selftests/bpf: De-flake test_tcpbpf

It looks like BPF program that handles BPF_SOCK_OPS_STATE_CB state
can race with the bpf_map_lookup_elem("global_map"); I sometimes
see the failures in this test and re-running helps.

Since we know that we expect the callback to be called 3 times (one
time for listener socket, two times for both ends of the connection),
let's export this number and add simple retry logic around that.

Also, let's make EXPECT_EQ() not return on failure, but continue
evaluating all conditions; that should make potential debugging
easier.

With this fix in place I don't observe the flakiness anymore.

Signed-off-by: Stanislav Fomichev <[email protected]>
Signed-off-by: Alexei Starovoitov <[email protected]>
Cc: Lawrence Brakmo <[email protected]>
Link: https://lore.kernel.org/bpf/[email protected]

tools/testing/selftests/bpf/progs/test_tcpbpf_kern.c

@@ -131,6 +131,7 @@ int bpf_testcb(struct bpf_sock_ops *skops)
 				g.bytes_received = skops->bytes_received;
 				g.bytes_acked = skops->bytes_acked;
 			}
+			g.num_close_events++;
 			bpf_map_update_elem(&global_map, &key, &g,
 					    BPF_ANY);
 		}

tools/testing/selftests/bpf/test_tcpbpf.h

@@ -13,5 +13,6 @@ struct tcpbpf_globals {
 	__u64 bytes_received;
 	__u64 bytes_acked;
 	__u32 num_listen;
+	__u32 num_close_events;
 };
 #endif

tools/testing/selftests/bpf/test_tcpbpf_user.c

@@ -16,20 +16,24 @@
 
 #include "test_tcpbpf.h"
 
+/* 3 comes from one listening socket + both ends of the connection */
+#define EXPECTED_CLOSE_EVENTS		3
+
 #define EXPECT_EQ(expected, actual, fmt)			\
 	do {							\
 		if ((expected) != (actual)) {			\
 			printf("  Value of: " #actual "\n"	\
 			       "    Actual: %" fmt "\n"		\
 			       "  Expected: %" fmt "\n",	\
 			       (actual), (expected));		\
-			goto err;				\
+			ret--;					\
 		}						\
 	} while (0)
 
 int verify_result(const struct tcpbpf_globals *result)
 {
 	__u32 expected_events;
+	int ret = 0;
 
 	expected_events = ((1 << BPF_SOCK_OPS_TIMEOUT_INIT) |
 			   (1 << BPF_SOCK_OPS_RWND_INIT) |
@@ -48,15 +52,15 @@ int verify_result(const struct tcpbpf_globals *result)
 	EXPECT_EQ(0x80, result->bad_cb_test_rv, PRIu32);
 	EXPECT_EQ(0, result->good_cb_test_rv, PRIu32);
 	EXPECT_EQ(1, result->num_listen, PRIu32);
+	EXPECT_EQ(EXPECTED_CLOSE_EVENTS, result->num_close_events, PRIu32);
 
-	return 0;
-err:
-	return -1;
+	return ret;
 }
 
 int verify_sockopt_result(int sock_map_fd)
 {
 	__u32 key = 0;
+	int ret = 0;
 	int res;
 	int rv;
 
@@ -69,9 +73,7 @@ int verify_sockopt_result(int sock_map_fd)
 	rv = bpf_map_lookup_elem(sock_map_fd, &key, &res);
 	EXPECT_EQ(0, rv, "d");
 	EXPECT_EQ(1, res, "d");
-	return 0;
-err:
-	return -1;
+	return ret;
 }
 
 static int bpf_find_map(const char *test, struct bpf_object *obj,
@@ -96,6 +98,7 @@ int main(int argc, char **argv)
 	int error = EXIT_FAILURE;
 	struct bpf_object *obj;
 	int cg_fd = -1;
+	int retry = 10;
 	__u32 key = 0;
 	int rv;
 
@@ -134,12 +137,20 @@ int main(int argc, char **argv)
 	if (sock_map_fd < 0)
 		goto err;
 
+retry_lookup:
 	rv = bpf_map_lookup_elem(map_fd, &key, &g);
 	if (rv != 0) {
 		printf("FAILED: bpf_map_lookup_elem returns %d\n", rv);
 		goto err;
 	}
 
+	if (g.num_close_events != EXPECTED_CLOSE_EVENTS && retry--) {
+		printf("Unexpected number of close events (%d), retrying!\n",
+		       g.num_close_events);
+		usleep(100);
+		goto retry_lookup;
+	}
+
 	if (verify_result(&g)) {
 		printf("FAILED: Wrong stats\n");
 		goto err;