Merge branch 'ipv6-defrag-rbtree'

Peter Oskolkov says:

====================
net: IP defrag: use rbtrees in IPv6 defragmentation

Currently, IPv6 defragmentation code drops non-last fragments that
are smaller than 1280 bytes: see
commit 0ed4229b08c1 ("ipv6: defrag: drop non-last frags smaller than min mtu")

This behavior is not specified in IPv6 RFCs and appears to break compatibility
with some IPv6 implementations, as reported here:
https://www.spinics.net/lists/netdev/msg543846.html

This patchset contains four patches:
- patch 1 moves rbtree-related code from IPv4 to files shared b/w
IPv4/IPv6
- patch 2 changes IPv6 defragmenation code to use rbtrees for defrag
queue
- patch 3 changes nf_conntrack IPv6 defragmentation code to use rbtrees
- patch 4 changes ip_defrag selftest to test changes made in the
previous three patches.

Along the way, the 1280-byte restrictions are removed.

I plan to introduce similar changes to 6lowpan defragmentation code
once I figure out how to test it.
====================

Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

2 files changed, 51 insertions, 34 deletions

diff --git a/tools/testing/selftests/net/ip_defrag.c b/tools/testing/selftests/net/ip_defrag.c
index 5d56cc0838f6..c0c9ecb891e1 100644
--- a/tools/testing/selftests/net/ip_defrag.c
+++ b/tools/testing/selftests/net/ip_defrag.c
@@ -20,6 +20,7 @@ static bool		cfg_do_ipv4;
 static bool		cfg_do_ipv6;
 static bool		cfg_verbose;
 static bool		cfg_overlap;
+static bool		cfg_permissive;
 static unsigned short	cfg_port = 9000;
 
 const struct in_addr addr4 = { .s_addr = __constant_htonl(INADDR_LOOPBACK + 2) };
@@ -35,7 +36,7 @@ const struct in6_addr addr6 = IN6ADDR_LOOPBACK_INIT;
 static int payload_len;
 static int max_frag_len;
 
-#define MSG_LEN_MAX	60000	/* Max UDP payload length. */
+#define MSG_LEN_MAX	10000	/* Max UDP payload length. */
 
 #define IP4_MF		(1u << 13)  /* IPv4 MF flag. */
 #define IP6_MF		(1)  /* IPv6 MF flag. */
@@ -59,13 +60,14 @@ static void recv_validate_udp(int fd_udp)
 	msg_counter++;
 
 	if (cfg_overlap) {
-		if (ret != -1)
-			error(1, 0, "recv: expected timeout; got %d",
-				(int)ret);
-		if (errno != ETIMEDOUT && errno != EAGAIN)
-			error(1, errno, "recv: expected timeout: %d",
-				 errno);
-		return;  /* OK */
+		if (ret == -1 && (errno == ETIMEDOUT || errno == EAGAIN))
+			return;  /* OK */
+		if (!cfg_permissive) {
+			if (ret != -1)
+				error(1, 0, "recv: expected timeout; got %d",
+					(int)ret);
+			error(1, errno, "recv: expected timeout: %d", errno);
+		}
 	}
 
 	if (ret == -1)
@@ -203,7 +205,6 @@ static void send_udp_frags(int fd_raw, struct sockaddr *addr,
 {
 	struct ip *iphdr = (struct ip *)ip_frame;
 	struct ip6_hdr *ip6hdr = (struct ip6_hdr *)ip_frame;
-	const bool ipv4 = !ipv6;
 	int res;
 	int offset;
 	int frag_len;
@@ -251,7 +252,7 @@ static void send_udp_frags(int fd_raw, struct sockaddr *addr,
 	}
 
 	/* Occasionally test IPv4 "runs" (see net/ipv4/ip_fragment.c) */
-	if (ipv4 && !cfg_overlap && (rand() % 100 < 20) &&
+	if (!cfg_overlap && (rand() % 100 < 20) &&
 			(payload_len > 9 * max_frag_len)) {
 		offset = 6 * max_frag_len;
 		while (offset < (UDP_HLEN + payload_len)) {
@@ -276,41 +277,38 @@ static void send_udp_frags(int fd_raw, struct sockaddr *addr,
 	while (offset < (UDP_HLEN + payload_len)) {
 		send_fragment(fd_raw, addr, alen, offset, ipv6);
 		/* IPv4 ignores duplicates, so randomly send a duplicate. */
-		if (ipv4 && (1 == rand() % 100))
+		if (rand() % 100 == 1)
 			send_fragment(fd_raw, addr, alen, offset, ipv6);
 		offset += 2 * max_frag_len;
 	}
 
 	if (cfg_overlap) {
-		/* Send an extra random fragment. */
+		/* Send an extra random fragment.
+		 *
+		 * Duplicates and some fragments completely inside
+		 * previously sent fragments are dropped/ignored. So
+		 * random offset and frag_len can result in a dropped
+		 * fragment instead of a dropped queue/packet. Thus we
+		 * hard-code offset and frag_len.
+		 */
+		if (max_frag_len * 4 < payload_len || max_frag_len < 16) {
+			/* not enough payload for random offset and frag_len. */
+			offset = 8;
+			frag_len = UDP_HLEN + max_frag_len;
+		} else {
+			offset = rand() % (payload_len / 2);
+			frag_len = 2 * max_frag_len + 1 + rand() % 256;
+		}
 		if (ipv6) {
 			struct ip6_frag *fraghdr = (struct ip6_frag *)(ip_frame + IP6_HLEN);
 			/* sendto() returns EINVAL if offset + frag_len is too small. */
-			offset = rand() % (UDP_HLEN + payload_len - 1);
-			frag_len = max_frag_len + rand() % 256;
 			/* In IPv6 if !!(frag_len % 8), the fragment is dropped. */
 			frag_len &= ~0x7;
 			fraghdr->ip6f_offlg = htons(offset / 8 | IP6_MF);
 			ip6hdr->ip6_plen = htons(frag_len);
 			frag_len += IP6_HLEN;
 		} else {
-			/* In IPv4, duplicates and some fragments completely inside
-			 * previously sent fragments are dropped/ignored. So
-			 * random offset and frag_len can result in a dropped
-			 * fragment instead of a dropped queue/packet. So we
-			 * hard-code offset and frag_len.
-			 *
-			 * See ade446403bfb ("net: ipv4: do not handle duplicate
-			 * fragments as overlapping").
-			 */
-			if (max_frag_len * 4 < payload_len || max_frag_len < 16) {
-				/* not enough payload to play with random offset and frag_len. */
-				offset = 8;
-				frag_len = IP4_HLEN + UDP_HLEN + max_frag_len;
-			} else {
-				offset = rand() % (payload_len / 2);
-				frag_len = 2 * max_frag_len + 1 + rand() % 256;
-			}
+			frag_len += IP4_HLEN;
 			iphdr->ip_off = htons(offset / 8 | IP4_MF);
 			iphdr->ip_len = htons(frag_len);
 		}
@@ -327,7 +325,7 @@ static void send_udp_frags(int fd_raw, struct sockaddr *addr,
 	while (offset < (UDP_HLEN + payload_len)) {
 		send_fragment(fd_raw, addr, alen, offset, ipv6);
 		/* IPv4 ignores duplicates, so randomly send a duplicate. */
-		if (ipv4 && (1 == rand() % 100))
+		if (rand() % 100 == 1)
 			send_fragment(fd_raw, addr, alen, offset, ipv6);
 		offset += 2 * max_frag_len;
 	}
@@ -342,7 +340,7 @@ static void run_test(struct sockaddr *addr, socklen_t alen, bool ipv6)
 	 */
 	struct timeval tv = { .tv_sec = 1, .tv_usec = 10 };
 	int idx;
-	int min_frag_len = ipv6 ? 1280 : 8;
+	int min_frag_len = 8;
 
 	/* Initialize the payload. */
 	for (idx = 0; idx < MSG_LEN_MAX; ++idx)
@@ -434,7 +432,7 @@ static void parse_opts(int argc, char **argv)
 {
 	int c;
 
-	while ((c = getopt(argc, argv, "46ov")) != -1) {
+	while ((c = getopt(argc, argv, "46opv")) != -1) {
 		switch (c) {
 		case '4':
 			cfg_do_ipv4 = true;
@@ -445,6 +443,9 @@ static void parse_opts(int argc, char **argv)
 		case 'o':
 			cfg_overlap = true;
 			break;
+		case 'p':
+			cfg_permissive = true;
+			break;
 		case 'v':
 			cfg_verbose = true;
 			break;
diff --git a/tools/testing/selftests/net/ip_defrag.sh b/tools/testing/selftests/net/ip_defrag.sh
index 7dd79a9efb17..15d3489ecd9c 100755
--- a/tools/testing/selftests/net/ip_defrag.sh
+++ b/tools/testing/selftests/net/ip_defrag.sh
@@ -20,6 +20,10 @@ setup() {
 	ip netns exec "${NETNS}" sysctl -w net.ipv6.ip6frag_low_thresh=7000000 >/dev/null 2>&1
 	ip netns exec "${NETNS}" sysctl -w net.ipv6.ip6frag_time=1 >/dev/null 2>&1
 
+	ip netns exec "${NETNS}" sysctl -w net.netfilter.nf_conntrack_frag6_high_thresh=9000000 >/dev/null 2>&1
+	ip netns exec "${NETNS}" sysctl -w net.netfilter.nf_conntrack_frag6_low_thresh=7000000  >/dev/null 2>&1
+	ip netns exec "${NETNS}" sysctl -w net.netfilter.nf_conntrack_frag6_timeout=1 >/dev/null 2>&1
+
 	# DST cache can get full with a lot of frags, with GC not keeping up with the test.
 	ip netns exec "${NETNS}" sysctl -w net.ipv6.route.max_size=65536 >/dev/null 2>&1
 }
@@ -43,4 +47,16 @@ ip netns exec "${NETNS}" ./ip_defrag -6
 echo "ipv6 defrag with overlaps"
 ip netns exec "${NETNS}" ./ip_defrag -6o
 
+# insert an nf_conntrack rule so that the codepath in nf_conntrack_reasm.c taken
+ip netns exec "${NETNS}" ip6tables -A INPUT  -m conntrack --ctstate INVALID -j ACCEPT
+
+echo "ipv6 nf_conntrack defrag"
+ip netns exec "${NETNS}" ./ip_defrag -6
+
+echo "ipv6 nf_conntrack defrag with overlaps"
+# netfilter will drop some invalid packets, so we run the test in
+# permissive mode: i.e. pass the test if the packet is correctly assembled
+# even if we sent an overlap
+ip netns exec "${NETNS}" ./ip_defrag -6op
+
 echo "all tests done"