netlink: Add functionality to copy routes from outer namespace

Instead of just fetching the default gateway and configuring a single
equivalent route in the target namespace, on 'pasta --config-net', it
might be desirable in some cases to copy the whole set of routes
corresponding to a given output interface.

For instance, in:
  https://github.com/containers/podman/issues/18539
  IPv4 Default Route Does Not Propagate to Pasta Containers on Hetzner VPSes

configuring the default gateway won't work without a gateway-less
route (specifying the output interface only), because the default
gateway is, somewhat dubiously, not on the same subnet as the
container.

This is a similar case to the one covered by commit 7656a6f88882
("conf: Adjust netmask on mismatch between IPv4 address/netmask and
gateway"), and I'm not exactly proud of that workaround.

We also have:
  https://bugs.passt.top/show_bug.cgi?id=49
  pasta does not work with tap-style interface

for which, eventually, we should be able to configure a gateway-less
route in the target namespace.

Introduce different operation modes for nl_route(), including a new
NL_DUP one, not exposed yet, which simply parrots back to the kernel
the route dump for a given interface from the outer namespace, fixing
up flags and interface indices on the way, and requesting to add the
same routes in the target namespace, on the interface we manage.

For n routes we want to duplicate, send n identical netlink requests
including the full dump: routes might depend on each other and the
kernel processes RTM_NEWROUTE messages sequentially, not atomically,
and repeating the full dump naturally resolves dependencies without
the need to actually calculate them.

I'm not kidding, it actually works pretty well.

Link: https://github.com/containers/podman/issues/18539
Link: https://bugs.passt.top/show_bug.cgi?id=49
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>

4 files changed, 68 insertions, 22 deletions

diff --git a/conf.c b/conf.c
index 984c3ce..1f6bbef 100644
--- a/conf.c
+++ b/conf.c
@@ -646,7 +646,7 @@ static unsigned int conf_ip4(unsigned int ifi,
 	}
 
 	if (IN4_IS_ADDR_UNSPECIFIED(&ip4->gw))
-		nl_route(0, ifi, AF_INET, &ip4->gw);
+		nl_route(NL_GET, ifi, 0, AF_INET, &ip4->gw);
 
 	if (IN4_IS_ADDR_UNSPECIFIED(&ip4->addr))
 		nl_addr(0, ifi, AF_INET, &ip4->addr, &ip4->prefix_len, NULL);
@@ -718,7 +718,7 @@ static unsigned int conf_ip6(unsigned int ifi,
 	}
 
 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->gw))
-		nl_route(0, ifi, AF_INET6, &ip6->gw);
+		nl_route(NL_GET, ifi, 0, AF_INET6, &ip6->gw);
 
 	nl_addr(0, ifi, AF_INET6,
 		IN6_IS_ADDR_UNSPECIFIED(&ip6->addr) ? &ip6->addr : NULL,
diff --git a/netlink.c b/netlink.c
index c07a13c..d93ecda 100644
--- a/netlink.c
+++ b/netlink.c
@@ -185,16 +185,16 @@ unsigned int nl_get_ext_if(sa_family_t af)
 }
 
 /**
- * nl_route() - Get/set default gateway for given interface and address family
- * @ns:		Use netlink socket in namespace
- * @ifi:	Interface index
+ * nl_route() - Get/set/copy routes for given interface and address family
+ * @op:		Requested operation
+ * @ifi:	Interface index in outer network namespace
+ * @ifi_ns:	Interface index in target namespace for NL_SET, NL_DUP
  * @af:		Address family
- * @gw:		Default gateway to fill if zero, to set if not
+ * @gw:		Default gateway to fill on NL_GET, to set on NL_SET
  */
-void nl_route(int ns, unsigned int ifi, sa_family_t af, void *gw)
+void nl_route(enum nl_op op, unsigned int ifi, unsigned int ifi_ns,
+	      sa_family_t af, void *gw)
 {
-	int set = (af == AF_INET6 && !IN6_IS_ADDR_UNSPECIFIED(gw)) ||
-		  (af == AF_INET && *(uint32_t *)gw);
 	struct req_t {
 		struct nlmsghdr nlh;
 		struct rtmsg rtm;
@@ -215,7 +215,7 @@ void nl_route(int ns, unsigned int ifi, sa_family_t af, void *gw)
 			} r4;
 		} set;
 	} req = {
-		.nlh.nlmsg_type	  = set ? RTM_NEWROUTE : RTM_GETROUTE,
+		.nlh.nlmsg_type	  = op == NL_SET ? RTM_NEWROUTE : RTM_GETROUTE,
 		.nlh.nlmsg_flags  = NLM_F_REQUEST,
 		.nlh.nlmsg_seq	  = nl_seq++,
 
@@ -228,14 +228,15 @@ void nl_route(int ns, unsigned int ifi, sa_family_t af, void *gw)
 		.rta.rta_len	  = RTA_LENGTH(sizeof(unsigned int)),
 		.ifi		  = ifi,
 	};
+	unsigned dup_routes = 0;
+	ssize_t n, nlmsgs_size;
 	struct nlmsghdr *nh;
 	struct rtattr *rta;
-	struct rtmsg *rtm;
 	char buf[NLBUFSIZ];
-	ssize_t n;
+	struct rtmsg *rtm;
 	size_t na;
 
-	if (set) {
+	if (op == NL_SET) {
 		if (af == AF_INET6) {
 			size_t rta_len = RTA_LENGTH(sizeof(req.set.r6.d));
 
@@ -269,31 +270,67 @@ void nl_route(int ns, unsigned int ifi, sa_family_t af, void *gw)
 		req.nlh.nlmsg_flags |= NLM_F_DUMP;
 	}
 
-	if ((n = nl_req(ns, buf, &req, req.nlh.nlmsg_len)) < 0 || set)
+	if ((n = nl_req(op == NL_SET, buf, &req, req.nlh.nlmsg_len)) < 0)
+		return;
+
+	if (op == NL_SET)
 		return;
 
 	nh = (struct nlmsghdr *)buf;
+	nlmsgs_size = n;
+
 	for ( ; NLMSG_OK(nh, n); nh = NLMSG_NEXT(nh, n)) {
 		if (nh->nlmsg_type != RTM_NEWROUTE)
 			goto next;
 
+		if (op == NL_DUP) {
+			nh->nlmsg_seq = nl_seq++;
+			nh->nlmsg_pid = 0;
+			nh->nlmsg_flags &= ~NLM_F_DUMP_FILTERED;
+			nh->nlmsg_flags |= NLM_F_REQUEST | NLM_F_ACK |
+					   NLM_F_CREATE;
+			dup_routes++;
+		}
+
 		rtm = (struct rtmsg *)NLMSG_DATA(nh);
-		if (rtm->rtm_dst_len)
+		if (op == NL_GET && rtm->rtm_dst_len)
 			continue;
 
 		for (rta = RTM_RTA(rtm), na = RTM_PAYLOAD(nh); RTA_OK(rta, na);
 		     rta = RTA_NEXT(rta, na)) {
-			if (rta->rta_type != RTA_GATEWAY)
-				continue;
+			if (op == NL_GET) {
+				if (rta->rta_type != RTA_GATEWAY)
+					continue;
 
-			memcpy(gw, RTA_DATA(rta), RTA_PAYLOAD(rta));
-			return;
+				memcpy(gw, RTA_DATA(rta), RTA_PAYLOAD(rta));
+				return;
+			}
+
+			if (op == NL_DUP && rta->rta_type == RTA_OIF)
+				*(unsigned int *)RTA_DATA(rta) = ifi_ns;
 		}
 
 next:
 		if (nh->nlmsg_type == NLMSG_DONE)
 			break;
 	}
+
+	if (op == NL_DUP) {
+		char resp[NLBUFSIZ];
+		unsigned i;
+
+		nh = (struct nlmsghdr *)buf;
+		/* Routes might have dependencies between each other, and the
+		 * kernel processes RTM_NEWROUTE messages sequentially. For n
+		 * valid routes, we might need to send up to n requests to get
+		 * all of them inserted. Routes that have been already inserted
+		 * won't cause the whole request to fail, so we can simply
+		 * repeat the whole request. This approach avoids the need to
+		 * calculate dependencies: let the kernel do that.
+		 */
+		for (i = 0; i < dup_routes; i++)
+			nl_req(1, resp, nh, nlmsgs_size);
+	}
 }
 
 /**
diff --git a/netlink.h b/netlink.h
index ca4d6ef..217cf1e 100644
--- a/netlink.h
+++ b/netlink.h
@@ -6,9 +6,16 @@
 #ifndef NETLINK_H
 #define NETLINK_H
 
+enum nl_op {
+	NL_GET,
+	NL_SET,
+	NL_DUP,
+};
+
 void nl_sock_init(const struct ctx *c, bool ns);
 unsigned int nl_get_ext_if(sa_family_t af);
-void nl_route(int ns, unsigned int ifi, sa_family_t af, void *gw);
+void nl_route(enum nl_op op, unsigned int ifi, unsigned int ifi_ns,
+	      sa_family_t af, void *gw);
 void nl_addr(int ns, unsigned int ifi, sa_family_t af,
 	     void *addr, int *prefix_len, void *addr_l);
 void nl_link(int ns, unsigned int ifi, void *mac, int up, int mtu);
diff --git a/pasta.c b/pasta.c
index 1ebc868..c9bf1d3 100644
--- a/pasta.c
+++ b/pasta.c
@@ -278,14 +278,16 @@ void pasta_ns_conf(struct ctx *c)
 		if (c->ifi4) {
 			nl_addr(1, c->pasta_ifi, AF_INET, &c->ip4.addr,
 				&c->ip4.prefix_len, NULL);
-			nl_route(1, c->pasta_ifi, AF_INET, &c->ip4.gw);
+			nl_route(NL_SET, c->ifi4, c->pasta_ifi, AF_INET,
+				 &c->ip4.gw);
 		}
 
 		if (c->ifi6) {
 			int prefix_len = 64;
 			nl_addr(1, c->pasta_ifi, AF_INET6, &c->ip6.addr,
 				&prefix_len, NULL);
-			nl_route(1, c->pasta_ifi, AF_INET6, &c->ip6.gw);
+			nl_route(NL_SET, c->ifi6, c->pasta_ifi, AF_INET6,
+				 &c->ip6.gw);
 		}
 	} else {
 		nl_link(1, c->pasta_ifi, c->mac_guest, 0, 0);