bpf, test_run: Propagate bpf_flow_dissect's retval to user's bpf_attr.test.retval
authorShmulik Ladkani <shmulik.ladkani@gmail.com>
Sun, 21 Aug 2022 11:35:18 +0000 (14:35 +0300)
committerDaniel Borkmann <daniel@iogearbox.net>
Tue, 23 Aug 2022 20:48:03 +0000 (22:48 +0200)
Formerly, a boolean denoting whether bpf_flow_dissect returned BPF_OK
was set into 'bpf_attr.test.retval'.

Augment this, so users can check the actual return code of the dissector
program under test.

Existing prog_tests/flow_dissector*.c tests were correspondingly changed
to check against each test's expected retval.

Also, tests' resulting 'flow_keys' are verified only in case the expected
retval is BPF_OK. This allows adding new tests that expect non BPF_OK.

Signed-off-by: Shmulik Ladkani <shmulik.ladkani@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Reviewed-by: Stanislav Fomichev <sdf@google.com>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Link: https://lore.kernel.org/bpf/20220821113519.116765-4-shmulik.ladkani@gmail.com
net/bpf/test_run.c
tools/testing/selftests/bpf/prog_tests/flow_dissector.c
tools/testing/selftests/bpf/prog_tests/flow_dissector_load_bytes.c

index 51c4794..25d8ecf 100644 (file)
@@ -1445,7 +1445,7 @@ int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog,
        bpf_test_timer_enter(&t);
        do {
                retval = bpf_flow_dissect(prog, &ctx, eth->h_proto, ETH_HLEN,
-                                         size, flags) == BPF_OK;
+                                         size, flags);
        } while (bpf_test_timer_continue(&t, 1, repeat, &ret, &duration));
        bpf_test_timer_leave(&t);
 
index 0c1661e..8fa3c45 100644 (file)
@@ -100,6 +100,7 @@ struct test {
        } pkt;
        struct bpf_flow_keys keys;
        __u32 flags;
+       __u32 retval;
 };
 
 #define VLAN_HLEN      4
@@ -126,6 +127,7 @@ struct test tests[] = {
                        .sport = 80,
                        .dport = 8080,
                },
+               .retval = BPF_OK,
        },
        {
                .name = "ipv6",
@@ -146,6 +148,7 @@ struct test tests[] = {
                        .sport = 80,
                        .dport = 8080,
                },
+               .retval = BPF_OK,
        },
        {
                .name = "802.1q-ipv4",
@@ -168,6 +171,7 @@ struct test tests[] = {
                        .sport = 80,
                        .dport = 8080,
                },
+               .retval = BPF_OK,
        },
        {
                .name = "802.1ad-ipv6",
@@ -191,6 +195,7 @@ struct test tests[] = {
                        .sport = 80,
                        .dport = 8080,
                },
+               .retval = BPF_OK,
        },
        {
                .name = "ipv4-frag",
@@ -217,6 +222,7 @@ struct test tests[] = {
                        .dport = 8080,
                },
                .flags = BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG,
+               .retval = BPF_OK,
        },
        {
                .name = "ipv4-no-frag",
@@ -239,6 +245,7 @@ struct test tests[] = {
                        .is_frag = true,
                        .is_first_frag = true,
                },
+               .retval = BPF_OK,
        },
        {
                .name = "ipv6-frag",
@@ -265,6 +272,7 @@ struct test tests[] = {
                        .dport = 8080,
                },
                .flags = BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG,
+               .retval = BPF_OK,
        },
        {
                .name = "ipv6-no-frag",
@@ -287,6 +295,7 @@ struct test tests[] = {
                        .is_frag = true,
                        .is_first_frag = true,
                },
+               .retval = BPF_OK,
        },
        {
                .name = "ipv6-flow-label",
@@ -309,6 +318,7 @@ struct test tests[] = {
                        .dport = 8080,
                        .flow_label = __bpf_constant_htonl(0xbeeef),
                },
+               .retval = BPF_OK,
        },
        {
                .name = "ipv6-no-flow-label",
@@ -331,6 +341,7 @@ struct test tests[] = {
                        .flow_label = __bpf_constant_htonl(0xbeeef),
                },
                .flags = BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL,
+               .retval = BPF_OK,
        },
        {
                .name = "ipip-encap",
@@ -359,6 +370,7 @@ struct test tests[] = {
                        .sport = 80,
                        .dport = 8080,
                },
+               .retval = BPF_OK,
        },
        {
                .name = "ipip-no-encap",
@@ -386,6 +398,7 @@ struct test tests[] = {
                        .is_encap = true,
                },
                .flags = BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP,
+               .retval = BPF_OK,
        },
 };
 
@@ -503,6 +516,10 @@ static void run_tests_skb_less(int tap_fd, struct bpf_map *keys)
                err = tx_tap(tap_fd, &tests[i].pkt, sizeof(tests[i].pkt));
                CHECK(err < 0, "tx_tap", "err %d errno %d\n", err, errno);
 
+               /* check the stored flow_keys only if BPF_OK expected */
+               if (tests[i].retval != BPF_OK)
+                       continue;
+
                err = bpf_map_lookup_elem(keys_fd, &key, &flow_keys);
                ASSERT_OK(err, "bpf_map_lookup_elem");
 
@@ -588,7 +605,11 @@ void test_flow_dissector(void)
 
                err = bpf_prog_test_run_opts(prog_fd, &topts);
                ASSERT_OK(err, "test_run");
-               ASSERT_EQ(topts.retval, 1, "test_run retval");
+               ASSERT_EQ(topts.retval, tests[i].retval, "test_run retval");
+
+               /* check the resulting flow_keys only if BPF_OK returned */
+               if (topts.retval != BPF_OK)
+                       continue;
                ASSERT_EQ(topts.data_size_out, sizeof(flow_keys),
                          "test_run data_size_out");
                CHECK_FLOW_KEYS(tests[i].name, flow_keys, tests[i].keys);
index 36afb40..c7a47b5 100644 (file)
@@ -44,7 +44,7 @@ void serial_test_flow_dissector_load_bytes(void)
        ASSERT_OK(err, "test_run");
        ASSERT_EQ(topts.data_size_out, sizeof(flow_keys),
                  "test_run data_size_out");
-       ASSERT_EQ(topts.retval, 1, "test_run retval");
+       ASSERT_EQ(topts.retval, BPF_OK, "test_run retval");
 
        if (fd >= -1)
                close(fd);