/* * Copyright 2014 The Android Open Source Project * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. * You may obtain a copy of the License at * * http://www.apache.org/licenses/LICENSE-2.0 * * Unless required by applicable law or agreed to in writing, software * distributed under the License is distributed on an "AS IS" BASIS, * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. * See the License for the specific language governing permissions and * limitations under the License. * * clatd_test.cpp - unit tests for clatd */ #include #include #include #include #include #include #include "netutils/ifc.h" #include "tun_interface.h" extern "C" { #include "clatd.h" #include "config.h" #include "getaddr.h" #include "netutils/checksum.h" #include "translate.h" } // For convenience. #define ARRAYSIZE(x) sizeof((x)) / sizeof((x)[0]) using android::net::TunInterface; // Default translation parameters. static const char kIPv4LocalAddr[] = "192.0.0.4"; static const char kIPv6LocalAddr[] = "2001:db8:0:b11::464"; static const char kIPv6PlatSubnet[] = "64:ff9b::"; // clang-format off // Test packet portions. Defined as macros because it's easy to concatenate them to make packets. #define IPV4_HEADER(p, c1, c2) \ 0x45, 0x00, 0, 41, /* Version=4, IHL=5, ToS=0x80, len=41 */ \ 0x00, 0x00, 0x40, 0x00, /* ID=0x0000, flags=IP_DF, offset=0 */ \ 55, (p), (c1), (c2), /* TTL=55, protocol=p, checksum=c1,c2 */ \ 192, 0, 0, 4, /* Src=192.0.0.4 */ \ 8, 8, 8, 8, /* Dst=8.8.8.8 */ #define IPV4_UDP_HEADER IPV4_HEADER(IPPROTO_UDP, 0x73, 0xb0) #define IPV4_ICMP_HEADER IPV4_HEADER(IPPROTO_ICMP, 0x73, 0xc0) #define IPV6_HEADER(p) \ 0x60, 0x00, 0, 0, /* Version=6, tclass=0x00, flowlabel=0 */ \ 0, 21, (p), 55, /* plen=11, nxthdr=p, hlim=55 */ \ 0x20, 0x01, 0x0d, 0xb8, /* Src=2001:db8:0:b11::464 */ \ 0x00, 0x00, 0x0b, 0x11, \ 0x00, 0x00, 0x00, 0x00, \ 0x00, 0x00, 0x04, 0x64, \ 0x00, 0x64, 0xff, 0x9b, /* Dst=64:ff9b::8.8.8.8 */ \ 0x00, 0x00, 0x00, 0x00, \ 0x00, 0x00, 0x00, 0x00, \ 0x08, 0x08, 0x08, 0x08, #define IPV6_UDP_HEADER IPV6_HEADER(IPPROTO_UDP) #define IPV6_ICMPV6_HEADER IPV6_HEADER(IPPROTO_ICMPV6) #define UDP_LEN 21 #define UDP_HEADER \ 0xc8, 0x8b, 0, 53, /* Port 51339->53 */ \ 0x00, UDP_LEN, 0, 0, /* Length 21, checksum empty for now */ #define PAYLOAD 'H', 'e', 'l', 'l', 'o', ' ', 0x4e, 0xb8, 0x96, 0xe7, 0x95, 0x8c, 0x00 #define IPV4_PING \ 0x08, 0x00, 0x88, 0xd0, /* Type 8, code 0, checksum 0x88d0 */ \ 0xd0, 0x0d, 0x00, 0x03, /* ID=0xd00d, seq=3 */ #define IPV6_PING \ 0x80, 0x00, 0xc3, 0x42, /* Type 128, code 0, checksum 0xc342 */ \ 0xd0, 0x0d, 0x00, 0x03, /* ID=0xd00d, seq=3 */ // Macros to return pseudo-headers from packets. #define IPV4_PSEUDOHEADER(ip, tlen) \ ip[12], ip[13], ip[14], ip[15], /* Source address */ \ ip[16], ip[17], ip[18], ip[19], /* Destination address */ \ 0, ip[9], /* 0, protocol */ \ ((tlen) >> 16) & 0xff, (tlen) & 0xff, /* Transport length */ #define IPV6_PSEUDOHEADER(ip6, protocol, tlen) \ ip6[8], ip6[9], ip6[10], ip6[11], /* Source address */ \ ip6[12], ip6[13], ip6[14], ip6[15], \ ip6[16], ip6[17], ip6[18], ip6[19], \ ip6[20], ip6[21], ip6[22], ip6[23], \ ip6[24], ip6[25], ip6[26], ip6[27], /* Destination address */ \ ip6[28], ip6[29], ip6[30], ip6[31], \ ip6[32], ip6[33], ip6[34], ip6[35], \ ip6[36], ip6[37], ip6[38], ip6[39], \ ((tlen) >> 24) & 0xff, /* Transport length */ \ ((tlen) >> 16) & 0xff, \ ((tlen) >> 8) & 0xff, \ (tlen) & 0xff, \ 0, 0, 0, (protocol), // A fragmented DNS request. static const uint8_t kIPv4Frag1[] = { 0x45, 0x00, 0x00, 0x24, 0xfe, 0x47, 0x20, 0x00, 0x40, 0x11, 0x8c, 0x6d, 0xc0, 0x00, 0x00, 0x04, 0x08, 0x08, 0x08, 0x08, 0x14, 0x5d, 0x00, 0x35, 0x00, 0x29, 0x68, 0xbb, 0x50, 0x47, 0x01, 0x00, 0x00, 0x01, 0x00, 0x00 }; static const uint8_t kIPv4Frag2[] = { 0x45, 0x00, 0x00, 0x24, 0xfe, 0x47, 0x20, 0x02, 0x40, 0x11, 0x8c, 0x6b, 0xc0, 0x00, 0x00, 0x04, 0x08, 0x08, 0x08, 0x08, 0x00, 0x00, 0x00, 0x00, 0x04, 0x69, 0x70, 0x76, 0x34, 0x06, 0x67, 0x6f, 0x6f, 0x67, 0x6c, 0x65 }; static const uint8_t kIPv4Frag3[] = { 0x45, 0x00, 0x00, 0x1d, 0xfe, 0x47, 0x00, 0x04, 0x40, 0x11, 0xac, 0x70, 0xc0, 0x00, 0x00, 0x04, 0x08, 0x08, 0x08, 0x08, 0x03, 0x63, 0x6f, 0x6d, 0x00, 0x00, 0x01, 0x00, 0x01 }; static const uint8_t *kIPv4Fragments[] = { kIPv4Frag1, kIPv4Frag2, kIPv4Frag3 }; static const size_t kIPv4FragLengths[] = { sizeof(kIPv4Frag1), sizeof(kIPv4Frag2), sizeof(kIPv4Frag3) }; static const uint8_t kIPv6Frag1[] = { 0x60, 0x00, 0x00, 0x00, 0x00, 0x18, 0x2c, 0x40, 0x20, 0x01, 0x0d, 0xb8, 0x00, 0x00, 0x0b, 0x11, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x04, 0x64, 0x00, 0x64, 0xff, 0x9b, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x08, 0x08, 0x08, 0x08, 0x11, 0x00, 0x00, 0x01, 0x00, 0x00, 0xfe, 0x47, 0x14, 0x5d, 0x00, 0x35, 0x00, 0x29, 0xeb, 0x91, 0x50, 0x47, 0x01, 0x00, 0x00, 0x01, 0x00, 0x00 }; static const uint8_t kIPv6Frag2[] = { 0x60, 0x00, 0x00, 0x00, 0x00, 0x18, 0x2c, 0x40, 0x20, 0x01, 0x0d, 0xb8, 0x00, 0x00, 0x0b, 0x11, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x04, 0x64, 0x00, 0x64, 0xff, 0x9b, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x08, 0x08, 0x08, 0x08, 0x11, 0x00, 0x00, 0x11, 0x00, 0x00, 0xfe, 0x47, 0x00, 0x00, 0x00, 0x00, 0x04, 0x69, 0x70, 0x76, 0x34, 0x06, 0x67, 0x6f, 0x6f, 0x67, 0x6c, 0x65 }; static const uint8_t kIPv6Frag3[] = { 0x60, 0x00, 0x00, 0x00, 0x00, 0x11, 0x2c, 0x40, 0x20, 0x01, 0x0d, 0xb8, 0x00, 0x00, 0x0b, 0x11, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x04, 0x64, 0x00, 0x64, 0xff, 0x9b, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x08, 0x08, 0x08, 0x08, 0x11, 0x00, 0x00, 0x20, 0x00, 0x00, 0xfe, 0x47, 0x03, 0x63, 0x6f, 0x6d, 0x00, 0x00, 0x01, 0x00, 0x01 }; static const uint8_t *kIPv6Fragments[] = { kIPv6Frag1, kIPv6Frag2, kIPv6Frag3 }; static const size_t kIPv6FragLengths[] = { sizeof(kIPv6Frag1), sizeof(kIPv6Frag2), sizeof(kIPv6Frag3) }; static const uint8_t kReassembledIPv4[] = { 0x45, 0x00, 0x00, 0x3d, 0xfe, 0x47, 0x00, 0x00, 0x40, 0x11, 0xac, 0x54, 0xc0, 0x00, 0x00, 0x04, 0x08, 0x08, 0x08, 0x08, 0x14, 0x5d, 0x00, 0x35, 0x00, 0x29, 0x68, 0xbb, 0x50, 0x47, 0x01, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x04, 0x69, 0x70, 0x76, 0x34, 0x06, 0x67, 0x6f, 0x6f, 0x67, 0x6c, 0x65, 0x03, 0x63, 0x6f, 0x6d, 0x00, 0x00, 0x01, 0x00, 0x01 }; // clang-format on // Expected checksums. static const uint32_t kUdpPartialChecksum = 0xd5c8; static const uint32_t kPayloadPartialChecksum = 0x31e9c; static const uint16_t kUdpV4Checksum = 0xd0c7; static const uint16_t kUdpV6Checksum = 0xa74a; uint8_t ip_version(const uint8_t *packet) { uint8_t version = packet[0] >> 4; return version; } int is_ipv4_fragment(struct iphdr *ip) { // A packet is a fragment if its fragment offset is nonzero or if the MF flag is set. return ntohs(ip->frag_off) & (IP_OFFMASK | IP_MF); } int is_ipv6_fragment(struct ip6_hdr *ip6, size_t len) { if (ip6->ip6_nxt != IPPROTO_FRAGMENT) { return 0; } struct ip6_frag *frag = (struct ip6_frag *)(ip6 + 1); return len >= sizeof(*ip6) + sizeof(*frag) && (frag->ip6f_offlg & (IP6F_OFF_MASK | IP6F_MORE_FRAG)); } int ipv4_fragment_offset(struct iphdr *ip) { return ntohs(ip->frag_off) & IP_OFFMASK; } int ipv6_fragment_offset(struct ip6_frag *frag) { return ntohs((frag->ip6f_offlg & IP6F_OFF_MASK) >> 3); } void check_packet(const uint8_t *packet, size_t len, const char *msg) { void *payload; size_t payload_length = 0; uint32_t pseudo_checksum = 0; uint8_t protocol = 0; int version = ip_version(packet); switch (version) { case 4: { struct iphdr *ip = (struct iphdr *)packet; ASSERT_GE(len, sizeof(*ip)) << msg << ": IPv4 packet shorter than IPv4 header\n"; EXPECT_EQ(5, ip->ihl) << msg << ": Unsupported IP header length\n"; EXPECT_EQ(len, ntohs(ip->tot_len)) << msg << ": Incorrect IPv4 length\n"; EXPECT_EQ(0, ip_checksum(ip, sizeof(*ip))) << msg << ": Incorrect IP checksum\n"; protocol = ip->protocol; payload = ip + 1; if (!is_ipv4_fragment(ip)) { payload_length = len - sizeof(*ip); pseudo_checksum = ipv4_pseudo_header_checksum(ip, payload_length); } ASSERT_TRUE(protocol == IPPROTO_TCP || protocol == IPPROTO_UDP || protocol == IPPROTO_ICMP) << msg << ": Unsupported IPv4 protocol " << protocol << "\n"; break; } case 6: { struct ip6_hdr *ip6 = (struct ip6_hdr *)packet; ASSERT_GE(len, sizeof(*ip6)) << msg << ": IPv6 packet shorter than IPv6 header\n"; EXPECT_EQ(len - sizeof(*ip6), htons(ip6->ip6_plen)) << msg << ": Incorrect IPv6 length\n"; if (ip6->ip6_nxt == IPPROTO_FRAGMENT) { struct ip6_frag *frag = (struct ip6_frag *)(ip6 + 1); ASSERT_GE(len, sizeof(*ip6) + sizeof(*frag)) << msg << ": IPv6 fragment: short fragment header\n"; protocol = frag->ip6f_nxt; payload = frag + 1; // Even though the packet has a Fragment header, it might not be a fragment. if (!is_ipv6_fragment(ip6, len)) { payload_length = len - sizeof(*ip6) - sizeof(*frag); } } else { // Since there are no extension headers except Fragment, this must be the payload. protocol = ip6->ip6_nxt; payload = ip6 + 1; payload_length = len - sizeof(*ip6); } ASSERT_TRUE(protocol == IPPROTO_TCP || protocol == IPPROTO_UDP || protocol == IPPROTO_ICMPV6) << msg << ": Unsupported IPv6 next header " << protocol; if (payload_length) { pseudo_checksum = ipv6_pseudo_header_checksum(ip6, payload_length, protocol); } break; } default: FAIL() << msg << ": Unsupported IP version " << version << "\n"; return; } // If we understand the payload, verify the checksum. if (payload_length) { uint16_t checksum; switch (protocol) { case IPPROTO_UDP: case IPPROTO_TCP: case IPPROTO_ICMPV6: checksum = ip_checksum_finish(ip_checksum_add(pseudo_checksum, payload, payload_length)); break; case IPPROTO_ICMP: checksum = ip_checksum(payload, payload_length); break; default: checksum = 0; // Don't check. break; } EXPECT_EQ(0, checksum) << msg << ": Incorrect transport checksum\n"; } if (protocol == IPPROTO_UDP) { struct udphdr *udp = (struct udphdr *)payload; EXPECT_NE(0, udp->check) << msg << ": UDP checksum 0 should be 0xffff"; // If this is not a fragment, check the UDP length field. if (payload_length) { EXPECT_EQ(payload_length, ntohs(udp->len)) << msg << ": Incorrect UDP length\n"; } } } void reassemble_packet(const uint8_t **fragments, const size_t lengths[], int numpackets, uint8_t *reassembled, size_t *reassembled_len, const char *msg) { struct iphdr *ip = nullptr; struct ip6_hdr *ip6 = nullptr; size_t total_length, pos = 0; uint8_t protocol = 0; uint8_t version = ip_version(fragments[0]); for (int i = 0; i < numpackets; i++) { const uint8_t *packet = fragments[i]; int len = lengths[i]; int headersize, payload_offset; ASSERT_EQ(ip_version(packet), version) << msg << ": Inconsistent fragment versions\n"; check_packet(packet, len, "Fragment sanity check"); switch (version) { case 4: { struct iphdr *ip_orig = (struct iphdr *)packet; headersize = sizeof(*ip_orig); ASSERT_TRUE(is_ipv4_fragment(ip_orig)) << msg << ": IPv4 fragment #" << i + 1 << " not a fragment\n"; ASSERT_EQ(pos, ipv4_fragment_offset(ip_orig) * 8 + ((i != 0) ? sizeof(*ip) : 0)) << msg << ": IPv4 fragment #" << i + 1 << ": inconsistent offset\n"; headersize = sizeof(*ip_orig); payload_offset = headersize; if (pos == 0) { ip = (struct iphdr *)reassembled; } break; } case 6: { struct ip6_hdr *ip6_orig = (struct ip6_hdr *)packet; struct ip6_frag *frag = (struct ip6_frag *)(ip6_orig + 1); ASSERT_TRUE(is_ipv6_fragment(ip6_orig, len)) << msg << ": IPv6 fragment #" << i + 1 << " not a fragment\n"; ASSERT_EQ(pos, ipv6_fragment_offset(frag) * 8 + ((i != 0) ? sizeof(*ip6) : 0)) << msg << ": IPv6 fragment #" << i + 1 << ": inconsistent offset\n"; headersize = sizeof(*ip6_orig); payload_offset = sizeof(*ip6_orig) + sizeof(*frag); if (pos == 0) { ip6 = (struct ip6_hdr *)reassembled; protocol = frag->ip6f_nxt; } break; } default: FAIL() << msg << ": Invalid IP version << " << version; } // If this is the first fragment, copy the header. if (pos == 0) { ASSERT_LT(headersize, (int)*reassembled_len) << msg << ": Reassembly buffer too small\n"; memcpy(reassembled, packet, headersize); total_length = headersize; pos += headersize; } // Copy the payload. int payload_length = len - payload_offset; total_length += payload_length; ASSERT_LT(total_length, *reassembled_len) << msg << ": Reassembly buffer too small\n"; memcpy(reassembled + pos, packet + payload_offset, payload_length); pos += payload_length; } // Fix up the reassembled headers to reflect fragmentation and length (and IPv4 checksum). ASSERT_EQ(total_length, pos) << msg << ": Reassembled packet length incorrect\n"; if (ip) { ip->frag_off &= ~htons(IP_MF); ip->tot_len = htons(total_length); ip->check = 0; ip->check = ip_checksum(ip, sizeof(*ip)); ASSERT_FALSE(is_ipv4_fragment(ip)) << msg << ": reassembled IPv4 packet is a fragment!\n"; } if (ip6) { ip6->ip6_nxt = protocol; ip6->ip6_plen = htons(total_length - sizeof(*ip6)); ASSERT_FALSE(is_ipv6_fragment(ip6, ip6->ip6_plen)) << msg << ": reassembled IPv6 packet is a fragment!\n"; } *reassembled_len = total_length; } void check_data_matches(const void *expected, const void *actual, size_t len, const char *msg) { if (memcmp(expected, actual, len)) { // Hex dump, 20 bytes per line, one space between bytes (1 byte = 3 chars), indented by 4. int hexdump_len = len * 3 + (len / 20 + 1) * 5; char expected_hexdump[hexdump_len], actual_hexdump[hexdump_len]; unsigned pos = 0; for (unsigned i = 0; i < len; i++) { if (i % 20 == 0) { snprintf(expected_hexdump + pos, hexdump_len - pos, "\n "); snprintf(actual_hexdump + pos, hexdump_len - pos, "\n "); pos += 4; } snprintf(expected_hexdump + pos, hexdump_len - pos, " %02x", ((uint8_t *)expected)[i]); snprintf(actual_hexdump + pos, hexdump_len - pos, " %02x", ((uint8_t *)actual)[i]); pos += 3; } FAIL() << msg << ": Data doesn't match" << "\n Expected:" << (char *) expected_hexdump << "\n Actual:" << (char *) actual_hexdump << "\n"; } } void fix_udp_checksum(uint8_t *packet) { uint32_t pseudo_checksum; uint8_t version = ip_version(packet); struct udphdr *udp; switch (version) { case 4: { struct iphdr *ip = (struct iphdr *)packet; udp = (struct udphdr *)(ip + 1); pseudo_checksum = ipv4_pseudo_header_checksum(ip, ntohs(udp->len)); break; } case 6: { struct ip6_hdr *ip6 = (struct ip6_hdr *)packet; udp = (struct udphdr *)(ip6 + 1); pseudo_checksum = ipv6_pseudo_header_checksum(ip6, ntohs(udp->len), IPPROTO_UDP); break; } default: FAIL() << "unsupported IP version" << version << "\n"; return; } udp->check = 0; udp->check = ip_checksum_finish(ip_checksum_add(pseudo_checksum, udp, ntohs(udp->len))); } // Testing stub for send_rawv6. The real version uses sendmsg() with a // destination IPv6 address, and attempting to call that on our test socketpair // fd results in EINVAL. extern "C" void send_rawv6(int fd, clat_packet out, int iov_len) { writev(fd, out, iov_len); } void do_translate_packet(const uint8_t *original, size_t original_len, uint8_t *out, size_t *outlen, const char *msg) { int fds[2]; if (socketpair(AF_UNIX, SOCK_DGRAM | SOCK_NONBLOCK, 0, fds)) { abort(); } char foo[512]; snprintf(foo, sizeof(foo), "%s: Invalid original packet", msg); check_packet(original, original_len, foo); int read_fd, write_fd; uint16_t expected_proto; int version = ip_version(original); switch (version) { case 4: expected_proto = htons(ETH_P_IPV6); read_fd = fds[1]; write_fd = fds[0]; break; case 6: expected_proto = htons(ETH_P_IP); read_fd = fds[0]; write_fd = fds[1]; break; default: FAIL() << msg << ": Unsupported IP version " << version << "\n"; break; } translate_packet(write_fd, (version == 4), original, original_len); snprintf(foo, sizeof(foo), "%s: Invalid translated packet", msg); if (version == 6) { // Translating to IPv4. Expect a tun header. struct tun_pi new_tun_header; struct iovec iov[] = { { &new_tun_header, sizeof(new_tun_header) }, { out, *outlen }, }; int len = readv(read_fd, iov, 2); if (len > (int)sizeof(new_tun_header)) { ASSERT_LT((size_t)len, *outlen) << msg << ": Translated packet buffer too small\n"; EXPECT_EQ(expected_proto, new_tun_header.proto) << msg << "Unexpected tun proto\n"; *outlen = len - sizeof(new_tun_header); check_packet(out, *outlen, msg); } else { FAIL() << msg << ": Packet was not translated: len=" << len; *outlen = 0; } } else { // Translating to IPv6. Expect raw packet. *outlen = read(read_fd, out, *outlen); check_packet(out, *outlen, msg); } } void check_translated_packet(const uint8_t *original, size_t original_len, const uint8_t *expected, size_t expected_len, const char *msg) { uint8_t translated[MAXMTU]; size_t translated_len = sizeof(translated); do_translate_packet(original, original_len, translated, &translated_len, msg); EXPECT_EQ(expected_len, translated_len) << msg << ": Translated packet length incorrect\n"; check_data_matches(expected, translated, translated_len, msg); } void check_fragment_translation(const uint8_t *original[], const size_t original_lengths[], const uint8_t *expected[], const size_t expected_lengths[], int numfragments, const char *msg) { for (int i = 0; i < numfragments; i++) { // Check that each of the fragments translates as expected. char frag_msg[512]; snprintf(frag_msg, sizeof(frag_msg), "%s: fragment #%d", msg, i + 1); check_translated_packet(original[i], original_lengths[i], expected[i], expected_lengths[i], frag_msg); } // Sanity check that reassembling the original and translated fragments produces valid packets. uint8_t reassembled[MAXMTU]; size_t reassembled_len = sizeof(reassembled); reassemble_packet(original, original_lengths, numfragments, reassembled, &reassembled_len, msg); check_packet(reassembled, reassembled_len, msg); uint8_t translated[MAXMTU]; size_t translated_len = sizeof(translated); do_translate_packet(reassembled, reassembled_len, translated, &translated_len, msg); check_packet(translated, translated_len, msg); } int get_transport_checksum(const uint8_t *packet) { struct iphdr *ip; struct ip6_hdr *ip6; uint8_t protocol; const void *payload; int version = ip_version(packet); switch (version) { case 4: ip = (struct iphdr *)packet; if (is_ipv4_fragment(ip)) { return -1; } protocol = ip->protocol; payload = ip + 1; break; case 6: ip6 = (struct ip6_hdr *)packet; protocol = ip6->ip6_nxt; payload = ip6 + 1; break; default: return -1; } switch (protocol) { case IPPROTO_UDP: return ((struct udphdr *)payload)->check; case IPPROTO_TCP: return ((struct tcphdr *)payload)->check; case IPPROTO_FRAGMENT: default: return -1; } } static tun_data makeTunData() { // Create some fake but realistic-looking sockets so update_clat_ipv6_address doesn't balk. return { .read_fd6 = socket(AF_PACKET, SOCK_DGRAM, htons(ETH_P_IPV6)), .write_fd6 = socket(AF_INET6, SOCK_RAW | SOCK_NONBLOCK, IPPROTO_RAW), .fd4 = socket(AF_UNIX, SOCK_DGRAM, 0), }; } void freeTunData(tun_data *tunnel) { close(tunnel->write_fd6); close(tunnel->read_fd6); close(tunnel->fd4); } struct clat_config Global_Clatd_Config; class ClatdTest : public ::testing::Test { protected: static TunInterface sTun; virtual void SetUp() { inet_pton(AF_INET, kIPv4LocalAddr, &Global_Clatd_Config.ipv4_local_subnet); inet_pton(AF_INET6, kIPv6PlatSubnet, &Global_Clatd_Config.plat_subnet); memset(&Global_Clatd_Config.ipv6_local_subnet, 0, sizeof(in6_addr)); Global_Clatd_Config.native_ipv6_interface = const_cast(sTun.name().c_str()); } // Static because setting up the tun interface takes about 40ms. static void SetUpTestCase() { ASSERT_EQ(0, sTun.init()); } // Closing the socket removes the interface and IP addresses. static void TearDownTestCase() { sTun.destroy(); } }; TunInterface ClatdTest::sTun; void expect_ipv6_addr_equal(struct in6_addr *expected, struct in6_addr *actual) { if (!IN6_ARE_ADDR_EQUAL(expected, actual)) { char expected_str[INET6_ADDRSTRLEN], actual_str[INET6_ADDRSTRLEN]; inet_ntop(AF_INET6, expected, expected_str, sizeof(expected_str)); inet_ntop(AF_INET6, actual, actual_str, sizeof(actual_str)); FAIL() << "Unexpected IPv6 address:: " << "\n Expected: " << expected_str << "\n Actual: " << actual_str << "\n"; } } TEST_F(ClatdTest, TestIPv6PrefixEqual) { EXPECT_TRUE(ipv6_prefix_equal(&Global_Clatd_Config.plat_subnet, &Global_Clatd_Config.plat_subnet)); EXPECT_FALSE(ipv6_prefix_equal(&Global_Clatd_Config.plat_subnet, &Global_Clatd_Config.ipv6_local_subnet)); struct in6_addr subnet2 = Global_Clatd_Config.ipv6_local_subnet; EXPECT_TRUE(ipv6_prefix_equal(&Global_Clatd_Config.ipv6_local_subnet, &subnet2)); EXPECT_TRUE(ipv6_prefix_equal(&subnet2, &Global_Clatd_Config.ipv6_local_subnet)); subnet2.s6_addr[6] = 0xff; EXPECT_FALSE(ipv6_prefix_equal(&Global_Clatd_Config.ipv6_local_subnet, &subnet2)); EXPECT_FALSE(ipv6_prefix_equal(&subnet2, &Global_Clatd_Config.ipv6_local_subnet)); } TEST_F(ClatdTest, DetectMtu) { // ::1 with bottom 32 bits set to 1 is still ::1 which routes via lo with mtu of 64KiB ASSERT_EQ(detect_mtu(&in6addr_loopback, htonl(1), 0 /*MARK_UNSET*/), 65536); } TEST_F(ClatdTest, ConfigureTunIpManual) { // Create an interface for configure_tun_ip to configure and bring up. TunInterface v4Iface; ASSERT_EQ(0, v4Iface.init()); struct tun_data tunnel = makeTunData(); strlcpy(tunnel.device4, v4Iface.name().c_str(), sizeof(tunnel.device4)); configure_tun_ip(&tunnel, "192.0.2.1" /* v4_addr */, 1472); EXPECT_EQ(inet_addr("192.0.2.1"), Global_Clatd_Config.ipv4_local_subnet.s_addr); union anyip *ip = getinterface_ip(v4Iface.name().c_str(), AF_INET); ASSERT_NE(nullptr, ip); EXPECT_EQ(inet_addr("192.0.2.1"), ip->ip4.s_addr); free(ip); v4Iface.destroy(); } TEST_F(ClatdTest, DataSanitycheck) { // Sanity checks the data. uint8_t v4_header[] = { IPV4_UDP_HEADER }; ASSERT_EQ(sizeof(struct iphdr), sizeof(v4_header)) << "Test IPv4 header: incorrect length\n"; uint8_t v6_header[] = { IPV6_UDP_HEADER }; ASSERT_EQ(sizeof(struct ip6_hdr), sizeof(v6_header)) << "Test IPv6 header: incorrect length\n"; uint8_t udp_header[] = { UDP_HEADER }; ASSERT_EQ(sizeof(struct udphdr), sizeof(udp_header)) << "Test UDP header: incorrect length\n"; // Sanity checks check_packet. struct udphdr *udp; uint8_t v4_udp_packet[] = { IPV4_UDP_HEADER UDP_HEADER PAYLOAD }; udp = (struct udphdr *)(v4_udp_packet + sizeof(struct iphdr)); fix_udp_checksum(v4_udp_packet); ASSERT_EQ(kUdpV4Checksum, udp->check) << "UDP/IPv4 packet checksum sanity check\n"; check_packet(v4_udp_packet, sizeof(v4_udp_packet), "UDP/IPv4 packet sanity check"); uint8_t v6_udp_packet[] = { IPV6_UDP_HEADER UDP_HEADER PAYLOAD }; udp = (struct udphdr *)(v6_udp_packet + sizeof(struct ip6_hdr)); fix_udp_checksum(v6_udp_packet); ASSERT_EQ(kUdpV6Checksum, udp->check) << "UDP/IPv6 packet checksum sanity check\n"; check_packet(v6_udp_packet, sizeof(v6_udp_packet), "UDP/IPv6 packet sanity check"); uint8_t ipv4_ping[] = { IPV4_ICMP_HEADER IPV4_PING PAYLOAD }; check_packet(ipv4_ping, sizeof(ipv4_ping), "IPv4 ping sanity check"); uint8_t ipv6_ping[] = { IPV6_ICMPV6_HEADER IPV6_PING PAYLOAD }; check_packet(ipv6_ping, sizeof(ipv6_ping), "IPv6 ping sanity check"); // Sanity checks reassemble_packet. uint8_t reassembled[MAXMTU]; size_t total_length = sizeof(reassembled); reassemble_packet(kIPv4Fragments, kIPv4FragLengths, ARRAYSIZE(kIPv4Fragments), reassembled, &total_length, "Reassembly sanity check"); check_packet(reassembled, total_length, "IPv4 Reassembled packet is valid"); ASSERT_EQ(sizeof(kReassembledIPv4), total_length) << "IPv4 reassembly sanity check: length\n"; ASSERT_TRUE(!is_ipv4_fragment((struct iphdr *)reassembled)) << "Sanity check: reassembled packet is a fragment!\n"; check_data_matches(kReassembledIPv4, reassembled, total_length, "IPv4 reassembly sanity check"); total_length = sizeof(reassembled); reassemble_packet(kIPv6Fragments, kIPv6FragLengths, ARRAYSIZE(kIPv6Fragments), reassembled, &total_length, "IPv6 reassembly sanity check"); ASSERT_TRUE(!is_ipv6_fragment((struct ip6_hdr *)reassembled, total_length)) << "Sanity check: reassembled packet is a fragment!\n"; check_packet(reassembled, total_length, "IPv6 Reassembled packet is valid"); } TEST_F(ClatdTest, PseudoChecksum) { uint32_t pseudo_checksum; uint8_t v4_header[] = { IPV4_UDP_HEADER }; uint8_t v4_pseudo_header[] = { IPV4_PSEUDOHEADER(v4_header, UDP_LEN) }; pseudo_checksum = ipv4_pseudo_header_checksum((struct iphdr *)v4_header, UDP_LEN); EXPECT_EQ(ip_checksum_finish(pseudo_checksum), ip_checksum(v4_pseudo_header, sizeof(v4_pseudo_header))) << "ipv4_pseudo_header_checksum incorrect\n"; uint8_t v6_header[] = { IPV6_UDP_HEADER }; uint8_t v6_pseudo_header[] = { IPV6_PSEUDOHEADER(v6_header, IPPROTO_UDP, UDP_LEN) }; pseudo_checksum = ipv6_pseudo_header_checksum((struct ip6_hdr *)v6_header, UDP_LEN, IPPROTO_UDP); EXPECT_EQ(ip_checksum_finish(pseudo_checksum), ip_checksum(v6_pseudo_header, sizeof(v6_pseudo_header))) << "ipv6_pseudo_header_checksum incorrect\n"; } TEST_F(ClatdTest, TransportChecksum) { uint8_t udphdr[] = { UDP_HEADER }; uint8_t payload[] = { PAYLOAD }; EXPECT_EQ(kUdpPartialChecksum, ip_checksum_add(0, udphdr, sizeof(udphdr))) << "UDP partial checksum\n"; EXPECT_EQ(kPayloadPartialChecksum, ip_checksum_add(0, payload, sizeof(payload))) << "Payload partial checksum\n"; uint8_t ip[] = { IPV4_UDP_HEADER }; uint8_t ip6[] = { IPV6_UDP_HEADER }; uint32_t ipv4_pseudo_sum = ipv4_pseudo_header_checksum((struct iphdr *)ip, UDP_LEN); uint32_t ipv6_pseudo_sum = ipv6_pseudo_header_checksum((struct ip6_hdr *)ip6, UDP_LEN, IPPROTO_UDP); EXPECT_NE(0, ipv4_pseudo_sum); EXPECT_NE(0, ipv6_pseudo_sum); EXPECT_EQ(0x3ad0U, ipv4_pseudo_sum % 0xFFFF) << "IPv4 pseudo-checksum sanity check\n"; EXPECT_EQ(0x644dU, ipv6_pseudo_sum % 0xFFFF) << "IPv6 pseudo-checksum sanity check\n"; EXPECT_EQ( kUdpV4Checksum, ip_checksum_finish(ipv4_pseudo_sum + kUdpPartialChecksum + kPayloadPartialChecksum)) << "Unexpected UDP/IPv4 checksum\n"; EXPECT_EQ( kUdpV6Checksum, ip_checksum_finish(ipv6_pseudo_sum + kUdpPartialChecksum + kPayloadPartialChecksum)) << "Unexpected UDP/IPv6 checksum\n"; EXPECT_EQ(kUdpV6Checksum, ip_checksum_adjust(kUdpV4Checksum, ipv4_pseudo_sum, ipv6_pseudo_sum)) << "Adjust IPv4/UDP checksum to IPv6\n"; EXPECT_EQ(kUdpV4Checksum, ip_checksum_adjust(kUdpV6Checksum, ipv6_pseudo_sum, ipv4_pseudo_sum)) << "Adjust IPv6/UDP checksum to IPv4\n"; } TEST_F(ClatdTest, AdjustChecksum) { struct checksum_data { uint16_t checksum; uint32_t old_hdr_sum; uint32_t new_hdr_sum; uint16_t result; } DATA[] = { { 0x1423, 0xb8ec, 0x2d757, 0xf5b5 }, { 0xf5b5, 0x2d757, 0xb8ec, 0x1423 }, { 0xdd2f, 0x5555, 0x3285, 0x0000 }, { 0x1215, 0x5560, 0x15560 + 20, 0x1200 }, { 0xd0c7, 0x3ad0, 0x2644b, 0xa74a }, }; unsigned i = 0; for (i = 0; i < ARRAYSIZE(DATA); i++) { struct checksum_data *data = DATA + i; uint16_t result = ip_checksum_adjust(data->checksum, data->old_hdr_sum, data->new_hdr_sum); EXPECT_EQ(result, data->result) << "Incorrect checksum" << std::showbase << std::hex << "\n Expected: " << data->result << "\n Actual: " << result << "\n checksum=" << data->checksum << " old_sum=" << data->old_hdr_sum << " new_sum=" << data->new_hdr_sum << "\n"; } } TEST_F(ClatdTest, Translate) { // This test uses hardcoded packets so the clatd address must be fixed. inet_pton(AF_INET6, kIPv6LocalAddr, &Global_Clatd_Config.ipv6_local_subnet); uint8_t udp_ipv4[] = { IPV4_UDP_HEADER UDP_HEADER PAYLOAD }; uint8_t udp_ipv6[] = { IPV6_UDP_HEADER UDP_HEADER PAYLOAD }; fix_udp_checksum(udp_ipv4); fix_udp_checksum(udp_ipv6); check_translated_packet(udp_ipv4, sizeof(udp_ipv4), udp_ipv6, sizeof(udp_ipv6), "UDP/IPv4 -> UDP/IPv6 translation"); check_translated_packet(udp_ipv6, sizeof(udp_ipv6), udp_ipv4, sizeof(udp_ipv4), "UDP/IPv6 -> UDP/IPv4 translation"); uint8_t ipv4_ping[] = { IPV4_ICMP_HEADER IPV4_PING PAYLOAD }; uint8_t ipv6_ping[] = { IPV6_ICMPV6_HEADER IPV6_PING PAYLOAD }; check_translated_packet(ipv4_ping, sizeof(ipv4_ping), ipv6_ping, sizeof(ipv6_ping), "ICMP->ICMPv6 translation"); check_translated_packet(ipv6_ping, sizeof(ipv6_ping), ipv4_ping, sizeof(ipv4_ping), "ICMPv6->ICMP translation"); } TEST_F(ClatdTest, Fragmentation) { // This test uses hardcoded packets so the clatd address must be fixed. inet_pton(AF_INET6, kIPv6LocalAddr, &Global_Clatd_Config.ipv6_local_subnet); check_fragment_translation(kIPv4Fragments, kIPv4FragLengths, kIPv6Fragments, kIPv6FragLengths, ARRAYSIZE(kIPv4Fragments), "IPv4->IPv6 fragment translation"); check_fragment_translation(kIPv6Fragments, kIPv6FragLengths, kIPv4Fragments, kIPv4FragLengths, ARRAYSIZE(kIPv6Fragments), "IPv6->IPv4 fragment translation"); } // picks a random interface ID that is checksum neutral with the IPv4 address and the NAT64 prefix void gen_random_iid(struct in6_addr *myaddr, struct in_addr *ipv4_local_subnet, struct in6_addr *plat_subnet) { // Fill last 8 bytes of IPv6 address with random bits. arc4random_buf(&myaddr->s6_addr[8], 8); // Make the IID checksum-neutral. That is, make it so that: // checksum(Local IPv4 | Remote IPv4) = checksum(Local IPv6 | Remote IPv6) // in other words (because remote IPv6 = NAT64 prefix | Remote IPv4): // checksum(Local IPv4) = checksum(Local IPv6 | NAT64 prefix) // Do this by adjusting the two bytes in the middle of the IID. uint16_t middlebytes = (myaddr->s6_addr[11] << 8) + myaddr->s6_addr[12]; uint32_t c1 = ip_checksum_add(0, ipv4_local_subnet, sizeof(*ipv4_local_subnet)); uint32_t c2 = ip_checksum_add(0, plat_subnet, sizeof(*plat_subnet)) + ip_checksum_add(0, myaddr, sizeof(*myaddr)); uint16_t delta = ip_checksum_adjust(middlebytes, c1, c2); myaddr->s6_addr[11] = delta >> 8; myaddr->s6_addr[12] = delta & 0xff; } void check_translate_checksum_neutral(const uint8_t *original, size_t original_len, size_t expected_len, const char *msg) { uint8_t translated[MAXMTU]; size_t translated_len = sizeof(translated); do_translate_packet(original, original_len, translated, &translated_len, msg); EXPECT_EQ(expected_len, translated_len) << msg << ": Translated packet length incorrect\n"; // do_translate_packet already checks packets for validity and verifies the checksum. int original_check = get_transport_checksum(original); int translated_check = get_transport_checksum(translated); ASSERT_NE(-1, original_check); ASSERT_NE(-1, translated_check); ASSERT_EQ(original_check, translated_check) << "Not checksum neutral: original and translated checksums differ\n"; } TEST_F(ClatdTest, TranslateChecksumNeutral) { // Generate a random clat IPv6 address and check that translation is checksum-neutral. ASSERT_TRUE(inet_pton(AF_INET6, "2001:db8:1:2:f076:ae99:124e:aa54", &Global_Clatd_Config.ipv6_local_subnet)); gen_random_iid(&Global_Clatd_Config.ipv6_local_subnet, &Global_Clatd_Config.ipv4_local_subnet, &Global_Clatd_Config.plat_subnet); ASSERT_NE(htonl((uint32_t)0x00000464), Global_Clatd_Config.ipv6_local_subnet.s6_addr32[3]); ASSERT_NE((uint32_t)0, Global_Clatd_Config.ipv6_local_subnet.s6_addr32[3]); // Check that translating UDP packets is checksum-neutral. First, IPv4. uint8_t udp_ipv4[] = { IPV4_UDP_HEADER UDP_HEADER PAYLOAD }; fix_udp_checksum(udp_ipv4); check_translate_checksum_neutral(udp_ipv4, sizeof(udp_ipv4), sizeof(udp_ipv4) + 20, "UDP/IPv4 -> UDP/IPv6 checksum neutral"); // Now try IPv6. uint8_t udp_ipv6[] = { IPV6_UDP_HEADER UDP_HEADER PAYLOAD }; // The test packet uses the static IID, not the random IID. Fix up the source address. struct ip6_hdr *ip6 = (struct ip6_hdr *)udp_ipv6; memcpy(&ip6->ip6_src, &Global_Clatd_Config.ipv6_local_subnet, sizeof(ip6->ip6_src)); fix_udp_checksum(udp_ipv6); check_translate_checksum_neutral(udp_ipv4, sizeof(udp_ipv4), sizeof(udp_ipv4) + 20, "UDP/IPv4 -> UDP/IPv6 checksum neutral"); } TEST_F(ClatdTest, GetInterfaceIp) { union anyip *ip = getinterface_ip(sTun.name().c_str(), AF_INET6); ASSERT_NE(nullptr, ip); in6_addr expected = sTun.srcAddr(); in6_addr actual = ip->ip6; expect_ipv6_addr_equal(&expected, &actual); } void expectSocketBound(int ifindex, int sock) { // Check that the packet socket is bound to the interface. We can't check the socket filter // because there is no way to fetch it from the kernel. sockaddr_ll sll; socklen_t len = sizeof(sll); ASSERT_EQ(0, getsockname(sock, reinterpret_cast(&sll), &len)); EXPECT_EQ(htons(ETH_P_IPV6), sll.sll_protocol); EXPECT_EQ(ifindex, sll.sll_ifindex); } TEST_F(ClatdTest, ConfigureIpv6Address) { struct tun_data tunnel = makeTunData(); ASSERT_TRUE(IN6_IS_ADDR_UNSPECIFIED(&Global_Clatd_Config.ipv6_local_subnet)); const char *addrStr = "2001:db8::f00"; in6_addr addr; ASSERT_EQ(1, inet_pton(AF_INET6, addrStr, &addr)); ASSERT_EQ(1, configure_clat_ipv6_address(&tunnel, sTun.name().c_str(), addrStr)); EXPECT_EQ(htonl(0x20010db8), Global_Clatd_Config.ipv6_local_subnet.s6_addr32[0]); EXPECT_EQ(htonl(0x00000000), Global_Clatd_Config.ipv6_local_subnet.s6_addr32[1]); EXPECT_EQ(htonl(0x00000000), Global_Clatd_Config.ipv6_local_subnet.s6_addr32[2]); EXPECT_EQ(htonl(0x00000f00), Global_Clatd_Config.ipv6_local_subnet.s6_addr32[3]); // Check that the packet socket is bound to the interface. We can't check the socket filter // because there is no way to fetch it from the kernel. sockaddr_ll sll; socklen_t len = sizeof(sll); ASSERT_EQ(0, getsockname(tunnel.read_fd6, reinterpret_cast(&sll), &len)); EXPECT_EQ(htons(ETH_P_IPV6), sll.sll_protocol); EXPECT_EQ(sll.sll_ifindex, sTun.ifindex()); expectSocketBound(sTun.ifindex(), tunnel.read_fd6); freeTunData(&tunnel); } TEST_F(ClatdTest, Ipv6AddressChanged) { // Configure the clat IPv6 address. struct tun_data tunnel = { .read_fd6 = socket(AF_PACKET, SOCK_DGRAM, htons(ETH_P_IPV6)), .write_fd6 = socket(AF_INET6, SOCK_RAW | SOCK_NONBLOCK, IPPROTO_RAW), }; const char *ifname = sTun.name().c_str(); in6_addr myaddr = sTun.srcAddr(); gen_random_iid(&myaddr, &Global_Clatd_Config.ipv4_local_subnet, &Global_Clatd_Config.plat_subnet); char addrstr[INET6_ADDRSTRLEN]; ASSERT_NE(nullptr, inet_ntop(AF_INET6, &myaddr, addrstr, sizeof(addrstr))); ASSERT_EQ(1, configure_clat_ipv6_address(&tunnel, ifname, addrstr)); EXPECT_EQ(0, ipv6_address_changed(ifname)); EXPECT_EQ(0, ipv6_address_changed(ifname)); // Change the IP address on the tun interface to a new prefix. char srcaddr[INET6_ADDRSTRLEN]; char dstaddr[INET6_ADDRSTRLEN]; ASSERT_NE(nullptr, inet_ntop(AF_INET6, &sTun.srcAddr(), srcaddr, sizeof(srcaddr))); ASSERT_NE(nullptr, inet_ntop(AF_INET6, &sTun.dstAddr(), dstaddr, sizeof(dstaddr))); EXPECT_EQ(0, ifc_del_address(ifname, srcaddr, 64)); EXPECT_EQ(0, ifc_del_address(ifname, dstaddr, 64)); // Check that we can tell that the address has changed. EXPECT_EQ(0, ifc_add_address(ifname, "2001:db8::1:2", 64)); EXPECT_EQ(1, ipv6_address_changed(ifname)); EXPECT_EQ(1, ipv6_address_changed(ifname)); // Restore the tun interface configuration. sTun.destroy(); ASSERT_EQ(0, sTun.init()); }