bpf: Add an example for bpf_get_listener_sock
authorMartin KaFai Lau <kafai@fb.com>
Tue, 12 Mar 2019 17:23:11 +0000 (10:23 -0700)
committerAlexei Starovoitov <ast@kernel.org>
Wed, 13 Mar 2019 19:04:35 +0000 (12:04 -0700)
This patch adds an example in using the new helper
bpf_get_listener_sock().

Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
tools/testing/selftests/bpf/bpf_helpers.h
tools/testing/selftests/bpf/progs/test_sock_fields_kern.c
tools/testing/selftests/bpf/test_sock_fields.c

index c9433a496d548daf9417a1f19af71753f93547c0..c81fc350f7ad46ad60d53ac3dd8121059020f9a6 100644 (file)
@@ -180,6 +180,8 @@ static struct bpf_sock *(*bpf_sk_fullsock)(struct bpf_sock *sk) =
        (void *) BPF_FUNC_sk_fullsock;
 static struct bpf_tcp_sock *(*bpf_tcp_sock)(struct bpf_sock *sk) =
        (void *) BPF_FUNC_tcp_sock;
+static struct bpf_sock *(*bpf_get_listener_sock)(struct bpf_sock *sk) =
+       (void *) BPF_FUNC_get_listener_sock;
 static int (*bpf_skb_ecn_set_ce)(void *ctx) =
        (void *) BPF_FUNC_skb_ecn_set_ce;
 
index de1a43e8f61070220c511b9aae698e4de7fb0659..37328f1485384b756c3120f428a3348206a9eb47 100644 (file)
@@ -8,38 +8,51 @@
 #include "bpf_helpers.h"
 #include "bpf_endian.h"
 
-enum bpf_array_idx {
-       SRV_IDX,
-       CLI_IDX,
-       __NR_BPF_ARRAY_IDX,
+enum bpf_addr_array_idx {
+       ADDR_SRV_IDX,
+       ADDR_CLI_IDX,
+       __NR_BPF_ADDR_ARRAY_IDX,
+};
+
+enum bpf_result_array_idx {
+       EGRESS_SRV_IDX,
+       EGRESS_CLI_IDX,
+       INGRESS_LISTEN_IDX,
+       __NR_BPF_RESULT_ARRAY_IDX,
+};
+
+enum bpf_linum_array_idx {
+       EGRESS_LINUM_IDX,
+       INGRESS_LINUM_IDX,
+       __NR_BPF_LINUM_ARRAY_IDX,
 };
 
 struct bpf_map_def SEC("maps") addr_map = {
        .type = BPF_MAP_TYPE_ARRAY,
        .key_size = sizeof(__u32),
        .value_size = sizeof(struct sockaddr_in6),
-       .max_entries = __NR_BPF_ARRAY_IDX,
+       .max_entries = __NR_BPF_ADDR_ARRAY_IDX,
 };
 
 struct bpf_map_def SEC("maps") sock_result_map = {
        .type = BPF_MAP_TYPE_ARRAY,
        .key_size = sizeof(__u32),
        .value_size = sizeof(struct bpf_sock),
-       .max_entries = __NR_BPF_ARRAY_IDX,
+       .max_entries = __NR_BPF_RESULT_ARRAY_IDX,
 };
 
 struct bpf_map_def SEC("maps") tcp_sock_result_map = {
        .type = BPF_MAP_TYPE_ARRAY,
        .key_size = sizeof(__u32),
        .value_size = sizeof(struct bpf_tcp_sock),
-       .max_entries = __NR_BPF_ARRAY_IDX,
+       .max_entries = __NR_BPF_RESULT_ARRAY_IDX,
 };
 
 struct bpf_map_def SEC("maps") linum_map = {
        .type = BPF_MAP_TYPE_ARRAY,
        .key_size = sizeof(__u32),
        .value_size = sizeof(__u32),
-       .max_entries = 1,
+       .max_entries = __NR_BPF_LINUM_ARRAY_IDX,
 };
 
 static bool is_loopback6(__u32 *a6)
@@ -100,18 +113,20 @@ static void tpcpy(struct bpf_tcp_sock *dst,
 
 #define RETURN {                                               \
        linum = __LINE__;                                       \
-       bpf_map_update_elem(&linum_map, &idx0, &linum, 0);      \
+       bpf_map_update_elem(&linum_map, &linum_idx, &linum, 0); \
        return 1;                                               \
 }
 
 SEC("cgroup_skb/egress")
-int read_sock_fields(struct __sk_buff *skb)
+int egress_read_sock_fields(struct __sk_buff *skb)
 {
-       __u32 srv_idx = SRV_IDX, cli_idx = CLI_IDX, idx;
+       __u32 srv_idx = ADDR_SRV_IDX, cli_idx = ADDR_CLI_IDX, result_idx;
        struct sockaddr_in6 *srv_sa6, *cli_sa6;
        struct bpf_tcp_sock *tp, *tp_ret;
        struct bpf_sock *sk, *sk_ret;
-       __u32 linum, idx0 = 0;
+       __u32 linum, linum_idx;
+
+       linum_idx = EGRESS_LINUM_IDX;
 
        sk = skb->sk;
        if (!sk || sk->state == 10)
@@ -132,14 +147,55 @@ int read_sock_fields(struct __sk_buff *skb)
                RETURN;
 
        if (sk->src_port == bpf_ntohs(srv_sa6->sin6_port))
-               idx = srv_idx;
+               result_idx = EGRESS_SRV_IDX;
        else if (sk->src_port == bpf_ntohs(cli_sa6->sin6_port))
-               idx = cli_idx;
+               result_idx = EGRESS_CLI_IDX;
        else
                RETURN;
 
-       sk_ret = bpf_map_lookup_elem(&sock_result_map, &idx);
-       tp_ret = bpf_map_lookup_elem(&tcp_sock_result_map, &idx);
+       sk_ret = bpf_map_lookup_elem(&sock_result_map, &result_idx);
+       tp_ret = bpf_map_lookup_elem(&tcp_sock_result_map, &result_idx);
+       if (!sk_ret || !tp_ret)
+               RETURN;
+
+       skcpy(sk_ret, sk);
+       tpcpy(tp_ret, tp);
+
+       RETURN;
+}
+
+SEC("cgroup_skb/ingress")
+int ingress_read_sock_fields(struct __sk_buff *skb)
+{
+       __u32 srv_idx = ADDR_SRV_IDX, result_idx = INGRESS_LISTEN_IDX;
+       struct bpf_tcp_sock *tp, *tp_ret;
+       struct bpf_sock *sk, *sk_ret;
+       struct sockaddr_in6 *srv_sa6;
+       __u32 linum, linum_idx;
+
+       linum_idx = INGRESS_LINUM_IDX;
+
+       sk = skb->sk;
+       if (!sk || sk->family != AF_INET6 || !is_loopback6(sk->src_ip6))
+               RETURN;
+
+       srv_sa6 = bpf_map_lookup_elem(&addr_map, &srv_idx);
+       if (!srv_sa6 || sk->src_port != bpf_ntohs(srv_sa6->sin6_port))
+               RETURN;
+
+       if (sk->state != 10 && sk->state != 12)
+               RETURN;
+
+       sk = bpf_get_listener_sock(sk);
+       if (!sk)
+               RETURN;
+
+       tp = bpf_tcp_sock(sk);
+       if (!tp)
+               RETURN;
+
+       sk_ret = bpf_map_lookup_elem(&sock_result_map, &result_idx);
+       tp_ret = bpf_map_lookup_elem(&tcp_sock_result_map, &result_idx);
        if (!sk_ret || !tp_ret)
                RETURN;
 
index bc8943938bf53933577c9be075c2f35314708eb3..dcae7f664dce0827f5ddda4bc5f693d191def9b0 100644 (file)
 #include "cgroup_helpers.h"
 #include "bpf_rlimit.h"
 
-enum bpf_array_idx {
-       SRV_IDX,
-       CLI_IDX,
-       __NR_BPF_ARRAY_IDX,
+enum bpf_addr_array_idx {
+       ADDR_SRV_IDX,
+       ADDR_CLI_IDX,
+       __NR_BPF_ADDR_ARRAY_IDX,
+};
+
+enum bpf_result_array_idx {
+       EGRESS_SRV_IDX,
+       EGRESS_CLI_IDX,
+       INGRESS_LISTEN_IDX,
+       __NR_BPF_RESULT_ARRAY_IDX,
+};
+
+enum bpf_linum_array_idx {
+       EGRESS_LINUM_IDX,
+       INGRESS_LINUM_IDX,
+       __NR_BPF_LINUM_ARRAY_IDX,
 };
 
 #define CHECK(condition, tag, format...) ({                            \
@@ -41,8 +54,16 @@ static int linum_map_fd;
 static int addr_map_fd;
 static int tp_map_fd;
 static int sk_map_fd;
-static __u32 srv_idx = SRV_IDX;
-static __u32 cli_idx = CLI_IDX;
+
+static __u32 addr_srv_idx = ADDR_SRV_IDX;
+static __u32 addr_cli_idx = ADDR_CLI_IDX;
+
+static __u32 egress_srv_idx = EGRESS_SRV_IDX;
+static __u32 egress_cli_idx = EGRESS_CLI_IDX;
+static __u32 ingress_listen_idx = INGRESS_LISTEN_IDX;
+
+static __u32 egress_linum_idx = EGRESS_LINUM_IDX;
+static __u32 ingress_linum_idx = INGRESS_LINUM_IDX;
 
 static void init_loopback6(struct sockaddr_in6 *sa6)
 {
@@ -93,29 +114,46 @@ static void print_tp(const struct bpf_tcp_sock *tp)
 
 static void check_result(void)
 {
-       struct bpf_tcp_sock srv_tp, cli_tp;
-       struct bpf_sock srv_sk, cli_sk;
-       __u32 linum, idx0 = 0;
+       struct bpf_tcp_sock srv_tp, cli_tp, listen_tp;
+       struct bpf_sock srv_sk, cli_sk, listen_sk;
+       __u32 ingress_linum, egress_linum;
        int err;
 
-       err = bpf_map_lookup_elem(linum_map_fd, &idx0, &linum);
+       err = bpf_map_lookup_elem(linum_map_fd, &egress_linum_idx,
+                                 &egress_linum);
        CHECK(err == -1, "bpf_map_lookup_elem(linum_map_fd)",
              "err:%d errno:%d", err, errno);
 
-       err = bpf_map_lookup_elem(sk_map_fd, &srv_idx, &srv_sk);
-       CHECK(err == -1, "bpf_map_lookup_elem(sk_map_fd, &srv_idx)",
+       err = bpf_map_lookup_elem(linum_map_fd, &ingress_linum_idx,
+                                 &ingress_linum);
+       CHECK(err == -1, "bpf_map_lookup_elem(linum_map_fd)",
+             "err:%d errno:%d", err, errno);
+
+       err = bpf_map_lookup_elem(sk_map_fd, &egress_srv_idx, &srv_sk);
+       CHECK(err == -1, "bpf_map_lookup_elem(sk_map_fd, &egress_srv_idx)",
+             "err:%d errno:%d", err, errno);
+       err = bpf_map_lookup_elem(tp_map_fd, &egress_srv_idx, &srv_tp);
+       CHECK(err == -1, "bpf_map_lookup_elem(tp_map_fd, &egress_srv_idx)",
+             "err:%d errno:%d", err, errno);
+
+       err = bpf_map_lookup_elem(sk_map_fd, &egress_cli_idx, &cli_sk);
+       CHECK(err == -1, "bpf_map_lookup_elem(sk_map_fd, &egress_cli_idx)",
              "err:%d errno:%d", err, errno);
-       err = bpf_map_lookup_elem(tp_map_fd, &srv_idx, &srv_tp);
-       CHECK(err == -1, "bpf_map_lookup_elem(tp_map_fd, &srv_idx)",
+       err = bpf_map_lookup_elem(tp_map_fd, &egress_cli_idx, &cli_tp);
+       CHECK(err == -1, "bpf_map_lookup_elem(tp_map_fd, &egress_cli_idx)",
              "err:%d errno:%d", err, errno);
 
-       err = bpf_map_lookup_elem(sk_map_fd, &cli_idx, &cli_sk);
-       CHECK(err == -1, "bpf_map_lookup_elem(sk_map_fd, &cli_idx)",
+       err = bpf_map_lookup_elem(sk_map_fd, &ingress_listen_idx, &listen_sk);
+       CHECK(err == -1, "bpf_map_lookup_elem(sk_map_fd, &ingress_listen_idx)",
              "err:%d errno:%d", err, errno);
-       err = bpf_map_lookup_elem(tp_map_fd, &cli_idx, &cli_tp);
-       CHECK(err == -1, "bpf_map_lookup_elem(tp_map_fd, &cli_idx)",
+       err = bpf_map_lookup_elem(tp_map_fd, &ingress_listen_idx, &listen_tp);
+       CHECK(err == -1, "bpf_map_lookup_elem(tp_map_fd, &ingress_listen_idx)",
              "err:%d errno:%d", err, errno);
 
+       printf("listen_sk: ");
+       print_sk(&listen_sk);
+       printf("\n");
+
        printf("srv_sk: ");
        print_sk(&srv_sk);
        printf("\n");
@@ -124,6 +162,10 @@ static void check_result(void)
        print_sk(&cli_sk);
        printf("\n");
 
+       printf("listen_tp: ");
+       print_tp(&listen_tp);
+       printf("\n");
+
        printf("srv_tp: ");
        print_tp(&srv_tp);
        printf("\n");
@@ -132,6 +174,19 @@ static void check_result(void)
        print_tp(&cli_tp);
        printf("\n");
 
+       CHECK(listen_sk.state != 10 ||
+             listen_sk.family != AF_INET6 ||
+             listen_sk.protocol != IPPROTO_TCP ||
+             memcmp(listen_sk.src_ip6, &in6addr_loopback,
+                    sizeof(listen_sk.src_ip6)) ||
+             listen_sk.dst_ip6[0] || listen_sk.dst_ip6[1] ||
+             listen_sk.dst_ip6[2] || listen_sk.dst_ip6[3] ||
+             listen_sk.src_port != ntohs(srv_sa6.sin6_port) ||
+             listen_sk.dst_port,
+             "Unexpected listen_sk",
+             "Check listen_sk output. ingress_linum:%u",
+             ingress_linum);
+
        CHECK(srv_sk.state == 10 ||
              !srv_sk.state ||
              srv_sk.family != AF_INET6 ||
@@ -142,7 +197,8 @@ static void check_result(void)
                     sizeof(srv_sk.dst_ip6)) ||
              srv_sk.src_port != ntohs(srv_sa6.sin6_port) ||
              srv_sk.dst_port != cli_sa6.sin6_port,
-             "Unexpected srv_sk", "Check srv_sk output. linum:%u", linum);
+             "Unexpected srv_sk", "Check srv_sk output. egress_linum:%u",
+             egress_linum);
 
        CHECK(cli_sk.state == 10 ||
              !cli_sk.state ||
@@ -154,21 +210,31 @@ static void check_result(void)
                     sizeof(cli_sk.dst_ip6)) ||
              cli_sk.src_port != ntohs(cli_sa6.sin6_port) ||
              cli_sk.dst_port != srv_sa6.sin6_port,
-             "Unexpected cli_sk", "Check cli_sk output. linum:%u", linum);
+             "Unexpected cli_sk", "Check cli_sk output. egress_linum:%u",
+             egress_linum);
+
+       CHECK(listen_tp.data_segs_out ||
+             listen_tp.data_segs_in ||
+             listen_tp.total_retrans ||
+             listen_tp.bytes_acked,
+             "Unexpected listen_tp", "Check listen_tp output. ingress_linum:%u",
+             ingress_linum);
 
        CHECK(srv_tp.data_segs_out != 1 ||
              srv_tp.data_segs_in ||
              srv_tp.snd_cwnd != 10 ||
              srv_tp.total_retrans ||
              srv_tp.bytes_acked != DATA_LEN,
-             "Unexpected srv_tp", "Check srv_tp output. linum:%u", linum);
+             "Unexpected srv_tp", "Check srv_tp output. egress_linum:%u",
+             egress_linum);
 
        CHECK(cli_tp.data_segs_out ||
              cli_tp.data_segs_in != 1 ||
              cli_tp.snd_cwnd != 10 ||
              cli_tp.total_retrans ||
              cli_tp.bytes_received != DATA_LEN,
-             "Unexpected cli_tp", "Check cli_tp output. linum:%u", linum);
+             "Unexpected cli_tp", "Check cli_tp output. egress_linum:%u",
+             egress_linum);
 }
 
 static void test(void)
@@ -211,10 +277,10 @@ static void test(void)
              err, errno);
 
        /* Update addr_map with srv_sa6 and cli_sa6 */
-       err = bpf_map_update_elem(addr_map_fd, &srv_idx, &srv_sa6, 0);
+       err = bpf_map_update_elem(addr_map_fd, &addr_srv_idx, &srv_sa6, 0);
        CHECK(err, "map_update", "err:%d errno:%d", err, errno);
 
-       err = bpf_map_update_elem(addr_map_fd, &cli_idx, &cli_sa6, 0);
+       err = bpf_map_update_elem(addr_map_fd, &addr_cli_idx, &cli_sa6, 0);
        CHECK(err, "map_update", "err:%d errno:%d", err, errno);
 
        /* Connect from cli_sa6 to srv_sa6 */
@@ -273,9 +339,9 @@ int main(int argc, char **argv)
        struct bpf_prog_load_attr attr = {
                .file = "test_sock_fields_kern.o",
                .prog_type = BPF_PROG_TYPE_CGROUP_SKB,
-               .expected_attach_type = BPF_CGROUP_INET_EGRESS,
        };
-       int cgroup_fd, prog_fd, err;
+       int cgroup_fd, egress_fd, ingress_fd, err;
+       struct bpf_program *ingress_prog;
        struct bpf_object *obj;
        struct bpf_map *map;
 
@@ -293,12 +359,24 @@ int main(int argc, char **argv)
        err = join_cgroup(TEST_CGROUP);
        CHECK(err, "join_cgroup", "err:%d errno:%d", err, errno);
 
-       err = bpf_prog_load_xattr(&attr, &obj, &prog_fd);
+       err = bpf_prog_load_xattr(&attr, &obj, &egress_fd);
        CHECK(err, "bpf_prog_load_xattr()", "err:%d", err);
 
-       err = bpf_prog_attach(prog_fd, cgroup_fd, BPF_CGROUP_INET_EGRESS, 0);
+       ingress_prog = bpf_object__find_program_by_title(obj,
+                                                        "cgroup_skb/ingress");
+       CHECK(!ingress_prog,
+             "bpf_object__find_program_by_title(cgroup_skb/ingress)",
+             "not found");
+       ingress_fd = bpf_program__fd(ingress_prog);
+
+       err = bpf_prog_attach(egress_fd, cgroup_fd, BPF_CGROUP_INET_EGRESS, 0);
        CHECK(err == -1, "bpf_prog_attach(CPF_CGROUP_INET_EGRESS)",
              "err:%d errno%d", err, errno);
+
+       err = bpf_prog_attach(ingress_fd, cgroup_fd,
+                             BPF_CGROUP_INET_INGRESS, 0);
+       CHECK(err == -1, "bpf_prog_attach(CPF_CGROUP_INET_INGRESS)",
+             "err:%d errno%d", err, errno);
        close(cgroup_fd);
 
        map = bpf_object__find_map_by_name(obj, "addr_map");