->p.cf);
-  struct radv_prefix_config *pc;
-
-  if (a->scope <= SCOPE_LINK)
-    return NULL;
-
-  WALK_LIST(pc, ifa->cf->pref_list)
-    if (net_in_netX(&a->prefix, (net_addr *) &pc->prefix))
-      return pc;
-
-  WALK_LIST(pc, cf->pref_list)
-    if (net_in_netX(&a->prefix, (net_addr *) &pc->prefix))
-      return pc;
-
-  return &default_prefix;
-}
-
 static int
 radv_prepare_rdnss(struct radv_iface *ifa, list *rdnss_list, char **buf, char *bufend)
 {
@@ -234,6 +204,36 @@ radv_prepare_dnssl(struct radv_iface *ifa, list *dnssl_list, char **buf, char *b
   return -1;
 }
 
+static int
+radv_prepare_prefix(struct radv_iface *ifa, struct radv_prefix *px,
+		    char **buf, char *bufend)
+{
+  struct radv_prefix_config *pc = px->cf;
+
+  if (*buf + sizeof(struct radv_opt_prefix) > bufend)
+  {
+    log(L_WARN "%s: Too many prefixes on interface %s",
+	ifa->ra->p.name, ifa->iface->name);
+    return -1;
+  }
+
+  struct radv_opt_prefix *op = (void *) *buf;
+  op->type = OPT_PREFIX;
+  op->length = 4;
+  op->pxlen = px->prefix.pxlen;
+  op->flags = (pc->onlink ? OPT_PX_ONLINK : 0) |
+    (pc->autonomous ? OPT_PX_AUTONOMOUS : 0);
+  op->valid_lifetime = (ifa->ra->active || !pc->valid_lifetime_sensitive) ?
+    htonl(pc->valid_lifetime) : 0;
+  op->preferred_lifetime = (ifa->ra->active || !pc->preferred_lifetime_sensitive) ?
+    htonl(pc->preferred_lifetime) : 0;
+  op->reserved = 0;
+  op->prefix = ip6_hton(px->prefix.prefix);
+  *buf += sizeof(*op);
+
+  return 0;
+}
+
 static void
 radv_prepare_ra(struct radv_iface *ifa)
 {
@@ -269,37 +269,11 @@ radv_prepare_ra(struct radv_iface *ifa)
     buf += sizeof (*om);
   }
 
-  struct ifa *addr;
+  struct radv_prefix *prefix;
-  WALK_LIST(addr, ifa->iface->addrs)
+  WALK_LIST(prefix, ifa->prefixes)
   {
-    if (addr->prefix.type != NET_IP6)
+    if (radv_prepare_prefix(ifa, prefix, &buf, bufend) < 0)
-      continue;
-
-    struct radv_prefix_config *pc;
-    pc = radv_prefix_match(ifa, addr);
-
-    if (!pc || pc->skip)
-      continue;
-
-    if (buf + sizeof(struct radv_opt_prefix) > bufend)
-    {
-      log(L_WARN "%s: Too many prefixes on interface %s", p->p.name, ifa->iface->name);
       goto done;
-    }
-
-    struct radv_opt_prefix *op = (void *) buf;
-    op->type = OPT_PREFIX;
-    op->length = 4;
-    op->pxlen = net6_pxlen(&addr->prefix);
-    op->flags = (pc->onlink ? OPT_PX_ONLINK : 0) |
-      (pc->autonomous ? OPT_PX_AUTONOMOUS : 0);
-    op->valid_lifetime = (p->active || !pc->valid_lifetime_sensitive) ?
-      htonl(pc->valid_lifetime) : 0;
-    op->preferred_lifetime = (p->active || !pc->preferred_lifetime_sensitive) ?
-      htonl(pc->preferred_lifetime) : 0;
-    op->reserved = 0;
-    op->prefix = ip6_hton(net6_prefix(&addr->prefix));
-    buf += sizeof(*op);
   }
 
   if (! ic->rdnss_local)
@@ -408,7 +382,7 @@ radv_err_hook(sock *sk, int err)
 int
 radv_sk_open(struct radv_iface *ifa)
 {
-  sock *sk = sk_new(ifa->ra->p.pool);
+  sock *sk = sk_new(ifa->pool);
   sk->type = SK_IP;
   sk->subtype = SK_IPV6;
   sk->dport = ICMPV6_PROTO;

proto/radv/radv.c

@@ -51,6 +51,16 @@ radv_timer(timer *tm)
 
   RADV_TRACE(D_EVENTS, "Timer fired on %s", ifa->iface->name);
 
+  /*
+   * If some dead prefixes expired, regenerate the prefix list and the packet.
+   * We do so by pretending there was a change on the interface.
+   *
+   * This sets the timer, but we replace it just at the end of this function
+   * (replacing a timer is fine).
+   */
+  if (ifa->prefix_expires && (ifa->prefix_expires <= current_time()))
+    radv_iface_notify(ifa, RA_EV_GC);
+
   radv_send_ra(ifa, 0);
 
   /* Update timer */
@@ -68,7 +78,136 @@ radv_timer(timer *tm)
   tm_start(ifa->timer, t);
 }
 
-static char* ev_name[] = { NULL, "Init", "Change", "RS" };
+static struct radv_prefix_config default_prefix = {
+  .onlink = 1,
+  .autonomous = 1,
+  .valid_lifetime = DEFAULT_VALID_LIFETIME,
+  .preferred_lifetime = DEFAULT_PREFERRED_LIFETIME
+};
+
+static struct radv_prefix_config dead_prefix = {
+};
+
+/* Find a corresponding config for the given prefix */
+static struct radv_prefix_config *
+radv_prefix_match(struct radv_iface *ifa, net_addr_ip6 *px)
+{
+  struct radv_proto *p = ifa->ra;
+  struct radv_config *cf = (struct radv_config *) (p->p.cf);
+  struct radv_prefix_config *pc;
+
+  WALK_LIST(pc, ifa->cf->pref_list)
+    if (net_in_net_ip6(px, &pc->prefix))
+      return pc;
+
+  WALK_LIST(pc, cf->pref_list)
+    if (net_in_net_ip6(px, &pc->prefix))
+      return pc;
+
+  return &default_prefix;
+}
+
+/*
+ * Go through the list of prefixes, compare them with configs and decide if we
+ * want them or not.
+ */
+static void
+radv_prepare_prefixes(struct radv_iface *ifa)
+{
+  struct radv_proto *p = ifa->ra;
+  struct radv_iface_config *cf = ifa->cf;
+  struct radv_prefix *pfx;
+
+  /* First mark all the prefixes as unused */
+  WALK_LIST(pfx, ifa->prefixes)
+    pfx->mark = 0;
+
+  /* Find all the prefixes we want to use and make sure they are in the list. */
+  struct ifa *addr;
+  WALK_LIST(addr, ifa->iface->addrs)
+  {
+    if ((addr->prefix.type != NET_IP6) ||
+	(addr->scope <= SCOPE_LINK))
+      continue;
+
+    net_addr_ip6 *prefix = (void *) &addr->prefix;
+    struct radv_prefix_config *pc = radv_prefix_match(ifa, prefix);
+
+    if (!pc || pc->skip)
+      continue;
+
+    /* Do we have it already? */
+    struct radv_prefix *existing = NULL;
+    WALK_LIST(pfx, ifa->prefixes)
+      if (net_equal_ip6(&pfx->prefix, prefix))
+      {
+	existing = pfx;
+	break;
+      }
+
+    if (!existing)
+    {
+      RADV_TRACE(D_EVENTS, "Adding new prefix %N on %s",
+		 prefix, ifa->iface->name);
+
+      existing = mb_allocz(ifa->pool, sizeof *existing);
+      net_copy_ip6(&existing->prefix, prefix);
+      add_tail(&ifa->prefixes, NODE existing);
+    }
+
+    /*
+     * Update the information (it may have changed, or even bring a prefix back
+     * to life).
+     */
+    existing->alive = 1;
+    existing->mark = 1;
+    existing->cf = pc;
+  }
+
+  /*
+   * Garbage-collect the prefixes. If something isn't used, it dies (but isn't
+   * dropped just yet). If something is dead and rots there for long enough,
+   * clean it up.
+   */
+  btime now_ = current_time();
+  btime expires = now_ + cf->linger_time S;
+  btime expires_min = 0;
+  struct radv_prefix *next;
+  WALK_LIST_DELSAFE(pfx, next, ifa->prefixes)
+  {
+    if (pfx->alive && !pfx->mark)
+    {
+      RADV_TRACE(D_EVENTS, "Marking prefix %N on %s as dead",
+		 pfx->prefix, ifa->iface->name);
+
+      pfx->alive = 0;
+      pfx->expires = expires;
+      pfx->cf = &dead_prefix;
+    }
+
+    if (!pfx->alive)
+    {
+      if (pfx->expires <= now_)
+      {
+	RADV_TRACE(D_EVENTS, "Removing prefix %N on %s",
+		   pfx->prefix, ifa->iface->name);
+
+	rem_node(NODE pfx);
+	mb_free(pfx);
+      }
+      else
+      {
+	/* Find minimum expiration time */
+	if (!expires_min || (pfx->expires < expires_min))
+	  expires_min = pfx->expires;
+      }
+    }
+  }
+
+  ifa->prefix_expires = expires_min;
+}
+
+static char* ev_name[] = { NULL, "Init", "Change", "RS", "Garbage collect" };
 
 void
 radv_iface_notify(struct radv_iface *ifa, int event)
@@ -83,6 +222,7 @@ radv_iface_notify(struct radv_iface *ifa, int event)
   switch (event)
   {
   case RA_EV_CHANGE:
+  case RA_EV_GC:
     ifa->plen = 0;
   case RA_EV_INIT:
     ifa->initial = MAX_INITIAL_RTR_ADVERTISEMENTS;
@@ -92,6 +232,8 @@ radv_iface_notify(struct radv_iface *ifa, int event)
     break;
   }
 
+  radv_prepare_prefixes(ifa);
+
   /* Update timer */
   btime t = ifa->last + ifa->cf->min_delay S - current_time();
   tm_start(ifa->timer, t);
@@ -137,16 +279,18 @@ radv_iface_add(struct object_lock *lock)
 static void
 radv_iface_new(struct radv_proto *p, struct iface *iface, struct radv_iface_config *cf)
 {
-  pool *pool = p->p.pool;
   struct radv_iface *ifa;
 
   RADV_TRACE(D_EVENTS, "Adding interface %s", iface->name);
 
+  pool *pool = rp_new(p->p.pool, iface->name);
   ifa = mb_allocz(pool, sizeof(struct radv_iface));
+  ifa->pool = pool;
   ifa->ra = p;
   ifa->cf = cf;
   ifa->iface = iface;
   ifa->addr = iface->llv6;
+  init_list(&ifa->prefixes);
 
   add_tail(&p->iface_list, NODE ifa);
 
@@ -172,11 +316,7 @@ radv_iface_remove(struct radv_iface *ifa)
 
   rem_node(NODE ifa);
 
-  rfree(ifa->sk);
+  rfree(ifa->pool);
-  rfree(ifa->timer);
-  rfree(ifa->lock);
-
-  mb_free(ifa);
 }
 
 static void

proto/radv/radv.h

@@ -35,6 +35,7 @@
 #define DEFAULT_MAX_RA_INT 600
 #define DEFAULT_MIN_DELAY 3
 #define DEFAULT_CURRENT_HOP_LIMIT 64
+#define DEFAULT_LINGER_TIME 300
 
 #define DEFAULT_VALID_LIFETIME 86400
 #define DEFAULT_PREFERRED_LIFETIME 14400
@@ -64,6 +65,9 @@ struct radv_iface_config
   u32 max_ra_int;
   u32 min_delay;
 
+  u32 linger_time;		/* How long a dead prefix should still be advertised with 0
+				   lifetime */
+
   u8 rdnss_local;		/* Global list is not used for RDNSS */
   u8 dnssl_local;		/* Global list is not used for DNSSL */
 
@@ -75,7 +79,7 @@ struct radv_iface_config
   u32 current_hop_limit;
   u32 default_lifetime;
   u8 default_lifetime_sensitive; /* Whether default_lifetime depends on trigger */
-  u8 default_preference;	/* Default Router Preference (RFC 4191) */
+  u8 default_preference;	 /* Default Router Preference (RFC 4191) */
 };
 
 struct radv_prefix_config
@@ -118,6 +122,19 @@ struct radv_proto
   u8 active;			/* Whether radv is active w.r.t. triggers */
 };
 
+struct radv_prefix		/* One prefix we advertise */
+{
+  node n;
+  net_addr_ip6 prefix;
+
+  u8 alive;			/* Is the prefix alive? If not, we advertise it
+				   with 0 lifetime, so clients stop using it */
+  u8 mark;			/* A temporary mark for processing */
+  btime expires;		/* The time when we drop this prefix from
+				   advertising. It is valid only if !alive. */
+  struct radv_prefix_config *cf; /* The config tied to this prefix */
+};
+
 struct radv_iface
 {
   node n;
@@ -125,6 +142,9 @@ struct radv_iface
   struct radv_iface_config *cf;	/* Related config, must be updated in reconfigure */
   struct iface *iface;
   struct ifa *addr;		/* Link-local address of iface */
+  struct pool *pool;		/* A pool for interface-specific things */
+  list prefixes;		/* The prefixes we advertise (struct radv_prefix) */
+  btime prefix_expires;		/* When the soonest prefix expires (0 = none dead) */
 
   timer *timer;
   struct object_lock *lock;
@@ -132,12 +152,13 @@ struct radv_iface
 
   btime last;			/* Time of last sending of RA */
   u16 plen;			/* Length of prepared RA in tbuf, or 0 if not valid */
-  byte initial;			/* List of active ifaces */
+  byte initial;			/* How many RAs are still to be sent as initial */
 };
 
 #define RA_EV_INIT 1		/* Switch to initial mode */
 #define RA_EV_CHANGE 2		/* Change of options or prefixes */
 #define RA_EV_RS 3		/* Received RS */
+#define RA_EV_GC 4		/* Internal garbage collection of prefixes */
 
 /* Default Router Preferences (RFC 4191) */
 #define RA_PREF_LOW	0x18

sysdep/bsd/krt-sock.c

@@ -165,7 +165,7 @@ struct ks_msg
 {
   struct rt_msghdr rtm;
   struct sockaddr_storage buf[RTAX_MAX];
-};
+} PACKED;
 
 #define ROUNDUP(a) \
         ((a) > 0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long))

sysdep/linux/netlink.c

@@ -83,22 +83,26 @@ struct rtvia {
 /*
  * Structure nl_parse_state keeps state of received route processing. Ideally,
  * we could just independently parse received Netlink messages and immediately
- * propagate received routes to the rest of BIRD, but Linux kernel represents
+ * propagate received routes to the rest of BIRD, but older Linux kernel (before
- * and announces IPv6 ECMP routes not as one route with multiple next hops (like
+ * version 4.11) represents and announces IPv6 ECMP routes not as one route with
- * RTA_MULTIPATH in IPv4 ECMP), but as a set of routes with the same prefix.
+ * multiple next hops (like RTA_MULTIPATH in IPv4 ECMP), but as a sequence of
+ * routes with the same prefix. More recent kernels work as with IPv4.
  *
  * Therefore, BIRD keeps currently processed route in nl_parse_state structure
  * and postpones its propagation until we expect it to be final; i.e., when
  * non-matching route is received or when the scan ends. When another matching
  * route is received, it is merged with the already processed route to form an
  * ECMP route. Note that merging is done only for IPv6 (merge == 1), but the
- * postponing is done in both cases (for simplicity). All IPv4 routes are just
+ * postponing is done in both cases (for simplicity). All IPv4 routes or IPv6
- * considered non-matching.
+ * routes with RTA_MULTIPATH set are just considered non-matching.
  *
  * This is ignored for asynchronous notifications (every notification is handled
  * as a separate route). It is not an issue for our routes, as we ignore such
  * notifications anyways. But importing alien IPv6 ECMP routes does not work
- * properly.
+ * properly with older kernels.
+ *
+ * Whatever the kernel version is, IPv6 ECMP routes are sent as multiple routes
+ * for the same prefix.
  */
 
 struct nl_parse_state
@@ -348,6 +352,12 @@ static struct nl_want_attrs nexthop_attr_want4[BIRD_RTA_MAX] = {
   [RTA_ENCAP]	  = { 1, 0, 0 },
 };
 
+static struct nl_want_attrs nexthop_attr_want6[BIRD_RTA_MAX] = {
+  [RTA_GATEWAY]	  = { 1, 1, sizeof(ip6_addr) },
+  [RTA_ENCAP_TYPE]= { 1, 1, sizeof(u16) },
+  [RTA_ENCAP]	  = { 1, 0, 0 },
+};
+
 static struct nl_want_attrs encap_mpls_want[BIRD_RTA_MAX] = {
   [RTA_DST]       = { 1, 0, 0 },
 };
@@ -374,6 +384,7 @@ static struct nl_want_attrs rtm_attr_want6[BIRD_RTA_MAX] = {
   [RTA_PRIORITY]  = { 1, 1, sizeof(u32) },
   [RTA_PREFSRC]	  = { 1, 1, sizeof(ip6_addr) },
   [RTA_METRICS]	  = { 1, 0, 0 },
+  [RTA_MULTIPATH] = { 1, 0, 0 },
   [RTA_FLOW]	  = { 1, 1, sizeof(u32) },
   [RTA_TABLE]	  = { 1, 1, sizeof(u32) },
   [RTA_ENCAP_TYPE]= { 1, 1, sizeof(u16) },
@@ -631,7 +642,7 @@ nl_add_multipath(struct nlmsghdr *h, uint bufsize, struct nexthop *nh, int af)
 }
 
 static struct nexthop *
-nl_parse_multipath(struct krt_proto *p, struct rtattr *ra)
+nl_parse_multipath(struct krt_proto *p, struct rtattr *ra, int af)
 {
   /* Temporary buffer for multicast nexthops */
   static struct nexthop *nh_buffer;
@@ -670,7 +681,22 @@ nl_parse_multipath(struct krt_proto *p, struct rtattr *ra)
 
       /* Nonexistent RTNH_PAYLOAD ?? */
       nl_attr_len = nh->rtnh_len - RTNH_LENGTH(0);
-      nl_parse_attrs(RTNH_DATA(nh), nexthop_attr_want4, a, sizeof(a));
+      switch (af)
+        {
+	case AF_INET:
+	  if (!nl_parse_attrs(RTNH_DATA(nh), nexthop_attr_want4, a, sizeof(a)))
+	    return NULL;
+	  break;
+
+	case AF_INET6:
+	  if (!nl_parse_attrs(RTNH_DATA(nh), nexthop_attr_want6, a, sizeof(a)))
+	    return NULL;
+	  break;
+
+	default:
+	  return NULL;
+	}
+
       if (a[RTA_GATEWAY])
 	{
 	  rv->gw = rta_get_ipa(a[RTA_GATEWAY]);
@@ -1520,9 +1546,9 @@ nl_parse_route(struct nl_parse_state *s, struct nlmsghdr *h)
     case RTN_UNICAST:
       ra->dest = RTD_UNICAST;
 
-      if (a[RTA_MULTIPATH] && (i->rtm_family == AF_INET))
+      if (a[RTA_MULTIPATH])
-	{
+        {
-	  struct nexthop *nh = nl_parse_multipath(p, a[RTA_MULTIPATH]);
+	  struct nexthop *nh = nl_parse_multipath(p, a[RTA_MULTIPATH], i->rtm_family);
 	  if (!nh)
 	    {
 	      log(L_ERR "KRT: Received strange multipath route %N", net->n.addr);
@@ -1699,8 +1725,10 @@ nl_parse_route(struct nl_parse_state *s, struct nlmsghdr *h)
 
   /*
    * Ideally, now we would send the received route to the rest of kernel code.
-   * But IPv6 ECMP routes are sent as a sequence of routes, so we postpone it
+   * But IPv6 ECMP routes before 4.11 are sent as a sequence of routes, so we
-   * and merge next hops until the end of the sequence.
+   * postpone it and merge next hops until the end of the sequence. Note that
+   * proper multipath updates are rejected by nl_mergable_route(), so it is
+   * always the first case for them.
    */
 
   if (!s->net)