bpf: selftest: Test batching and bpf_(get|set)sockopt in bpf tcp iter

This patch adds tests for the batching and bpf_(get|set)sockopt in
bpf tcp iter.

It first creates:
a) 1 non SO_REUSEPORT listener in lhash2.
b) 256 passive and active fds connected to the listener in (a).
c) 256 SO_REUSEPORT listeners in one of the lhash2 bucket.

The test sets all listeners and connections to bpf_cubic before
running the bpf iter.

The bpf iter then calls setsockopt(TCP_CONGESTION) to switch
each listener and connection from bpf_cubic to bpf_dctcp.

The bpf iter has a random_retry mode such that it can return EAGAIN
to the usespace in the middle of a batch.

Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Acked-by: Kuniyuki Iwashima <kuniyu@amazon.co.jp>
Acked-by: Yonghong Song <yhs@fb.com>
Link: https://lore.kernel.org/bpf/20210701200625.1036874-1-kafai@fb.com

1 files changed, 72 insertions, 0 deletions

diff --git a/tools/testing/selftests/bpf/progs/bpf_iter_setsockopt.c b/tools/testing/selftests/bpf/progs/bpf_iter_setsockopt.c
new file mode 100644
index 000000000000..b77adfd55d73
--- /dev/null
+++ b/tools/testing/selftests/bpf/progs/bpf_iter_setsockopt.c
@@ -0,0 +1,72 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2021 Facebook */
+#include "bpf_iter.h"
+#include "bpf_tracing_net.h"
+#include <bpf/bpf_helpers.h>
+#include <bpf/bpf_endian.h>
+
+#define bpf_tcp_sk(skc)	({				\
+	struct sock_common *_skc = skc;			\
+	sk = NULL;					\
+	tp = NULL;					\
+	if (_skc) {					\
+		tp = bpf_skc_to_tcp_sock(_skc);		\
+		sk = (struct sock *)tp;			\
+	}						\
+	tp;						\
+})
+
+unsigned short reuse_listen_hport = 0;
+unsigned short listen_hport = 0;
+char cubic_cc[TCP_CA_NAME_MAX] = "bpf_cubic";
+char dctcp_cc[TCP_CA_NAME_MAX] = "bpf_dctcp";
+bool random_retry = false;
+
+static bool tcp_cc_eq(const char *a, const char *b)
+{
+	int i;
+
+	for (i = 0; i < TCP_CA_NAME_MAX; i++) {
+		if (a[i] != b[i])
+			return false;
+		if (!a[i])
+			break;
+	}
+
+	return true;
+}
+
+SEC("iter/tcp")
+int change_tcp_cc(struct bpf_iter__tcp *ctx)
+{
+	char cur_cc[TCP_CA_NAME_MAX];
+	struct tcp_sock *tp;
+	struct sock *sk;
+	int ret;
+
+	if (!bpf_tcp_sk(ctx->sk_common))
+		return 0;
+
+	if (sk->sk_family != AF_INET6 ||
+	    (sk->sk_state != TCP_LISTEN &&
+	     sk->sk_state != TCP_ESTABLISHED) ||
+	    (sk->sk_num != reuse_listen_hport &&
+	     sk->sk_num != listen_hport &&
+	     bpf_ntohs(sk->sk_dport) != listen_hport))
+		return 0;
+
+	if (bpf_getsockopt(tp, SOL_TCP, TCP_CONGESTION,
+			   cur_cc, sizeof(cur_cc)))
+		return 0;
+
+	if (!tcp_cc_eq(cur_cc, cubic_cc))
+		return 0;
+
+	if (random_retry && bpf_get_prandom_u32() % 4 == 1)
+		return 1;
+
+	bpf_setsockopt(tp, SOL_TCP, TCP_CONGESTION, dctcp_cc, sizeof(dctcp_cc));
+	return 0;
+}
+
+char _license[] SEC("license") = "GPL";