From: Jakub Sitnicki Date: Thu, 12 Dec 2019 10:22:58 +0000 (+0100) Subject: selftests/bpf: Move reuseport tests under prog_tests/ X-Git-Url: http://git.lede-project.org./?a=commitdiff_plain;h=415bb4e125b68a3aaccb8381baaa4a06dd1dcb27;p=openwrt%2Fstaging%2Fblogic.git selftests/bpf: Move reuseport tests under prog_tests/ Do a pure move the show the actual work needed to adapt the tests in subsequent patch at the cost of breaking test_progs build for the moment. Signed-off-by: Jakub Sitnicki Signed-off-by: Alexei Starovoitov Link: https://lore.kernel.org/bpf/20191212102259.418536-10-jakub@cloudflare.com --- diff --git a/tools/testing/selftests/bpf/Makefile b/tools/testing/selftests/bpf/Makefile index e0fe01d9ec33..90de7d75b5c6 100644 --- a/tools/testing/selftests/bpf/Makefile +++ b/tools/testing/selftests/bpf/Makefile @@ -28,7 +28,7 @@ LDLIBS += -lcap -lelf -lrt -lpthread TEST_GEN_PROGS = test_verifier test_tag test_maps test_lru_map test_lpm_map test_progs \ test_align test_verifier_log test_dev_cgroup test_tcpbpf_user \ test_sock test_btf test_sockmap get_cgroup_id_user test_socket_cookie \ - test_cgroup_storage test_select_reuseport \ + test_cgroup_storage \ test_netcnt test_tcpnotify_user test_sock_fields test_sysctl test_hashmap \ test_cgroup_attach test_progs-no_alu32 diff --git a/tools/testing/selftests/bpf/prog_tests/select_reuseport.c b/tools/testing/selftests/bpf/prog_tests/select_reuseport.c new file mode 100644 index 000000000000..0d5687feb689 --- /dev/null +++ b/tools/testing/selftests/bpf/prog_tests/select_reuseport.c @@ -0,0 +1,811 @@ +// SPDX-License-Identifier: GPL-2.0 +/* Copyright (c) 2018 Facebook */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include "bpf_rlimit.h" +#include "bpf_util.h" +#include "test_select_reuseport_common.h" + +#define MIN_TCPHDR_LEN 20 +#define UDPHDR_LEN 8 + +#define TCP_SYNCOOKIE_SYSCTL "/proc/sys/net/ipv4/tcp_syncookies" +#define TCP_FO_SYSCTL "/proc/sys/net/ipv4/tcp_fastopen" +#define REUSEPORT_ARRAY_SIZE 32 + +static int result_map, tmp_index_ovr_map, linum_map, data_check_map; +static enum result expected_results[NR_RESULTS]; +static int sk_fds[REUSEPORT_ARRAY_SIZE]; +static int reuseport_array = -1, outer_map = -1; +static int select_by_skb_data_prog; +static int saved_tcp_syncookie; +static struct bpf_object *obj; +static int saved_tcp_fo; +static __u32 index_zero; +static int epfd; + +static union sa46 { + struct sockaddr_in6 v6; + struct sockaddr_in v4; + sa_family_t family; +} srv_sa; + +#define CHECK(condition, tag, format...) ({ \ + int __ret = !!(condition); \ + if (__ret) { \ + printf("%s(%d):FAIL:%s ", __func__, __LINE__, tag); \ + printf(format); \ + exit(-1); \ + } \ +}) + +#define RET_ERR(condition, tag, format...) ({ \ + int __ret = !!(condition); \ + if (__ret) { \ + printf("%s(%d):FAIL:%s ", __func__, __LINE__, tag); \ + printf(format); \ + return -1; \ + } \ +}) + +static int create_maps(void) +{ + struct bpf_create_map_attr attr = {}; + + /* Creating reuseport_array */ + attr.name = "reuseport_array"; + attr.map_type = BPF_MAP_TYPE_REUSEPORT_SOCKARRAY; + attr.key_size = sizeof(__u32); + attr.value_size = sizeof(__u32); + attr.max_entries = REUSEPORT_ARRAY_SIZE; + + reuseport_array = bpf_create_map_xattr(&attr); + RET_ERR(reuseport_array == -1, "creating reuseport_array", + "reuseport_array:%d errno:%d\n", reuseport_array, errno); + + /* Creating outer_map */ + attr.name = "outer_map"; + attr.map_type = BPF_MAP_TYPE_ARRAY_OF_MAPS; + attr.key_size = sizeof(__u32); + attr.value_size = sizeof(__u32); + attr.max_entries = 1; + attr.inner_map_fd = reuseport_array; + outer_map = bpf_create_map_xattr(&attr); + RET_ERR(outer_map == -1, "creating outer_map", + "outer_map:%d errno:%d\n", outer_map, errno); + + return 0; +} + +static int prepare_bpf_obj(void) +{ + struct bpf_program *prog; + struct bpf_map *map; + int err; + + obj = bpf_object__open("test_select_reuseport_kern.o"); + RET_ERR(IS_ERR_OR_NULL(obj), "open test_select_reuseport_kern.o", + "obj:%p PTR_ERR(obj):%ld\n", obj, PTR_ERR(obj)); + + map = bpf_object__find_map_by_name(obj, "outer_map"); + RET_ERR(!map, "find outer_map", "!map\n"); + err = bpf_map__reuse_fd(map, outer_map); + RET_ERR(err, "reuse outer_map", "err:%d\n", err); + + err = bpf_object__load(obj); + RET_ERR(err, "load bpf_object", "err:%d\n", err); + + prog = bpf_program__next(NULL, obj); + RET_ERR(!prog, "get first bpf_program", "!prog\n"); + select_by_skb_data_prog = bpf_program__fd(prog); + RET_ERR(select_by_skb_data_prog == -1, "get prog fd", + "select_by_skb_data_prog:%d\n", select_by_skb_data_prog); + + map = bpf_object__find_map_by_name(obj, "result_map"); + RET_ERR(!map, "find result_map", "!map\n"); + result_map = bpf_map__fd(map); + RET_ERR(result_map == -1, "get result_map fd", + "result_map:%d\n", result_map); + + map = bpf_object__find_map_by_name(obj, "tmp_index_ovr_map"); + RET_ERR(!map, "find tmp_index_ovr_map\n", "!map"); + tmp_index_ovr_map = bpf_map__fd(map); + RET_ERR(tmp_index_ovr_map == -1, "get tmp_index_ovr_map fd", + "tmp_index_ovr_map:%d\n", tmp_index_ovr_map); + + map = bpf_object__find_map_by_name(obj, "linum_map"); + RET_ERR(!map, "find linum_map", "!map\n"); + linum_map = bpf_map__fd(map); + RET_ERR(linum_map == -1, "get linum_map fd", + "linum_map:%d\n", linum_map); + + map = bpf_object__find_map_by_name(obj, "data_check_map"); + RET_ERR(!map, "find data_check_map", "!map\n"); + data_check_map = bpf_map__fd(map); + RET_ERR(data_check_map == -1, "get data_check_map fd", + "data_check_map:%d\n", data_check_map); + + return 0; +} + +static void sa46_init_loopback(union sa46 *sa, sa_family_t family) +{ + memset(sa, 0, sizeof(*sa)); + sa->family = family; + if (sa->family == AF_INET6) + sa->v6.sin6_addr = in6addr_loopback; + else + sa->v4.sin_addr.s_addr = htonl(INADDR_LOOPBACK); +} + +static void sa46_init_inany(union sa46 *sa, sa_family_t family) +{ + memset(sa, 0, sizeof(*sa)); + sa->family = family; + if (sa->family == AF_INET6) + sa->v6.sin6_addr = in6addr_any; + else + sa->v4.sin_addr.s_addr = INADDR_ANY; +} + +static int read_int_sysctl(const char *sysctl) +{ + char buf[16]; + int fd, ret; + + fd = open(sysctl, 0); + RET_ERR(fd == -1, "open(sysctl)", + "sysctl:%s fd:%d errno:%d\n", sysctl, fd, errno); + + ret = read(fd, buf, sizeof(buf)); + RET_ERR(ret <= 0, "read(sysctl)", + "sysctl:%s ret:%d errno:%d\n", sysctl, ret, errno); + + close(fd); + return atoi(buf); +} + +static int write_int_sysctl(const char *sysctl, int v) +{ + int fd, ret, size; + char buf[16]; + + fd = open(sysctl, O_RDWR); + RET_ERR(fd == -1, "open(sysctl)", + "sysctl:%s fd:%d errno:%d\n", sysctl, fd, errno); + + size = snprintf(buf, sizeof(buf), "%d", v); + ret = write(fd, buf, size); + RET_ERR(ret != size, "write(sysctl)", + "sysctl:%s ret:%d size:%d errno:%d\n", + sysctl, ret, size, errno); + + close(fd); + return 0; +} + +static void restore_sysctls(void) +{ + write_int_sysctl(TCP_FO_SYSCTL, saved_tcp_fo); + write_int_sysctl(TCP_SYNCOOKIE_SYSCTL, saved_tcp_syncookie); +} + +static int enable_fastopen(void) +{ + int fo; + + fo = read_int_sysctl(TCP_FO_SYSCTL); + if (fo < 0) + return -1; + + return write_int_sysctl(TCP_FO_SYSCTL, fo | 7); +} + +static int enable_syncookie(void) +{ + return write_int_sysctl(TCP_SYNCOOKIE_SYSCTL, 2); +} + +static int disable_syncookie(void) +{ + return write_int_sysctl(TCP_SYNCOOKIE_SYSCTL, 0); +} + +static __u32 get_linum(void) +{ + __u32 linum; + int err; + + err = bpf_map_lookup_elem(linum_map, &index_zero, &linum); + CHECK(err == -1, "lookup_elem(linum_map)", "err:%d errno:%d\n", + err, errno); + + return linum; +} + +static void check_data(int type, sa_family_t family, const struct cmd *cmd, + int cli_fd) +{ + struct data_check expected = {}, result; + union sa46 cli_sa; + socklen_t addrlen; + int err; + + addrlen = sizeof(cli_sa); + err = getsockname(cli_fd, (struct sockaddr *)&cli_sa, + &addrlen); + CHECK(err == -1, "getsockname(cli_fd)", "err:%d errno:%d\n", + err, errno); + + err = bpf_map_lookup_elem(data_check_map, &index_zero, &result); + CHECK(err == -1, "lookup_elem(data_check_map)", "err:%d errno:%d\n", + err, errno); + + if (type == SOCK_STREAM) { + expected.len = MIN_TCPHDR_LEN; + expected.ip_protocol = IPPROTO_TCP; + } else { + expected.len = UDPHDR_LEN; + expected.ip_protocol = IPPROTO_UDP; + } + + if (family == AF_INET6) { + expected.eth_protocol = htons(ETH_P_IPV6); + expected.bind_inany = !srv_sa.v6.sin6_addr.s6_addr32[3] && + !srv_sa.v6.sin6_addr.s6_addr32[2] && + !srv_sa.v6.sin6_addr.s6_addr32[1] && + !srv_sa.v6.sin6_addr.s6_addr32[0]; + + memcpy(&expected.skb_addrs[0], cli_sa.v6.sin6_addr.s6_addr32, + sizeof(cli_sa.v6.sin6_addr)); + memcpy(&expected.skb_addrs[4], &in6addr_loopback, + sizeof(in6addr_loopback)); + expected.skb_ports[0] = cli_sa.v6.sin6_port; + expected.skb_ports[1] = srv_sa.v6.sin6_port; + } else { + expected.eth_protocol = htons(ETH_P_IP); + expected.bind_inany = !srv_sa.v4.sin_addr.s_addr; + + expected.skb_addrs[0] = cli_sa.v4.sin_addr.s_addr; + expected.skb_addrs[1] = htonl(INADDR_LOOPBACK); + expected.skb_ports[0] = cli_sa.v4.sin_port; + expected.skb_ports[1] = srv_sa.v4.sin_port; + } + + if (memcmp(&result, &expected, offsetof(struct data_check, + equal_check_end))) { + printf("unexpected data_check\n"); + printf(" result: (0x%x, %u, %u)\n", + result.eth_protocol, result.ip_protocol, + result.bind_inany); + printf("expected: (0x%x, %u, %u)\n", + expected.eth_protocol, expected.ip_protocol, + expected.bind_inany); + CHECK(1, "data_check result != expected", + "bpf_prog_linum:%u\n", get_linum()); + } + + CHECK(!result.hash, "data_check result.hash empty", + "result.hash:%u", result.hash); + + expected.len += cmd ? sizeof(*cmd) : 0; + if (type == SOCK_STREAM) + CHECK(expected.len > result.len, "expected.len > result.len", + "expected.len:%u result.len:%u bpf_prog_linum:%u\n", + expected.len, result.len, get_linum()); + else + CHECK(expected.len != result.len, "expected.len != result.len", + "expected.len:%u result.len:%u bpf_prog_linum:%u\n", + expected.len, result.len, get_linum()); +} + +static void check_results(void) +{ + __u32 results[NR_RESULTS]; + __u32 i, broken = 0; + int err; + + for (i = 0; i < NR_RESULTS; i++) { + err = bpf_map_lookup_elem(result_map, &i, &results[i]); + CHECK(err == -1, "lookup_elem(result_map)", + "i:%u err:%d errno:%d\n", i, err, errno); + } + + for (i = 0; i < NR_RESULTS; i++) { + if (results[i] != expected_results[i]) { + broken = i; + break; + } + } + + if (i == NR_RESULTS) + return; + + printf("unexpected result\n"); + printf(" result: ["); + printf("%u", results[0]); + for (i = 1; i < NR_RESULTS; i++) + printf(", %u", results[i]); + printf("]\n"); + + printf("expected: ["); + printf("%u", expected_results[0]); + for (i = 1; i < NR_RESULTS; i++) + printf(", %u", expected_results[i]); + printf("]\n"); + + CHECK(expected_results[broken] != results[broken], + "unexpected result", + "expected_results[%u] != results[%u] bpf_prog_linum:%u\n", + broken, broken, get_linum()); +} + +static int send_data(int type, sa_family_t family, void *data, size_t len, + enum result expected) +{ + union sa46 cli_sa; + int fd, err; + + fd = socket(family, type, 0); + CHECK(fd == -1, "socket()", "fd:%d errno:%d\n", fd, errno); + + sa46_init_loopback(&cli_sa, family); + err = bind(fd, (struct sockaddr *)&cli_sa, sizeof(cli_sa)); + CHECK(fd == -1, "bind(cli_sa)", "err:%d errno:%d\n", err, errno); + + err = sendto(fd, data, len, MSG_FASTOPEN, (struct sockaddr *)&srv_sa, + sizeof(srv_sa)); + CHECK(err != len && expected >= PASS, + "sendto()", "family:%u err:%d errno:%d expected:%d\n", + family, err, errno, expected); + + return fd; +} + +static void do_test(int type, sa_family_t family, struct cmd *cmd, + enum result expected) +{ + int nev, srv_fd, cli_fd; + struct epoll_event ev; + struct cmd rcv_cmd; + ssize_t nread; + + cli_fd = send_data(type, family, cmd, cmd ? sizeof(*cmd) : 0, + expected); + nev = epoll_wait(epfd, &ev, 1, expected >= PASS ? 5 : 0); + CHECK((nev <= 0 && expected >= PASS) || + (nev > 0 && expected < PASS), + "nev <> expected", + "nev:%d expected:%d type:%d family:%d data:(%d, %d)\n", + nev, expected, type, family, + cmd ? cmd->reuseport_index : -1, + cmd ? cmd->pass_on_failure : -1); + check_results(); + check_data(type, family, cmd, cli_fd); + + if (expected < PASS) + return; + + CHECK(expected != PASS_ERR_SK_SELECT_REUSEPORT && + cmd->reuseport_index != ev.data.u32, + "check cmd->reuseport_index", + "cmd:(%u, %u) ev.data.u32:%u\n", + cmd->pass_on_failure, cmd->reuseport_index, ev.data.u32); + + srv_fd = sk_fds[ev.data.u32]; + if (type == SOCK_STREAM) { + int new_fd = accept(srv_fd, NULL, 0); + + CHECK(new_fd == -1, "accept(srv_fd)", + "ev.data.u32:%u new_fd:%d errno:%d\n", + ev.data.u32, new_fd, errno); + + nread = recv(new_fd, &rcv_cmd, sizeof(rcv_cmd), MSG_DONTWAIT); + CHECK(nread != sizeof(rcv_cmd), + "recv(new_fd)", + "ev.data.u32:%u nread:%zd sizeof(rcv_cmd):%zu errno:%d\n", + ev.data.u32, nread, sizeof(rcv_cmd), errno); + + close(new_fd); + } else { + nread = recv(srv_fd, &rcv_cmd, sizeof(rcv_cmd), MSG_DONTWAIT); + CHECK(nread != sizeof(rcv_cmd), + "recv(sk_fds)", + "ev.data.u32:%u nread:%zd sizeof(rcv_cmd):%zu errno:%d\n", + ev.data.u32, nread, sizeof(rcv_cmd), errno); + } + + close(cli_fd); +} + +static void test_err_inner_map(int type, sa_family_t family) +{ + struct cmd cmd = { + .reuseport_index = 0, + .pass_on_failure = 0, + }; + + expected_results[DROP_ERR_INNER_MAP]++; + do_test(type, family, &cmd, DROP_ERR_INNER_MAP); + printf("OK\n"); +} + +static void test_err_skb_data(int type, sa_family_t family) +{ + expected_results[DROP_ERR_SKB_DATA]++; + do_test(type, family, NULL, DROP_ERR_SKB_DATA); + printf("OK\n"); +} + +static void test_err_sk_select_port(int type, sa_family_t family) +{ + struct cmd cmd = { + .reuseport_index = REUSEPORT_ARRAY_SIZE, + .pass_on_failure = 0, + }; + + expected_results[DROP_ERR_SK_SELECT_REUSEPORT]++; + do_test(type, family, &cmd, DROP_ERR_SK_SELECT_REUSEPORT); + printf("OK\n"); +} + +static void test_pass(int type, sa_family_t family) +{ + struct cmd cmd; + int i; + + cmd.pass_on_failure = 0; + for (i = 0; i < REUSEPORT_ARRAY_SIZE; i++) { + expected_results[PASS]++; + cmd.reuseport_index = i; + do_test(type, family, &cmd, PASS); + } + printf("OK\n"); +} + +static void test_syncookie(int type, sa_family_t family) +{ + int err, tmp_index = 1; + struct cmd cmd = { + .reuseport_index = 0, + .pass_on_failure = 0, + }; + + if (type != SOCK_STREAM) + return; + + /* + * +1 for TCP-SYN and + * +1 for the TCP-ACK (ack the syncookie) + */ + expected_results[PASS] += 2; + enable_syncookie(); + /* + * Simulate TCP-SYN and TCP-ACK are handled by two different sk: + * TCP-SYN: select sk_fds[tmp_index = 1] tmp_index is from the + * tmp_index_ovr_map + * TCP-ACK: select sk_fds[reuseport_index = 0] reuseport_index + * is from the cmd.reuseport_index + */ + err = bpf_map_update_elem(tmp_index_ovr_map, &index_zero, + &tmp_index, BPF_ANY); + CHECK(err == -1, "update_elem(tmp_index_ovr_map, 0, 1)", + "err:%d errno:%d\n", err, errno); + do_test(type, family, &cmd, PASS); + err = bpf_map_lookup_elem(tmp_index_ovr_map, &index_zero, + &tmp_index); + CHECK(err == -1 || tmp_index != -1, + "lookup_elem(tmp_index_ovr_map)", + "err:%d errno:%d tmp_index:%d\n", + err, errno, tmp_index); + disable_syncookie(); + printf("OK\n"); +} + +static void test_pass_on_err(int type, sa_family_t family) +{ + struct cmd cmd = { + .reuseport_index = REUSEPORT_ARRAY_SIZE, + .pass_on_failure = 1, + }; + + expected_results[PASS_ERR_SK_SELECT_REUSEPORT] += 1; + do_test(type, family, &cmd, PASS_ERR_SK_SELECT_REUSEPORT); + printf("OK\n"); +} + +static void test_detach_bpf(int type, sa_family_t family) +{ +#ifdef SO_DETACH_REUSEPORT_BPF + __u32 nr_run_before = 0, nr_run_after = 0, tmp, i; + struct epoll_event ev; + int cli_fd, err, nev; + struct cmd cmd = {}; + int optvalue = 0; + + err = setsockopt(sk_fds[0], SOL_SOCKET, SO_DETACH_REUSEPORT_BPF, + &optvalue, sizeof(optvalue)); + CHECK(err == -1, "setsockopt(SO_DETACH_REUSEPORT_BPF)", + "err:%d errno:%d\n", err, errno); + + err = setsockopt(sk_fds[1], SOL_SOCKET, SO_DETACH_REUSEPORT_BPF, + &optvalue, sizeof(optvalue)); + CHECK(err == 0 || errno != ENOENT, "setsockopt(SO_DETACH_REUSEPORT_BPF)", + "err:%d errno:%d\n", err, errno); + + for (i = 0; i < NR_RESULTS; i++) { + err = bpf_map_lookup_elem(result_map, &i, &tmp); + CHECK(err == -1, "lookup_elem(result_map)", + "i:%u err:%d errno:%d\n", i, err, errno); + nr_run_before += tmp; + } + + cli_fd = send_data(type, family, &cmd, sizeof(cmd), PASS); + nev = epoll_wait(epfd, &ev, 1, 5); + CHECK(nev <= 0, "nev <= 0", + "nev:%d expected:1 type:%d family:%d data:(0, 0)\n", + nev, type, family); + + for (i = 0; i < NR_RESULTS; i++) { + err = bpf_map_lookup_elem(result_map, &i, &tmp); + CHECK(err == -1, "lookup_elem(result_map)", + "i:%u err:%d errno:%d\n", i, err, errno); + nr_run_after += tmp; + } + + CHECK(nr_run_before != nr_run_after, + "nr_run_before != nr_run_after", + "nr_run_before:%u nr_run_after:%u\n", + nr_run_before, nr_run_after); + + printf("OK\n"); + close(cli_fd); +#else + printf("SKIP\n"); +#endif +} + +static void prepare_sk_fds(int type, sa_family_t family, bool inany) +{ + const int first = REUSEPORT_ARRAY_SIZE - 1; + int i, err, optval = 1; + struct epoll_event ev; + socklen_t addrlen; + + if (inany) + sa46_init_inany(&srv_sa, family); + else + sa46_init_loopback(&srv_sa, family); + addrlen = sizeof(srv_sa); + + /* + * The sk_fds[] is filled from the back such that the order + * is exactly opposite to the (struct sock_reuseport *)reuse->socks[]. + */ + for (i = first; i >= 0; i--) { + sk_fds[i] = socket(family, type, 0); + CHECK(sk_fds[i] == -1, "socket()", "sk_fds[%d]:%d errno:%d\n", + i, sk_fds[i], errno); + err = setsockopt(sk_fds[i], SOL_SOCKET, SO_REUSEPORT, + &optval, sizeof(optval)); + CHECK(err == -1, "setsockopt(SO_REUSEPORT)", + "sk_fds[%d] err:%d errno:%d\n", + i, err, errno); + + if (i == first) { + err = setsockopt(sk_fds[i], SOL_SOCKET, + SO_ATTACH_REUSEPORT_EBPF, + &select_by_skb_data_prog, + sizeof(select_by_skb_data_prog)); + CHECK(err == -1, "setsockopt(SO_ATTACH_REUEPORT_EBPF)", + "err:%d errno:%d\n", err, errno); + } + + err = bind(sk_fds[i], (struct sockaddr *)&srv_sa, addrlen); + CHECK(err == -1, "bind()", "sk_fds[%d] err:%d errno:%d\n", + i, err, errno); + + if (type == SOCK_STREAM) { + err = listen(sk_fds[i], 10); + CHECK(err == -1, "listen()", + "sk_fds[%d] err:%d errno:%d\n", + i, err, errno); + } + + err = bpf_map_update_elem(reuseport_array, &i, &sk_fds[i], + BPF_NOEXIST); + CHECK(err == -1, "update_elem(reuseport_array)", + "sk_fds[%d] err:%d errno:%d\n", i, err, errno); + + if (i == first) { + socklen_t addrlen = sizeof(srv_sa); + + err = getsockname(sk_fds[i], (struct sockaddr *)&srv_sa, + &addrlen); + CHECK(err == -1, "getsockname()", + "sk_fds[%d] err:%d errno:%d\n", i, err, errno); + } + } + + epfd = epoll_create(1); + CHECK(epfd == -1, "epoll_create(1)", + "epfd:%d errno:%d\n", epfd, errno); + + ev.events = EPOLLIN; + for (i = 0; i < REUSEPORT_ARRAY_SIZE; i++) { + ev.data.u32 = i; + err = epoll_ctl(epfd, EPOLL_CTL_ADD, sk_fds[i], &ev); + CHECK(err, "epoll_ctl(EPOLL_CTL_ADD)", "sk_fds[%d]\n", i); + } +} + +static void setup_per_test(int type, sa_family_t family, bool inany, + bool no_inner_map) +{ + int ovr = -1, err; + + prepare_sk_fds(type, family, inany); + err = bpf_map_update_elem(tmp_index_ovr_map, &index_zero, &ovr, + BPF_ANY); + CHECK(err == -1, "update_elem(tmp_index_ovr_map, 0, -1)", + "err:%d errno:%d\n", err, errno); + + /* Install reuseport_array to outer_map? */ + if (no_inner_map) + return; + + err = bpf_map_update_elem(outer_map, &index_zero, &reuseport_array, + BPF_ANY); + CHECK(err == -1, "update_elem(outer_map, 0, reuseport_array)", + "err:%d errno:%d\n", err, errno); +} + +static void cleanup_per_test(bool no_inner_map) +{ + int i, err; + + for (i = 0; i < REUSEPORT_ARRAY_SIZE; i++) + close(sk_fds[i]); + close(epfd); + + /* Delete reuseport_array from outer_map? */ + if (no_inner_map) + return; + + err = bpf_map_delete_elem(outer_map, &index_zero); + CHECK(err == -1, "delete_elem(outer_map)", + "err:%d errno:%d\n", err, errno); +} + +static void cleanup(void) +{ + if (outer_map != -1) + close(outer_map); + if (reuseport_array != -1) + close(reuseport_array); + if (obj) + bpf_object__close(obj); +} + +static const char *family_str(sa_family_t family) +{ + switch (family) { + case AF_INET: + return "IPv4"; + case AF_INET6: + return "IPv6"; + default: + return "unknown"; + } +} + +static const char *sotype_str(int sotype) +{ + switch (sotype) { + case SOCK_STREAM: + return "TCP"; + case SOCK_DGRAM: + return "UDP"; + default: + return "unknown"; + } +} + +#define TEST_INIT(fn, ...) { fn, #fn, __VA_ARGS__ } + +static void test_config(int type, sa_family_t family, bool inany) +{ + const struct test { + void (*fn)(int sotype, sa_family_t family); + const char *name; + bool no_inner_map; + } tests[] = { + TEST_INIT(test_err_inner_map, true /* no_inner_map */), + TEST_INIT(test_err_skb_data), + TEST_INIT(test_err_sk_select_port), + TEST_INIT(test_pass), + TEST_INIT(test_syncookie), + TEST_INIT(test_pass_on_err), + TEST_INIT(test_detach_bpf), + }; + const struct test *t; + + printf("######## %s/%s %s ########\n", + family_str(family), sotype_str(type), + inany ? " INANY " : "LOOPBACK"); + + for (t = tests; t < tests + ARRAY_SIZE(tests); t++) { + setup_per_test(type, family, inany, t->no_inner_map); + printf("%s: ", t->name); + t->fn(type, family); + cleanup_per_test(t->no_inner_map); + } + + printf("\n"); +} + +#define BIND_INANY true + +static void test_all(void) +{ + const struct config { + int sotype; + sa_family_t family; + bool inany; + } configs[] = { + { SOCK_STREAM, AF_INET }, + { SOCK_STREAM, AF_INET, BIND_INANY }, + { SOCK_STREAM, AF_INET6 }, + { SOCK_STREAM, AF_INET6, BIND_INANY }, + { SOCK_DGRAM, AF_INET }, + { SOCK_DGRAM, AF_INET6 }, + }; + const struct config *c; + + for (c = configs; c < configs + ARRAY_SIZE(configs); c++) + test_config(c->sotype, c->family, c->inany); +} + +int main(int argc, const char **argv) +{ + int ret = EXIT_FAILURE; + + if (create_maps()) + goto out; + if (prepare_bpf_obj()) + goto out; + + saved_tcp_fo = read_int_sysctl(TCP_FO_SYSCTL); + saved_tcp_syncookie = read_int_sysctl(TCP_SYNCOOKIE_SYSCTL); + if (saved_tcp_syncookie < 0 || saved_tcp_syncookie < 0) + goto out; + atexit(restore_sysctls); + + if (enable_fastopen()) + goto out; + if (disable_syncookie()) + goto out; + + test_all(); + + ret = EXIT_SUCCESS; +out: + cleanup(); + return ret; +} diff --git a/tools/testing/selftests/bpf/test_select_reuseport.c b/tools/testing/selftests/bpf/test_select_reuseport.c deleted file mode 100644 index 0d5687feb689..000000000000 --- a/tools/testing/selftests/bpf/test_select_reuseport.c +++ /dev/null @@ -1,811 +0,0 @@ -// SPDX-License-Identifier: GPL-2.0 -/* Copyright (c) 2018 Facebook */ - -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include "bpf_rlimit.h" -#include "bpf_util.h" -#include "test_select_reuseport_common.h" - -#define MIN_TCPHDR_LEN 20 -#define UDPHDR_LEN 8 - -#define TCP_SYNCOOKIE_SYSCTL "/proc/sys/net/ipv4/tcp_syncookies" -#define TCP_FO_SYSCTL "/proc/sys/net/ipv4/tcp_fastopen" -#define REUSEPORT_ARRAY_SIZE 32 - -static int result_map, tmp_index_ovr_map, linum_map, data_check_map; -static enum result expected_results[NR_RESULTS]; -static int sk_fds[REUSEPORT_ARRAY_SIZE]; -static int reuseport_array = -1, outer_map = -1; -static int select_by_skb_data_prog; -static int saved_tcp_syncookie; -static struct bpf_object *obj; -static int saved_tcp_fo; -static __u32 index_zero; -static int epfd; - -static union sa46 { - struct sockaddr_in6 v6; - struct sockaddr_in v4; - sa_family_t family; -} srv_sa; - -#define CHECK(condition, tag, format...) ({ \ - int __ret = !!(condition); \ - if (__ret) { \ - printf("%s(%d):FAIL:%s ", __func__, __LINE__, tag); \ - printf(format); \ - exit(-1); \ - } \ -}) - -#define RET_ERR(condition, tag, format...) ({ \ - int __ret = !!(condition); \ - if (__ret) { \ - printf("%s(%d):FAIL:%s ", __func__, __LINE__, tag); \ - printf(format); \ - return -1; \ - } \ -}) - -static int create_maps(void) -{ - struct bpf_create_map_attr attr = {}; - - /* Creating reuseport_array */ - attr.name = "reuseport_array"; - attr.map_type = BPF_MAP_TYPE_REUSEPORT_SOCKARRAY; - attr.key_size = sizeof(__u32); - attr.value_size = sizeof(__u32); - attr.max_entries = REUSEPORT_ARRAY_SIZE; - - reuseport_array = bpf_create_map_xattr(&attr); - RET_ERR(reuseport_array == -1, "creating reuseport_array", - "reuseport_array:%d errno:%d\n", reuseport_array, errno); - - /* Creating outer_map */ - attr.name = "outer_map"; - attr.map_type = BPF_MAP_TYPE_ARRAY_OF_MAPS; - attr.key_size = sizeof(__u32); - attr.value_size = sizeof(__u32); - attr.max_entries = 1; - attr.inner_map_fd = reuseport_array; - outer_map = bpf_create_map_xattr(&attr); - RET_ERR(outer_map == -1, "creating outer_map", - "outer_map:%d errno:%d\n", outer_map, errno); - - return 0; -} - -static int prepare_bpf_obj(void) -{ - struct bpf_program *prog; - struct bpf_map *map; - int err; - - obj = bpf_object__open("test_select_reuseport_kern.o"); - RET_ERR(IS_ERR_OR_NULL(obj), "open test_select_reuseport_kern.o", - "obj:%p PTR_ERR(obj):%ld\n", obj, PTR_ERR(obj)); - - map = bpf_object__find_map_by_name(obj, "outer_map"); - RET_ERR(!map, "find outer_map", "!map\n"); - err = bpf_map__reuse_fd(map, outer_map); - RET_ERR(err, "reuse outer_map", "err:%d\n", err); - - err = bpf_object__load(obj); - RET_ERR(err, "load bpf_object", "err:%d\n", err); - - prog = bpf_program__next(NULL, obj); - RET_ERR(!prog, "get first bpf_program", "!prog\n"); - select_by_skb_data_prog = bpf_program__fd(prog); - RET_ERR(select_by_skb_data_prog == -1, "get prog fd", - "select_by_skb_data_prog:%d\n", select_by_skb_data_prog); - - map = bpf_object__find_map_by_name(obj, "result_map"); - RET_ERR(!map, "find result_map", "!map\n"); - result_map = bpf_map__fd(map); - RET_ERR(result_map == -1, "get result_map fd", - "result_map:%d\n", result_map); - - map = bpf_object__find_map_by_name(obj, "tmp_index_ovr_map"); - RET_ERR(!map, "find tmp_index_ovr_map\n", "!map"); - tmp_index_ovr_map = bpf_map__fd(map); - RET_ERR(tmp_index_ovr_map == -1, "get tmp_index_ovr_map fd", - "tmp_index_ovr_map:%d\n", tmp_index_ovr_map); - - map = bpf_object__find_map_by_name(obj, "linum_map"); - RET_ERR(!map, "find linum_map", "!map\n"); - linum_map = bpf_map__fd(map); - RET_ERR(linum_map == -1, "get linum_map fd", - "linum_map:%d\n", linum_map); - - map = bpf_object__find_map_by_name(obj, "data_check_map"); - RET_ERR(!map, "find data_check_map", "!map\n"); - data_check_map = bpf_map__fd(map); - RET_ERR(data_check_map == -1, "get data_check_map fd", - "data_check_map:%d\n", data_check_map); - - return 0; -} - -static void sa46_init_loopback(union sa46 *sa, sa_family_t family) -{ - memset(sa, 0, sizeof(*sa)); - sa->family = family; - if (sa->family == AF_INET6) - sa->v6.sin6_addr = in6addr_loopback; - else - sa->v4.sin_addr.s_addr = htonl(INADDR_LOOPBACK); -} - -static void sa46_init_inany(union sa46 *sa, sa_family_t family) -{ - memset(sa, 0, sizeof(*sa)); - sa->family = family; - if (sa->family == AF_INET6) - sa->v6.sin6_addr = in6addr_any; - else - sa->v4.sin_addr.s_addr = INADDR_ANY; -} - -static int read_int_sysctl(const char *sysctl) -{ - char buf[16]; - int fd, ret; - - fd = open(sysctl, 0); - RET_ERR(fd == -1, "open(sysctl)", - "sysctl:%s fd:%d errno:%d\n", sysctl, fd, errno); - - ret = read(fd, buf, sizeof(buf)); - RET_ERR(ret <= 0, "read(sysctl)", - "sysctl:%s ret:%d errno:%d\n", sysctl, ret, errno); - - close(fd); - return atoi(buf); -} - -static int write_int_sysctl(const char *sysctl, int v) -{ - int fd, ret, size; - char buf[16]; - - fd = open(sysctl, O_RDWR); - RET_ERR(fd == -1, "open(sysctl)", - "sysctl:%s fd:%d errno:%d\n", sysctl, fd, errno); - - size = snprintf(buf, sizeof(buf), "%d", v); - ret = write(fd, buf, size); - RET_ERR(ret != size, "write(sysctl)", - "sysctl:%s ret:%d size:%d errno:%d\n", - sysctl, ret, size, errno); - - close(fd); - return 0; -} - -static void restore_sysctls(void) -{ - write_int_sysctl(TCP_FO_SYSCTL, saved_tcp_fo); - write_int_sysctl(TCP_SYNCOOKIE_SYSCTL, saved_tcp_syncookie); -} - -static int enable_fastopen(void) -{ - int fo; - - fo = read_int_sysctl(TCP_FO_SYSCTL); - if (fo < 0) - return -1; - - return write_int_sysctl(TCP_FO_SYSCTL, fo | 7); -} - -static int enable_syncookie(void) -{ - return write_int_sysctl(TCP_SYNCOOKIE_SYSCTL, 2); -} - -static int disable_syncookie(void) -{ - return write_int_sysctl(TCP_SYNCOOKIE_SYSCTL, 0); -} - -static __u32 get_linum(void) -{ - __u32 linum; - int err; - - err = bpf_map_lookup_elem(linum_map, &index_zero, &linum); - CHECK(err == -1, "lookup_elem(linum_map)", "err:%d errno:%d\n", - err, errno); - - return linum; -} - -static void check_data(int type, sa_family_t family, const struct cmd *cmd, - int cli_fd) -{ - struct data_check expected = {}, result; - union sa46 cli_sa; - socklen_t addrlen; - int err; - - addrlen = sizeof(cli_sa); - err = getsockname(cli_fd, (struct sockaddr *)&cli_sa, - &addrlen); - CHECK(err == -1, "getsockname(cli_fd)", "err:%d errno:%d\n", - err, errno); - - err = bpf_map_lookup_elem(data_check_map, &index_zero, &result); - CHECK(err == -1, "lookup_elem(data_check_map)", "err:%d errno:%d\n", - err, errno); - - if (type == SOCK_STREAM) { - expected.len = MIN_TCPHDR_LEN; - expected.ip_protocol = IPPROTO_TCP; - } else { - expected.len = UDPHDR_LEN; - expected.ip_protocol = IPPROTO_UDP; - } - - if (family == AF_INET6) { - expected.eth_protocol = htons(ETH_P_IPV6); - expected.bind_inany = !srv_sa.v6.sin6_addr.s6_addr32[3] && - !srv_sa.v6.sin6_addr.s6_addr32[2] && - !srv_sa.v6.sin6_addr.s6_addr32[1] && - !srv_sa.v6.sin6_addr.s6_addr32[0]; - - memcpy(&expected.skb_addrs[0], cli_sa.v6.sin6_addr.s6_addr32, - sizeof(cli_sa.v6.sin6_addr)); - memcpy(&expected.skb_addrs[4], &in6addr_loopback, - sizeof(in6addr_loopback)); - expected.skb_ports[0] = cli_sa.v6.sin6_port; - expected.skb_ports[1] = srv_sa.v6.sin6_port; - } else { - expected.eth_protocol = htons(ETH_P_IP); - expected.bind_inany = !srv_sa.v4.sin_addr.s_addr; - - expected.skb_addrs[0] = cli_sa.v4.sin_addr.s_addr; - expected.skb_addrs[1] = htonl(INADDR_LOOPBACK); - expected.skb_ports[0] = cli_sa.v4.sin_port; - expected.skb_ports[1] = srv_sa.v4.sin_port; - } - - if (memcmp(&result, &expected, offsetof(struct data_check, - equal_check_end))) { - printf("unexpected data_check\n"); - printf(" result: (0x%x, %u, %u)\n", - result.eth_protocol, result.ip_protocol, - result.bind_inany); - printf("expected: (0x%x, %u, %u)\n", - expected.eth_protocol, expected.ip_protocol, - expected.bind_inany); - CHECK(1, "data_check result != expected", - "bpf_prog_linum:%u\n", get_linum()); - } - - CHECK(!result.hash, "data_check result.hash empty", - "result.hash:%u", result.hash); - - expected.len += cmd ? sizeof(*cmd) : 0; - if (type == SOCK_STREAM) - CHECK(expected.len > result.len, "expected.len > result.len", - "expected.len:%u result.len:%u bpf_prog_linum:%u\n", - expected.len, result.len, get_linum()); - else - CHECK(expected.len != result.len, "expected.len != result.len", - "expected.len:%u result.len:%u bpf_prog_linum:%u\n", - expected.len, result.len, get_linum()); -} - -static void check_results(void) -{ - __u32 results[NR_RESULTS]; - __u32 i, broken = 0; - int err; - - for (i = 0; i < NR_RESULTS; i++) { - err = bpf_map_lookup_elem(result_map, &i, &results[i]); - CHECK(err == -1, "lookup_elem(result_map)", - "i:%u err:%d errno:%d\n", i, err, errno); - } - - for (i = 0; i < NR_RESULTS; i++) { - if (results[i] != expected_results[i]) { - broken = i; - break; - } - } - - if (i == NR_RESULTS) - return; - - printf("unexpected result\n"); - printf(" result: ["); - printf("%u", results[0]); - for (i = 1; i < NR_RESULTS; i++) - printf(", %u", results[i]); - printf("]\n"); - - printf("expected: ["); - printf("%u", expected_results[0]); - for (i = 1; i < NR_RESULTS; i++) - printf(", %u", expected_results[i]); - printf("]\n"); - - CHECK(expected_results[broken] != results[broken], - "unexpected result", - "expected_results[%u] != results[%u] bpf_prog_linum:%u\n", - broken, broken, get_linum()); -} - -static int send_data(int type, sa_family_t family, void *data, size_t len, - enum result expected) -{ - union sa46 cli_sa; - int fd, err; - - fd = socket(family, type, 0); - CHECK(fd == -1, "socket()", "fd:%d errno:%d\n", fd, errno); - - sa46_init_loopback(&cli_sa, family); - err = bind(fd, (struct sockaddr *)&cli_sa, sizeof(cli_sa)); - CHECK(fd == -1, "bind(cli_sa)", "err:%d errno:%d\n", err, errno); - - err = sendto(fd, data, len, MSG_FASTOPEN, (struct sockaddr *)&srv_sa, - sizeof(srv_sa)); - CHECK(err != len && expected >= PASS, - "sendto()", "family:%u err:%d errno:%d expected:%d\n", - family, err, errno, expected); - - return fd; -} - -static void do_test(int type, sa_family_t family, struct cmd *cmd, - enum result expected) -{ - int nev, srv_fd, cli_fd; - struct epoll_event ev; - struct cmd rcv_cmd; - ssize_t nread; - - cli_fd = send_data(type, family, cmd, cmd ? sizeof(*cmd) : 0, - expected); - nev = epoll_wait(epfd, &ev, 1, expected >= PASS ? 5 : 0); - CHECK((nev <= 0 && expected >= PASS) || - (nev > 0 && expected < PASS), - "nev <> expected", - "nev:%d expected:%d type:%d family:%d data:(%d, %d)\n", - nev, expected, type, family, - cmd ? cmd->reuseport_index : -1, - cmd ? cmd->pass_on_failure : -1); - check_results(); - check_data(type, family, cmd, cli_fd); - - if (expected < PASS) - return; - - CHECK(expected != PASS_ERR_SK_SELECT_REUSEPORT && - cmd->reuseport_index != ev.data.u32, - "check cmd->reuseport_index", - "cmd:(%u, %u) ev.data.u32:%u\n", - cmd->pass_on_failure, cmd->reuseport_index, ev.data.u32); - - srv_fd = sk_fds[ev.data.u32]; - if (type == SOCK_STREAM) { - int new_fd = accept(srv_fd, NULL, 0); - - CHECK(new_fd == -1, "accept(srv_fd)", - "ev.data.u32:%u new_fd:%d errno:%d\n", - ev.data.u32, new_fd, errno); - - nread = recv(new_fd, &rcv_cmd, sizeof(rcv_cmd), MSG_DONTWAIT); - CHECK(nread != sizeof(rcv_cmd), - "recv(new_fd)", - "ev.data.u32:%u nread:%zd sizeof(rcv_cmd):%zu errno:%d\n", - ev.data.u32, nread, sizeof(rcv_cmd), errno); - - close(new_fd); - } else { - nread = recv(srv_fd, &rcv_cmd, sizeof(rcv_cmd), MSG_DONTWAIT); - CHECK(nread != sizeof(rcv_cmd), - "recv(sk_fds)", - "ev.data.u32:%u nread:%zd sizeof(rcv_cmd):%zu errno:%d\n", - ev.data.u32, nread, sizeof(rcv_cmd), errno); - } - - close(cli_fd); -} - -static void test_err_inner_map(int type, sa_family_t family) -{ - struct cmd cmd = { - .reuseport_index = 0, - .pass_on_failure = 0, - }; - - expected_results[DROP_ERR_INNER_MAP]++; - do_test(type, family, &cmd, DROP_ERR_INNER_MAP); - printf("OK\n"); -} - -static void test_err_skb_data(int type, sa_family_t family) -{ - expected_results[DROP_ERR_SKB_DATA]++; - do_test(type, family, NULL, DROP_ERR_SKB_DATA); - printf("OK\n"); -} - -static void test_err_sk_select_port(int type, sa_family_t family) -{ - struct cmd cmd = { - .reuseport_index = REUSEPORT_ARRAY_SIZE, - .pass_on_failure = 0, - }; - - expected_results[DROP_ERR_SK_SELECT_REUSEPORT]++; - do_test(type, family, &cmd, DROP_ERR_SK_SELECT_REUSEPORT); - printf("OK\n"); -} - -static void test_pass(int type, sa_family_t family) -{ - struct cmd cmd; - int i; - - cmd.pass_on_failure = 0; - for (i = 0; i < REUSEPORT_ARRAY_SIZE; i++) { - expected_results[PASS]++; - cmd.reuseport_index = i; - do_test(type, family, &cmd, PASS); - } - printf("OK\n"); -} - -static void test_syncookie(int type, sa_family_t family) -{ - int err, tmp_index = 1; - struct cmd cmd = { - .reuseport_index = 0, - .pass_on_failure = 0, - }; - - if (type != SOCK_STREAM) - return; - - /* - * +1 for TCP-SYN and - * +1 for the TCP-ACK (ack the syncookie) - */ - expected_results[PASS] += 2; - enable_syncookie(); - /* - * Simulate TCP-SYN and TCP-ACK are handled by two different sk: - * TCP-SYN: select sk_fds[tmp_index = 1] tmp_index is from the - * tmp_index_ovr_map - * TCP-ACK: select sk_fds[reuseport_index = 0] reuseport_index - * is from the cmd.reuseport_index - */ - err = bpf_map_update_elem(tmp_index_ovr_map, &index_zero, - &tmp_index, BPF_ANY); - CHECK(err == -1, "update_elem(tmp_index_ovr_map, 0, 1)", - "err:%d errno:%d\n", err, errno); - do_test(type, family, &cmd, PASS); - err = bpf_map_lookup_elem(tmp_index_ovr_map, &index_zero, - &tmp_index); - CHECK(err == -1 || tmp_index != -1, - "lookup_elem(tmp_index_ovr_map)", - "err:%d errno:%d tmp_index:%d\n", - err, errno, tmp_index); - disable_syncookie(); - printf("OK\n"); -} - -static void test_pass_on_err(int type, sa_family_t family) -{ - struct cmd cmd = { - .reuseport_index = REUSEPORT_ARRAY_SIZE, - .pass_on_failure = 1, - }; - - expected_results[PASS_ERR_SK_SELECT_REUSEPORT] += 1; - do_test(type, family, &cmd, PASS_ERR_SK_SELECT_REUSEPORT); - printf("OK\n"); -} - -static void test_detach_bpf(int type, sa_family_t family) -{ -#ifdef SO_DETACH_REUSEPORT_BPF - __u32 nr_run_before = 0, nr_run_after = 0, tmp, i; - struct epoll_event ev; - int cli_fd, err, nev; - struct cmd cmd = {}; - int optvalue = 0; - - err = setsockopt(sk_fds[0], SOL_SOCKET, SO_DETACH_REUSEPORT_BPF, - &optvalue, sizeof(optvalue)); - CHECK(err == -1, "setsockopt(SO_DETACH_REUSEPORT_BPF)", - "err:%d errno:%d\n", err, errno); - - err = setsockopt(sk_fds[1], SOL_SOCKET, SO_DETACH_REUSEPORT_BPF, - &optvalue, sizeof(optvalue)); - CHECK(err == 0 || errno != ENOENT, "setsockopt(SO_DETACH_REUSEPORT_BPF)", - "err:%d errno:%d\n", err, errno); - - for (i = 0; i < NR_RESULTS; i++) { - err = bpf_map_lookup_elem(result_map, &i, &tmp); - CHECK(err == -1, "lookup_elem(result_map)", - "i:%u err:%d errno:%d\n", i, err, errno); - nr_run_before += tmp; - } - - cli_fd = send_data(type, family, &cmd, sizeof(cmd), PASS); - nev = epoll_wait(epfd, &ev, 1, 5); - CHECK(nev <= 0, "nev <= 0", - "nev:%d expected:1 type:%d family:%d data:(0, 0)\n", - nev, type, family); - - for (i = 0; i < NR_RESULTS; i++) { - err = bpf_map_lookup_elem(result_map, &i, &tmp); - CHECK(err == -1, "lookup_elem(result_map)", - "i:%u err:%d errno:%d\n", i, err, errno); - nr_run_after += tmp; - } - - CHECK(nr_run_before != nr_run_after, - "nr_run_before != nr_run_after", - "nr_run_before:%u nr_run_after:%u\n", - nr_run_before, nr_run_after); - - printf("OK\n"); - close(cli_fd); -#else - printf("SKIP\n"); -#endif -} - -static void prepare_sk_fds(int type, sa_family_t family, bool inany) -{ - const int first = REUSEPORT_ARRAY_SIZE - 1; - int i, err, optval = 1; - struct epoll_event ev; - socklen_t addrlen; - - if (inany) - sa46_init_inany(&srv_sa, family); - else - sa46_init_loopback(&srv_sa, family); - addrlen = sizeof(srv_sa); - - /* - * The sk_fds[] is filled from the back such that the order - * is exactly opposite to the (struct sock_reuseport *)reuse->socks[]. - */ - for (i = first; i >= 0; i--) { - sk_fds[i] = socket(family, type, 0); - CHECK(sk_fds[i] == -1, "socket()", "sk_fds[%d]:%d errno:%d\n", - i, sk_fds[i], errno); - err = setsockopt(sk_fds[i], SOL_SOCKET, SO_REUSEPORT, - &optval, sizeof(optval)); - CHECK(err == -1, "setsockopt(SO_REUSEPORT)", - "sk_fds[%d] err:%d errno:%d\n", - i, err, errno); - - if (i == first) { - err = setsockopt(sk_fds[i], SOL_SOCKET, - SO_ATTACH_REUSEPORT_EBPF, - &select_by_skb_data_prog, - sizeof(select_by_skb_data_prog)); - CHECK(err == -1, "setsockopt(SO_ATTACH_REUEPORT_EBPF)", - "err:%d errno:%d\n", err, errno); - } - - err = bind(sk_fds[i], (struct sockaddr *)&srv_sa, addrlen); - CHECK(err == -1, "bind()", "sk_fds[%d] err:%d errno:%d\n", - i, err, errno); - - if (type == SOCK_STREAM) { - err = listen(sk_fds[i], 10); - CHECK(err == -1, "listen()", - "sk_fds[%d] err:%d errno:%d\n", - i, err, errno); - } - - err = bpf_map_update_elem(reuseport_array, &i, &sk_fds[i], - BPF_NOEXIST); - CHECK(err == -1, "update_elem(reuseport_array)", - "sk_fds[%d] err:%d errno:%d\n", i, err, errno); - - if (i == first) { - socklen_t addrlen = sizeof(srv_sa); - - err = getsockname(sk_fds[i], (struct sockaddr *)&srv_sa, - &addrlen); - CHECK(err == -1, "getsockname()", - "sk_fds[%d] err:%d errno:%d\n", i, err, errno); - } - } - - epfd = epoll_create(1); - CHECK(epfd == -1, "epoll_create(1)", - "epfd:%d errno:%d\n", epfd, errno); - - ev.events = EPOLLIN; - for (i = 0; i < REUSEPORT_ARRAY_SIZE; i++) { - ev.data.u32 = i; - err = epoll_ctl(epfd, EPOLL_CTL_ADD, sk_fds[i], &ev); - CHECK(err, "epoll_ctl(EPOLL_CTL_ADD)", "sk_fds[%d]\n", i); - } -} - -static void setup_per_test(int type, sa_family_t family, bool inany, - bool no_inner_map) -{ - int ovr = -1, err; - - prepare_sk_fds(type, family, inany); - err = bpf_map_update_elem(tmp_index_ovr_map, &index_zero, &ovr, - BPF_ANY); - CHECK(err == -1, "update_elem(tmp_index_ovr_map, 0, -1)", - "err:%d errno:%d\n", err, errno); - - /* Install reuseport_array to outer_map? */ - if (no_inner_map) - return; - - err = bpf_map_update_elem(outer_map, &index_zero, &reuseport_array, - BPF_ANY); - CHECK(err == -1, "update_elem(outer_map, 0, reuseport_array)", - "err:%d errno:%d\n", err, errno); -} - -static void cleanup_per_test(bool no_inner_map) -{ - int i, err; - - for (i = 0; i < REUSEPORT_ARRAY_SIZE; i++) - close(sk_fds[i]); - close(epfd); - - /* Delete reuseport_array from outer_map? */ - if (no_inner_map) - return; - - err = bpf_map_delete_elem(outer_map, &index_zero); - CHECK(err == -1, "delete_elem(outer_map)", - "err:%d errno:%d\n", err, errno); -} - -static void cleanup(void) -{ - if (outer_map != -1) - close(outer_map); - if (reuseport_array != -1) - close(reuseport_array); - if (obj) - bpf_object__close(obj); -} - -static const char *family_str(sa_family_t family) -{ - switch (family) { - case AF_INET: - return "IPv4"; - case AF_INET6: - return "IPv6"; - default: - return "unknown"; - } -} - -static const char *sotype_str(int sotype) -{ - switch (sotype) { - case SOCK_STREAM: - return "TCP"; - case SOCK_DGRAM: - return "UDP"; - default: - return "unknown"; - } -} - -#define TEST_INIT(fn, ...) { fn, #fn, __VA_ARGS__ } - -static void test_config(int type, sa_family_t family, bool inany) -{ - const struct test { - void (*fn)(int sotype, sa_family_t family); - const char *name; - bool no_inner_map; - } tests[] = { - TEST_INIT(test_err_inner_map, true /* no_inner_map */), - TEST_INIT(test_err_skb_data), - TEST_INIT(test_err_sk_select_port), - TEST_INIT(test_pass), - TEST_INIT(test_syncookie), - TEST_INIT(test_pass_on_err), - TEST_INIT(test_detach_bpf), - }; - const struct test *t; - - printf("######## %s/%s %s ########\n", - family_str(family), sotype_str(type), - inany ? " INANY " : "LOOPBACK"); - - for (t = tests; t < tests + ARRAY_SIZE(tests); t++) { - setup_per_test(type, family, inany, t->no_inner_map); - printf("%s: ", t->name); - t->fn(type, family); - cleanup_per_test(t->no_inner_map); - } - - printf("\n"); -} - -#define BIND_INANY true - -static void test_all(void) -{ - const struct config { - int sotype; - sa_family_t family; - bool inany; - } configs[] = { - { SOCK_STREAM, AF_INET }, - { SOCK_STREAM, AF_INET, BIND_INANY }, - { SOCK_STREAM, AF_INET6 }, - { SOCK_STREAM, AF_INET6, BIND_INANY }, - { SOCK_DGRAM, AF_INET }, - { SOCK_DGRAM, AF_INET6 }, - }; - const struct config *c; - - for (c = configs; c < configs + ARRAY_SIZE(configs); c++) - test_config(c->sotype, c->family, c->inany); -} - -int main(int argc, const char **argv) -{ - int ret = EXIT_FAILURE; - - if (create_maps()) - goto out; - if (prepare_bpf_obj()) - goto out; - - saved_tcp_fo = read_int_sysctl(TCP_FO_SYSCTL); - saved_tcp_syncookie = read_int_sysctl(TCP_SYNCOOKIE_SYSCTL); - if (saved_tcp_syncookie < 0 || saved_tcp_syncookie < 0) - goto out; - atexit(restore_sysctls); - - if (enable_fastopen()) - goto out; - if (disable_syncookie()) - goto out; - - test_all(); - - ret = EXIT_SUCCESS; -out: - cleanup(); - return ret; -}