Add RTR support to OpenBGPD. Add basic support for the protocol.
authorclaudio <claudio@openbsd.org>
Tue, 16 Feb 2021 08:29:16 +0000 (08:29 +0000)
committerclaudio <claudio@openbsd.org>
Tue, 16 Feb 2021 08:29:16 +0000 (08:29 +0000)
The RTR client runs in a new process where the protocol handling is done
and when new data is available all sources are merged into one ROA set
which is then loaded into the RDE. The roa-set from the config is also
handled by the new RTR engine.
Tested by and ok job@

15 files changed:
usr.sbin/bgpd/Makefile
usr.sbin/bgpd/bgpd.8
usr.sbin/bgpd/bgpd.c
usr.sbin/bgpd/bgpd.conf.5
usr.sbin/bgpd/bgpd.h
usr.sbin/bgpd/config.c
usr.sbin/bgpd/control.c
usr.sbin/bgpd/parse.y
usr.sbin/bgpd/printconf.c
usr.sbin/bgpd/rde.c
usr.sbin/bgpd/rtr.c [new file with mode: 0644]
usr.sbin/bgpd/rtr_proto.c [new file with mode: 0644]
usr.sbin/bgpd/session.c
usr.sbin/bgpd/session.h
usr.sbin/bgpd/util.c

index 55d6f47..61c73d5 100644 (file)
@@ -1,11 +1,11 @@
-#      $OpenBSD: Makefile,v 1.36 2020/01/01 07:25:04 claudio Exp $
+#      $OpenBSD: Makefile,v 1.37 2021/02/16 08:29:16 claudio Exp $
 
 PROG=  bgpd
 SRCS=  bgpd.c session.c log.c logmsg.c parse.y config.c \
        rde.c rde_rib.c rde_decide.c rde_prefix.c mrt.c kroute.c control.c \
        pfkey.c rde_update.c rde_attr.c rde_community.c printconf.c \
        rde_filter.c rde_sets.c rde_trie.c pftable.c name2id.c \
-       util.c carp.c timer.c rde_peer.c
+       util.c carp.c timer.c rde_peer.c rtr.c rtr_proto.c
 CFLAGS+= -Wall -I${.CURDIR}
 CFLAGS+= -Wstrict-prototypes -Wmissing-prototypes
 CFLAGS+= -Wmissing-declarations
index 3c2cb0a..26439ce 100644 (file)
@@ -1,4 +1,4 @@
-.\" $OpenBSD: bgpd.8,v 1.63 2021/02/01 07:39:59 jmc Exp $
+.\" $OpenBSD: bgpd.8,v 1.64 2021/02/16 08:29:16 claudio Exp $
 .\"
 .\" Copyright (c) 2003, 2004 Henning Brauer <henning@openbsd.org>
 .\"
@@ -14,7 +14,7 @@
 .\" ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
 .\" OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
 .\"
-.Dd $Mdocdate: February 1 2021 $
+.Dd $Mdocdate: February 16 2021 $
 .Dt BGPD 8
 .Os
 .Sh NAME
@@ -413,6 +413,14 @@ has been started.
 .Re
 .Pp
 .Rs
+.%A R. Bush
+.%A R. Austein
+.%D September 2017
+.%R RFC 8210
+.%T The Resource Public Key Infrastructure (RPKI) to Router Protocol, Version 1
+.Re
+.Pp
+.Rs
 .%A J. Mauch
 .%A J. Snijders
 .%A G. Hankins
index 63e4768..e6bf4c8 100644 (file)
@@ -1,4 +1,4 @@
-/*     $OpenBSD: bgpd.c,v 1.233 2021/01/04 17:44:14 claudio Exp $ */
+/*     $OpenBSD: bgpd.c,v 1.234 2021/02/16 08:29:16 claudio Exp $ */
 
 /*
  * Copyright (c) 2003, 2004 Henning Brauer <henning@openbsd.org>
@@ -47,7 +47,9 @@ int           send_config(struct bgpd_config *);
 int            dispatch_imsg(struct imsgbuf *, int, struct bgpd_config *);
 int            control_setup(struct bgpd_config *);
 static void    getsockpair(int [2]);
-int            imsg_send_sockets(struct imsgbuf *, struct imsgbuf *);
+int            imsg_send_sockets(struct imsgbuf *, struct imsgbuf *,
+                   struct imsgbuf *);
+void           bgpd_rtr_connect(struct rtr_config *);
 
 int                     cflags;
 volatile sig_atomic_t   mrtdump;
@@ -57,6 +59,7 @@ pid_t                  reconfpid;
 int                     reconfpending;
 struct imsgbuf         *ibuf_se;
 struct imsgbuf         *ibuf_rde;
+struct imsgbuf         *ibuf_rtr;
 struct rib_names        ribnames = SIMPLEQ_HEAD_INITIALIZER(ribnames);
 char                   *cname;
 char                   *rcname;
@@ -91,9 +94,10 @@ usage(void)
 
 #define PFD_PIPE_SESSION       0
 #define PFD_PIPE_RDE           1
-#define PFD_SOCK_ROUTE         2
-#define PFD_SOCK_PFKEY         3
-#define POLL_MAX               4
+#define PFD_PIPE_RTR           2
+#define PFD_SOCK_ROUTE         3
+#define PFD_SOCK_PFKEY         4
+#define POLL_MAX               5
 #define MAX_TIMEOUT            3600
 
 int     cmd_opts;
@@ -107,7 +111,7 @@ main(int argc, char *argv[])
        struct peer             *p;
        struct pollfd            pfd[POLL_MAX];
        time_t                   timeout;
-       pid_t                    se_pid = 0, rde_pid = 0, pid;
+       pid_t                    se_pid = 0, rde_pid = 0, rtr_pid = 0, pid;
        char                    *conffile;
        char                    *saved_argv0;
        int                      debug = 0;
@@ -115,6 +119,7 @@ main(int argc, char *argv[])
        int                      ch, status;
        int                      pipe_m2s[2];
        int                      pipe_m2r[2];
+       int                      pipe_m2roa[2];
 
        conffile = CONFFILE;
 
@@ -126,7 +131,7 @@ main(int argc, char *argv[])
        if (saved_argv0 == NULL)
                saved_argv0 = "bgpd";
 
-       while ((ch = getopt(argc, argv, "cdD:f:nRSv")) != -1) {
+       while ((ch = getopt(argc, argv, "cdD:f:nRSTv")) != -1) {
                switch (ch) {
                case 'c':
                        cmd_opts |= BGPD_OPT_FORCE_DEMOTE;
@@ -156,6 +161,9 @@ main(int argc, char *argv[])
                case 'S':
                        proc = PROC_SE;
                        break;
+               case 'T':
+                       proc = PROC_RTR;
+                       break;
                default:
                        usage();
                        /* NOTREACHED */
@@ -168,7 +176,7 @@ main(int argc, char *argv[])
                usage();
 
        if (cmd_opts & BGPD_OPT_NOACTION) {
-               if ((conf = parse_config(conffile, NULL)) == NULL)
+               if ((conf = parse_config(conffile, NULL, NULL)) == NULL)
                        exit(1);
 
                if (cmd_opts & BGPD_OPT_VERBOSE)
@@ -193,6 +201,9 @@ main(int argc, char *argv[])
        case PROC_SE:
                session_main(debug, cmd_opts & BGPD_OPT_VERBOSE);
                /* NOTREACHED */
+       case PROC_RTR:
+               rtr_main(debug, cmd_opts & BGPD_OPT_VERBOSE);
+               /* NOTREACHED */
        }
 
        if (geteuid())
@@ -201,7 +212,7 @@ main(int argc, char *argv[])
        if (getpwnam(BGPD_USER) == NULL)
                errx(1, "unknown user %s", BGPD_USER);
 
-       if ((conf = parse_config(conffile, NULL)) == NULL) {
+       if ((conf = parse_config(conffile, NULL, NULL)) == NULL) {
                log_warnx("config file %s has errors", conffile);
                exit(1);
        }
@@ -219,12 +230,15 @@ main(int argc, char *argv[])
 
        getsockpair(pipe_m2s);
        getsockpair(pipe_m2r);
+       getsockpair(pipe_m2roa);
 
        /* fork children */
        rde_pid = start_child(PROC_RDE, saved_argv0, pipe_m2r[1], debug,
            cmd_opts & BGPD_OPT_VERBOSE);
        se_pid = start_child(PROC_SE, saved_argv0, pipe_m2s[1], debug,
            cmd_opts & BGPD_OPT_VERBOSE);
+       rtr_pid = start_child(PROC_RTR, saved_argv0, pipe_m2roa[1], debug,
+           cmd_opts & BGPD_OPT_VERBOSE);
 
        signal(SIGTERM, sighdlr);
        signal(SIGINT, sighdlr);
@@ -234,10 +248,12 @@ main(int argc, char *argv[])
        signal(SIGPIPE, SIG_IGN);
 
        if ((ibuf_se = malloc(sizeof(struct imsgbuf))) == NULL ||
-           (ibuf_rde = malloc(sizeof(struct imsgbuf))) == NULL)
+           (ibuf_rde = malloc(sizeof(struct imsgbuf))) == NULL ||
+           (ibuf_rtr = malloc(sizeof(struct imsgbuf))) == NULL)
                fatal(NULL);
        imsg_init(ibuf_se, pipe_m2s[0]);
        imsg_init(ibuf_rde, pipe_m2r[0]);
+       imsg_init(ibuf_rtr, pipe_m2roa[0]);
        mrt_init(ibuf_rde, ibuf_se);
        if (kr_init(&rfd) == -1)
                quit = 1;
@@ -262,7 +278,7 @@ BROKEN      if (pledge("stdio rpath wpath cpath fattr unix route recvfd sendfd",
                fatal("pledge");
 #endif
 
-       if (imsg_send_sockets(ibuf_se, ibuf_rde))
+       if (imsg_send_sockets(ibuf_se, ibuf_rde, ibuf_rtr))
                fatal("could not establish imsg links");
        /* control setup needs to happen late since it sends imsgs */
        if (control_setup(conf) == -1)
@@ -285,6 +301,7 @@ BROKEN      if (pledge("stdio rpath wpath cpath fattr unix route recvfd sendfd",
 
                set_pollfd(&pfd[PFD_PIPE_SESSION], ibuf_se);
                set_pollfd(&pfd[PFD_PIPE_RDE], ibuf_rde);
+               set_pollfd(&pfd[PFD_PIPE_RTR], ibuf_rtr);
 
                if (timeout < 0 || timeout > MAX_TIMEOUT)
                        timeout = MAX_TIMEOUT;
@@ -313,8 +330,18 @@ BROKEN     if (pledge("stdio rpath wpath cpath fattr unix route recvfd sendfd",
                        ibuf_rde = NULL;
                        quit = 1;
                } else {
-                       if (dispatch_imsg(ibuf_rde, PFD_PIPE_RDE, conf) ==
-                           -1)
+                       if (dispatch_imsg(ibuf_rde, PFD_PIPE_RDE, conf) == -1)
+                               quit = 1;
+               }
+
+               if (handle_pollfd(&pfd[PFD_PIPE_RTR], ibuf_rtr) == -1) {
+                       log_warnx("main: Lost connection to RTR");
+                       msgbuf_clear(&ibuf_rtr->w);
+                       free(ibuf_rtr);
+                       ibuf_rtr = NULL;
+                       quit = 1;
+               } else {
+                       if (dispatch_imsg(ibuf_rtr, PFD_PIPE_RTR, conf) == -1)
                                quit = 1;
                }
 
@@ -377,6 +404,12 @@ BROKEN     if (pledge("stdio rpath wpath cpath fattr unix route recvfd sendfd",
                free(ibuf_rde);
                ibuf_rde = NULL;
        }
+       if (ibuf_rtr) {
+               msgbuf_clear(&ibuf_rtr->w);
+               close(ibuf_rtr->fd);
+               free(ibuf_rtr);
+               ibuf_rtr = NULL;
+       }
 
        /* cleanup kernel data structures */
        carp_demote_shutdown();
@@ -404,6 +437,8 @@ BROKEN      if (pledge("stdio rpath wpath cpath fattr unix route recvfd sendfd",
                                name = "route decision engine";
                        else if (pid == se_pid)
                                name = "session engine";
+                       else if (pid == rtr_pid)
+                               name = "rtr engine";
                        log_warnx("%s terminated; signal %d", name,
                                WTERMSIG(status));
                }
@@ -449,6 +484,9 @@ start_child(enum bgpd_process p, char *argv0, int fd, int debug, int verbose)
        case PROC_SE:
                argv[argc++] = "-S";
                break;
+       case PROC_RTR:
+               argv[argc++] = "-T";
+               break;
        }
        if (debug)
                argv[argc++] = "-d";
@@ -481,7 +519,8 @@ reconfigure(char *conffile, struct bgpd_config *conf)
                return (2);
 
        log_info("rereading config");
-       if ((new_conf = parse_config(conffile, &conf->peers)) == NULL)
+       if ((new_conf = parse_config(conffile, &conf->peers,
+           &conf->rtrs)) == NULL)
                return (1);
 
        merge_config(conf, new_conf);
@@ -509,8 +548,9 @@ send_config(struct bgpd_config *conf)
        struct prefixset        *ps;
        struct prefixset_item   *psi, *npsi;
        struct roa              *roa, *nroa;
+       struct rtr_config       *rtr;
 
-       reconfpending = 2;      /* one per child */
+       reconfpending = 3;      /* one per child */
 
        expand_networks(conf);
 
@@ -523,6 +563,9 @@ send_config(struct bgpd_config *conf)
        if (imsg_compose(ibuf_rde, IMSG_RECONF_CONF, 0, 0, -1,
            conf, sizeof(*conf)) == -1)
                return (-1);
+       if (imsg_compose(ibuf_rtr, IMSG_RECONF_CONF, 0, 0, -1,
+           conf, sizeof(*conf)) == -1)
+               return (-1);
 
        TAILQ_FOREACH(la, conf->listen_addrs, entry) {
                if (imsg_compose(ibuf_se, IMSG_RECONF_LISTENER, 0, 0, la->fd,
@@ -595,17 +638,18 @@ send_config(struct bgpd_config *conf)
                free(ps);
        }
 
-       if (!RB_EMPTY(&conf->roa)) {
-               if (imsg_compose(ibuf_rde, IMSG_RECONF_ROA_SET, 0, 0, -1,
-                   NULL, 0) == -1)
+       /* roa table and rtr config are sent to the RTR engine */
+       RB_FOREACH_SAFE(roa, roa_tree, &conf->roa, nroa) {
+               RB_REMOVE(roa_tree, &conf->roa, roa);
+               if (imsg_compose(ibuf_rtr, IMSG_RECONF_ROA_ITEM, 0, 0,
+                   -1, roa, sizeof(*roa)) == -1)
+                       return (-1);
+               free(roa);
+       }
+       SIMPLEQ_FOREACH(rtr, &conf->rtrs, entry) {
+               if (imsg_compose(ibuf_rtr, IMSG_RECONF_RTR_CONFIG, rtr->id,
+                   0, -1, rtr->descr, sizeof(rtr->descr)) == -1)
                        return (-1);
-               RB_FOREACH_SAFE(roa, roa_tree, &conf->roa, nroa) {
-                       RB_REMOVE(roa_tree, &conf->roa, roa);
-                       if (imsg_compose(ibuf_rde, IMSG_RECONF_ROA_ITEM, 0, 0,
-                           -1, roa, sizeof(*roa)) == -1)
-                               return (-1);
-                       free(roa);
-               }
        }
 
        /* as-sets for filters in the RDE */
@@ -695,6 +739,8 @@ send_config(struct bgpd_config *conf)
                return (-1);
        if (imsg_compose(ibuf_rde, IMSG_RECONF_DRAIN, 0, 0, -1, NULL, 0) == -1)
                return (-1);
+       if (imsg_compose(ibuf_rtr, IMSG_RECONF_DRAIN, 0, 0, -1, NULL, 0) == -1)
+               return (-1);
 
        /* mrt changes can be sent out of bound */
        mrt_reconfigure(conf->mrt);
@@ -706,6 +752,7 @@ dispatch_imsg(struct imsgbuf *ibuf, int idx, struct bgpd_config *conf)
 {
        struct imsg              imsg;
        struct peer             *p;
+       struct rtr_config       *r;
        ssize_t                  n;
        int                      rv, verbose;
 
@@ -874,6 +921,9 @@ dispatch_imsg(struct imsgbuf *ibuf, int idx, struct bgpd_config *conf)
                                break;
                        }
                        if (idx == PFD_PIPE_SESSION) {
+                               imsg_compose(ibuf_rtr, IMSG_RECONF_DONE, 0,
+                                   0, -1, NULL, 0);
+                       } else if (idx == PFD_PIPE_RTR) {
                                imsg_compose(ibuf_rde, IMSG_RECONF_DONE, 0,
                                    0, -1, NULL, 0);
 
@@ -898,9 +948,66 @@ dispatch_imsg(struct imsgbuf *ibuf, int idx, struct bgpd_config *conf)
                                 */
                                imsg_compose(ibuf_se, IMSG_RECONF_DONE, 0,
                                    0, -1, NULL, 0);
-                               reconfpending = 2; /* expecting 2 DONE msg */
+                               reconfpending = 3; /* expecting 2 DONE msg */
+                       }
+                       break;
+               case IMSG_SOCKET_CONN:
+                       if (idx != PFD_PIPE_RTR) {
+                               log_warnx("connect request not from RTR");
+                       } else {
+                               SIMPLEQ_FOREACH(r, &conf->rtrs, entry) {
+                                       if (imsg.hdr.peerid == r->id)
+                                               break;
+                               }
+                               if (r == NULL)
+                                       log_warnx("unknown rtr id %d",
+                                           imsg.hdr.peerid);
+                               else
+                                       bgpd_rtr_connect(r);
                        }
                        break;
+               case IMSG_CTL_SHOW_RTR:
+                       if (idx == PFD_PIPE_SESSION) {
+                               SIMPLEQ_FOREACH(r, &conf->rtrs, entry) {
+                                       imsg_compose(ibuf_rtr, imsg.hdr.type,
+                                           r->id, imsg.hdr.pid, -1, NULL, 0);
+                               }
+                               imsg_compose(ibuf_rtr, IMSG_CTL_END,
+                                   0, imsg.hdr.pid, -1, NULL, 0);
+                       } else if (imsg.hdr.len != IMSG_HEADER_SIZE +
+                           sizeof(struct ctl_show_rtr)) {
+                               log_warnx("IMSG_CTL_SHOW_RTR with wrong len");
+                       } else if (idx == PFD_PIPE_RTR) {
+                               SIMPLEQ_FOREACH(r, &conf->rtrs, entry) {
+                                       if (imsg.hdr.peerid == r->id)
+                                               break;
+                               }
+                               if (r != NULL) {
+                                       struct ctl_show_rtr *msg;
+                                       msg = imsg.data;
+                                       strlcpy(msg->descr, r->descr,
+                                           sizeof(msg->descr));
+                                       msg->local_addr = r->local_addr;
+                                       msg->remote_addr = r->remote_addr;
+                                       msg->remote_port = r->remote_port;
+
+                                       imsg_compose(ibuf_se, imsg.hdr.type,
+                                           imsg.hdr.peerid, imsg.hdr.pid,
+                                           -1, imsg.data,
+                                           imsg.hdr.len - IMSG_HEADER_SIZE);
+                               }
+                       }
+                       break;
+               case IMSG_CTL_END:
+               case IMSG_CTL_SHOW_TIMER:
+                       if (idx != PFD_PIPE_RTR) {
+                               log_warnx("connect request not from RTR");
+                               break;
+                       }
+                       imsg_compose(ibuf_se, imsg.hdr.type, imsg.hdr.peerid,
+                           imsg.hdr.pid, -1, imsg.data,
+                           imsg.hdr.len - IMSG_HEADER_SIZE);
+                       break;
                default:
                        break;
                }
@@ -1117,13 +1224,15 @@ getsockpair(int pipe[2])
 }
 
 int
-imsg_send_sockets(struct imsgbuf *se, struct imsgbuf *rde)
+imsg_send_sockets(struct imsgbuf *se, struct imsgbuf *rde, struct imsgbuf *roa)
 {
        int pipe_s2r[2];
        int pipe_s2r_ctl[2];
+       int pipe_r2r[2];
 
        getsockpair(pipe_s2r);
        getsockpair(pipe_s2r_ctl);
+       getsockpair(pipe_r2r);
 
        if (imsg_compose(se, IMSG_SOCKET_CONN, 0, 0, pipe_s2r[0],
            NULL, 0) == -1)
@@ -1139,5 +1248,44 @@ imsg_send_sockets(struct imsgbuf *se, struct imsgbuf *rde)
            NULL, 0) == -1)
                return (-1);
 
+       if (imsg_compose(roa, IMSG_SOCKET_CONN_RTR, 0, 0, pipe_r2r[0],
+           NULL, 0) == -1)
+               return (-1);
+       if (imsg_compose(rde, IMSG_SOCKET_CONN_RTR, 0, 0, pipe_r2r[1],
+           NULL, 0) == -1)
+               return (-1);
+
        return (0);
 }
+
+void
+bgpd_rtr_connect(struct rtr_config *r)
+{
+       socklen_t len;
+       int fd;
+
+       /* XXX should be non-blocking */
+       fd = socket(aid2af(r->remote_addr.aid), SOCK_STREAM, 0);
+       if (fd == -1) {
+               log_warn("rtr %s", r->descr);
+               return;
+       }
+       if (r->local_addr.aid != AID_UNSPEC) {
+               if (bind(fd,  addr2sa(&r->local_addr, 0, &len), len) == -1) {
+                       log_warn("rtr %s: bind to %s", r->descr,
+                           log_addr(&r->local_addr));
+                       close(fd);
+                       return;
+               }
+       }
+
+       if (connect(fd, addr2sa(&r->remote_addr, r->remote_port, &len), len) ==
+           -1) {
+               log_warn("rtr %s: connect to %s:%u", r->descr,
+                   log_addr(&r->remote_addr), r->remote_port);
+               close(fd);
+               return;
+       }
+
+       imsg_compose(ibuf_rtr, IMSG_SOCKET_CONN, r->id, 0, fd, NULL, 0);
+}
index eb93741..a1e21a7 100644 (file)
@@ -1,4 +1,4 @@
-.\" $OpenBSD: bgpd.conf.5,v 1.207 2021/01/27 14:57:29 jmc Exp $
+.\" $OpenBSD: bgpd.conf.5,v 1.208 2021/02/16 08:29:16 claudio Exp $
 .\"
 .\" Copyright (c) 2004 Claudio Jeker <claudio@openbsd.org>
 .\" Copyright (c) 2003, 2004 Henning Brauer <henning@openbsd.org>
@@ -16,7 +16,7 @@
 .\" ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
 .\" OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
 .\"
-.Dd $Mdocdate: January 27 2021 $
+.Dd $Mdocdate: February 16 2021 $
 .Dt BGPD.CONF 5
 .Os
 .Sh NAME
@@ -494,6 +494,39 @@ and the Origin Validation State (OVS) is set.
 roa-set { 192.0.2.0/24 maxlen 24 source-as 64511
           203.0.113.0/24 source-as 64496 }
 .Ed
+.Pp
+.It Xo
+.Ic rtr Ar address
+.Ic { Ar ... Ic }
+.Xc
+The
+.Ic rtr
+block specifies a
+.Em RPKI to Router Protocol
+session.
+The rtr session properties are as follows:
+.Pp
+.Bl -tag -width Ds -compact
+.It Ic descr Ar description
+Add a description.
+The description is used in logging and status reports, but has no further
+meaning for
+.Xr bgpd 8 .
+.Pp
+.It Ic local-address Ar address
+Bind to the specific IP address before opening the TCP connection to the
+.Em rtr
+server.
+.Pp
+.It Ic port Ar number
+Specify the TCP destination port for the
+.Em rtr
+session.
+If not specified the default
+.Ic port
+is
+.Em 323 .
+.El
 .El
 .Sh NETWORK ANNOUNCEMENTS
 .Ic network
index 97f20cd..03578e9 100644 (file)
@@ -1,4 +1,4 @@
-/*     $OpenBSD: bgpd.h,v 1.411 2021/01/25 09:15:23 claudio Exp $ */
+/*     $OpenBSD: bgpd.h,v 1.412 2021/02/16 08:29:16 claudio Exp $ */
 
 /*
  * Copyright (c) 2003, 2004 Henning Brauer <henning@openbsd.org>
 enum bgpd_process {
        PROC_MAIN,
        PROC_SE,
-       PROC_RDE
+       PROC_RDE,
+       PROC_RTR,
 };
 
 enum reconf_action {
@@ -266,6 +267,9 @@ SIMPLEQ_HEAD(as_set_head, as_set);
 struct filter_rule;
 TAILQ_HEAD(filter_head, filter_rule);
 
+struct rtr_config;
+SIMPLEQ_HEAD(rtr_config_head, rtr_config);
+
 struct bgpd_config {
        struct peer_head                         peers;
        struct l3vpn_head                        l3vpns;
@@ -278,8 +282,8 @@ struct bgpd_config {
        struct roa_tree                          roa;
        struct rde_prefixset_head                rde_prefixsets;
        struct rde_prefixset_head                rde_originsets;
-       struct rde_prefixset                     rde_roa;
        struct as_set_head                       as_sets;
+       struct rtr_config_head                   rtrs;
        char                                    *csock;
        char                                    *rcsock;
        int                                      flags;
@@ -433,6 +437,44 @@ struct network {
        TAILQ_ENTRY(network)            entry;
 };
 
+enum rtr_error {
+       NO_ERROR = -1,
+       CORRUPT_DATA = 0,
+       INTERNAL_ERROR,
+       NO_DATA_AVAILABLE,
+       INVALID_REQUEST,
+       UNSUPP_PROTOCOL_VERS,
+       UNSUPP_PDU_TYPE,
+       UNK_REC_WDRAWL,
+       DUP_REC_RECV,
+       UNEXP_PROTOCOL_VERS,
+};
+
+struct rtr_config {
+       SIMPLEQ_ENTRY(rtr_config)       entry;
+       char                            descr[PEER_DESCR_LEN];
+       struct bgpd_addr                remote_addr;
+       struct bgpd_addr                local_addr;
+       u_int32_t                       id;
+       in_addr_t                       remote_port;
+};
+
+struct ctl_show_rtr {
+       char                    descr[PEER_DESCR_LEN];
+       struct bgpd_addr        remote_addr;
+       struct bgpd_addr        local_addr;
+       uint32_t                serial;
+       uint32_t                refresh;
+       uint32_t                retry;
+       uint32_t                expire;
+       int                     session_id;
+       in_addr_t               remote_port;
+       enum rtr_error          last_sent_error;
+       enum rtr_error          last_recv_error;
+       char                    last_sent_msg[REASON_LEN];
+       char                    last_recv_msg[REASON_LEN];
+};
+
 enum imsg_type {
        IMSG_NONE,
        IMSG_CTL_END,
@@ -462,6 +504,7 @@ enum imsg_type {
        IMSG_CTL_LOG_VERBOSE,
        IMSG_CTL_SHOW_FIB_TABLES,
        IMSG_CTL_SHOW_SET,
+       IMSG_CTL_SHOW_RTR,
        IMSG_CTL_TERMINATE,
        IMSG_NETWORK_ADD,
        IMSG_NETWORK_ASPATH,
@@ -472,6 +515,7 @@ enum imsg_type {
        IMSG_FILTER_SET,
        IMSG_SOCKET_CONN,
        IMSG_SOCKET_CONN_CTL,
+       IMSG_SOCKET_CONN_RTR,
        IMSG_RECONF_CONF,
        IMSG_RECONF_RIB,
        IMSG_RECONF_PEER,
@@ -490,6 +534,7 @@ enum imsg_type {
        IMSG_RECONF_ORIGIN_SET,
        IMSG_RECONF_ROA_SET,
        IMSG_RECONF_ROA_ITEM,
+       IMSG_RECONF_RTR_CONFIG,
        IMSG_RECONF_DRAIN,
        IMSG_RECONF_DONE,
        IMSG_UPDATE,
@@ -1194,11 +1239,13 @@ void            free_prefixsets(struct prefixset_head *);
 void           free_rde_prefixsets(struct rde_prefixset_head *);
 void           free_prefixtree(struct prefixset_tree *);
 void           free_roatree(struct roa_tree *);
+void           free_rtrs(struct rtr_config_head *);
 void           filterlist_free(struct filter_head *);
 int            host(const char *, struct bgpd_addr *, u_int8_t *);
 u_int32_t      get_bgpid(void);
 void           expand_networks(struct bgpd_config *);
 RB_PROTOTYPE(prefixset_tree, prefixset_item, entry, prefixset_cmp);
+int            roa_cmp(struct roa *, struct roa *);
 RB_PROTOTYPE(roa_tree, roa, entry, roa_cmp);
 
 /* kroute.c */
@@ -1262,8 +1309,10 @@ void              pftable_unref(u_int16_t);
 u_int16_t       pftable_ref(u_int16_t);
 
 /* parse.y */
-int             cmdline_symset(char *);
-struct prefixset *find_prefixset(char *, struct prefixset_head *);
+int                    cmdline_symset(char *);
+struct prefixset       *find_prefixset(char *, struct prefixset_head *);
+struct bgpd_config     *parse_config(char *, struct peer_head *,
+                           struct rtr_config_head *);
 
 /* pftable.c */
 int    pftable_exists(const char *);
@@ -1320,6 +1369,7 @@ const char        *log_as(u_int32_t);
 const char     *log_rd(u_int64_t);
 const char     *log_ext_subtype(short, u_int8_t);
 const char     *log_reason(const char *);
+const char     *log_rtr_error(enum rtr_error);
 int             aspath_snprint(char *, size_t, void *, u_int16_t);
 int             aspath_asprint(char **, void *, u_int16_t);
 size_t          aspath_strlen(void *, u_int16_t);
@@ -1356,7 +1406,8 @@ const char *       get_baudrate(unsigned long long, char *);
 static const char * const log_procnames[] = {
        "parent",
        "SE",
-       "RDE"
+       "RDE",
+       "RTR"
 };
 
 /* logmsg.c and needed by bgpctl */
@@ -1485,6 +1536,10 @@ static const char * const timernames[] = {
        "IdleHoldTimer",
        "IdleHoldResetTimer",
        "CarpUndemoteTimer",
+       "RestartTimer",
+       "RTR RefreshTimer",
+       "RTR RetryTimer",
+       "RTR ExpireTimer",
        ""
 };
 
index 1ccb45e..d5b326e 100644 (file)
@@ -1,4 +1,4 @@
-/*     $OpenBSD: config.c,v 1.98 2021/01/04 13:42:11 claudio Exp $ */
+/*     $OpenBSD: config.c,v 1.99 2021/02/16 08:29:16 claudio Exp $ */
 
 /*
  * Copyright (c) 2003, 2004, 2005 Henning Brauer <henning@openbsd.org>
@@ -60,6 +60,7 @@ new_config(void)
        SIMPLEQ_INIT(&conf->rde_originsets);
        RB_INIT(&conf->roa);
        SIMPLEQ_INIT(&conf->as_sets);
+       SIMPLEQ_INIT(&conf->rtrs);
 
        TAILQ_INIT(conf->filters);
        TAILQ_INIT(conf->listen_addrs);
@@ -162,6 +163,18 @@ free_roatree(struct roa_tree *r)
        }
 }
 
+void
+free_rtrs(struct rtr_config_head *rh)
+{
+       struct rtr_config       *r;
+
+       while (!SIMPLEQ_EMPTY(rh)) {
+               r = SIMPLEQ_FIRST(rh);
+               SIMPLEQ_REMOVE_HEAD(rh, entry);
+               free(r);
+       }
+}
+
 void
 free_config(struct bgpd_config *conf)
 {
@@ -178,6 +191,7 @@ free_config(struct bgpd_config *conf)
        free_rde_prefixsets(&conf->rde_originsets);
        as_sets_free(&conf->as_sets);
        free_roatree(&conf->roa);
+       free_rtrs(&conf->rtrs);
 
        while ((la = TAILQ_FIRST(conf->listen_addrs)) != NULL) {
                TAILQ_REMOVE(conf->listen_addrs, la, entry);
@@ -246,6 +260,10 @@ merge_config(struct bgpd_config *xconf, struct bgpd_config *conf)
        RB_ROOT(&xconf->roa) = RB_ROOT(&conf->roa);
        RB_ROOT(&conf->roa) = NULL;
 
+       /* switch the rtr_configs, first remove the old ones */
+       free_rtrs(&xconf->rtrs);
+       SIMPLEQ_CONCAT(&xconf->rtrs, &conf->rtrs);
+
        /* switch the prefixsets, first remove the old ones */
        free_prefixsets(&xconf->prefixsets);
        SIMPLEQ_CONCAT(&xconf->prefixsets, &conf->prefixsets);
@@ -582,7 +600,7 @@ prefixset_cmp(struct prefixset_item *a, struct prefixset_item *b)
 
 RB_GENERATE(prefixset_tree, prefixset_item, entry, prefixset_cmp);
 
-static inline int
+int
 roa_cmp(struct roa *a, struct roa *b)
 {
        int i;
index 4430362..1e7bbda 100644 (file)
@@ -1,4 +1,4 @@
-/*     $OpenBSD: control.c,v 1.103 2020/12/30 07:29:56 claudio Exp $ */
+/*     $OpenBSD: control.c,v 1.104 2021/02/16 08:29:16 claudio Exp $ */
 
 /*
  * Copyright (c) 2003, 2004 Henning Brauer <henning@openbsd.org>
@@ -448,6 +448,7 @@ control_dispatch_msg(struct pollfd *pfd, struct peer_head *peers)
                case IMSG_CTL_RELOAD:
                case IMSG_CTL_SHOW_INTERFACE:
                case IMSG_CTL_SHOW_FIB_TABLES:
+               case IMSG_CTL_SHOW_RTR:
                        c->ibuf.pid = imsg.hdr.pid;
                        imsg_ctl_parent(imsg.hdr.type, 0, imsg.hdr.pid,
                            imsg.data, imsg.hdr.len - IMSG_HEADER_SIZE);
index f280240..7dd7f7c 100644 (file)
@@ -1,4 +1,4 @@
-/*     $OpenBSD: parse.y,v 1.412 2021/01/25 09:15:23 claudio Exp $ */
+/*     $OpenBSD: parse.y,v 1.413 2021/02/16 08:29:16 claudio Exp $ */
 
 /*
  * Copyright (c) 2002, 2003, 2004 Henning Brauer <henning@openbsd.org>
@@ -89,12 +89,14 @@ char                *symget(const char *);
 static struct bgpd_config      *conf;
 static struct network_head     *netconf;
 static struct peer_head                *new_peers, *cur_peers;
+static struct rtr_config_head  *cur_rtrs;
 static struct peer             *curpeer;
 static struct peer             *curgroup;
 static struct rde_rib          *currib;
 static struct l3vpn            *curvpn;
 static struct prefixset                *curpset, *curoset;
 static struct roa_tree         *curroatree;
+static struct rtr_config       *currtr;
 static struct filter_head      *filter_l;
 static struct filter_head      *peerfilter_l;
 static struct filter_head      *groupfilter_l;
@@ -162,6 +164,8 @@ static void  add_as_set(u_int32_t);
 static void     done_as_set(void);
 static struct prefixset        *new_prefix_set(char *, int);
 static void     add_roa_set(struct prefixset_item *, u_int32_t, u_int8_t);
+static struct rtr_config       *get_rtr(struct bgpd_addr *);
+static int      insert_rtr(struct rtr_config *);
 
 typedef struct {
        union {
@@ -196,7 +200,7 @@ typedef struct {
 
 %token AS ROUTERID HOLDTIME YMIN LISTEN ON FIBUPDATE FIBPRIORITY RTABLE
 %token NONE UNICAST VPN RD EXPORT EXPORTTRGT IMPORTTRGT DEFAULTROUTE
-%token RDE RIB EVALUATE IGNORE COMPARE
+%token RDE RIB EVALUATE IGNORE COMPARE RTR PORT
 %token GROUP NEIGHBOR NETWORK
 %token EBGP IBGP
 %token LOCALAS REMOTEAS DESCR LOCALADDR MULTIHOP PASSIVE MAXPREFIX RESTART
@@ -252,6 +256,7 @@ grammar             : /* empty */
                | grammar prefixset '\n'
                | grammar roa_set '\n'
                | grammar origin_set '\n'
+               | grammar rtr '\n'
                | grammar rib '\n'
                | grammar conf_main '\n'
                | grammar l3vpn '\n'
@@ -551,6 +556,59 @@ roa_set_l  : prefixset_item SOURCEAS as4number_any                 {
                }
                ;
 
+rtr            : RTR address   {
+                       currtr = get_rtr(&$2);
+                       currtr->remote_port = 323;
+                       if (insert_rtr(currtr) == -1) {
+                               free(currtr);
+                               YYERROR;
+                       }
+                       currtr = NULL;
+               }
+               | RTR address   {
+                       currtr = get_rtr(&$2);
+                       currtr->remote_port = 323;
+               } '{' optnl rtropt_l optnl '}' {
+                       if (insert_rtr(currtr) == -1) {
+                               free(currtr);
+                               YYERROR;
+                       }
+                       currtr = NULL;
+               }
+               ;
+
+rtropt_l       : rtropt
+               | rtropt_l optnl rtropt
+
+rtropt         : DESCR STRING          {
+                       if (strlcpy(currtr->descr, $2,
+                           sizeof(currtr->descr)) >=
+                           sizeof(currtr->descr)) {
+                               yyerror("descr \"%s\" too long: max %zu",
+                                   $2, sizeof(currtr->descr) - 1);
+                               free($2);
+                               YYERROR;
+                       }
+                       free($2);
+               }
+               | LOCALADDR address     {
+                       if ($2.aid != currtr->remote_addr.aid) {
+                               yyerror("Bad address family %s for "
+                                   "local-addr", aid2str($2.aid));
+                               YYERROR;
+                       }
+                       currtr->local_addr = $2;
+               }
+               | PORT NUMBER {
+                       if ($2 < 1 || $2 > USHRT_MAX) {
+                               yyerror("local-port must be between %u and %u",
+                                   1, USHRT_MAX);
+                               YYERROR;
+                       }
+                       currtr->remote_port = $2;
+               }
+               ;
+
 conf_main      : AS as4number          {
                        conf->as = $2;
                        if ($2 > USHRT_MAX)
@@ -2866,6 +2924,7 @@ lookup(char *s)
                { "password",           PASSWORD},
                { "peer-as",            PEERAS},
                { "pftable",            PFTABLE},
+               { "port",               PORT},
                { "prefix",             PREFIX},
                { "prefix-set",         PREFIXSET},
                { "prefixlen",          PREFIXLEN},
@@ -2887,6 +2946,7 @@ lookup(char *s)
                { "router-id",          ROUTERID},
                { "rtable",             RTABLE},
                { "rtlabel",            RTLABEL},
+               { "rtr",                RTR},
                { "self",               SELF},
                { "set",                SET},
                { "socket",             SOCKET },
@@ -3285,7 +3345,7 @@ init_config(struct bgpd_config *c)
 }
 
 struct bgpd_config *
-parse_config(char *filename, struct peer_head *ph)
+parse_config(char *filename, struct peer_head *ph, struct rtr_config_head *rh)
 {
        struct sym              *sym, *next;
        struct rde_rib          *rr;
@@ -3309,6 +3369,7 @@ parse_config(char *filename, struct peer_head *ph)
        curgroup = NULL;
 
        cur_peers = ph;
+       cur_rtrs = rh;
        new_peers = &conf->peers;
        netconf = &conf->networks;
 
@@ -4568,3 +4629,57 @@ add_roa_set(struct prefixset_item *npsi, u_int32_t as, u_int8_t max)
                /* just ignore duplicates */
                free(roa);
 }
+
+static struct rtr_config *
+get_rtr(struct bgpd_addr *addr)
+{
+       struct rtr_config *n;
+
+       n = calloc(1, sizeof(*n));
+       if (n == NULL) {
+               yyerror("out of memory");
+               return NULL;
+       }
+
+       n->remote_addr = *addr;
+       strlcpy(n->descr, log_addr(addr), sizeof(currtr->descr));
+
+       return n;
+}
+
+static int
+insert_rtr(struct rtr_config *new)
+{
+       static uint32_t id;
+       struct rtr_config *r;
+
+       if (id == UINT32_MAX) {
+               yyerror("out of rtr session IDs");
+               return -1;
+       }
+
+       SIMPLEQ_FOREACH(r, &conf->rtrs, entry)
+               if (memcmp(&r->remote_addr, &new->remote_addr,
+                   sizeof(r->remote_addr)) == 0 &&
+                   r->remote_port == new->remote_port) {
+                       yyerror("duplicate rtr session to %s:%u",
+                           log_addr(&new->remote_addr), new->remote_port);
+                       return -1;
+               }
+
+       if (cur_rtrs)
+               SIMPLEQ_FOREACH(r, cur_rtrs, entry)
+                       if (memcmp(&r->remote_addr, &new->remote_addr,
+                           sizeof(r->remote_addr)) == 0 &&
+                           r->remote_port == new->remote_port) {
+                               new->id = r->id;
+                               break;
+                       }
+
+       if (new->id == 0)
+               new->id = ++id;
+
+       SIMPLEQ_INSERT_TAIL(&conf->rtrs, currtr, entry);
+
+       return 0;
+}
index 58cff60..122ae26 100644 (file)
@@ -1,4 +1,4 @@
-/*     $OpenBSD: printconf.c,v 1.145 2021/01/25 09:15:23 claudio Exp $ */
+/*     $OpenBSD: printconf.c,v 1.146 2021/02/16 08:29:16 claudio Exp $ */
 
 /*
  * Copyright (c) 2003, 2004 Henning Brauer <henning@openbsd.org>
@@ -42,6 +42,7 @@ void           print_as_sets(struct as_set_head *);
 void            print_prefixsets(struct prefixset_head *);
 void            print_originsets(struct prefixset_head *);
 void            print_roa(struct roa_tree *);
+void            print_rtrs(struct rtr_config_head *);
 void            print_peer(struct peer_config *, struct bgpd_config *,
                    const char *);
 const char     *print_auth_alg(u_int8_t);
@@ -578,6 +579,21 @@ print_roa(struct roa_tree *r)
        printf("\n}\n\n");
 }
 
+void
+print_rtrs(struct rtr_config_head *rh)
+{
+       struct rtr_config *r;
+
+       SIMPLEQ_FOREACH(r, rh, entry) {
+               printf("rtr %s {\n", log_addr(&r->remote_addr));
+               printf("\tdescr \"%s\"\n", r->descr);
+               printf("\tport %u\n", r->remote_port);
+               if (r->local_addr.aid != AID_UNSPEC)
+                       printf("local-addr %s\n", log_addr(&r->local_addr));
+               printf("}\n\n");
+       }
+}
+
 void
 print_peer(struct peer_config *p, struct bgpd_config *conf, const char *c)
 {
@@ -1010,6 +1026,7 @@ print_config(struct bgpd_config *conf, struct rib_names *rib_l)
        struct l3vpn            *vpn;
 
        print_mainconf(conf);
+       print_rtrs(&conf->rtrs);
        print_roa(&conf->roa);
        print_as_sets(&conf->as_sets);
        print_prefixsets(&conf->prefixsets);
index f720b84..ca6a515 100644 (file)
@@ -1,4 +1,4 @@
-/*     $OpenBSD: rde.c,v 1.514 2021/01/25 09:15:24 claudio Exp $ */
+/*     $OpenBSD: rde.c,v 1.515 2021/02/16 08:29:16 claudio Exp $ */
 
 /*
  * Copyright (c) 2003, 2004 Henning Brauer <henning@openbsd.org>
@@ -32,7 +32,6 @@
 #include <string.h>
 #include <syslog.h>
 #include <unistd.h>
-#include <err.h>
 
 #include "bgpd.h"
 #include "rde.h"
 #define PFD_PIPE_MAIN          0
 #define PFD_PIPE_SESSION       1
 #define PFD_PIPE_SESSION_CTL   2
-#define PFD_PIPE_COUNT         3
+#define PFD_PIPE_ROA           3
+#define PFD_PIPE_COUNT         4
 
 void            rde_sighdlr(int);
 void            rde_dispatch_imsg_session(struct imsgbuf *);
 void            rde_dispatch_imsg_parent(struct imsgbuf *);
+void            rde_dispatch_imsg_rtr(struct imsgbuf *);
 void            rde_dispatch_imsg_peer(struct rde_peer *, void *);
 void            rde_update_dispatch(struct rde_peer *, struct imsg *);
 int             rde_update_update(struct rde_peer *, struct filterstate *,
@@ -79,6 +80,7 @@ static void    rde_softreconfig_in(struct rib_entry *, void *);
 static void     rde_softreconfig_sync_reeval(struct rib_entry *, void *);
 static void     rde_softreconfig_sync_fib(struct rib_entry *, void *);
 static void     rde_softreconfig_sync_done(void *, u_int8_t);
+static void     rde_roa_reload(void);
 static int      rde_no_as_set(struct rde_peer *);
 int             rde_update_queue_pending(void);
 void            rde_update_queue_runner(void);
@@ -102,8 +104,10 @@ int                 ovs_match(struct prefix *, u_int32_t);
 
 static struct imsgbuf          *ibuf_se;
 static struct imsgbuf          *ibuf_se_ctl;
+static struct imsgbuf          *ibuf_rtr;
 static struct imsgbuf          *ibuf_main;
 static struct bgpd_config      *conf, *nconf;
+static struct rde_prefixset     rde_roa, roa_new;
 
 volatile sig_atomic_t   rde_quit = 0;
 struct filter_head     *out_rules, *out_rules_tmp;
@@ -231,6 +235,7 @@ rde_main(int debug, int verbose)
                set_pollfd(&pfd[PFD_PIPE_MAIN], ibuf_main);
                set_pollfd(&pfd[PFD_PIPE_SESSION], ibuf_se);
                set_pollfd(&pfd[PFD_PIPE_SESSION_CTL], ibuf_se_ctl);
+               set_pollfd(&pfd[PFD_PIPE_ROA], ibuf_rtr);
 
                i = PFD_PIPE_COUNT;
                for (mctx = LIST_FIRST(&rde_mrts); mctx != 0; mctx = xmctx) {
@@ -282,6 +287,14 @@ rde_main(int debug, int verbose)
                } else
                        rde_dispatch_imsg_session(ibuf_se_ctl);
 
+               if (handle_pollfd(&pfd[PFD_PIPE_ROA], ibuf_rtr) == -1) {
+                       log_warnx("RDE: Lost connection to ROA");
+                       msgbuf_clear(&ibuf_rtr->w);
+                       free(ibuf_rtr);
+                       ibuf_rtr = NULL;
+               } else
+                       rde_dispatch_imsg_rtr(ibuf_rtr);
+
                for (j = PFD_PIPE_COUNT, mctx = LIST_FIRST(&rde_mrts);
                    j < i && mctx != 0; j++) {
                        if (pfd[j].fd == mctx->mrt.wbuf.fd &&
@@ -578,7 +591,7 @@ badnetdel:
                        break;
                case IMSG_CTL_SHOW_SET:
                        /* first roa set */
-                       pset = &conf->rde_roa;
+                       pset = &rde_roa;
                        memset(&cset, 0, sizeof(cset));
                        cset.type = ROA_SET;
                        strlcpy(cset.name, "RPKI ROA", sizeof(cset.name));
@@ -668,7 +681,6 @@ rde_dispatch_imsg_parent(struct imsgbuf *ibuf)
        static struct l3vpn     *vpn;
        struct imsg              imsg;
        struct mrt               xmrt;
-       struct roa               roa;
        struct rde_rib           rr;
        struct filterstate       state;
        struct imsgbuf          *i;
@@ -693,15 +705,17 @@ rde_dispatch_imsg_parent(struct imsgbuf *ibuf)
                switch (imsg.hdr.type) {
                case IMSG_SOCKET_CONN:
                case IMSG_SOCKET_CONN_CTL:
+               case IMSG_SOCKET_CONN_RTR:
                        if ((fd = imsg.fd) == -1) {
-                               log_warnx("expected to receive imsg fd to "
-                                   "SE but didn't receive any");
+                               log_warnx("expected to receive imsg fd "
+                                   "but didn't receive any");
                                break;
                        }
                        if ((i = malloc(sizeof(struct imsgbuf))) == NULL)
                                fatal(NULL);
                        imsg_init(i, fd);
-                       if (imsg.hdr.type == IMSG_SOCKET_CONN) {
+                       switch (imsg.hdr.type) {
+                       case IMSG_SOCKET_CONN:
                                if (ibuf_se) {
                                        log_warnx("Unexpected imsg connection "
                                            "to SE received");
@@ -709,7 +723,8 @@ rde_dispatch_imsg_parent(struct imsgbuf *ibuf)
                                        free(ibuf_se);
                                }
                                ibuf_se = i;
-                       } else {
+                               break;
+                       case IMSG_SOCKET_CONN_CTL:
                                if (ibuf_se_ctl) {
                                        log_warnx("Unexpected imsg ctl "
                                            "connection to SE received");
@@ -717,6 +732,16 @@ rde_dispatch_imsg_parent(struct imsgbuf *ibuf)
                                        free(ibuf_se_ctl);
                                }
                                ibuf_se_ctl = i;
+                               break;
+                       case IMSG_SOCKET_CONN_RTR:
+                               if (ibuf_rtr) {
+                                       log_warnx("Unexpected imsg ctl "
+                                           "connection to ROA received");
+                                       msgbuf_clear(&ibuf_rtr->w);
+                                       free(ibuf_rtr);
+                               }
+                               ibuf_rtr = i;
+                               break;
                        }
                        break;
                case IMSG_NETWORK_ADD:
@@ -865,17 +890,6 @@ rde_dispatch_imsg_parent(struct imsgbuf *ibuf)
                        }
                        last_prefixset = ps;
                        break;
-               case IMSG_RECONF_ROA_SET:
-                       strlcpy(nconf->rde_roa.name, "RPKI ROA",
-                           sizeof(nconf->rde_roa.name));
-                       last_prefixset = &nconf->rde_roa;
-                       break;
-               case IMSG_RECONF_ROA_ITEM:
-                       if (imsg.hdr.len - IMSG_HEADER_SIZE != sizeof(roa))
-                               fatalx("IMSG_RECONF_ROA_ITEM bad len");
-                       memcpy(&roa, imsg.data, sizeof(roa));
-                       rv = trie_roa_add(&last_prefixset->th, &roa);
-                       break;
                case IMSG_RECONF_PREFIX_SET_ITEM:
                        if (imsg.hdr.len - IMSG_HEADER_SIZE != sizeof(psi))
                                fatalx("IMSG_RECONF_PREFIX_SET_ITEM bad len");
@@ -886,7 +900,7 @@ rde_dispatch_imsg_parent(struct imsgbuf *ibuf)
                            &psi.p.addr, psi.p.len,
                            psi.p.len_min, psi.p.len_max);
                        if (rv == -1)
-                               log_warnx("trie_add(%s) %s/%u) failed",
+                               log_warnx("trie_add(%s) %s/%u failed",
                                    last_prefixset->name, log_addr(&psi.p.addr),
                                    psi.p.len);
                        break;
@@ -991,6 +1005,46 @@ rde_dispatch_imsg_parent(struct imsgbuf *ibuf)
                        /* ignore end message because a dump is atomic */
                        break;
                default:
+                       fatalx("unhandled IMSG %u", imsg.hdr.type);
+               }
+               imsg_free(&imsg);
+       }
+}
+
+void
+rde_dispatch_imsg_rtr(struct imsgbuf *ibuf)
+{
+       struct imsg      imsg;
+       struct roa       roa;
+       int              n;
+
+       while (ibuf) {
+               if ((n = imsg_get(ibuf, &imsg)) == -1)
+                       fatal("rde_dispatch_imsg_parent: imsg_get error");
+               if (n == 0)
+                       break;
+
+               switch (imsg.hdr.type) {
+               case IMSG_RECONF_ROA_SET:
+                       /* start of update */
+                       break;
+               case IMSG_RECONF_ROA_ITEM:
+                       if (imsg.hdr.len - IMSG_HEADER_SIZE !=
+                           sizeof(roa))
+                               fatalx("IMSG_RECONF_ROA_ITEM bad len");
+                       memcpy(&roa, imsg.data, sizeof(roa));
+                       if (trie_roa_add(&roa_new.th, &roa) != 0) {
+                               struct bgpd_addr p = {
+                                       .aid = roa.aid,
+                                       .v6 = roa.prefix.inet6
+                               };
+                               log_warnx("trie_roa_add %s/%u failed",
+                                   log_addr(&p), roa.prefixlen);
+                       }
+                       break;
+               case IMSG_RECONF_DONE:
+                       /* end of update */
+                       rde_roa_reload();
                        break;
                }
                imsg_free(&imsg);
@@ -1439,7 +1493,7 @@ rde_update_update(struct rde_peer *peer, struct filterstate *in,
        const char              *wmsg = "filtered, withdraw";
 
        peer->prefix_rcvd_update++;
-       vstate = rde_roa_validity(&conf->rde_roa, prefix, prefixlen,
+       vstate = rde_roa_validity(&rde_roa, prefix, prefixlen,
            aspath_origin(in->aspath.aspath));
 
        /* add original path to the Adj-RIB-In */
@@ -3153,7 +3207,6 @@ rde_reload_done(void)
        struct filter_head      *fh;
        struct rde_prefixset_head prefixsets_old;
        struct rde_prefixset_head originsets_old;
-       struct rde_prefixset     roa_old;
        struct as_set_head       as_sets_old;
        u_int16_t                rid;
        int                      reload = 0;
@@ -3166,7 +3219,6 @@ rde_reload_done(void)
        SIMPLEQ_CONCAT(&prefixsets_old, &conf->rde_prefixsets);
        SIMPLEQ_CONCAT(&originsets_old, &conf->rde_originsets);
        SIMPLEQ_CONCAT(&as_sets_old, &conf->as_sets);
-       roa_old = conf->rde_roa;
 
        /* merge the main config */
        copy_config(conf, nconf);
@@ -3176,10 +3228,6 @@ rde_reload_done(void)
        SIMPLEQ_CONCAT(&conf->rde_originsets, &nconf->rde_originsets);
        SIMPLEQ_CONCAT(&conf->as_sets, &nconf->as_sets);
 
-       conf->rde_roa = nconf->rde_roa;
-       conf->rde_roa.lastchange = roa_old.lastchange;
-       memset(&nconf->rde_roa, 0, sizeof(nconf->rde_roa));
-
        /* apply new set of l3vpn, sync will be done later */
        free_l3vpns(&conf->l3vpns);
        SIMPLEQ_CONCAT(&conf->l3vpns, &nconf->l3vpns);
@@ -3197,16 +3245,6 @@ rde_reload_done(void)
        peerself->conf.remote_masklen = 32;
        peerself->short_as = conf->short_as;
 
-       /* check if roa changed */
-       if (trie_equal(&conf->rde_roa.th, &roa_old.th) == 0) {
-               log_debug("roa change: reloading Adj-RIB-In");
-               conf->rde_roa.dirty = 1;
-               conf->rde_roa.lastchange = getmonotime();
-               reload++;       /* run softreconf in */
-       }
-
-       trie_free(&roa_old.th); /* old roa no longer needed */
-
        rde_mark_prefixsets_dirty(&prefixsets_old, &conf->rde_prefixsets);
        rde_mark_prefixsets_dirty(&originsets_old, &conf->rde_originsets);
        as_sets_mark_dirty(&as_sets_old, &conf->as_sets);
@@ -3443,8 +3481,6 @@ rde_softreconfig_in(struct rib_entry *re, void *bula)
        struct rde_aspath       *asp;
        enum filter_actions      action;
        struct bgpd_addr         prefix;
-       int                      force_eval;
-       u_int8_t                 vstate;
        u_int16_t                i;
 
        pt = re->prefix;
@@ -3452,17 +3488,6 @@ rde_softreconfig_in(struct rib_entry *re, void *bula)
        LIST_FOREACH(p, &re->prefix_h, entry.list.rib) {
                asp = prefix_aspath(p);
                peer = prefix_peer(p);
-               force_eval = 0;
-
-               if (conf->rde_roa.dirty) {
-                       /* ROA validation state update */
-                       vstate = rde_roa_validity(&conf->rde_roa,
-                           &prefix, pt->prefixlen, aspath_origin(asp->aspath));
-                       if (vstate != p->validation_state) {
-                               force_eval = 1;
-                               p->validation_state = vstate;
-                       }
-               }
 
                /* skip announced networks, they are never filtered */
                if (asp->flags & F_PREFIX_ANNOUNCED)
@@ -3473,7 +3498,7 @@ rde_softreconfig_in(struct rib_entry *re, void *bula)
                        if (rib == NULL)
                                continue;
 
-                       if (rib->state != RECONF_RELOAD && !force_eval)
+                       if (rib->state != RECONF_RELOAD)
                                continue;
 
                        rde_filterstate_prep(&state, asp, prefix_communities(p),
@@ -3574,6 +3599,99 @@ rde_softreconfig_sync_done(void *arg, u_int8_t aid)
                rde_softreconfig_done();
 }
 
+/*
+ * ROA specific functions. The roa set is updated independent of the config
+ * so this runs outside of the softreconfig handlers.
+ */
+static void
+rde_roa_softreload(struct rib_entry *re, void *bula)
+{
+       struct filterstate       state;
+       struct rib              *rib;
+       struct prefix           *p;
+       struct pt_entry         *pt;
+       struct rde_peer         *peer;
+       struct rde_aspath       *asp;
+       enum filter_actions      action;
+       struct bgpd_addr         prefix;
+       u_int8_t                 vstate;
+       u_int16_t                i;
+
+       pt = re->prefix;
+       pt_getaddr(pt, &prefix);
+       LIST_FOREACH(p, &re->prefix_h, entry.list.rib) {
+               asp = prefix_aspath(p);
+               peer = prefix_peer(p);
+
+               /* ROA validation state update */
+               vstate = rde_roa_validity(&rde_roa,
+                   &prefix, pt->prefixlen, aspath_origin(asp->aspath));
+               if (vstate == p->validation_state)
+                       continue;
+               p->validation_state = vstate;
+
+               /* skip announced networks, they are never filtered */
+               if (asp->flags & F_PREFIX_ANNOUNCED)
+                       continue;
+
+               for (i = RIB_LOC_START; i < rib_size; i++) {
+                       rib = rib_byid(i);
+                       if (rib == NULL)
+                               continue;
+
+                       rde_filterstate_prep(&state, asp, prefix_communities(p),
+                           prefix_nexthop(p), prefix_nhflags(p));
+                       action = rde_filter(rib->in_rules, peer, peer, &prefix,
+                           pt->prefixlen, p->validation_state, &state);
+
+                       if (action == ACTION_ALLOW) {
+                               /* update Local-RIB */
+                               prefix_update(rib, peer, &state, &prefix,
+                                   pt->prefixlen, p->validation_state);
+                       } else if (action == ACTION_DENY) {
+                               /* remove from Local-RIB */
+                               prefix_withdraw(rib, peer, &prefix,
+                                   pt->prefixlen);
+                       }
+
+                       rde_filterstate_clean(&state);
+               }
+       }
+}
+
+static void
+rde_roa_softreload_done(void *arg, u_int8_t aid)
+{
+       /* the roa update is done */
+       log_info("ROA softreload done");
+}
+
+static void
+rde_roa_reload(void)
+{
+       struct rde_prefixset roa_old;
+
+       roa_old = rde_roa;
+       rde_roa = roa_new;
+       memset(&roa_new, 0, sizeof(roa_new));
+
+       /* check if roa changed */
+       if (trie_equal(&rde_roa.th, &roa_old.th)) {
+               rde_roa.lastchange = roa_old.lastchange;
+               trie_free(&roa_old.th); /* old roa no longer needed */
+               return;
+       }
+
+       rde_roa.lastchange = getmonotime();
+       trie_free(&roa_old.th); /* old roa no longer needed */
+
+       log_debug("ROA change: reloading Adj-RIB-In");
+       if (rib_dump_new(RIB_ADJ_IN, AID_UNSPEC, RDE_RUNNER_ROUNDS,
+           rib_byid(RIB_ADJ_IN), rde_roa_softreload,
+           rde_roa_softreload_done, NULL) == -1)
+               fatal("%s: rib_dump_new", __func__);
+}
+
 /*
  * generic helper function
  */
@@ -3737,7 +3855,7 @@ network_add(struct network_config *nc, struct filterstate *state)
                rde_apply_set(vpnset, peerself, peerself, state,
                    nc->prefix.aid);
 
-       vstate = rde_roa_validity(&conf->rde_roa, &nc->prefix,
+       vstate = rde_roa_validity(&rde_roa, &nc->prefix,
            nc->prefixlen, aspath_origin(state->aspath.aspath));
        if (prefix_update(rib_byid(RIB_ADJ_IN), peerself, state, &nc->prefix,
            nc->prefixlen, vstate) == 1)
diff --git a/usr.sbin/bgpd/rtr.c b/usr.sbin/bgpd/rtr.c
new file mode 100644 (file)
index 0000000..dc4d9ca
--- /dev/null
@@ -0,0 +1,334 @@
+/*     $OpenBSD: rtr.c,v 1.1 2021/02/16 08:29:16 claudio Exp $ */
+
+/*
+ * Copyright (c) 2020 Claudio Jeker <claudio@openbsd.org>
+ *
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose with or without fee is hereby granted, provided that the above
+ * copyright notice and this permission notice appear in all copies.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+ * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
+ * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+ * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+ */
+#include <sys/tree.h>
+#include <errno.h>
+#include <poll.h>
+#include <pwd.h>
+#include <signal.h>
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <syslog.h>
+#include <unistd.h>
+
+#include "bgpd.h"
+#include "session.h"
+#include "log.h"
+
+static void    rtr_dispatch_imsg_parent(struct imsgbuf *);
+static void    rtr_dispatch_imsg_rde(struct imsgbuf *);
+
+volatile sig_atomic_t           rtr_quit;
+static struct imsgbuf          *ibuf_main;
+static struct imsgbuf          *ibuf_rde;
+static struct bgpd_config      *conf, *nconf;
+
+static void
+rtr_sighdlr(int sig)
+{
+       switch (sig) {
+       case SIGINT:
+       case SIGTERM:
+               rtr_quit = 1;
+               break;
+       }
+}
+
+#define PFD_PIPE_MAIN  0
+#define PFD_PIPE_RDE   1
+#define PFD_PIPE_COUNT 2
+
+void
+roa_insert(struct roa_tree *rt, struct roa *in)
+{
+       struct roa *roa;
+
+       if ((roa = malloc(sizeof(*roa))) == NULL)
+               fatal("roa alloc");
+       memcpy(roa, in, sizeof(*roa));
+       if (RB_INSERT(roa_tree, rt, roa) != NULL)
+               /* just ignore duplicates */
+               free(roa);
+}
+
+void
+rtr_main(int debug, int verbose)
+{
+       struct passwd           *pw;
+       struct pollfd           *pfd = NULL;
+       void                    *newp;
+       size_t                   pfd_elms = 0, i;
+       time_t                   now, timeout;
+
+       log_init(debug, LOG_DAEMON);
+       log_setverbose(verbose);
+
+       log_procinit(log_procnames[PROC_RTR]);
+
+       if ((pw = getpwnam(BGPD_USER)) == NULL)
+               fatal("getpwnam");
+
+       if (chroot(pw->pw_dir) == -1)
+               fatal("chroot");
+       if (chdir("/") == -1)
+               fatal("chdir(\"/\")");
+
+       setproctitle("rtr engine");
+
+       if (setgroups(1, &pw->pw_gid) ||
+           setresgid(pw->pw_gid, pw->pw_gid, pw->pw_gid) ||
+           setresuid(pw->pw_uid, pw->pw_uid, pw->pw_uid))
+               fatal("can't drop privileges");
+
+       if (pledge("stdio recvfd", NULL) == -1)
+               fatal("pledge");
+
+       signal(SIGTERM, rtr_sighdlr);
+       signal(SIGINT, rtr_sighdlr);
+       signal(SIGPIPE, SIG_IGN);
+       signal(SIGHUP, SIG_IGN);
+       signal(SIGALRM, SIG_IGN);
+       signal(SIGUSR1, SIG_IGN);
+
+       if ((ibuf_main = malloc(sizeof(struct imsgbuf))) == NULL)
+               fatal(NULL);
+       imsg_init(ibuf_main, 3);
+
+       conf = new_config();
+       log_info("rtr engine ready");
+
+       while (rtr_quit == 0) {
+               i = rtr_count();
+               if (pfd_elms < PFD_PIPE_COUNT + i) {
+                       if ((newp = reallocarray(pfd,
+                           PFD_PIPE_COUNT + i,
+                           sizeof(struct pollfd))) == NULL)
+                               fatal("realloc pollfd");
+                       pfd = newp;
+                       pfd_elms = PFD_PIPE_COUNT + i;
+               }
+               timeout = 240;  /* loop every 240s at least */
+               bzero(pfd, sizeof(struct pollfd) * pfd_elms);
+
+               set_pollfd(&pfd[PFD_PIPE_MAIN], ibuf_main);
+               set_pollfd(&pfd[PFD_PIPE_RDE], ibuf_rde);
+
+               now = getmonotime();
+               i = PFD_PIPE_COUNT;
+               i += rtr_poll_events(pfd + i, pfd_elms - i, &timeout);
+
+               if (poll(pfd, i, timeout * 1000) == -1) {
+                       if (errno != EINTR)
+                               fatal("poll error");
+                       continue;
+               }
+
+               if (handle_pollfd(&pfd[PFD_PIPE_MAIN], ibuf_main) == -1)
+                       fatalx("Lost connection to parent");
+               else
+                       rtr_dispatch_imsg_parent(ibuf_main);
+
+               if (handle_pollfd(&pfd[PFD_PIPE_RDE], ibuf_rde) == -1) {
+                       log_warnx("RTR: Lost connection to RDE");
+                       msgbuf_clear(&ibuf_rde->w);
+                       free(ibuf_rde);
+                       ibuf_rde = NULL;
+               } else
+                       rtr_dispatch_imsg_rde(ibuf_rde);
+
+               now = getmonotime();
+               i = PFD_PIPE_COUNT;
+               rtr_check_events(pfd + i, pfd_elms - i);
+       }
+
+       rtr_shutdown();
+
+       free_config(conf);
+       free(pfd);
+
+       /* close pipes */
+       if (ibuf_rde) {
+               msgbuf_clear(&ibuf_rde->w);
+               close(ibuf_rde->fd);
+               free(ibuf_rde);
+       }
+       msgbuf_clear(&ibuf_main->w);
+       close(ibuf_main->fd);
+       free(ibuf_main);
+
+       log_info("rtr engine exiting");
+       exit(0);
+}
+
+static void
+rtr_dispatch_imsg_parent(struct imsgbuf *ibuf)
+{
+       struct imsg     imsg;
+       struct roa      *roa;
+       struct rtr_session *rs;
+       int             n, fd;
+
+       while (ibuf) {
+               if ((n = imsg_get(ibuf, &imsg)) == -1)
+                       fatal("%s: imsg_get error", __func__);
+               if (n == 0)
+                       break;
+
+               switch (imsg.hdr.type) {
+               case IMSG_SOCKET_CONN_RTR:
+                       if ((fd = imsg.fd) == -1) {
+                               log_warnx("expected to receive imsg fd "
+                                   "but didn't receive any");
+                               break;
+                       }
+                       if (ibuf_rde) {
+                               log_warnx("Unexpected imsg ctl "
+                                   "connection to RDE received");
+                               msgbuf_clear(&ibuf_rde->w);
+                               free(ibuf_rde);
+                       }
+                       if ((ibuf_rde = malloc(sizeof(struct imsgbuf))) == NULL)
+                               fatal(NULL);
+                       imsg_init(ibuf_rde, fd);
+                       break;
+               case IMSG_SOCKET_CONN:
+                       if ((fd = imsg.fd) == -1) {
+                               log_warnx("expected to receive imsg fd "
+                                   "but didn't receive any");
+                               break;
+                       }
+                       if ((rs = rtr_get(imsg.hdr.peerid)) == NULL) {
+                               log_warnx("IMSG_SOCKET_CONN: unknown rtr id %d",
+                                   imsg.hdr.peerid);
+                               break;
+                       }
+                       rtr_open(rs, fd);
+                       break;
+               case IMSG_RECONF_CONF:
+                       if (imsg.hdr.len - IMSG_HEADER_SIZE !=
+                           sizeof(struct bgpd_config))
+                               fatalx("IMSG_RECONF_CONF bad len");
+                       nconf = new_config();
+                       copy_config(nconf, imsg.data);
+                       rtr_config_prep();
+                       break;
+               case IMSG_RECONF_ROA_ITEM:
+                       if (imsg.hdr.len - IMSG_HEADER_SIZE !=
+                           sizeof(*roa))
+                               fatalx("IMSG_RECONF_ROA_ITEM bad len");
+                       roa_insert(&nconf->roa, imsg.data);
+                       break;
+               case IMSG_RECONF_RTR_CONFIG:
+                       if (imsg.hdr.len - IMSG_HEADER_SIZE != PEER_DESCR_LEN)
+                               fatalx("IMSG_RECONF_RTR_CONFIG bad len");
+                       rs = rtr_get(imsg.hdr.peerid);
+                       if (rs == NULL)
+                               rtr_new(imsg.hdr.peerid, imsg.data);
+                       else
+                               rtr_config_keep(rs);
+                       break;
+               case IMSG_RECONF_DRAIN:
+                       imsg_compose(ibuf_main, IMSG_RECONF_DRAIN, 0, 0,
+                           -1, NULL, 0);
+                       break;
+               case IMSG_RECONF_DONE:
+                       if (nconf == NULL)
+                               fatalx("got IMSG_RECONF_DONE but no config");
+                       copy_config(conf, nconf);
+                       /* switch the roa, first remove the old one */
+                       free_roatree(&conf->roa);
+                       /* then move the RB tree root */
+                       RB_ROOT(&conf->roa) = RB_ROOT(&nconf->roa);
+                       RB_ROOT(&nconf->roa) = NULL;
+                       /* finally merge the rtr session */
+                       rtr_config_merge();
+                       rtr_recalc();
+                       log_info("RTR engine reconfigured");
+                       imsg_compose(ibuf_main, IMSG_RECONF_DONE, 0, 0,
+                           -1, NULL, 0);
+                       free_config(nconf);
+                       nconf = NULL;
+                       break;
+               case IMSG_CTL_SHOW_RTR:
+                       if ((rs = rtr_get(imsg.hdr.peerid)) == NULL) {
+                               log_warnx("IMSG_CTL_SHOW_RTR: "
+                                   "unknown rtr id %d", imsg.hdr.peerid);
+                               break;
+                       }
+                       rtr_show(rs, imsg.hdr.pid);
+                       break;
+               case IMSG_CTL_END:
+                       imsg_compose(ibuf_main, IMSG_CTL_END, 0, imsg.hdr.pid,
+                           -1, NULL, 0);
+                       break;
+               }
+               imsg_free(&imsg);
+       }
+}
+
+static void
+rtr_dispatch_imsg_rde(struct imsgbuf *ibuf)
+{
+       struct imsg     imsg;
+       int             n;
+
+       while (ibuf) {
+               if ((n = imsg_get(ibuf, &imsg)) == -1)
+                       fatal("%s: imsg_get error", __func__);
+               if (n == 0)
+                       break;
+
+               /* NOTHING */
+
+               imsg_free(&imsg);
+       }
+}
+
+void
+rtr_imsg_compose(int type, uint32_t id, pid_t pid, void *data, size_t datalen)
+{
+       imsg_compose(ibuf_main, type, id, pid, -1, data, datalen);
+}
+
+/*
+ * Merge all RPKI ROA trees into one as one big union.
+ * Simply try to add all roa entries into a new RB tree.
+ * This could be made a fair bit faster but for now this is good enough.
+ */
+void
+rtr_recalc(void)
+{
+       struct roa_tree rt;
+       struct roa *roa, *nr;
+
+       RB_INIT(&rt);
+
+       RB_FOREACH(roa, roa_tree, &conf->roa)
+               roa_insert(&rt, roa);
+       rtr_roa_merge(&rt);
+
+       imsg_compose(ibuf_rde, IMSG_RECONF_ROA_SET, 0, 0, -1, NULL, 0);
+       RB_FOREACH_SAFE(roa, roa_tree, &rt, nr) {
+               RB_REMOVE(roa_tree, &rt, roa);
+               imsg_compose(ibuf_rde, IMSG_RECONF_ROA_ITEM, 0, 0, -1,
+                   roa, sizeof(*roa));
+               free(roa);
+       }
+       imsg_compose(ibuf_rde, IMSG_RECONF_DONE, 0, 0, -1, NULL, 0);
+}
diff --git a/usr.sbin/bgpd/rtr_proto.c b/usr.sbin/bgpd/rtr_proto.c
new file mode 100644 (file)
index 0000000..4354a58
--- /dev/null
@@ -0,0 +1,1157 @@
+/*     $OpenBSD: rtr_proto.c,v 1.1 2021/02/16 08:29:16 claudio Exp $ */
+
+/*
+ * Copyright (c) 2020 Claudio Jeker <claudio@openbsd.org>
+ *
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose with or without fee is hereby granted, provided that the above
+ * copyright notice and this permission notice appear in all copies.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+ * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
+ * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+ * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+ */
+#include <sys/tree.h>
+#include <errno.h>
+#include <stdint.h>
+#include <poll.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <unistd.h>
+
+#include "bgpd.h"
+#include "session.h"
+#include "log.h"
+
+struct rtr_header {
+       uint8_t         version;
+       uint8_t         type;
+       uint16_t        session_id; /* or error code */
+       uint32_t        length;
+};
+
+#define RTR_MAX_LEN            2048
+#define RTR_DEFAULT_REFRESH    3600
+#define RTR_DEFAULT_RETRY      600
+#define RTR_DEFAULT_EXPIRE     7200
+
+enum rtr_pdu_type {
+       SERIAL_NOTIFY = 0,
+       SERIAL_QUERY,
+       RESET_QUERY,
+       CACHE_RESPONSE,
+       IPV4_PREFIX,
+       IPV6_PREFIX = 6,
+       END_OF_DATA = 7,
+       CACHE_RESET = 8,
+       ROUTER_KEY = 9,
+       ERROR_REPORT = 10,
+};
+
+#define FLAG_ANNOUNCE  0x1
+#define FLAG_MASK      FLAG_ANNOUNCE
+struct rtr_ipv4 {
+       uint8_t         flags;
+       uint8_t         prefixlen;
+       uint8_t         maxlen;
+       uint8_t         zero;
+       uint32_t        prefix;
+       uint32_t        asnum;
+};
+
+struct rtr_ipv6 {
+       uint8_t         flags;
+       uint8_t         prefixlen;
+       uint8_t         maxlen;
+       uint8_t         zero;
+       uint32_t        prefix[4];
+       uint32_t        asnum;
+};
+
+struct rtr_endofdata {
+       uint32_t        serial;
+       uint32_t        refresh;
+       uint32_t        retry;
+       uint32_t        expire;
+};
+
+enum rtr_event {
+       RTR_EVNT_START,
+       RTR_EVNT_CON_OPEN,
+       RTR_EVNT_CON_CLOSED,
+       RTR_EVNT_TIMER_REFRESH,
+       RTR_EVNT_TIMER_RETRY,
+       RTR_EVNT_TIMER_EXPIRE,
+       RTR_EVNT_SEND_ERROR,
+       RTR_EVNT_SERIAL_NOTIFY,
+       RTR_EVNT_CACHE_RESPONSE,
+       RTR_EVNT_END_OF_DATA,
+       RTR_EVNT_CACHE_RESET,
+       RTR_EVNT_NO_DATA,
+};
+
+static const char *rtr_eventnames[] = {
+       "start",
+       "connection open",
+       "connection closed",
+       "refresh timer expired",
+       "retry timer expired",
+       "expire timer expired",
+       "sent error",
+       "serial notify received",
+       "cache response received",
+       "end of data received",
+       "cache reset received",
+       "no data"
+};
+
+enum rtr_state {
+       RTR_STATE_CLOSED,
+       RTR_STATE_ERROR,
+       RTR_STATE_IDLE,
+       RTR_STATE_ACTIVE,
+};
+
+static const char *rtr_statenames[] = {
+       "closed",
+       "error",
+       "idle",
+       "active"
+};
+
+struct rtr_session {
+       TAILQ_ENTRY(rtr_session)        entry;
+       char                            descr[PEER_DESCR_LEN];
+       struct roa_tree                 roa_set;
+       struct ibuf_read                r;
+       struct msgbuf                   w;
+       struct timer_head               timers;
+       uint32_t                        id;             /* rtr_config id */
+       uint32_t                        serial;
+       uint32_t                        refresh;
+       uint32_t                        retry;
+       uint32_t                        expire;
+       int                             session_id;
+       int                             fd;
+       enum rtr_state                  state;
+       enum reconf_action              reconf_action;
+       enum rtr_error                  last_sent_error;
+       enum rtr_error                  last_recv_error;
+       char                            last_sent_msg[REASON_LEN];
+       char                            last_recv_msg[REASON_LEN];
+};
+
+TAILQ_HEAD(, rtr_session) rtrs = TAILQ_HEAD_INITIALIZER(rtrs);
+
+static void    rtr_fsm(struct rtr_session *, enum rtr_event);
+
+static const char *
+log_rtr(struct rtr_session *rs)
+{
+       return rs->descr;
+}
+
+static const char *
+log_rtr_type(enum rtr_pdu_type type)
+{
+       static char buf[20];
+
+       switch (type) {
+       case SERIAL_NOTIFY:
+               return "serial notify";
+       case SERIAL_QUERY:
+               return "serial query";
+       case RESET_QUERY:
+               return "reset query";
+       case CACHE_RESPONSE:
+               return "cache response";
+       case IPV4_PREFIX:
+               return "IPv4 prefix";
+       case IPV6_PREFIX:
+               return "IPv6 prefix";
+       case END_OF_DATA:
+               return "end of data";
+       case CACHE_RESET:
+               return "cache reset";
+       case ROUTER_KEY:
+               return "router key";
+       case ERROR_REPORT:
+               return "error report";
+       default:
+               snprintf(buf, sizeof(buf), "unknown %u", type);
+               return buf;
+       }
+};
+
+static struct ibuf *
+rtr_newmsg(enum rtr_pdu_type type, uint32_t len, u_int16_t session_id)
+{
+       struct ibuf *buf;
+       struct rtr_header rh;
+
+       if (len > RTR_MAX_LEN) {
+               errno = ERANGE;
+               return NULL;
+       }
+       len += sizeof(rh);
+       if ((buf = ibuf_open(len)) == NULL)
+               return NULL;
+
+       memset(&rh, 0, sizeof(rh));
+       rh.version = 1;
+       rh.type = type;
+       rh.session_id = htons(session_id);
+       rh.length = htonl(len);
+
+       /* can not fail with fixed buffers */
+       ibuf_add(buf, &rh, sizeof(rh));
+       return buf;
+}
+
+/*
+ * Try to send an error PDU to cache, put connection into error
+ * state.
+ */
+static void
+rtr_send_error(struct rtr_session *rs, enum rtr_error err, char *msg,
+    void *pdu, size_t len)
+{
+       struct ibuf *buf;
+       size_t mlen = 0;
+       uint32_t hdrlen;
+
+       rs->last_sent_error = err;
+       if (msg) {
+               mlen = strlen(msg);
+               strlcpy(rs->last_sent_msg, msg, sizeof(rs->last_sent_msg));
+       } else
+               memset(rs->last_sent_msg, 0, sizeof(rs->last_sent_msg));
+
+       rtr_fsm(rs, RTR_EVNT_SEND_ERROR);
+
+       buf = rtr_newmsg(ERROR_REPORT, 2 * sizeof(hdrlen) + len + mlen, err);
+       if (buf == NULL) {
+               log_warn("rtr %s: send error report", log_rtr(rs));
+               return;
+       }
+
+       /* can not fail with fixed buffers */
+       hdrlen = ntohl(len);
+       ibuf_add(buf, &hdrlen, sizeof(hdrlen));
+       ibuf_add(buf, pdu, len);
+       hdrlen = ntohl(mlen);
+       ibuf_add(buf, &hdrlen, sizeof(hdrlen));
+       ibuf_add(buf, msg, mlen);
+       ibuf_close(&rs->w, buf);
+
+       log_warnx("%s: sending error report[%u] %s", log_rtr(rs), err,
+           msg ? msg : "");
+}
+
+static void
+rtr_reset_query(struct rtr_session *rs)
+{
+       struct ibuf *buf;
+
+       buf = rtr_newmsg(RESET_QUERY, 0, 0);
+       if (buf == NULL) {
+               log_warn("rtr %s: send reset query", log_rtr(rs));
+               rtr_send_error(rs, INTERNAL_ERROR, "out of memory", NULL, 0);
+               return;
+       }
+       ibuf_close(&rs->w, buf);
+}
+
+static void
+rtr_serial_query(struct rtr_session *rs)
+{
+       struct ibuf *buf;
+       uint32_t s;
+
+       buf = rtr_newmsg(SERIAL_QUERY, sizeof(s), rs->session_id);
+       if (buf == NULL) {
+               log_warn("rtr %s: send serial query", log_rtr(rs));
+               rtr_send_error(rs, INTERNAL_ERROR, "out of memory", NULL, 0);
+               return;
+       }
+
+       /* can not fail with fixed buffers */
+       s = htonl(rs->serial);
+       ibuf_add(buf, &s, sizeof(s));
+       ibuf_close(&rs->w, buf);
+}
+
+/*
+ * Validate the common rtr header (first 8 bytes) including the
+ * included length field.
+ * Returns -1 on failure. On success msgtype and msglen are set
+ * and the function return 0.
+ */
+static int
+rtr_parse_header(struct rtr_session *rs, void *buf,
+    size_t *msglen, enum rtr_pdu_type *msgtype)
+{
+       struct rtr_header rh;
+       uint32_t len = 16;      /* default for ERROR_REPORT */
+       int session_id;
+
+       memcpy(&rh, buf, sizeof(rh));
+
+       if (rh.version != 1) {
+               log_warnx("rtr %s: received message with unsupported version",
+                   log_rtr(rs));
+               rtr_send_error(rs, UNEXP_PROTOCOL_VERS, NULL, &rh, sizeof(rh));
+               return -1;
+       }
+
+       *msgtype = rh.type;
+       *msglen = ntohl(rh.length);
+
+       switch (*msgtype) {
+       case SERIAL_NOTIFY:
+               session_id = rs->session_id;
+               len = 12;
+               break;
+       case CACHE_RESPONSE:
+               /* set session_id if not yet happened */
+               if (rs->session_id == -1)
+                       rs->session_id = ntohs(rh.session_id);
+               session_id = rs->session_id;
+               len = 8;
+               break;
+       case IPV4_PREFIX:
+               session_id = 0;
+               len = 20;
+               break;
+       case IPV6_PREFIX:
+               session_id = 0;
+               len = 32;
+               break;
+       case END_OF_DATA:
+               session_id = rs->session_id;
+               len = 24;
+               break;
+       case CACHE_RESET:
+               session_id = 0;
+               len = 8;
+               break;
+       case ROUTER_KEY:
+               len = 36;       /* XXX probably too small, but we ignore it */
+               /* FALLTHROUGH */
+       case ERROR_REPORT:
+               if (*msglen > RTR_MAX_LEN) {
+                       log_warnx("rtr %s: received %s: msg too big: %zu byte",
+                           log_rtr(rs), log_rtr_type(*msgtype), *msglen);
+                       rtr_send_error(rs, CORRUPT_DATA, "too big",
+                           &rh, sizeof(rh));
+                       return -1;
+               }
+               if (*msglen < len) {
+                       log_warnx("rtr %s: received %s: msg too small: "
+                           "%zu byte", log_rtr(rs), log_rtr_type(*msgtype),
+                           *msglen);
+                       rtr_send_error(rs, CORRUPT_DATA, "too small",
+                           &rh, sizeof(rh));
+                       return -1;
+               }
+               /*
+                * session_id check ommitted since ROUTER_KEY and ERROR_REPORT
+                * use the field for different things.
+                */
+               return 0;
+       default:
+               log_warnx("rtr %s: received unknown message: type %u",
+                   log_rtr(rs), *msgtype);
+               rtr_send_error(rs, UNSUPP_PDU_TYPE, NULL, &rh, sizeof(rh));
+               return -1;
+       }
+
+       if (len != *msglen) {
+               log_warnx("rtr %s: received %s: illegal len: %zu byte not %u",
+                   log_rtr(rs), log_rtr_type(*msgtype), *msglen, len);
+               rtr_send_error(rs, CORRUPT_DATA, "bad length",
+                   &rh, sizeof(rh));
+               return -1;
+       }
+
+       if (session_id != ntohs(rh.session_id)) {
+               /* ignore SERIAL_NOTIFY during startup */
+               if (rs->session_id == -1 && *msgtype == SERIAL_NOTIFY)
+                       return 0;
+
+               log_warnx("rtr %s: received %s: bad session_id: %d != %d",
+                   log_rtr(rs), log_rtr_type(*msgtype), ntohs(rh.session_id),
+                   session_id);
+               rtr_send_error(rs, CORRUPT_DATA, "bad session_id",
+                   &rh, sizeof(rh));
+               return -1;
+       }
+
+       return 0;
+}
+
+static int
+rtr_parse_notify(struct rtr_session *rs, uint8_t *buf, size_t len)
+{
+       if (rs->state == RTR_STATE_ACTIVE) {
+               log_warnx("rtr %s: received %s: while active (ignored)",
+                   log_rtr(rs), log_rtr_type(SERIAL_NOTIFY));
+               return 0;
+       }
+
+       rtr_fsm(rs, RTR_EVNT_SERIAL_NOTIFY);
+       return 0;
+}
+
+static int
+rtr_parse_cache_response(struct rtr_session *rs, uint8_t *buf, size_t len)
+{
+       if (rs->state != RTR_STATE_IDLE) {
+               log_warnx("rtr %s: received %s: out of context",
+                   log_rtr(rs), log_rtr_type(CACHE_RESPONSE));
+               return -1;
+       }
+
+       rtr_fsm(rs, RTR_EVNT_CACHE_RESPONSE);
+       return 0;
+}
+
+static int
+rtr_parse_ipv4_prefix(struct rtr_session *rs, uint8_t *buf, size_t len)
+{
+       struct rtr_ipv4 ip4;
+       struct roa *roa;
+
+       if (len != sizeof(struct rtr_header) + sizeof(ip4)) {
+               log_warnx("rtr %s: received %s: bad pdu len",
+                   log_rtr(rs), log_rtr_type(IPV4_PREFIX));
+               rtr_send_error(rs, CORRUPT_DATA, "bad len", buf, len);
+               return -1;
+       }
+
+       if (rs->state != RTR_STATE_ACTIVE) {
+               log_warnx("rtr %s: received %s: out of context",
+                   log_rtr(rs), log_rtr_type(IPV4_PREFIX));
+               rtr_send_error(rs, CORRUPT_DATA, NULL, buf, len);
+               return -1;
+       }
+
+       memcpy(&ip4, buf + sizeof(struct rtr_header), sizeof(ip4));
+
+       if ((roa = calloc(1, sizeof(*roa))) == NULL) {
+               log_warn("rtr %s: received %s",
+                   log_rtr(rs), log_rtr_type(IPV4_PREFIX));
+               rtr_send_error(rs, INTERNAL_ERROR, "out of memory", NULL, 0);
+               return -1;
+       }
+       roa->aid = AID_INET;
+       roa->prefixlen = ip4.prefixlen;
+       roa->maxlen = ip4.maxlen;
+       roa->asnum = ntohl(ip4.asnum);
+       roa->prefix.inet.s_addr = ip4.prefix;
+
+       if (ip4.flags & FLAG_ANNOUNCE) {
+               if (RB_INSERT(roa_tree, &rs->roa_set, roa) != NULL) {
+                       log_warnx("rtr %s: received %s: duplicate announcement",
+                           log_rtr(rs), log_rtr_type(IPV4_PREFIX));
+                       rtr_send_error(rs, DUP_REC_RECV, NULL, buf, len);
+                       free(roa);
+                       return -1;
+               }
+       } else {
+               struct roa *r;
+
+               r = RB_FIND(roa_tree, &rs->roa_set, roa);
+               if (r == NULL) {
+                       log_warnx("rtr %s: received %s: unknown withdrawl",
+                           log_rtr(rs), log_rtr_type(IPV4_PREFIX));
+                       rtr_send_error(rs, UNK_REC_WDRAWL, NULL, buf, len);
+                       free(roa);
+                       return -1;
+               }
+               RB_REMOVE(roa_tree, &rs->roa_set, r);
+               free(r);
+               free(roa);
+       }
+
+       return 0;
+}
+
+static int
+rtr_parse_ipv6_prefix(struct rtr_session *rs, uint8_t *buf, size_t len)
+{
+       struct rtr_ipv6 ip6;
+       struct roa *roa;
+
+       if (len != sizeof(struct rtr_header) + sizeof(ip6)) {
+               log_warnx("rtr %s: received %s: bad pdu len",
+                   log_rtr(rs), log_rtr_type(IPV6_PREFIX));
+               rtr_send_error(rs, CORRUPT_DATA, "bad len", buf, len);
+               return -1;
+       }
+
+       if (rs->state != RTR_STATE_ACTIVE) {
+               log_warnx("rtr %s: received %s: out of context",
+                   log_rtr(rs), log_rtr_type(IPV6_PREFIX));
+               rtr_send_error(rs, CORRUPT_DATA, NULL, buf, len);
+               return -1;
+       }
+
+       memcpy(&ip6, buf + sizeof(struct rtr_header), sizeof(ip6));
+
+       if ((roa = calloc(1, sizeof(*roa))) == NULL) {
+               log_warn("rtr %s: received %s",
+                   log_rtr(rs), log_rtr_type(IPV6_PREFIX));
+               rtr_send_error(rs, INTERNAL_ERROR, "out of memory", NULL, 0);
+               return -1;
+       }
+       roa->aid = AID_INET6;
+       roa->prefixlen = ip6.prefixlen;
+       roa->maxlen = ip6.maxlen;
+       roa->asnum = ntohl(ip6.asnum);
+       memcpy(&roa->prefix.inet6, ip6.prefix, sizeof(roa->prefix.inet6));
+
+       if (ip6.flags & FLAG_ANNOUNCE) {
+               if (RB_INSERT(roa_tree, &rs->roa_set, roa) != NULL) {
+                       log_warnx("rtr %s: received %s: duplicate announcement",
+                           log_rtr(rs), log_rtr_type(IPV6_PREFIX));
+                       rtr_send_error(rs, DUP_REC_RECV, NULL, buf, len);
+                       free(roa);
+                       return -1;
+               }
+       } else {
+               struct roa *r;
+
+               r = RB_FIND(roa_tree, &rs->roa_set, roa);
+               if (r == NULL) {
+                       log_warnx("rtr %s: received %s: unknown withdrawl",
+                           log_rtr(rs), log_rtr_type(IPV6_PREFIX));
+                       rtr_send_error(rs, UNK_REC_WDRAWL, NULL, buf, len);
+                       free(roa);
+                       return -1;
+               }
+               RB_REMOVE(roa_tree, &rs->roa_set, r);
+               free(r);
+               free(roa);
+       }
+       return 0;
+}
+
+static int
+rtr_parse_end_of_data(struct rtr_session *rs, uint8_t *buf, size_t len)
+{
+       struct rtr_endofdata eod;
+       uint32_t t;
+
+       buf += sizeof(struct rtr_header);
+       len -= sizeof(struct rtr_header);
+
+       if (len != sizeof(eod)) {
+               log_warnx("rtr %s: received %s: bad pdu len",
+                   log_rtr(rs), log_rtr_type(END_OF_DATA));
+               return -1;
+       }
+
+       memcpy(&eod, buf, sizeof(eod));
+
+       if (rs->state != RTR_STATE_ACTIVE) {
+               log_warnx("rtr %s: received %s: out of context",
+                   log_rtr(rs), log_rtr_type(END_OF_DATA));
+               return -1;
+       }
+
+       rs->serial = ntohl(eod.serial);
+       /* validate timer values to be in the right range */
+       t = ntohl(eod.refresh);
+       if (t < 1 || t > 86400)
+               goto bad;
+       rs->refresh = t;
+       t = ntohl(eod.retry);
+       if (t < 1 || t > 7200)
+               goto bad;
+       rs->retry = t;
+       t = ntohl(eod.expire);
+       if (t < 600 || t > 172800)
+               goto bad;
+       if (t <= rs->retry || t <= rs->refresh)
+               goto bad;
+       rs->expire = t;
+
+       rtr_fsm(rs, RTR_EVNT_END_OF_DATA);
+       return 0;
+
+bad:
+       log_warnx("rtr %s: received %s: bad timeout values",
+           log_rtr(rs), log_rtr_type(END_OF_DATA));
+       return -1;
+}
+
+static int
+rtr_parse_cache_reset(struct rtr_session *rs, uint8_t *buf, size_t len)
+{
+       if (rs->state != RTR_STATE_IDLE) {
+               log_warnx("rtr %s: received %s: out of context",
+                   log_rtr(rs), log_rtr_type(CACHE_RESET));
+               return -1;
+       }
+
+       rtr_fsm(rs, RTR_EVNT_CACHE_RESET);
+       return 0;
+}
+
+/*
+ * Parse an Error Response message. This function behaves a bit different
+ * from other parse functions since on error the connection needs to be
+ * dropped without sending an error response back.
+ */
+static int
+rtr_parse_error(struct rtr_session *rs, uint8_t *buf, size_t len)
+{
+       struct rtr_header rh;
+       uint32_t pdu_len, msg_len;
+       uint8_t *msg;
+       char *str = NULL;
+       uint16_t errcode;
+
+       memcpy(&rh, buf, sizeof(rh));
+       buf += sizeof(struct rtr_header);
+       len -= sizeof(struct rtr_header);
+       errcode = ntohs(rh.session_id);
+
+       memcpy(&pdu_len, buf, sizeof(pdu_len));
+       pdu_len = ntohs(pdu_len);
+
+       if (len < pdu_len + sizeof(pdu_len)) {
+               log_warnx("rtr %s: received %s: bad pdu len: %u byte",
+                   log_rtr(rs), log_rtr_type(ERROR_REPORT), pdu_len);
+               rtr_fsm(rs, RTR_EVNT_CON_CLOSED);
+               return -1;
+       }
+
+       /* for now just ignore the embedded pdu */
+       buf += pdu_len + sizeof(pdu_len);
+       len -= pdu_len + sizeof(pdu_len);
+
+       memcpy(&msg_len, buf, sizeof(msg_len));
+       msg_len = ntohs(msg_len);
+
+       if (len < msg_len + sizeof(msg_len)) {
+               log_warnx("rtr %s: received %s: bad msg len: %u byte",
+                   log_rtr(rs), log_rtr_type(ERROR_REPORT), msg_len);
+               rtr_fsm(rs, RTR_EVNT_CON_CLOSED);
+               return -1;
+       }
+
+       msg = buf + sizeof(msg_len);
+       if (msg_len != 0)
+               /* optional error msg, no need to check for failure */
+               str = strndup(msg, msg_len);
+
+       log_warnx("rtr %s: received error: %s%s%s", log_rtr(rs),
+           log_rtr_error(errcode), str ? ": " : "", str ? str : "");
+
+       if (errcode == NO_DATA_AVAILABLE) {
+               rtr_fsm(rs, RTR_EVNT_NO_DATA);
+       } else {
+               rtr_fsm(rs, RTR_EVNT_CON_CLOSED);
+               rs->last_recv_error = errcode;
+               if (str)
+                       strlcpy(rs->last_recv_msg, str,
+                           sizeof(rs->last_recv_msg));
+               else
+                       memset(rs->last_recv_msg, 0,
+                           sizeof(rs->last_recv_msg));
+
+               free(str);
+               return -1;
+       }
+       free(str);
+
+       return 0;
+}
+
+/*
+ * Try to process received rtr message, it is possible that not a full
+ * message is in the buffer. In that case stop, once new data is available
+ * a retry will be done.
+ */
+static void
+rtr_process_msg(struct rtr_session *rs)
+{
+       size_t rpos, av, left;
+       void *rptr;
+       size_t msglen;
+       enum rtr_pdu_type msgtype;
+
+       rpos = 0;
+       av = rs->r.wpos;
+
+       for (;;) {
+               if (rpos + sizeof(struct rtr_header) > av)
+                       break;
+               rptr = rs->r.buf + rpos;
+               if (rtr_parse_header(rs, rptr, &msglen, &msgtype) == -1)
+                       return;
+
+               /* missing data */
+               if (rpos + msglen > av)
+                       break;
+
+               switch (msgtype) {
+               case SERIAL_NOTIFY:
+                       if (rtr_parse_notify(rs, rptr, msglen) == -1) {
+                               rtr_send_error(rs, CORRUPT_DATA, NULL,
+                                   rptr, msglen);
+                               return;
+                       }
+                       break;
+               case CACHE_RESPONSE:
+                       if (rtr_parse_cache_response(rs, rptr, msglen) == -1) {
+                               rtr_send_error(rs, CORRUPT_DATA, NULL,
+                                   rptr, msglen);
+                               return;
+                       }
+                       break;
+               case IPV4_PREFIX:
+                       if (rtr_parse_ipv4_prefix(rs, rptr, msglen) == -1) {
+                               return;
+                       }
+                       break;
+               case IPV6_PREFIX:
+                       if (rtr_parse_ipv6_prefix(rs, rptr, msglen) == -1) {
+                               return;
+                       }
+                       break;
+               case END_OF_DATA:
+                       if (rtr_parse_end_of_data(rs, rptr, msglen) == -1) {
+                               rtr_send_error(rs, CORRUPT_DATA, NULL,
+                                   rptr, msglen);
+                               return;
+                       }
+                       break;
+               case CACHE_RESET:
+                       if (rtr_parse_cache_reset(rs, rptr, msglen) == -1) {
+                               rtr_send_error(rs, CORRUPT_DATA, NULL,
+                                   rptr, msglen);
+                               return;
+                       }
+                       break;
+               case ROUTER_KEY:
+                       /* silently ignore router key */
+                       break;
+               case ERROR_REPORT:
+                       if (rtr_parse_error(rs, rptr, msglen) == -1)
+                               /* no need to send back an error */
+                               return;
+                       break;
+               default:
+                       log_warnx("rtr %s: received %s: unexpected pdu type",
+                           log_rtr(rs), log_rtr_type(msgtype));
+                       rtr_send_error(rs, INVALID_REQUEST, NULL, rptr, msglen);
+                       return;
+               }
+               rpos += msglen;
+       }
+
+       left = av - rpos;
+       memmove(&rs->r.buf, rs->r.buf + rpos, left);
+       rs->r.wpos = left;
+}
+
+/*
+ * Simple FSM for RTR sessions
+ */
+static void
+rtr_fsm(struct rtr_session *rs, enum rtr_event event)
+{
+       enum rtr_state prev_state = rs->state;
+
+       switch (event) {
+       case RTR_EVNT_CON_CLOSED:
+               if (rs->state != RTR_STATE_CLOSED) {
+                       /* flush buffers */
+                       msgbuf_clear(&rs->w);
+                       rs->r.wpos = 0;
+                       close(rs->fd);
+                       rs->fd = -1;
+               }
+               rs->state = RTR_STATE_CLOSED;
+               /* try to reopen session */
+               timer_set(&rs->timers, Timer_Rtr_Retry,
+                   arc4random_uniform(10));
+               break;
+       case RTR_EVNT_START:
+       case RTR_EVNT_TIMER_RETRY:
+               switch (rs->state) {
+               case RTR_STATE_ERROR:
+                       rtr_fsm(rs, RTR_EVNT_CON_CLOSED);
+                       return;
+               case RTR_STATE_CLOSED:
+                       timer_set(&rs->timers, Timer_Rtr_Retry, rs->retry);
+                       rtr_imsg_compose(IMSG_SOCKET_CONN, rs->id, 0, NULL, 0);
+                       return;
+               default:
+                       break;
+               }
+               /* FALLTHROUGH */
+       case RTR_EVNT_CON_OPEN:
+               timer_stop(&rs->timers, Timer_Rtr_Retry);
+               if (rs->session_id == -1)
+                       rtr_reset_query(rs);
+               else
+                       rtr_serial_query(rs);
+               break;
+       case RTR_EVNT_SERIAL_NOTIFY:
+               /* schedule a refresh after a quick wait */
+               timer_set(&rs->timers, Timer_Rtr_Refresh,
+                   arc4random_uniform(10));
+               break;
+       case RTR_EVNT_TIMER_REFRESH:
+               /* send serial query */
+               rtr_serial_query(rs);
+               break;
+       case RTR_EVNT_TIMER_EXPIRE:
+               free_roatree(&rs->roa_set);
+               rtr_recalc();
+               break;
+       case RTR_EVNT_CACHE_RESPONSE:
+               rs->state = RTR_STATE_ACTIVE;
+               timer_stop(&rs->timers, Timer_Rtr_Refresh);
+               timer_stop(&rs->timers, Timer_Rtr_Retry);
+               break;
+       case RTR_EVNT_END_OF_DATA:
+               /* start refresh and expire timers */
+               timer_set(&rs->timers, Timer_Rtr_Refresh, rs->refresh);
+               timer_set(&rs->timers, Timer_Rtr_Expire, rs->expire);
+               rs->state = RTR_STATE_IDLE;
+               rtr_recalc();
+               break;
+       case RTR_EVNT_CACHE_RESET:
+               /* reset session and retry after a quick wait */
+               rs->session_id = -1;
+               free_roatree(&rs->roa_set);
+               timer_set(&rs->timers, Timer_Rtr_Retry,
+                   arc4random_uniform(10));
+               break;
+       case RTR_EVNT_NO_DATA:
+               /* start retry timer */
+               timer_set(&rs->timers, Timer_Rtr_Retry, rs->retry);
+               /* stop refresh timer just to be sure */
+               timer_stop(&rs->timers, Timer_Rtr_Refresh);
+               break;
+       case RTR_EVNT_SEND_ERROR:
+               rs->state = RTR_STATE_ERROR;
+               /* flush receive buffer */
+               rs->r.wpos = 0;
+               break;
+       }
+
+       log_info("rtr %s: state change %s -> %s, reason: %s",
+           log_rtr(rs), rtr_statenames[prev_state], rtr_statenames[rs->state],
+           rtr_eventnames[event]);
+}
+
+/*
+ * IO handler for RTR sessions
+ */
+static void
+rtr_dispatch_msg(struct pollfd *pfd, struct rtr_session *rs)
+{
+       ssize_t n;
+       int error;
+
+       if (pfd->revents & POLLHUP) {
+               log_warnx("rtr %s: Connection closed, hangup", log_rtr(rs));
+               rtr_fsm(rs, RTR_EVNT_CON_CLOSED);
+               return;
+       }
+       if (pfd->revents & (POLLERR|POLLNVAL)) {
+               log_warnx("rtr %s: Connection closed, error", log_rtr(rs));
+               rtr_fsm(rs, RTR_EVNT_CON_CLOSED);
+               return;
+       }
+       if (pfd->revents & POLLOUT && rs->w.queued) {
+               if ((error = ibuf_write(&rs->w)) <= 0 && errno != EAGAIN) {
+                       if (error == 0)
+                               log_warnx("rtr %s: Connection closed",
+                                   log_rtr(rs));
+                       else if (error == -1)
+                               log_warn("rtr %s: write error", log_rtr(rs));
+                       rtr_fsm(rs, RTR_EVNT_CON_CLOSED);
+                       return;
+               }
+               if (rs->w.queued == 0 && rs->state == RTR_STATE_ERROR)
+                       rtr_fsm(rs, RTR_EVNT_CON_CLOSED);
+       }
+       if (pfd->revents & POLLIN) {
+               if ((n = read(rs->fd, rs->r.buf + rs->r.wpos,
+                   sizeof(rs->r.buf) - rs->r.wpos)) == -1) {
+                       if (errno != EINTR && errno != EAGAIN) {
+                               log_warn("rtr %s: read error", log_rtr(rs));
+                               rtr_fsm(rs, RTR_EVNT_CON_CLOSED);
+                       }
+                       return;
+               }
+               if (n == 0) {
+                       log_warnx("rtr %s: Connection closed", log_rtr(rs));
+                       rtr_fsm(rs, RTR_EVNT_CON_CLOSED);
+                       return;
+               }
+               rs->r.wpos += n;
+
+               /* new data arrived, try to process it */
+               rtr_process_msg(rs);
+       }
+
+}
+
+void
+rtr_check_events(struct pollfd *pfds, size_t npfds)
+{
+       struct rtr_session *rs;
+       struct timer *t;
+       time_t now;
+       size_t i = 0;
+
+       for (i = 0; i < npfds; i++) {
+               if (pfds[i].revents == 0)
+                       continue;
+               TAILQ_FOREACH(rs, &rtrs, entry)
+                       if (rs->fd == pfds[i].fd) {
+                               rtr_dispatch_msg(&pfds[i], rs);
+                               break;
+                       }
+               if (rs == NULL)
+                       log_warnx("%s: unknown fd in pollfds", __func__);
+       }
+
+       /* run all timers */
+       now = getmonotime();
+       TAILQ_FOREACH(rs, &rtrs, entry)
+               if ((t = timer_nextisdue(&rs->timers, now)) != NULL) {
+                       log_debug("rtr %s: %s triggered", log_rtr(rs),
+                           timernames[t->type]);
+                       /* stop timer so it does not trigger again */
+                       timer_stop(&rs->timers, t->type);
+                       switch (t->type) {
+                       case Timer_Rtr_Refresh:
+                               rtr_fsm(rs, RTR_EVNT_TIMER_REFRESH);
+                               break;
+                       case Timer_Rtr_Retry:
+                               rtr_fsm(rs, RTR_EVNT_TIMER_RETRY);
+                               break;
+                       case Timer_Rtr_Expire:
+                               rtr_fsm(rs, RTR_EVNT_TIMER_EXPIRE);
+                               break;
+                       default:
+                               fatalx("King Bula lost in time");
+                       }
+               }
+}
+
+size_t
+rtr_count(void)
+{
+       struct rtr_session *rs;
+       size_t count = 0;
+
+       TAILQ_FOREACH(rs, &rtrs, entry)
+               count++;
+       return count;
+}
+
+size_t
+rtr_poll_events(struct pollfd *pfds, size_t npfds, time_t *timeout)
+{
+       struct rtr_session *rs;
+       time_t now = getmonotime();
+       size_t i = 0;
+
+       TAILQ_FOREACH(rs, &rtrs, entry) {
+               time_t nextaction;
+               struct pollfd *pfd = pfds + i++;
+
+               if (i > npfds)
+                       fatalx("%s: too many sessions for pollfd", __func__);
+
+               if ((nextaction = timer_nextduein(&rs->timers, now)) != -1 &&
+                   nextaction < *timeout)
+                       *timeout = nextaction;
+
+               if (rs->state == RTR_STATE_CLOSED) {
+                       pfd->fd = -1;
+                       continue;
+               }
+
+               pfd->fd = rs->fd;
+               pfd->events = 0;
+
+               if (rs->w.queued)
+                       pfd->events |= POLLOUT;
+               if (rs->state >= RTR_STATE_IDLE)
+                       pfd->events |= POLLIN;
+       }
+
+       return i;
+}
+
+struct rtr_session *
+rtr_new(uint32_t id, char *descr)
+{
+       struct rtr_session *rs;
+       
+       if ((rs = calloc(1, sizeof(*rs))) == NULL)
+               fatal("RTR session %s", descr);
+
+       RB_INIT(&rs->roa_set);
+       TAILQ_INIT(&rs->timers);
+       msgbuf_init(&rs->w);
+
+       strlcpy(rs->descr, descr, sizeof(rs->descr));
+       rs->id = id;
+       rs->session_id = -1;
+       rs->refresh = RTR_DEFAULT_REFRESH;
+       rs->retry = RTR_DEFAULT_RETRY;
+       rs->expire = RTR_DEFAULT_EXPIRE;
+       rs->state = RTR_STATE_CLOSED;
+       rs->reconf_action = RECONF_REINIT;
+       rs->last_recv_error = NO_ERROR;
+       rs->last_sent_error = NO_ERROR;
+
+       /* make sure that some timer is running to abort bad sessions */
+       timer_set(&rs->timers, Timer_Rtr_Expire, rs->expire);
+
+       log_debug("rtr %s: new session, start", log_rtr(rs));
+       TAILQ_INSERT_TAIL(&rtrs, rs, entry);
+       rtr_fsm(rs, RTR_EVNT_START);
+
+       return rs;
+}
+
+struct rtr_session *
+rtr_get(uint32_t id)
+{
+       struct rtr_session *rs;
+
+       TAILQ_FOREACH(rs, &rtrs, entry)
+               if (rs->id == id)
+                       return rs;
+       return NULL;
+}
+
+void
+rtr_free(struct rtr_session *rs)
+{
+       if (rs == NULL)
+               return;
+
+       rtr_fsm(rs, RTR_EVNT_CON_CLOSED);
+       timer_remove_all(&rs->timers);
+       free_roatree(&rs->roa_set);
+       free(rs);
+}
+
+void
+rtr_open(struct rtr_session *rs, int fd)
+{
+       if (rs->state != RTR_STATE_CLOSED) {
+               log_warnx("rtr %s: bad session state", log_rtr(rs));
+               rtr_fsm(rs, RTR_EVNT_CON_CLOSED);
+       }
+
+       log_debug("rtr %s: connection opened", log_rtr(rs));
+
+       rs->fd = rs->w.fd = fd;
+       rs->state = RTR_STATE_IDLE;
+       rtr_fsm(rs, RTR_EVNT_CON_OPEN);
+}
+
+void
+rtr_config_prep(void)
+{
+       struct rtr_session *rs;
+
+       TAILQ_FOREACH(rs, &rtrs, entry)
+               rs->reconf_action = RECONF_DELETE;
+}
+
+void
+rtr_config_merge(void)
+{
+       struct rtr_session *rs, *nrs;
+
+       TAILQ_FOREACH_SAFE(rs, &rtrs, entry, nrs)
+               if (rs->reconf_action == RECONF_DELETE) {
+                       TAILQ_REMOVE(&rtrs, rs, entry);
+                       rtr_free(rs);
+               }
+}
+
+void
+rtr_config_keep(struct rtr_session *rs)
+{
+       rs->reconf_action = RECONF_KEEP;
+}
+
+void
+rtr_roa_merge(struct roa_tree *rt)
+{
+       struct rtr_session *rs;
+       struct roa *roa;
+
+       TAILQ_FOREACH(rs, &rtrs, entry) {
+               RB_FOREACH(roa, roa_tree, &rs->roa_set)
+                       roa_insert(rt, roa);
+       }
+}
+
+void
+rtr_shutdown(void)
+{
+       struct rtr_session *rs, *nrs;
+
+       TAILQ_FOREACH_SAFE(rs, &rtrs, entry, nrs)
+               rtr_free(rs);
+}
+
+void
+rtr_show(struct rtr_session *rs, pid_t pid)
+{
+       struct ctl_show_rtr msg;
+       struct ctl_timer ct;
+       u_int i;
+       time_t d;
+
+       memset(&msg, 0, sizeof(msg));
+
+       /* descr, remote_addr, local_addr and remote_port set by parent */
+       msg.serial = rs->serial;
+       msg.refresh = rs->refresh;
+       msg.retry = rs->retry;
+       msg.expire = rs->expire;
+       msg.session_id = rs->session_id;
+       msg.last_sent_error = rs->last_sent_error;
+       msg.last_recv_error = rs->last_recv_error;
+       strlcpy(msg.last_sent_msg, rs->last_sent_msg,
+           sizeof(msg.last_sent_msg));
+       strlcpy(msg.last_recv_msg, rs->last_recv_msg,
+           sizeof(msg.last_recv_msg));
+
+       /* send back imsg */
+       rtr_imsg_compose(IMSG_CTL_SHOW_RTR, rs->id, pid, &msg, sizeof(msg));
+
+       /* send back timer imsgs */
+       for (i = 1; i < Timer_Max; i++) {
+               if (!timer_running(&rs->timers, i, &d))
+                       continue;
+               ct.type = i;
+               ct.val = d;
+               rtr_imsg_compose(IMSG_CTL_SHOW_TIMER, rs->id, pid,
+                   &ct, sizeof(ct));
+       }
+}
index 93db42d..6539101 100644 (file)
@@ -1,4 +1,4 @@
-/*     $OpenBSD: session.c,v 1.410 2021/01/05 10:02:44 claudio Exp $ */
+/*     $OpenBSD: session.c,v 1.411 2021/02/16 08:29:16 claudio Exp $ */
 
 /*
  * Copyright (c) 2003, 2004, 2005 Henning Brauer <henning@openbsd.org>
@@ -2829,6 +2829,8 @@ session_dispatch_imsg(struct imsgbuf *ibuf, int idx, u_int *listener_cnt)
                case IMSG_CTL_SHOW_NEXTHOP:
                case IMSG_CTL_SHOW_INTERFACE:
                case IMSG_CTL_SHOW_FIB_TABLES:
+               case IMSG_CTL_SHOW_RTR:
+               case IMSG_CTL_SHOW_TIMER:
                        if (idx != PFD_PIPE_MAIN)
                                fatalx("ctl kroute request not from parent");
                        control_imsg_relay(&imsg);
index c3c8530..ddf1caf 100644 (file)
@@ -1,4 +1,4 @@
-/*     $OpenBSD: session.h,v 1.149 2020/12/23 13:20:48 claudio Exp $ */
+/*     $OpenBSD: session.h,v 1.150 2021/02/16 08:29:16 claudio Exp $ */
 
 /*
  * Copyright (c) 2003, 2004 Henning Brauer <henning@openbsd.org>
@@ -186,6 +186,9 @@ enum Timer {
        Timer_IdleHoldReset,
        Timer_CarpUndemote,
        Timer_RestartTimeout,
+       Timer_Rtr_Refresh,
+       Timer_Rtr_Retry,
+       Timer_Rtr_Expire,
        Timer_Max
 };
 
@@ -281,9 +284,6 @@ void                 mrt_dump_state(struct mrt *, u_int16_t, u_int16_t,
                     struct peer *);
 void            mrt_done(struct mrt *);
 
-/* parse.y */
-struct bgpd_config *parse_config(char *, struct peer_head *);
-
 /* pfkey.c */
 int    pfkey_read(int, struct sadb_msg *);
 int    pfkey_establish(struct peer *);
@@ -299,7 +299,30 @@ void       tcp_md5_del_listener(struct bgpd_config *, struct peer *);
 void   print_config(struct bgpd_config *, struct rib_names *);
 
 /* rde.c */
-void    rde_main(int, int);
+void   rde_main(int, int);
+
+/* rtr_proto.c */
+struct rtr_session;
+size_t                  rtr_count(void);
+void                    rtr_check_events(struct pollfd *, size_t);
+size_t                  rtr_poll_events(struct pollfd *, size_t, time_t *);
+struct rtr_session     *rtr_new(uint32_t, char *);
+struct rtr_session     *rtr_get(uint32_t);
+void                    rtr_free(struct rtr_session *);
+void                    rtr_open(struct rtr_session *, int);
+struct roa_tree                *rtr_get_roa(struct rtr_session *);
+void                    rtr_config_prep(void);
+void                    rtr_config_merge(void);
+void                    rtr_config_keep(struct rtr_session *);
+void                    rtr_roa_merge(struct roa_tree *);
+void                    rtr_shutdown(void);
+void                    rtr_show(struct rtr_session *, pid_t);
+
+/* rtr.c */
+void   roa_insert(struct roa_tree *, struct roa *);
+void   rtr_main(int, int);
+void   rtr_imsg_compose(int, uint32_t, pid_t, void *, size_t);
+void   rtr_recalc(void);
 
 /* session.c */
 RB_PROTOTYPE(peer_head, peer, entry, peer_compare);
index 5d677b0..09d82e4 100644 (file)
@@ -1,4 +1,4 @@
-/*     $OpenBSD: util.c,v 1.60 2021/01/25 09:15:24 claudio Exp $ */
+/*     $OpenBSD: util.c,v 1.61 2021/02/16 08:29:16 claudio Exp $ */
 
 /*
  * Copyright (c) 2006 Claudio Jeker <claudio@openbsd.org>
@@ -161,6 +161,38 @@ log_reason(const char *communication) {
        return buf;
 }
 
+const char *
+log_rtr_error(enum rtr_error err)
+{
+       static char buf[20];
+
+       switch (err) {
+       case NO_ERROR:
+               return "No Error";
+       case CORRUPT_DATA:
+               return "Corrupt Data";
+       case INTERNAL_ERROR:
+               return "Internal Error";
+       case NO_DATA_AVAILABLE:
+               return "No Data Available";
+       case INVALID_REQUEST:
+               return "Invalid Request";
+       case UNSUPP_PROTOCOL_VERS:
+               return "Unsupported Protocol Version";
+       case UNSUPP_PDU_TYPE:
+               return "Unsupported PDU Type";
+       case UNK_REC_WDRAWL:
+               return "Withdrawl of Unknown Record";
+       case DUP_REC_RECV:
+               return "Duplicate Announcement Received";
+       case UNEXP_PROTOCOL_VERS:
+               return "Unexpected Protocol Version";
+       default:
+               snprintf(buf, sizeof(buf), "unknown %u", err);
+               return buf;
+       }
+}
+
 const char *
 aspath_delim(u_int8_t seg_type, int closing)
 {