When a local peer sends a UDP message to a non-existing port on an
existing remote host, that host will return an ICMP message containing
the error code ICMP_PORT_UNREACH, plus the header and the first eight
bytes of the original message. If the sender socket has been connected,
it uses this message to issue a "Connection Refused" event to the user.

Until now, we have only read such events from the externally facing
socket, but we don't forward them back to the local sender because
we cannot read the ICMP message directly to user space. Because of
this, the local peer will hang and wait for a response that never
arrives.

We now fix this for IPv4 by recreating and forwarding a correct ICMP
message back to the internal sender. We synthesize the message based
on the information in the extended error structure, plus the returned
part of the original message body.

Note that for the sake of completeness, we even produce ICMP messages
for other error codes. We have noticed that at least ICMP_PROT_UNREACH
is propagated as an error event back to the user.

Reviewed-by: David Gibson <david(a)gibson.dropbear.id.au>
Signed-off-by: Jon Maloy <jmaloy(a)redhat.com>
---
v2: - Updated the ICMP creation to use the new function tap_push_uh4().
    - Added logics to find correct flow, depending on origin.
    - All done after feedback from David Gibson.
v3: - Passing parameter 'now' along with call to udp_sock_errs() call.
    - Corrected lookup of flow from listener socket
    - All done after feedback from David Gibson.
v4: - Eliminate any attempt to handle ICMPs from listener sockets.
    - After feedback from David Gibson.
v5: - Small reshuffle in anticipation of following commits.
v6: - Introduced macro ICMP4_MAX_DLEN
v7: - Introduced ASSERT() on data lenght size
    - Added MTU for ICMPv4 ICMP_FRAG_NEEDED messages
---
 tap.c          |  4 +--
 tap.h          |  2 ++
 udp.c          | 78 +++++++++++++++++++++++++++++++++++++++++++-------
 udp_internal.h |  2 +-
 udp_vu.c       |  4 +--
 5 files changed, 75 insertions(+), 15 deletions(-)

diff --git a/tap.c b/tap.c
index d5c8edd..97416ac 100644
--- a/tap.c
+++ b/tap.c
@@ -143,8 +143,8 @@ static void *tap_push_l2h(const struct ctx *c, void *buf, uint16_t
proto)
  *
  * Return: pointer at which to write the packet's payload
  */
-static void *tap_push_ip4h(struct iphdr *ip4h, struct in_addr src,
-			   struct in_addr dst, size_t l4len, uint8_t proto)
+void *tap_push_ip4h(struct iphdr *ip4h, struct in_addr src,
+		    struct in_addr dst, size_t l4len, uint8_t proto)
 {
 	uint16_t l3len = l4len + sizeof(*ip4h);
 
diff --git a/tap.h b/tap.h
index 0b3eb93..37da21d 100644
--- a/tap.h
+++ b/tap.h
@@ -47,6 +47,8 @@ static inline void tap_hdr_update(struct tap_hdr *thdr, size_t l2len)
 void *tap_push_uh4(struct udphdr *uh, struct in_addr src, in_port_t sport,
 		   struct in_addr dst, in_port_t dport,
 		   const void *in, size_t dlen);
+void *tap_push_ip4h(struct iphdr *ip4h, struct in_addr src,
+		    struct in_addr dst, size_t l4len, uint8_t proto);
 void tap_udp4_send(const struct ctx *c, struct in_addr src, in_port_t sport,
 		   struct in_addr dst, in_port_t dport,
 		   const void *in, size_t dlen);
diff --git a/udp.c b/udp.c
index 923cc38..8b53475 100644
--- a/udp.c
+++ b/udp.c
@@ -87,6 +87,7 @@
 #include <netinet/in.h>
 #include <netinet/ip.h>
 #include <netinet/udp.h>
+#include <netinet/ip_icmp.h>
 #include <stdint.h>
 #include <stddef.h>
 #include <string.h>
@@ -112,6 +113,9 @@
 #include "udp_internal.h"
 #include "udp_vu.h"
 
+/* Maximum UDP data to be returned in ICMP messages */
+#define ICMP4_MAX_DLEN 8
+
 /* "Spliced" sockets indexed by bound port (host order) */
 static int udp_splice_ns  [IP_VERSIONS][NUM_PORTS];
 static int udp_splice_init[IP_VERSIONS][NUM_PORTS];
@@ -402,25 +406,72 @@ static void udp_tap_prepare(const struct mmsghdr *mmh,
 	(*tap_iov)[UDP_IOV_PAYLOAD].iov_len = l4len;
 }
 
+/**
+ * udp_send_conn_fail_icmp4() - Construct and send ICMPv4 to local peer
+ * @c:		Execution context
+ * @ee:	Extended error descriptor
+ * @ref:	epoll reference
+ * @in:	First bytes (max 8) of original UDP message body
+ * @dlen:	Length of the read part of original UDP message body
+ */
+static void udp_send_conn_fail_icmp4(const struct ctx *c,
+				     const struct sock_extended_err *ee,
+				     const struct flowside *toside,
+				     void *in, size_t dlen)
+{
+	struct in_addr oaddr = toside->oaddr.v4mapped.a4;
+	struct in_addr eaddr = toside->eaddr.v4mapped.a4;
+	in_port_t eport = toside->eport;
+	in_port_t oport = toside->oport;
+	struct {
+		struct icmphdr icmp4h;
+		struct iphdr ip4h;
+		struct udphdr uh;
+		char data[ICMP4_MAX_DLEN];
+	} __attribute__((packed, aligned(__alignof__(max_align_t)))) msg;
+	size_t msglen = sizeof(msg) - sizeof(msg.data) + dlen;
+
+	ASSERT(dlen <= ICMP4_MAX_DLEN);
+	memset(&msg, 0, sizeof(msg));
+	msg.icmp4h.type = ee->ee_type;
+	msg.icmp4h.code = ee->ee_code;
+	if (ee->ee_type == ICMP_DEST_UNREACH && ee->ee_code == ICMP_FRAG_NEEDED)
+		msg.icmp4h.un.frag.mtu = htons((uint16_t) ee->ee_info);
+
+	/* Reconstruct the original headers as returned in the ICMP message */
+	tap_push_ip4h(&msg.ip4h, eaddr, oaddr, dlen, IPPROTO_UDP);
+	tap_push_uh4(&msg.uh, eaddr, eport, oaddr, oport, in, dlen);
+	memcpy(&msg.data, in, dlen);
+
+	tap_icmp4_send(c, oaddr, eaddr, &msg, msglen);
+}
+
 /**
  * udp_sock_recverr() - Receive and clear an error from a socket
- * @s:		Socket to receive from
+ * @c:		Execution context
+ * @ref:	epoll reference
  *
  * Return: 1 if error received and processed, 0 if no more errors in queue, < 0
  *         if there was an error reading the queue
  *
  * #syscalls recvmsg
  */
-static int udp_sock_recverr(int s)
+static int udp_sock_recverr(const struct ctx *c, union epoll_ref ref)
 {
 	const struct sock_extended_err *ee;
 	const struct cmsghdr *hdr;
 	char buf[CMSG_SPACE(sizeof(*ee))];
+	char data[ICMP4_MAX_DLEN];
+	int s = ref.fd;
+	struct iovec iov = {
+		.iov_base = data,
+		.iov_len = sizeof(data)
+	};
 	struct msghdr mh = {
 		.msg_name = NULL,
 		.msg_namelen = 0,
-		.msg_iov = NULL,
-		.msg_iovlen = 0,
+		.msg_iov = &iov,
+		.msg_iovlen = 1,
 		.msg_control = buf,
 		.msg_controllen = sizeof(buf),
 	};
@@ -450,8 +501,14 @@ static int udp_sock_recverr(int s)
 	}
 
 	ee = (const struct sock_extended_err *)CMSG_DATA(hdr);
+	if (ref.type == EPOLL_TYPE_UDP_REPLY) {
+		flow_sidx_t sidx = flow_sidx_opposite(ref.flowside);
+		const struct flowside *toside = flowside_at_sidx(sidx);
 
-	/* TODO: When possible propagate and otherwise handle errors */
+		udp_send_conn_fail_icmp4(c, ee, toside, data, rc);
+	} else {
+		trace("Ignoring received IP_RECVERR cmsg on listener socket");
+	}
 	debug("%s error on UDP socket %i: %s",
 	      str_ee_origin(ee), s, strerror_(ee->ee_errno));
 
@@ -461,15 +518,16 @@ static int udp_sock_recverr(int s)
 /**
  * udp_sock_errs() - Process errors on a socket
  * @c:		Execution context
- * @s:		Socket to receive from
+ * @ref:	epoll reference
  * @events:	epoll events bitmap
  *
  * Return: Number of errors handled, or < 0 if we have an unrecoverable error
  */
-int udp_sock_errs(const struct ctx *c, int s, uint32_t events)
+int udp_sock_errs(const struct ctx *c, union epoll_ref ref, uint32_t events)
 {
 	unsigned n_err = 0;
 	socklen_t errlen;
+	int s = ref.fd;
 	int rc, err;
 
 	ASSERT(!c->no_udp);
@@ -478,7 +536,7 @@ int udp_sock_errs(const struct ctx *c, int s, uint32_t events)
 		return 0; /* Nothing to do */
 
 	/* Empty the error queue */
-	while ((rc = udp_sock_recverr(s)) > 0)
+	while ((rc = udp_sock_recverr(c, ref)) > 0)
 		n_err += rc;
 
 	if (rc < 0)
@@ -558,7 +616,7 @@ static void udp_buf_listen_sock_handler(const struct ctx *c,
 	const socklen_t sasize = sizeof(udp_meta[0].s_in);
 	int n, i;
 
-	if (udp_sock_errs(c, ref.fd, events) < 0) {
+	if (udp_sock_errs(c, ref, events) < 0) {
 		err("UDP: Unrecoverable error on listening socket:"
 		    " (%s port %hu)", pif_name(ref.udp.pif), ref.udp.port);
 		/* FIXME: what now?  close/re-open socket? */
@@ -661,7 +719,7 @@ static void udp_buf_reply_sock_handler(const struct ctx *c, union
epoll_ref ref,
 
 	from_s = uflow->s[ref.flowside.sidei];
 
-	if (udp_sock_errs(c, from_s, events) < 0) {
+	if (udp_sock_errs(c, ref, events) < 0) {
 		flow_err(uflow, "Unrecoverable error on reply socket");
 		flow_err_details(uflow);
 		udp_flow_close(c, uflow);
diff --git a/udp_internal.h b/udp_internal.h
index cc80e30..3b081f5 100644
--- a/udp_internal.h
+++ b/udp_internal.h
@@ -30,5 +30,5 @@ size_t udp_update_hdr4(struct iphdr *ip4h, struct udp_payload_t *bp,
 size_t udp_update_hdr6(struct ipv6hdr *ip6h, struct udp_payload_t *bp,
                        const struct flowside *toside, size_t dlen,
 		       bool no_udp_csum);
-int udp_sock_errs(const struct ctx *c, int s, uint32_t events);
+int udp_sock_errs(const struct ctx *c, union epoll_ref ref, uint32_t events);
 #endif /* UDP_INTERNAL_H */
diff --git a/udp_vu.c b/udp_vu.c
index 4123510..c26a223 100644
--- a/udp_vu.c
+++ b/udp_vu.c
@@ -227,7 +227,7 @@ void udp_vu_listen_sock_handler(const struct ctx *c, union epoll_ref
ref,
 	struct vu_virtq *vq = &vdev->vq[VHOST_USER_RX_QUEUE];
 	int i;
 
-	if (udp_sock_errs(c, ref.fd, events) < 0) {
+	if (udp_sock_errs(c, ref, events) < 0) {
 		err("UDP: Unrecoverable error on listening socket:"
 		    " (%s port %hu)", pif_name(ref.udp.pif), ref.udp.port);
 		return;
@@ -302,7 +302,7 @@ void udp_vu_reply_sock_handler(const struct ctx *c, union epoll_ref
ref,
 
 	ASSERT(!c->no_udp);
 
-	if (udp_sock_errs(c, from_s, events) < 0) {
+	if (udp_sock_errs(c, ref, events) < 0) {
 		flow_err(uflow, "Unrecoverable error on reply socket");
 		flow_err_details(uflow);
 		udp_flow_close(c, uflow);
-- 
2.48.1

When a local peer sends a UDP message to a non-existing port on an
existing remote host, that host will return an ICMPv6 message containing
the error code ICMP6_DST_UNREACH_NOPORT, plus the IPv6 header, UDP header
and the first 1232 bytes of the original message, if any. If the sender
socket has been connected, it uses this message to issue a
"Connection Refused" event to the user.

Until now, we have only read such events from the externally facing
socket, but we don't forward them back to the local sender because
we cannot read the ICMP message directly to user space. Because of
this, the local peer will hang and wait for a response that never
arrives.

We now fix this for IPv6 by recreating and forwarding a correct ICMP
message back to the internal sender. We synthesize the message based
on the information in the extended error structure, plus the returned
part of the original message body.

Note that for the sake of completeness, we even produce ICMP messages
for other error types and codes. We have noticed that at least
ICMP_PROT_UNREACH is propagated as an error event back to the user.

Reviewed-by: David Gibson <david(a)gibson.dropbear.id.au>
Signed-off-by: Jon Maloy <jmaloy(a)redhat.com>
---
v2: - Handling all types or ICMP types and codes
    - Returning up to 1232 bytes of user data as per RFC 4884.
    Both suggested by anonymous PASST user.
v3: - Added ASSERT() in the ICMPv6 message creation function.
      Suggested by David Gibson.
---
 tap.c |  8 ++++----
 tap.h |  4 ++++
 udp.c | 60 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++---
 3 files changed, 65 insertions(+), 7 deletions(-)

diff --git a/tap.c b/tap.c
index 8aa47bd..7e4bc00 100644
--- a/tap.c
+++ b/tap.c
@@ -247,10 +247,10 @@ void tap_icmp4_send(const struct ctx *c, struct in_addr src, struct
in_addr dst,
  *
  * Return: pointer at which to write the packet's payload
  */
-static void *tap_push_ip6h(struct ipv6hdr *ip6h,
-			   const struct in6_addr *src,
-			   const struct in6_addr *dst,
-			   size_t l4len, uint8_t proto, uint32_t flow)
+void *tap_push_ip6h(struct ipv6hdr *ip6h,
+		    const struct in6_addr *src,
+		    const struct in6_addr *dst,
+		    size_t l4len, uint8_t proto, uint32_t flow)
 {
 	ip6h->payload_len = htons(l4len);
 	ip6h->priority = 0;
diff --git a/tap.h b/tap.h
index b7b8cef..67aa9e6 100644
--- a/tap.h
+++ b/tap.h
@@ -53,6 +53,10 @@ void *tap_push_uh6(struct udphdr *uh,
 		   void *in, size_t dlen);
 void *tap_push_ip4h(struct iphdr *ip4h, struct in_addr src,
 		    struct in_addr dst, size_t l4len, uint8_t proto);
+void *tap_push_ip6h(struct ipv6hdr *ip6h,
+		    const struct in6_addr *src,
+		    const struct in6_addr *dst,
+		    size_t l4len, uint8_t proto, uint32_t flow);
 void tap_udp4_send(const struct ctx *c, struct in_addr src, in_port_t sport,
 		   struct in_addr dst, in_port_t dport,
 		   const void *in, size_t dlen);
diff --git a/udp.c b/udp.c
index 8b53475..3e17fda 100644
--- a/udp.c
+++ b/udp.c
@@ -88,6 +88,7 @@
 #include <netinet/ip.h>
 #include <netinet/udp.h>
 #include <netinet/ip_icmp.h>
+#include <netinet/icmp6.h>
 #include <stdint.h>
 #include <stddef.h>
 #include <string.h>
@@ -115,6 +116,9 @@
 
 /* Maximum UDP data to be returned in ICMP messages */
 #define ICMP4_MAX_DLEN 8
+#define ICMP6_MAX_DLEN (IPV6_MIN_MTU			\
+			- sizeof(struct udphdr)	\
+			- sizeof(struct ipv6hdr))
 
 /* "Spliced" sockets indexed by bound port (host order) */
 static int udp_splice_ns  [IP_VERSIONS][NUM_PORTS];
@@ -446,6 +450,49 @@ static void udp_send_conn_fail_icmp4(const struct ctx *c,
 	tap_icmp4_send(c, oaddr, eaddr, &msg, msglen);
 }
 
+
+/**
+ * udp_send_conn_fail_icmp6() - Construct and send ICMPv6 to local peer
+ * @c:		Execution context
+ * @ee:	Extended error descriptor
+ * @ref:	epoll reference
+ * @in:	First bytes (max 1232) of original UDP message body
+ * @dlen:	Length of the read part of original UDP message body
+ * @flow:	IPv6 flow identifier
+ */
+static void udp_send_conn_fail_icmp6(const struct ctx *c,
+				     const struct sock_extended_err *ee,
+				     const struct flowside *toside,
+				     void *in, size_t dlen, uint32_t flow)
+{
+	const struct in6_addr *oaddr = &toside->oaddr.a6;
+	const struct in6_addr *eaddr = &toside->eaddr.a6;
+	in_port_t eport = toside->eport;
+	in_port_t oport = toside->oport;
+	struct {
+		struct icmp6_hdr icmp6h;
+		struct ipv6hdr ip6h;
+		struct udphdr uh;
+		char data[ICMP6_MAX_DLEN];
+	} __attribute__((packed, aligned(__alignof__(max_align_t)))) msg;
+	size_t msglen = sizeof(msg) - sizeof(msg.data) + dlen;
+	size_t l4len = dlen + sizeof(struct udphdr);
+
+	ASSERT(dlen <= ICMP6_MAX_DLEN);
+	memset(&msg, 0, sizeof(msg));
+	msg.icmp6h.icmp6_type = ee->ee_type;
+	msg.icmp6h.icmp6_code = ee->ee_code;
+	if (ee->ee_type == ICMP6_PACKET_TOO_BIG)
+		msg.icmp6h.icmp6_dataun.icmp6_un_data32[0] = htonl(ee->ee_info);
+
+	/* Reconstruct the original headers as returned in the ICMP message */
+	tap_push_ip6h(&msg.ip6h, eaddr, oaddr, l4len, IPPROTO_UDP, flow);
+	tap_push_uh6(&msg.uh, eaddr, eport, oaddr, oport, in, dlen);
+	memcpy(&msg.data, in, dlen);
+
+	tap_icmp6_send(c, oaddr, eaddr, &msg, msglen);
+}
+
 /**
  * udp_sock_recverr() - Receive and clear an error from a socket
  * @c:		Execution context
@@ -461,7 +508,7 @@ static int udp_sock_recverr(const struct ctx *c, union epoll_ref ref)
 	const struct sock_extended_err *ee;
 	const struct cmsghdr *hdr;
 	char buf[CMSG_SPACE(sizeof(*ee))];
-	char data[ICMP4_MAX_DLEN];
+	char data[ICMP6_MAX_DLEN];
 	int s = ref.fd;
 	struct iovec iov = {
 		.iov_base = data,
@@ -504,8 +551,15 @@ static int udp_sock_recverr(const struct ctx *c, union epoll_ref
ref)
 	if (ref.type == EPOLL_TYPE_UDP_REPLY) {
 		flow_sidx_t sidx = flow_sidx_opposite(ref.flowside);
 		const struct flowside *toside = flowside_at_sidx(sidx);
-
-		udp_send_conn_fail_icmp4(c, ee, toside, data, rc);
+		size_t dlen = rc;
+
+		if (hdr->cmsg_level == IPPROTO_IP) {
+			dlen = MIN(dlen, ICMP4_MAX_DLEN);
+			udp_send_conn_fail_icmp4(c, ee, toside, data, dlen);
+		} else if (hdr->cmsg_level == IPPROTO_IPV6) {
+			udp_send_conn_fail_icmp6(c, ee, toside, data,
+						 dlen, sidx.flowi);
+		}
 	} else {
 		trace("Ignoring received IP_RECVERR cmsg on listener socket");
 	}
-- 
2.48.1