X-Git-Url: https://git.proxmox.com/?a=blobdiff_plain;f=zebra%2Frt_socket.c;h=539370a498c79a30f5e7c7d107162acef4bf203d;hb=18d10d885461eaf44fba87ed57a347b565e94f55;hp=346699198f9b8c15f22f724da4fdeead4fec6e22;hpb=0be9d44862758e039771982f347b8af4142ee876;p=mirror_frr.git

diff --git a/zebra/rt_socket.c b/zebra/rt_socket.c
index 346699198..539370a49 100644
--- a/zebra/rt_socket.c
+++ b/zebra/rt_socket.c
@@ -33,12 +33,14 @@
 #include "log.h"
 #include "privs.h"
 #include "vxlan.h"
+#include "lib_errors.h"
 
 #include "zebra/debug.h"
 #include "zebra/rib.h"
 #include "zebra/rt.h"
 #include "zebra/kernel_socket.h"
 #include "zebra/zebra_mpls.h"
+#include "zebra/zebra_errors.h"
 
 extern struct zebra_privs_t zserv_privs;
 
@@ -71,10 +73,10 @@ static int kernel_rtm_add_labels(struct mpls_label_stack *nh_label,
 				 struct sockaddr_mpls *smpls)
 {
 	if (nh_label->num_labels > 1) {
-		zlog_warn(
-			"%s: can't push %u labels at "
-			"once (maximum is 1)",
-			__func__, nh_label->num_labels);
+		flog_warn(EC_ZEBRA_MAX_LABELS_PUSH,
+			  "%s: can't push %u labels at "
+			  "once (maximum is 1)",
+			  __func__, nh_label->num_labels);
 		return -1;
 	}
 
@@ -87,167 +89,6 @@ static int kernel_rtm_add_labels(struct mpls_label_stack *nh_label,
 }
 #endif
 
-/* Interface between zebra message and rtm message. */
-static int kernel_rtm_ipv4(int cmd, const struct prefix *p,
-			   struct route_entry *re)
-
-{
-	struct sockaddr_in *mask = NULL;
-	struct sockaddr_in sin_dest, sin_mask, sin_gate;
-#ifdef __OpenBSD__
-	struct sockaddr_mpls smpls;
-#endif
-	union sockunion *smplsp = NULL;
-	struct nexthop *nexthop;
-	int nexthop_num = 0;
-	ifindex_t ifindex = 0;
-	int gate = 0;
-	int error;
-	char prefix_buf[PREFIX_STRLEN];
-	enum blackhole_type bh_type = BLACKHOLE_UNSPEC;
-
-	if (IS_ZEBRA_DEBUG_RIB)
-		prefix2str(p, prefix_buf, sizeof(prefix_buf));
-	memset(&sin_dest, 0, sizeof(struct sockaddr_in));
-	sin_dest.sin_family = AF_INET;
-#ifdef HAVE_STRUCT_SOCKADDR_IN_SIN_LEN
-	sin_dest.sin_len = sizeof(struct sockaddr_in);
-#endif /* HAVE_STRUCT_SOCKADDR_IN_SIN_LEN */
-	sin_dest.sin_addr = p->u.prefix4;
-
-	memset(&sin_mask, 0, sizeof(struct sockaddr_in));
-
-	memset(&sin_gate, 0, sizeof(struct sockaddr_in));
-	sin_gate.sin_family = AF_INET;
-#ifdef HAVE_STRUCT_SOCKADDR_IN_SIN_LEN
-	sin_gate.sin_len = sizeof(struct sockaddr_in);
-#endif /* HAVE_STRUCT_SOCKADDR_IN_SIN_LEN */
-
-	/* Make gateway. */
-	for (ALL_NEXTHOPS(re->ng, nexthop)) {
-		if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE))
-			continue;
-
-		gate = 0;
-		char gate_buf[INET_ADDRSTRLEN] = "NULL";
-
-		/*
-		 * XXX We need to refrain from kernel operations in some cases,
-		 * but this if statement seems overly cautious - what about
-		 * other than ADD and DELETE?
-		 */
-		if ((cmd == RTM_ADD && NEXTHOP_IS_ACTIVE(nexthop->flags))
-		    || (cmd == RTM_DELETE
-			&& CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB))) {
-			if (nexthop->type == NEXTHOP_TYPE_IPV4
-			    || nexthop->type == NEXTHOP_TYPE_IPV4_IFINDEX) {
-				sin_gate.sin_addr = nexthop->gate.ipv4;
-				gate = 1;
-			}
-			if (nexthop->type == NEXTHOP_TYPE_IFINDEX
-			    || nexthop->type == NEXTHOP_TYPE_IPV4_IFINDEX)
-				ifindex = nexthop->ifindex;
-			if (nexthop->type == NEXTHOP_TYPE_BLACKHOLE) {
-				struct in_addr loopback;
-				loopback.s_addr = htonl(INADDR_LOOPBACK);
-				sin_gate.sin_addr = loopback;
-				bh_type = nexthop->bh_type;
-				gate = 1;
-			}
-
-			if (gate && p->prefixlen == 32)
-				mask = NULL;
-			else {
-				masklen2ip(p->prefixlen, &sin_mask.sin_addr);
-				sin_mask.sin_family = AF_INET;
-#ifdef HAVE_STRUCT_SOCKADDR_IN_SIN_LEN
-				sin_mask.sin_len =
-					sin_masklen(sin_mask.sin_addr);
-#endif /* HAVE_STRUCT_SOCKADDR_IN_SIN_LEN */
-				mask = &sin_mask;
-			}
-
-#ifdef __OpenBSD__
-			if (nexthop->nh_label
-			    && !kernel_rtm_add_labels(nexthop->nh_label,
-						      &smpls))
-				continue;
-			smplsp = (union sockunion *)&smpls;
-#endif
-
-			error = rtm_write(cmd, (union sockunion *)&sin_dest,
-					  (union sockunion *)mask,
-					  gate ? (union sockunion *)&sin_gate
-					       : NULL,
-					  smplsp, ifindex, bh_type, re->metric);
-
-			if (IS_ZEBRA_DEBUG_RIB) {
-				if (!gate) {
-					zlog_debug(
-						"%s: %s: attention! gate not found for re %p",
-						__func__, prefix_buf, re);
-					route_entry_dump(p, NULL, re);
-				} else
-					inet_ntop(AF_INET, &sin_gate.sin_addr,
-						  gate_buf, INET_ADDRSTRLEN);
-			}
-
-			switch (error) {
-			/* We only flag nexthops as being in FIB if rtm_write()
-			 * did its work. */
-			case ZEBRA_ERR_NOERROR:
-				nexthop_num++;
-				if (IS_ZEBRA_DEBUG_RIB)
-					zlog_debug(
-						"%s: %s: successfully did NH %s",
-						__func__, prefix_buf, gate_buf);
-				break;
-
-			/* The only valid case for this error is kernel's
-			 * failure to install
-			 * a multipath route, which is common for FreeBSD. This
-			 * should be
-			 * ignored silently, but logged as an error otherwise.
-			 */
-			case ZEBRA_ERR_RTEXIST:
-				if (cmd != RTM_ADD)
-					zlog_err(
-						"%s: rtm_write() returned %d for command %d",
-						__func__, error, cmd);
-				continue;
-				break;
-
-			/* Given that our NEXTHOP_FLAG_FIB matches real kernel
-			 * FIB, it isn't
-			 * normal to get any other messages in ANY case.
-			 */
-			case ZEBRA_ERR_RTNOEXIST:
-			case ZEBRA_ERR_RTUNREACH:
-			default:
-				zlog_err(
-					"%s: %s: rtm_write() unexpectedly returned %d for command %s",
-					__func__,
-					prefix2str(p, prefix_buf,
-						   sizeof(prefix_buf)),
-					error,
-					lookup_msg(rtm_type_str, cmd, NULL));
-				break;
-			}
-		} /* if (cmd and flags make sense) */
-		else if (IS_ZEBRA_DEBUG_RIB)
-			zlog_debug("%s: odd command %s for flags %d", __func__,
-				   lookup_msg(rtm_type_str, cmd, NULL),
-				   nexthop->flags);
-	} /* for (ALL_NEXTHOPS(...))*/
-
-	/* If there was no useful nexthop, then complain. */
-	if (nexthop_num == 0 && IS_ZEBRA_DEBUG_KERNEL)
-		zlog_debug("%s: No useful nexthops were found in RIB entry %p",
-			   __func__, re);
-
-	return 0; /*XXX*/
-}
-
 #ifdef SIN6_LEN
 /* Calculate sin6_len value for netmask socket value. */
 static int sin6_masklen(struct in6_addr mask)
@@ -273,11 +114,11 @@ static int sin6_masklen(struct in6_addr mask)
 #endif /* SIN6_LEN */
 
 /* Interface between zebra message and rtm message. */
-static int kernel_rtm_ipv6(int cmd, const struct prefix *p,
-			   struct route_entry *re)
+static int kernel_rtm(int cmd, const struct prefix *p,
+		      const struct nexthop_group *ng, uint32_t metric)
+
 {
-	struct sockaddr_in6 *mask;
-	struct sockaddr_in6 sin_dest, sin_mask, sin_gate;
+	union sockunion sin_dest, sin_mask, sin_gate;
 #ifdef __OpenBSD__
 	struct sockaddr_mpls smpls;
 #endif
@@ -287,67 +128,129 @@ static int kernel_rtm_ipv6(int cmd, const struct prefix *p,
 	ifindex_t ifindex = 0;
 	int gate = 0;
 	int error;
+	char prefix_buf[PREFIX_STRLEN];
 	enum blackhole_type bh_type = BLACKHOLE_UNSPEC;
 
-	memset(&sin_dest, 0, sizeof(struct sockaddr_in6));
-	sin_dest.sin6_family = AF_INET6;
-#ifdef SIN6_LEN
-	sin_dest.sin6_len = sizeof(struct sockaddr_in6);
-#endif /* SIN6_LEN */
-	sin_dest.sin6_addr = p->u.prefix6;
+	if (IS_ZEBRA_DEBUG_RIB)
+		prefix2str(p, prefix_buf, sizeof(prefix_buf));
+
+	/*
+	 * We only have the ability to ADD or DELETE at this point
+	 * in time.
+	 */
+	if (cmd != RTM_ADD && cmd != RTM_DELETE) {
+		if (IS_ZEBRA_DEBUG_KERNEL)
+			zlog_debug("%s: %s odd command %s for flags %d",
+				   __func__, prefix_buf,
+				   lookup_msg(rtm_type_str, cmd, NULL),
+				   nexthop->flags);
+		return 0;
+	}
 
-	memset(&sin_mask, 0, sizeof(struct sockaddr_in6));
+	memset(&sin_dest, 0, sizeof(sin_dest));
+	memset(&sin_gate, 0, sizeof(sin_gate));
+	memset(&sin_mask, 0, sizeof(sin_mask));
 
-	memset(&sin_gate, 0, sizeof(struct sockaddr_in6));
-	sin_gate.sin6_family = AF_INET6;
+	switch (p->family) {
+	case AF_INET:
+		sin_dest.sin.sin_family = AF_INET;
+#ifdef HAVE_STRUCT_SOCKADDR_IN_SIN_LEN
+		sin_dest.sin.sin_len = sizeof(sin_dest);
+		sin_gate.sin.sin_len = sizeof(sin_gate);
+#endif /* HAVE_STRUCT_SOCKADDR_IN_SIN_LEN */
+		sin_dest.sin.sin_addr = p->u.prefix4;
+		sin_gate.sin.sin_family = AF_INET;
+		break;
+	case AF_INET6:
+		sin_dest.sin6.sin6_family = AF_INET6;
+#ifdef SIN6_LEN
+		sin_dest.sin6.sin6_len = sizeof(sin_dest);
+#endif /* SIN6_LEN */
+		sin_dest.sin6.sin6_addr = p->u.prefix6;
+		sin_gate.sin6.sin6_family = AF_INET6;
 #ifdef HAVE_STRUCT_SOCKADDR_IN_SIN_LEN
-	sin_gate.sin6_len = sizeof(struct sockaddr_in6);
+		sin_gate.sin6.sin6_len = sizeof(sin_gate);
 #endif /* HAVE_STRUCT_SOCKADDR_IN_SIN_LEN */
+		break;
+	}
 
 	/* Make gateway. */
-	for (ALL_NEXTHOPS(re->ng, nexthop)) {
-		if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE))
+	for (ALL_NEXTHOPS_PTR(ng, nexthop)) {
+		/*
+		 * We only want to use the actual good nexthops
+		 */
+		if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE) ||
+		    !CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE))
 			continue;
 
 		gate = 0;
+		char gate_buf[INET_ADDRSTRLEN] = "NULL";
 
-		if ((cmd == RTM_ADD && NEXTHOP_IS_ACTIVE(nexthop->flags))
-		    || (cmd == RTM_DELETE)) {
-			if (nexthop->type == NEXTHOP_TYPE_IPV6
-			    || nexthop->type == NEXTHOP_TYPE_IPV6_IFINDEX) {
-				sin_gate.sin6_addr = nexthop->gate.ipv6;
-				gate = 1;
-			}
-			if (nexthop->type == NEXTHOP_TYPE_IFINDEX
-			    || nexthop->type == NEXTHOP_TYPE_IPV6_IFINDEX)
-				ifindex = nexthop->ifindex;
-
-			if (nexthop->type == NEXTHOP_TYPE_BLACKHOLE)
-				bh_type = nexthop->bh_type;
-		}
-
-/* Under kame set interface index to link local address. */
+		switch (nexthop->type) {
+		case NEXTHOP_TYPE_IPV4:
+		case NEXTHOP_TYPE_IPV4_IFINDEX:
+			sin_gate.sin.sin_addr = nexthop->gate.ipv4;
+			sin_gate.sin.sin_family = AF_INET;
+			ifindex = nexthop->ifindex;
+			gate = 1;
+			break;
+		case NEXTHOP_TYPE_IPV6:
+		case NEXTHOP_TYPE_IPV6_IFINDEX:
+			sin_gate.sin6.sin6_addr = nexthop->gate.ipv6;
+			sin_gate.sin6.sin6_family = AF_INET6;
+			ifindex = nexthop->ifindex;
+/* Under kame set interface index to link local address */
 #ifdef KAME
 
 #define SET_IN6_LINKLOCAL_IFINDEX(a, i)                                        \
-	do {                                                                   \
-		(a).s6_addr[2] = ((i) >> 8) & 0xff;                            \
-		(a).s6_addr[3] = (i)&0xff;                                     \
-	} while (0)
-
-		if (gate && IN6_IS_ADDR_LINKLOCAL(&sin_gate.sin6_addr))
-			SET_IN6_LINKLOCAL_IFINDEX(sin_gate.sin6_addr, ifindex);
+        do {                                                                   \
+                (a).s6_addr[2] = ((i) >> 8) & 0xff;                            \
+                (a).s6_addr[3] = (i)&0xff;                                     \
+        } while (0)
+
+			if (IN6_IS_ADDR_LINKLOCAL(&sin_gate.sin6.sin6_addr))
+				SET_IN6_LINKLOCAL_IFINDEX(
+					sin_gate.sin6.sin6_addr,
+					ifindex);
 #endif /* KAME */
 
-		if (gate && p->prefixlen == 128)
-			mask = NULL;
-		else {
-			masklen2ip6(p->prefixlen, &sin_mask.sin6_addr);
-			sin_mask.sin6_family = AF_INET6;
+			gate = 1;
+			break;
+		case NEXTHOP_TYPE_IFINDEX:
+			ifindex = nexthop->ifindex;
+			break;
+		case NEXTHOP_TYPE_BLACKHOLE:
+			bh_type = nexthop->bh_type;
+			switch (p->family) {
+			case AFI_IP: {
+				struct in_addr loopback;
+				loopback.s_addr = htonl(INADDR_LOOPBACK);
+				sin_gate.sin.sin_addr = loopback;
+				gate = 1;
+			}
+				break;
+			case AFI_IP6:
+				break;
+			}
+		}
+
+		switch (p->family) {
+		case AF_INET:
+			masklen2ip(p->prefixlen, &sin_mask.sin.sin_addr);
+			sin_mask.sin.sin_family = AF_INET;
+#ifdef HAVE_STRUCT_SOCKADDR_IN_SIN_LEN
+			sin_mask.sin.sin_len = sin_masklen(
+				sin_mask.sin.sin_addr);
+#endif /* HAVE_STRUCT_SOCKADDR_IN_SIN_LEN */
+			break;
+		case AF_INET6:
+			masklen2ip6(p->prefixlen, &sin_mask.sin6.sin6_addr);
+			sin_mask.sin6.sin6_family = AF_INET6;
 #ifdef SIN6_LEN
-			sin_mask.sin6_len = sin6_masklen(sin_mask.sin6_addr);
+			sin_mask.sin6.sin6_len = sin6_masklen(
+				sin_mask.sin6.sin6_addr);
 #endif /* SIN6_LEN */
-			mask = &sin_mask;
+			break;
 		}
 
 #ifdef __OpenBSD__
@@ -356,75 +259,115 @@ static int kernel_rtm_ipv6(int cmd, const struct prefix *p,
 			continue;
 		smplsp = (union sockunion *)&smpls;
 #endif
+		error = rtm_write(cmd, &sin_dest, &sin_mask,
+				  gate ? &sin_gate : NULL, smplsp,
+				  ifindex, bh_type, metric);
+
+		if (IS_ZEBRA_DEBUG_KERNEL) {
+			if (!gate) {
+				zlog_debug("%s: %s: attention! gate not found for re",
+					   __func__, prefix_buf);
+			} else
+				inet_ntop(p->family == AFI_IP ? AF_INET
+					  : AF_INET6,
+					  &sin_gate.sin.sin_addr,
+					  gate_buf, INET_ADDRSTRLEN);
+		}
+		switch (error) {
+			/* We only flag nexthops as being in FIB if
+			 * rtm_write() did its work. */
+		case ZEBRA_ERR_NOERROR:
+			nexthop_num++;
+			if (IS_ZEBRA_DEBUG_KERNEL)
+				zlog_debug("%s: %s: successfully did NH %s",
+					   __func__, prefix_buf, gate_buf);
+			if (cmd == RTM_ADD)
+				SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB);
+			break;
+
+			/* The only valid case for this error is
+			 * kernel's failure to install a multipath
+			 * route, which is common for FreeBSD. This
+			 * should be ignored silently, but logged as an error
+			 * otherwise.
+			 */
+		case ZEBRA_ERR_RTEXIST:
+			if (cmd != RTM_ADD)
+				flog_err(EC_LIB_SYSTEM_CALL,
+					 "%s: rtm_write() returned %d for command %d",
+					 __func__, error, cmd);
+			continue;
 
-		error = rtm_write(cmd, (union sockunion *)&sin_dest,
-				  (union sockunion *)mask,
-				  gate ? (union sockunion *)&sin_gate : NULL,
-				  smplsp, ifindex, bh_type, re->metric);
-		(void)error;
-
-		nexthop_num++;
-	}
+			/* Note any unexpected status returns */
+		default:
+			flog_err(EC_LIB_SYSTEM_CALL,
+				 "%s: %s: rtm_write() unexpectedly returned %d for command %s",
+				 __func__,
+				 prefix2str(p, prefix_buf,
+					    sizeof(prefix_buf)),
+				 error, lookup_msg(rtm_type_str, cmd, NULL));
+			break;
+		}
+	} /* for (ALL_NEXTHOPS(...))*/
 
-	/* If there is no useful nexthop then return. */
+	/* If there was no useful nexthop, then complain. */
 	if (nexthop_num == 0) {
 		if (IS_ZEBRA_DEBUG_KERNEL)
-			zlog_debug("kernel_rtm_ipv6(): No useful nexthop.");
-		return 0;
+			zlog_debug("%s: No useful nexthops were found in RIB prefix %s",
+				   __func__, prefix2str(p, prefix_buf,
+							sizeof(prefix_buf)));
+		return 1;
 	}
 
 	return 0; /*XXX*/
 }
 
-static int kernel_rtm(int cmd, const struct prefix *p, struct route_entry *re)
-{
-	switch (PREFIX_FAMILY(p)) {
-	case AF_INET:
-		return kernel_rtm_ipv4(cmd, p, re);
-	case AF_INET6:
-		return kernel_rtm_ipv6(cmd, p, re);
-	}
-	return 0;
-}
-
-enum dp_req_result kernel_route_rib(struct route_node *rn,
-				    const struct prefix *p,
-				    const struct prefix *src_p,
-				    struct route_entry *old,
-				    struct route_entry *new)
+/*
+ * Update or delete a prefix from the kernel,
+ * using info from a dataplane context struct.
+ */
+enum zebra_dplane_result kernel_route_update(struct zebra_dplane_ctx *ctx)
 {
-	int route = 0;
+	enum zebra_dplane_result res = ZEBRA_DPLANE_REQUEST_SUCCESS;
 
-	if (src_p && src_p->prefixlen) {
+	if (dplane_ctx_get_src(ctx) != NULL) {
 		zlog_err("route add: IPv6 sourcedest routes unsupported!");
-		return DP_REQUEST_FAILURE;
+		res = ZEBRA_DPLANE_REQUEST_FAILURE;
+		goto done;
 	}
 
-	if (zserv_privs.change(ZPRIVS_RAISE))
-		zlog_err("Can't raise privileges");
-
-	if (old)
-		route |= kernel_rtm(RTM_DELETE, p, old);
-
-	if (new)
-		route |= kernel_rtm(RTM_ADD, p, new);
-
-	if (zserv_privs.change(ZPRIVS_LOWER))
-		zlog_err("Can't lower privileges");
+	frr_elevate_privs(&zserv_privs) {
+
+		if (dplane_ctx_get_op(ctx) == DPLANE_OP_ROUTE_DELETE)
+			kernel_rtm(RTM_DELETE, dplane_ctx_get_dest(ctx),
+				   dplane_ctx_get_ng(ctx),
+				   dplane_ctx_get_metric(ctx));
+		else if (dplane_ctx_get_op(ctx) == DPLANE_OP_ROUTE_INSTALL)
+			kernel_rtm(RTM_ADD, dplane_ctx_get_dest(ctx),
+				   dplane_ctx_get_ng(ctx),
+				   dplane_ctx_get_metric(ctx));
+		else if (dplane_ctx_get_op(ctx) == DPLANE_OP_ROUTE_UPDATE) {
+			/* Must do delete and add separately -
+			 * no update available
+			 */
+			kernel_rtm(RTM_DELETE, dplane_ctx_get_dest(ctx),
+				   dplane_ctx_get_old_ng(ctx),
+				   dplane_ctx_get_old_metric(ctx));
+
+			kernel_rtm(RTM_ADD, dplane_ctx_get_dest(ctx),
+				   dplane_ctx_get_ng(ctx),
+				   dplane_ctx_get_metric(ctx));
+		} else {
+			zlog_err("Invalid routing socket update op %s (%u)",
+				 dplane_op2str(dplane_ctx_get_op(ctx)),
+				 dplane_ctx_get_op(ctx));
+			res = ZEBRA_DPLANE_REQUEST_FAILURE;
+		}
+	} /* Elevated privs */
 
-	if (new) {
-		kernel_route_rib_pass_fail(
-			rn, p, new,
-			(!route) ? DP_INSTALL_SUCCESS
-				 : DP_INSTALL_FAILURE);
-	} else {
-		kernel_route_rib_pass_fail(rn, p, old,
-					   (!route)
-						   ? DP_DELETE_SUCCESS
-						   : DP_DELETE_FAILURE);
-	}
+done:
 
-	return DP_REQUEST_SUCCESS;
+	return res;
 }
 
 int kernel_neigh_update(int add, int ifindex, uint32_t addr, char *lla,
@@ -450,13 +393,13 @@ int kernel_del_vtep(vni_t vni, struct interface *ifp, struct in_addr *vtep_ip)
 }
 
 int kernel_add_mac(struct interface *ifp, vlanid_t vid, struct ethaddr *mac,
-		   struct in_addr vtep_ip, uint8_t sticky)
+		   struct in_addr vtep_ip, bool sticky)
 {
 	return 0;
 }
 
 int kernel_del_mac(struct interface *ifp, vlanid_t vid, struct ethaddr *mac,
-		   struct in_addr vtep_ip, int local)
+		   struct in_addr vtep_ip)
 {
 	return 0;
 }