[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <ae95a774-2218-4ddc-b2e0-d7bac2b731fd@linux.dev>
Date: Tue, 27 May 2025 17:51:37 -0700
From: Martin KaFai Lau <martin.lau@...ux.dev>
To: Jordan Rife <jordan@...fe.io>
Cc: Daniel Borkmann <daniel@...earbox.net>,
Willem de Bruijn <willemdebruijn.kernel@...il.com>,
Kuniyuki Iwashima <kuniyu@...zon.com>,
Alexei Starovoitov <alexei.starovoitov@...il.com>, netdev@...r.kernel.org,
bpf@...r.kernel.org
Subject: Re: [PATCH v1 bpf-next 10/10] selftests/bpf: Add tests for bucket
resume logic in established sockets
On 5/20/25 7:50 AM, Jordan Rife wrote:
> +static bool close_and_wait(int fd, struct bpf_link *link)
> +{
> + static const int us_per_ms = 1000;
> + __u64 cookie = socket_cookie(fd);
> + struct iter_out out;
> + bool exists = true;
> + int iter_fd, nread;
> + int waits = 20; /* 2 seconds */
> +
> + close(fd);
> +
> + /* Wait for socket to disappear from the ehash table. */
> + while (waits--) {
> + iter_fd = bpf_iter_create(bpf_link__fd(link));
> + if (!ASSERT_OK_FD(iter_fd, "bpf_iter_create"))
> + return false;
> +
> + /* Is it still there? */
> + do {
> + nread = read(iter_fd, &out, sizeof(out));
> + if (!ASSERT_GE(nread, 0, "nread")) {
> + close(iter_fd);
> + return false;
> + }
> + exists = nread && cookie == out.cookie;
> + } while (!exists && nread);
> +
> + close(iter_fd);
> +
> + if (!exists)
> + break;
> +
> + usleep(100 * us_per_ms);
Instead of retrying with the bpf_iter_tcp to confirm the sk is gone from the
ehash table, I think the bpf_sock_destroy() can help here.
> + }
> +
> + return !exists;
> +}
> +
> static int get_seen_count(int fd, struct sock_count counts[], int n)
> {
> __u64 cookie = socket_cookie(fd);
> @@ -241,6 +279,43 @@ static void remove_seen(int family, int sock_type, const char *addr, __u16 port,
> counts_len);
> }
>
> +static void remove_seen_established(int family, int sock_type, const char *addr,
> + __u16 port, int *listen_socks,
> + int listen_socks_len, int *established_socks,
> + int established_socks_len,
> + struct sock_count *counts, int counts_len,
> + struct bpf_link *link, int iter_fd)
> +{
> + int close_idx;
> +
> + /* Iterate through all listening sockets. */
> + read_n(iter_fd, listen_socks_len, counts, counts_len);
> +
> + /* Make sure we saw all listening sockets exactly once. */
> + check_n_were_seen_once(listen_socks, listen_socks_len, listen_socks_len,
> + counts, counts_len);
> +
> + /* Leave one established socket. */
> + read_n(iter_fd, established_socks_len - 1, counts, counts_len);
hmm... In the "SEC("iter/tcp") int iter_tcp_soreuse(...)" bpf prog, there is a
"sk->sk_state != TCP_LISTEN" check and the established sk should have been
skipped. Does it have an existing bug? I suspect it is missing a "()" around
"sk->sk_family == AF_INET6 ? !ipv6_addr_loopback(...) : ...".
Powered by blists - more mailing lists