Re: [PATCH bpf-next v4 4/4] selftests/bpf: Verify bpf-tcp-cc rejects TCP_NODELAY

[sashiko-bot@kernel.org]

Sashiko AI review found 2 potential issue(s):
- [Medium] Test logic allows vacuous passes if BPF callbacks are never executed.
- [Low] Type mismatch for `bpf_setsockopt` return value.
--

commit 8a4ced8128582d3995376457598fd25202c0468a
Author: KaFai Wan <kafai.wan@linux.dev>

selftests/bpf: Verify bpf-tcp-cc rejects TCP_NODELAY

This commit adds a bpf_tcp_ca selftest to verify that the TCP_NODELAY
restriction is properly enforced in bpf-tcp-cc. It updates bpf_cubic to
exercise the init() and cwnd_event_tx_start() callbacks, ensuring both
reject the option with -EOPNOTSUPP.

> diff --git a/tools/testing/selftests/bpf/progs/bpf_cubic.c b/tools/testing/selftests/bpf/progs/bpf_cubic.c
> index ce18a4db813fa..b941ab3ebad57 100644
> --- a/tools/testing/selftests/bpf/progs/bpf_cubic.c
> +++ b/tools/testing/selftests/bpf/progs/bpf_cubic.c

[ ... ]

> @@ -170,10 +171,17 @@ static void bictcp_hystart_reset(struct sock *sk)
>  	ca->sample_cnt = 0;
>  }
>  
> +bool nodelay_init_reject = true;
> +bool nodelay_cwnd_event_tx_start_reject = true;

Can this test pass vacuously? Since these boolean variables are statically
initialized to true and the callbacks update them using a bitwise AND, the
userspace assertions will pass even if the callbacks are never executed.

Would it be safer to add explicit execution counters (like the existing
bpf_cubic_acked_called) to ensure the callbacks actually run during the
test workload?

> +
>  SEC("struct_ops")
>  void BPF_PROG(bpf_cubic_init, struct sock *sk)
>  {
>  	struct bpf_bictcp *ca = inet_csk_ca(sk);
> +	int true_val = 1, ret;
> +
> +	ret = bpf_setsockopt(sk, SOL_TCP, TCP_NODELAY, &true_val, sizeof(true_val));

This isn't a bug since we only expect a small negative constant here, but
bpf_setsockopt() returns a long. Could assigning the return value to an int
cause unintended truncation issues in the future?

> +	nodelay_init_reject &= ret == -EOPNOTSUPP;
>  
>  	bictcp_reset(ca);
>

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260421155804.135786-1-kafai.wan@linux.dev?part=4