[PATCH bpf-next 0/5] selftests/bpf: Extend sockmap_redir to test no-redir SK_DROP/SK_PASS

[PATCH bpf-next 0/5] selftests/bpf: Extend sockmap_redir to test no-redir SK_DROP/SK_PASS

[Michal Luczaj]

sockmap_redir was introduced to comprehensively test the BPF redirection.
This series strives to increase the tested sockmap/sockhash code coverage;
adds support for skipping the actual redirect part, allowing to simply
SK_DROP or SK_PASS the packet.

BPF_MAP_TYPE_SOCKMAP
BPF_MAP_TYPE_SOCKHASH
	x
sk_msg-to-egress
sk_msg-to-ingress
sk_skb-to-egress
sk_skb-to-ingress
	x
AF_INET, SOCK_STREAM
AF_INET6, SOCK_STREAM
AF_INET, SOCK_DGRAM
AF_INET6, SOCK_DGRAM
AF_UNIX, SOCK_STREAM
AF_UNIX, SOCK_DGRAM
AF_VSOCK, SOCK_STREAM
AF_VSOCK, SOCK_SEQPACKET
	x
SK_REDIRECT
SK_DROP
SK_PASS

Patch 5 ("Support no-redirect SK_DROP/SK_PASS") implements the feature.
Patches 3 ("Rename functions") and 4 ("Let test specify skel's
redirect_type") make preparatory changes.

I also took the opportunity to clean up (Patch 1, "Simplify try_recv()")
and improve a bit (Patch 2, "Fix OOB handling").

$ cd tools/testing/selftests/bpf
$ make
$ sudo ./test_progs -t sockmap_redir
...
Summary: 1/720 PASSED, 0 SKIPPED, 0 FAILED

Signed-off-by: Michal Luczaj <mhal@rbox.co>
---
Michal Luczaj (5):
      selftests/bpf: sockmap_redir: Simplify try_recv()
      selftests/bpf: sockmap_redir: Fix OOB handling
      selftests/bpf: sockmap_redir: Rename functions
      selftests/bpf: sockmap_redir: Let test specify skel's redirect_type
      selftests/bpf: sockmap_redir: Support no-redirect SK_DROP/SK_PASS

 .../selftests/bpf/prog_tests/sockmap_redir.c       | 143 +++++++++++++++------
 1 file changed, 105 insertions(+), 38 deletions(-)
---
base-commit: e8a6a9d3e8cc539d281e77b9df2439f223ec8153
change-id: 20250523-redir-test-pass-drop-2f2a5edca6e1

Best regards,
-- 
Michal Luczaj <mhal@rbox.co>

[PATCH bpf-next 1/5] selftests/bpf: sockmap_redir: Simplify try_recv()

[Michal Luczaj]

try_recv() was meant to support both @expect_success cases, but all the
callers use @expect_success=false anyway. Drop the unused logic and fold in
MSG_DONTWAIT. Adapt callers.

Subtle change here: recv() return value of 0 will also be considered (an
unexpected) success.

Signed-off-by: Michal Luczaj <mhal@rbox.co>
---
 .../selftests/bpf/prog_tests/sockmap_redir.c       | 25 +++++++++-------------
 1 file changed, 10 insertions(+), 15 deletions(-)

diff --git a/tools/testing/selftests/bpf/prog_tests/sockmap_redir.c b/tools/testing/selftests/bpf/prog_tests/sockmap_redir.c
index 9c461d93113db20de65ac353f92dfdbe32ffbd3b..c1bf1076e8152b7d83c3e07e2dce746b5a39cf7e 100644
--- a/tools/testing/selftests/bpf/prog_tests/sockmap_redir.c
+++ b/tools/testing/selftests/bpf/prog_tests/sockmap_redir.c
@@ -144,17 +144,14 @@ static void get_redir_params(struct redir_spec *redir,
 		*redirect_flags = 0;
 }
 
-static void try_recv(const char *prefix, int fd, int flags, bool expect_success)
+static void fail_recv(const char *prefix, int fd, int more_flags)
 {
 	ssize_t n;
 	char buf;
 
-	errno = 0;
-	n = recv(fd, &buf, 1, flags);
-	if (n < 0 && expect_success)
-		FAIL_ERRNO("%s: unexpected failure: retval=%zd", prefix, n);
-	if (!n && !expect_success)
-		FAIL("%s: expected failure: retval=%zd", prefix, n);
+	n = recv(fd, &buf, 1, MSG_DONTWAIT | more_flags);
+	if (n >= 0)
+		FAIL("%s: unexpected success: retval=%zd", prefix, n);
 }
 
 static void handle_unsupported(int sd_send, int sd_peer, int sd_in, int sd_out,
@@ -188,13 +185,13 @@ static void handle_unsupported(int sd_send, int sd_peer, int sd_in, int sd_out,
 	}
 
 	/* Ensure queues are empty */
-	try_recv("bpf.recv(sd_send)", sd_send, MSG_DONTWAIT, false);
+	fail_recv("bpf.recv(sd_send)", sd_send, 0);
 	if (sd_in != sd_send)
-		try_recv("bpf.recv(sd_in)", sd_in, MSG_DONTWAIT, false);
+		fail_recv("bpf.recv(sd_in)", sd_in, 0);
 
-	try_recv("bpf.recv(sd_out)", sd_out, MSG_DONTWAIT, false);
+	fail_recv("bpf.recv(sd_out)", sd_out, 0);
 	if (sd_recv != sd_out)
-		try_recv("bpf.recv(sd_recv)", sd_recv, MSG_DONTWAIT, false);
+		fail_recv("bpf.recv(sd_recv)", sd_recv, 0);
 }
 
 static void test_send_redir_recv(int sd_send, int send_flags, int sd_peer,
@@ -257,15 +254,13 @@ static void test_send_redir_recv(int sd_send, int send_flags, int sd_peer,
 
 	if (send_flags & MSG_OOB) {
 		/* Fail reading OOB while in sockmap */
-		try_recv("bpf.recv(sd_out, MSG_OOB)", sd_out,
-			 MSG_OOB | MSG_DONTWAIT, false);
+		fail_recv("bpf.recv(sd_out, MSG_OOB)", sd_out, MSG_OOB);
 
 		/* Remove sd_out from sockmap */
 		xbpf_map_delete_elem(maps->out, &u32(0));
 
 		/* Check that OOB was dropped on redirect */
-		try_recv("recv(sd_out, MSG_OOB)", sd_out,
-			 MSG_OOB | MSG_DONTWAIT, false);
+		fail_recv("recv(sd_out, MSG_OOB)", sd_out, MSG_OOB);
 
 		goto del_in;
 	}

-- 
2.50.1

[PATCH bpf-next 2/5] selftests/bpf: sockmap_redir: Fix OOB handling

[Michal Luczaj]

In some test cases, OOB packets might have been left unread. Flush them out
and introduce additional checks.

Signed-off-by: Michal Luczaj <mhal@rbox.co>
---
 tools/testing/selftests/bpf/prog_tests/sockmap_redir.c | 16 ++++++++++++++++
 1 file changed, 16 insertions(+)

diff --git a/tools/testing/selftests/bpf/prog_tests/sockmap_redir.c b/tools/testing/selftests/bpf/prog_tests/sockmap_redir.c
index c1bf1076e8152b7d83c3e07e2dce746b5a39cf7e..4997e72c14345b274367f3f2f4115c39d1ae48c9 100644
--- a/tools/testing/selftests/bpf/prog_tests/sockmap_redir.c
+++ b/tools/testing/selftests/bpf/prog_tests/sockmap_redir.c
@@ -184,6 +184,19 @@ static void handle_unsupported(int sd_send, int sd_peer, int sd_in, int sd_out,
 			FAIL_ERRNO("unsupported: packet missing, retval=%zd", n);
 	}
 
+	/* af_unix send("ab", MSG_OOB) spits out 2 packets, but only the latter
+	 * ("b") is designated OOB. If the peer is in a sockmap, the OOB packet
+	 * will be silently dropped. Otherwise OOB stays in the queue and should
+	 * be taken care of.
+	 */
+	if ((send_flags & MSG_OOB) && !pass && !drop) {
+		errno = 0;
+		n = recv_timeout(sd_peer, &recv_buf, 1, MSG_OOB, IO_TIMEOUT_SEC);
+		/* Ignore unsupported sk_msg error */
+		if (n != 1 && errno != EOPNOTSUPP)
+			FAIL_ERRNO("recv(OOB): retval=%zd", n);
+	}
+
 	/* Ensure queues are empty */
 	fail_recv("bpf.recv(sd_send)", sd_send, 0);
 	if (sd_in != sd_send)
@@ -192,6 +205,9 @@ static void handle_unsupported(int sd_send, int sd_peer, int sd_in, int sd_out,
 	fail_recv("bpf.recv(sd_out)", sd_out, 0);
 	if (sd_recv != sd_out)
 		fail_recv("bpf.recv(sd_recv)", sd_recv, 0);
+
+	fail_recv("recv(sd_peer, OOB)", sd_peer, MSG_OOB);
+	fail_recv("recv(sd_out, OOB)", sd_out, MSG_OOB);
 }
 
 static void test_send_redir_recv(int sd_send, int send_flags, int sd_peer,

-- 
2.50.1

[PATCH bpf-next 3/5] selftests/bpf: sockmap_redir: Rename functions

[Michal Luczaj]

Preparatory patch before adding SK_PASS/SK_DROP support:

test_redir() => test_sockets()
test_socket() => test_redir()
test_send_redir_recv() => test_send_recv()

After the change (and the following patch) the call stack will be:

serial_test_sockmap_redir
  test_map
    test_sockets
      test_redir
        test_send_recv
      (test_verdict)
        (test_send_recv)

Signed-off-by: Michal Luczaj <mhal@rbox.co>
---
 .../selftests/bpf/prog_tests/sockmap_redir.c       | 25 +++++++++++-----------
 1 file changed, 12 insertions(+), 13 deletions(-)

diff --git a/tools/testing/selftests/bpf/prog_tests/sockmap_redir.c b/tools/testing/selftests/bpf/prog_tests/sockmap_redir.c
index 4997e72c14345b274367f3f2f4115c39d1ae48c9..9fc5b4dd323d8554e6dc89d06be9054612e41020 100644
--- a/tools/testing/selftests/bpf/prog_tests/sockmap_redir.c
+++ b/tools/testing/selftests/bpf/prog_tests/sockmap_redir.c
@@ -210,9 +210,8 @@ static void handle_unsupported(int sd_send, int sd_peer, int sd_in, int sd_out,
 	fail_recv("recv(sd_out, OOB)", sd_out, MSG_OOB);
 }
 
-static void test_send_redir_recv(int sd_send, int send_flags, int sd_peer,
-				 int sd_in, int sd_out, int sd_recv,
-				 struct maps *maps, int status)
+static void test_send_recv(int sd_send, int send_flags, int sd_peer, int sd_in,
+			   int sd_out, int sd_recv, struct maps *maps, int status)
 {
 	unsigned int drop, pass;
 	char *send_buf = "ab";
@@ -335,9 +334,9 @@ static int get_support_status(enum prog_type type, const char *in,
 	return status;
 }
 
-static void test_socket(enum bpf_map_type type, struct redir_spec *redir,
-			struct maps *maps, struct socket_spec *s_in,
-			struct socket_spec *s_out)
+static void test_redir(enum bpf_map_type type, struct redir_spec *redir,
+		       struct maps *maps, struct socket_spec *s_in,
+		       struct socket_spec *s_out)
 {
 	int fd_in, fd_out, fd_send, fd_peer, fd_recv, flags, status;
 	const char *in_str, *out_str;
@@ -367,12 +366,12 @@ static void test_socket(enum bpf_map_type type, struct redir_spec *redir,
 	if (!test__start_subtest(s))
 		return;
 
-	test_send_redir_recv(fd_send, flags, fd_peer, fd_in, fd_out, fd_recv,
-			     maps, status);
+	test_send_recv(fd_send, flags, fd_peer, fd_in, fd_out, fd_recv, maps,
+		       status);
 }
 
-static void test_redir(enum bpf_map_type type, struct redir_spec *redir,
-		       struct maps *maps)
+static void test_sockets(enum bpf_map_type type, struct redir_spec *redir,
+			 struct maps *maps)
 {
 	struct socket_spec *s, sockets[] = {
 		{ AF_INET, SOCK_STREAM },
@@ -395,7 +394,7 @@ static void test_redir(enum bpf_map_type type, struct redir_spec *redir,
 
 	/* Intra-proto */
 	for (s = sockets; s < sockets + ARRAY_SIZE(sockets); s++)
-		test_socket(type, redir, maps, s, s);
+		test_redir(type, redir, maps, s, s);
 
 	/* Cross-proto */
 	for (int i = 0; i < ARRAY_SIZE(sockets); i++) {
@@ -409,7 +408,7 @@ static void test_redir(enum bpf_map_type type, struct redir_spec *redir,
 			     in->sotype == out->sotype))
 				continue;
 
-			test_socket(type, redir, maps, in, out);
+			test_redir(type, redir, maps, in, out);
 		}
 	}
 out:
@@ -460,7 +459,7 @@ static void test_map(enum bpf_map_type type)
 		if (xbpf_prog_attach(prog_fd, maps.in, attach_type, 0))
 			return;
 
-		test_redir(type, r, &maps);
+		test_sockets(type, r, &maps);
 
 		if (xbpf_prog_detach2(prog_fd, maps.in, attach_type))
 			return;

-- 
2.50.1

[PATCH bpf-next 4/5] selftests/bpf: sockmap_redir: Let test specify skel's redirect_type

[Michal Luczaj]

Preparatory patch before adding SK_PASS/SK_DROP support: allow to
dynamically switch BPF program's redirect_type. This way, after setting up
for a redirection, test can make the BPF program skip the actual
bpf_{sk,msg}_redirect_{map,hash} part and return a specified verdict.

Signed-off-by: Michal Luczaj <mhal@rbox.co>
---
 tools/testing/selftests/bpf/prog_tests/sockmap_redir.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/tools/testing/selftests/bpf/prog_tests/sockmap_redir.c b/tools/testing/selftests/bpf/prog_tests/sockmap_redir.c
index 9fc5b4dd323d8554e6dc89d06be9054612e41020..f89df1ca1174800d801f221a819b099abf911450 100644
--- a/tools/testing/selftests/bpf/prog_tests/sockmap_redir.c
+++ b/tools/testing/selftests/bpf/prog_tests/sockmap_redir.c
@@ -371,7 +371,7 @@ static void test_redir(enum bpf_map_type type, struct redir_spec *redir,
 }
 
 static void test_sockets(enum bpf_map_type type, struct redir_spec *redir,
-			 struct maps *maps)
+			 struct maps *maps, int *skel_redir_type)
 {
 	struct socket_spec *s, sockets[] = {
 		{ AF_INET, SOCK_STREAM },
@@ -392,6 +392,8 @@ static void test_sockets(enum bpf_map_type type, struct redir_spec *redir,
 		if (socket_spec_pairs(s))
 			goto out;
 
+	*skel_redir_type = type;
+
 	/* Intra-proto */
 	for (s = sockets; s < sockets + ARRAY_SIZE(sockets); s++)
 		test_redir(type, redir, maps, s, s);
@@ -451,7 +453,6 @@ static void test_map(enum bpf_map_type type)
 			return;
 		}
 
-		skel->bss->redirect_type = type;
 		maps.verd = bpf_map__fd(skel->maps.verdict_map);
 		get_redir_params(r, skel, &prog_fd, &attach_type,
 				 &skel->bss->redirect_flags);
@@ -459,7 +460,7 @@ static void test_map(enum bpf_map_type type)
 		if (xbpf_prog_attach(prog_fd, maps.in, attach_type, 0))
 			return;
 
-		test_sockets(type, r, &maps);
+		test_sockets(type, r, &maps, &skel->bss->redirect_type);
 
 		if (xbpf_prog_detach2(prog_fd, maps.in, attach_type))
 			return;

-- 
2.50.1

[PATCH bpf-next 5/5] selftests/bpf: sockmap_redir: Support no-redirect SK_DROP/SK_PASS

[Michal Luczaj]

Add tests that make the BPF programs skip the actual redirect and
immediately return SK_DROP/SK_PASS.

Suggested-by: Jiayuan Chen <mrpre@163.com>
Signed-off-by: Michal Luczaj <mhal@rbox.co>
---
 .../selftests/bpf/prog_tests/sockmap_redir.c       | 74 +++++++++++++++++++---
 1 file changed, 65 insertions(+), 9 deletions(-)

diff --git a/tools/testing/selftests/bpf/prog_tests/sockmap_redir.c b/tools/testing/selftests/bpf/prog_tests/sockmap_redir.c
index f89df1ca1174800d801f221a819b099abf911450..4c752000bbd8c84684b2df81eb389be953d8ba02 100644
--- a/tools/testing/selftests/bpf/prog_tests/sockmap_redir.c
+++ b/tools/testing/selftests/bpf/prog_tests/sockmap_redir.c
@@ -18,6 +18,10 @@
  * AF_UNIX, SOCK_DGRAM
  * AF_VSOCK, SOCK_STREAM
  * AF_VSOCK, SOCK_SEQPACKET
+ *	x
+ * SK_REDIRECT
+ * SK_DROP
+ * SK_PASS
  */
 
 #include <errno.h>
@@ -65,6 +69,10 @@
  */
 #define UNSUPPORTED_RACY_VERD	_BITUL(1)
 
+/* Mark for an immediate SK_DROP/SK_PASS, i.e. BPF program will not redirect.
+ */
+#define NO_REDIRECT		_BITUL(2)
+
 enum prog_type {
 	SK_MSG_EGRESS,
 	SK_MSG_INGRESS,
@@ -154,8 +162,9 @@ static void fail_recv(const char *prefix, int fd, int more_flags)
 		FAIL("%s: unexpected success: retval=%zd", prefix, n);
 }
 
-static void handle_unsupported(int sd_send, int sd_peer, int sd_in, int sd_out,
-			       int sd_recv, int map_verd, int status)
+static void handle_unsupported(int sd_send, int send_flags, int sd_peer,
+			       int sd_in, int sd_out, int sd_recv,
+			       int map_verd, int status)
 {
 	unsigned int drop, pass;
 	char recv_buf;
@@ -171,7 +180,7 @@ static void handle_unsupported(int sd_send, int sd_peer, int sd_in, int sd_out,
 		goto get_verdict;
 	}
 
-	if (pass != 0) {
+	if (pass && !(status & NO_REDIRECT)) {
 		FAIL("unsupported: wanted verdict pass 0, have %u", pass);
 		return;
 	}
@@ -237,14 +246,14 @@ static void test_send_recv(int sd_send, int send_flags, int sd_peer, int sd_in,
 		FAIL("incomplete send");
 	if (n < 0) {
 		/* sk_msg redirect combo not supported? */
-		if (status & SUPPORTED || errno != EACCES)
+		if (errno != EACCES)
 			FAIL_ERRNO("send");
 		goto out;
 	}
 
-	if (!(status & SUPPORTED)) {
-		handle_unsupported(sd_send, sd_peer, sd_in, sd_out, sd_recv,
-				   maps->verd, status);
+	if (!(status & SUPPORTED) || (status & NO_REDIRECT)) {
+		handle_unsupported(sd_send, send_flags, sd_peer, sd_in, sd_out,
+				   sd_recv, maps->verd, status);
 		goto out;
 	}
 
@@ -326,9 +335,10 @@ static int is_redir_supported(enum prog_type type, const char *in,
 static int get_support_status(enum prog_type type, const char *in,
 			      const char *out)
 {
-	int status = is_redir_supported(type, in, out);
+	int status = in ? is_redir_supported(type, in, out) : 0;
 
-	if (type == SK_SKB_INGRESS && strstarts(out, "v_"))
+	if ((type == SK_SKB_INGRESS || type == SK_SKB_EGRESS) &&
+	    strstarts(out, "v_"))
 		status |= UNSUPPORTED_RACY_VERD;
 
 	return status;
@@ -370,6 +380,41 @@ static void test_redir(enum bpf_map_type type, struct redir_spec *redir,
 		       status);
 }
 
+static void test_verdict(enum bpf_map_type type, struct redir_spec *redir,
+			 struct maps *maps, struct socket_spec *s_in,
+			 enum sk_action action)
+{
+	int fd_in, fd_out, fd_send, fd_peer, fd_recv, flags, status;
+	char s[MAX_TEST_NAME];
+	const char *s_str;
+
+	fd_in = s_in->in[0];
+	fd_out = s_in->in[1];
+	fd_send = s_in->in[redir->idx_send];
+	fd_peer = s_in->in[redir->idx_send ^ 1];
+	fd_recv = s_in->in[redir->idx_send ^ 1];
+	flags = s_in->send_flags;
+
+	s_str = socket_kind_to_str(fd_in);
+	status = get_support_status(redir->prog_type, NULL, s_str);
+	status |= NO_REDIRECT;
+
+	snprintf(s, sizeof(s),
+		 "%-4s %-17s %-7s %-5s%6s",
+		 /* hash sk_skb-to-ingress u_str pass (OOB) */
+		 type == BPF_MAP_TYPE_SOCKMAP ? "map" : "hash",
+		 redir->name,
+		 s_str,
+		 action == SK_PASS ? "pass" : "drop",
+		 flags & MSG_OOB ? "(OOB)" : "");
+
+	if (!test__start_subtest(s))
+		return;
+
+	test_send_recv(fd_send, flags, fd_peer, fd_in, fd_out, fd_recv, maps,
+		       status);
+}
+
 static void test_sockets(enum bpf_map_type type, struct redir_spec *redir,
 			 struct maps *maps, int *skel_redir_type)
 {
@@ -413,6 +458,17 @@ static void test_sockets(enum bpf_map_type type, struct redir_spec *redir,
 			test_redir(type, redir, maps, in, out);
 		}
 	}
+
+	/* No redirect: SK_DROP */
+	*skel_redir_type = __MAX_BPF_MAP_TYPE + SK_DROP;
+	for (s = sockets; s < sockets + ARRAY_SIZE(sockets); s++)
+		test_verdict(type, redir, maps, s, SK_DROP);
+
+	/* No redirect: SK_PASS */
+	*skel_redir_type = __MAX_BPF_MAP_TYPE + SK_PASS;
+	for (s = sockets; s < sockets + ARRAY_SIZE(sockets); s++)
+		test_verdict(type, redir, maps, s, SK_PASS);
+
 out:
 	while (--s >= sockets)
 		socket_spec_close(s);

-- 
2.50.1

Re: [PATCH bpf-next 2/5] selftests/bpf: sockmap_redir: Fix OOB handling

[Michal Luczaj]

On 9/5/25 13:11, Michal Luczaj wrote:
> In some test cases, OOB packets might have been left unread. Flush them out
> and introduce additional checks.
> 
> Signed-off-by: Michal Luczaj <mhal@rbox.co>

Sorry, this should also have:

Fixes: f0709263a07e ("selftests/bpf: Add selftest for sockmap/hashmap
redirection")

Re: [PATCH bpf-next 1/5] selftests/bpf: sockmap_redir: Simplify try_recv()

[Jakub Sitnicki]

On Fri, Sep 05, 2025 at 01:11 PM +02, Michal Luczaj wrote:
> try_recv() was meant to support both @expect_success cases, but all the
> callers use @expect_success=false anyway. Drop the unused logic and fold in
> MSG_DONTWAIT. Adapt callers.
>
> Subtle change here: recv() return value of 0 will also be considered (an
> unexpected) success.
>
> Signed-off-by: Michal Luczaj <mhal@rbox.co>
> ---
>  .../selftests/bpf/prog_tests/sockmap_redir.c       | 25 +++++++++-------------
>  1 file changed, 10 insertions(+), 15 deletions(-)
>
> diff --git a/tools/testing/selftests/bpf/prog_tests/sockmap_redir.c b/tools/testing/selftests/bpf/prog_tests/sockmap_redir.c
> index 9c461d93113db20de65ac353f92dfdbe32ffbd3b..c1bf1076e8152b7d83c3e07e2dce746b5a39cf7e 100644
> --- a/tools/testing/selftests/bpf/prog_tests/sockmap_redir.c
> +++ b/tools/testing/selftests/bpf/prog_tests/sockmap_redir.c
> @@ -144,17 +144,14 @@ static void get_redir_params(struct redir_spec *redir,
>  		*redirect_flags = 0;
>  }
>  
> -static void try_recv(const char *prefix, int fd, int flags, bool expect_success)
> +static void fail_recv(const char *prefix, int fd, int more_flags)
>  {
>  	ssize_t n;
>  	char buf;
>  
> -	errno = 0;
> -	n = recv(fd, &buf, 1, flags);
> -	if (n < 0 && expect_success)
> -		FAIL_ERRNO("%s: unexpected failure: retval=%zd", prefix, n);
> -	if (!n && !expect_success)
> -		FAIL("%s: expected failure: retval=%zd", prefix, n);
> +	n = recv(fd, &buf, 1, MSG_DONTWAIT | more_flags);
> +	if (n >= 0)
> +		FAIL("%s: unexpected success: retval=%zd", prefix, n);
>  }

This bit, which you highlighted in the description, I don't get.

If we're expecting to receive exactly one byte, why treat a short read
as a succcess? Why not make it a strict "n != 1" check?

[...]

Re: [PATCH bpf-next 1/5] selftests/bpf: sockmap_redir: Simplify try_recv()

[Jakub Sitnicki]

On Tue, Sep 09, 2025 at 11:51 AM +02, Jakub Sitnicki wrote:
> On Fri, Sep 05, 2025 at 01:11 PM +02, Michal Luczaj wrote:
>> try_recv() was meant to support both @expect_success cases, but all the
>> callers use @expect_success=false anyway. Drop the unused logic and fold in
>> MSG_DONTWAIT. Adapt callers.
>>
>> Subtle change here: recv() return value of 0 will also be considered (an
>> unexpected) success.
>>
>> Signed-off-by: Michal Luczaj <mhal@rbox.co>
>> ---
>>  .../selftests/bpf/prog_tests/sockmap_redir.c       | 25 +++++++++-------------
>>  1 file changed, 10 insertions(+), 15 deletions(-)
>>
>> diff --git a/tools/testing/selftests/bpf/prog_tests/sockmap_redir.c b/tools/testing/selftests/bpf/prog_tests/sockmap_redir.c
>> index 9c461d93113db20de65ac353f92dfdbe32ffbd3b..c1bf1076e8152b7d83c3e07e2dce746b5a39cf7e 100644
>> --- a/tools/testing/selftests/bpf/prog_tests/sockmap_redir.c
>> +++ b/tools/testing/selftests/bpf/prog_tests/sockmap_redir.c
>> @@ -144,17 +144,14 @@ static void get_redir_params(struct redir_spec *redir,
>>  		*redirect_flags = 0;
>>  }
>>  
>> -static void try_recv(const char *prefix, int fd, int flags, bool expect_success)
>> +static void fail_recv(const char *prefix, int fd, int more_flags)
>>  {
>>  	ssize_t n;
>>  	char buf;
>>  
>> -	errno = 0;
>> -	n = recv(fd, &buf, 1, flags);
>> -	if (n < 0 && expect_success)
>> -		FAIL_ERRNO("%s: unexpected failure: retval=%zd", prefix, n);
>> -	if (!n && !expect_success)
>> -		FAIL("%s: expected failure: retval=%zd", prefix, n);
>> +	n = recv(fd, &buf, 1, MSG_DONTWAIT | more_flags);
>> +	if (n >= 0)
>> +		FAIL("%s: unexpected success: retval=%zd", prefix, n);
>>  }
>
> This bit, which you highlighted in the description, I don't get.
>
> If we're expecting to receive exactly one byte, why treat a short read
> as a succcess? Why not make it a strict "n != 1" check?
>
> [...]

Nevermind. It makes sense now. We do want to report a failure for 0-len
msg recv as well. You're effectively checking if the rcv queue is empty.

I'd add MSG_PEEK, to signal that we're _just checking_ if the socket is
readable, and turn the check into the below to succeed only when
queue is empty:

        (n != -1 || (errno != EAGAIN && errno != EWOULDBLOCK))

It's a minor thing. Leaving it up to you. Either way:

Reviewed-by: Jakub Sitnicki <jakub@cloudflare.com>

Re: [PATCH bpf-next 1/5] selftests/bpf: sockmap_redir: Simplify try_recv()

[Michal Luczaj]

On 9/9/25 12:15, Jakub Sitnicki wrote:
> On Tue, Sep 09, 2025 at 11:51 AM +02, Jakub Sitnicki wrote:
>> On Fri, Sep 05, 2025 at 01:11 PM +02, Michal Luczaj wrote:
>>> try_recv() was meant to support both @expect_success cases, but all the
>>> callers use @expect_success=false anyway. Drop the unused logic and fold in
>>> MSG_DONTWAIT. Adapt callers.
>>>
>>> Subtle change here: recv() return value of 0 will also be considered (an
>>> unexpected) success.
>>>
>>> Signed-off-by: Michal Luczaj <mhal@rbox.co>
>>> ---
>>>  .../selftests/bpf/prog_tests/sockmap_redir.c       | 25 +++++++++-------------
>>>  1 file changed, 10 insertions(+), 15 deletions(-)
>>>
>>> diff --git a/tools/testing/selftests/bpf/prog_tests/sockmap_redir.c b/tools/testing/selftests/bpf/prog_tests/sockmap_redir.c
>>> index 9c461d93113db20de65ac353f92dfdbe32ffbd3b..c1bf1076e8152b7d83c3e07e2dce746b5a39cf7e 100644
>>> --- a/tools/testing/selftests/bpf/prog_tests/sockmap_redir.c
>>> +++ b/tools/testing/selftests/bpf/prog_tests/sockmap_redir.c
>>> @@ -144,17 +144,14 @@ static void get_redir_params(struct redir_spec *redir,
>>>  		*redirect_flags = 0;
>>>  }
>>>  
>>> -static void try_recv(const char *prefix, int fd, int flags, bool expect_success)
>>> +static void fail_recv(const char *prefix, int fd, int more_flags)
>>>  {
>>>  	ssize_t n;
>>>  	char buf;
>>>  
>>> -	errno = 0;
>>> -	n = recv(fd, &buf, 1, flags);
>>> -	if (n < 0 && expect_success)
>>> -		FAIL_ERRNO("%s: unexpected failure: retval=%zd", prefix, n);
>>> -	if (!n && !expect_success)
>>> -		FAIL("%s: expected failure: retval=%zd", prefix, n);
>>> +	n = recv(fd, &buf, 1, MSG_DONTWAIT | more_flags);
>>> +	if (n >= 0)
>>> +		FAIL("%s: unexpected success: retval=%zd", prefix, n);
>>>  }
>>
>> This bit, which you highlighted in the description, I don't get.
>>
>> If we're expecting to receive exactly one byte, why treat a short read
>> as a succcess? Why not make it a strict "n != 1" check?
>>
>> [...]
> 
> Nevermind. It makes sense now. We do want to report a failure for 0-len
> msg recv as well. You're effectively checking if the rcv queue is empty.
> 
> I'd add MSG_PEEK, to signal that we're _just checking_ if the socket is
> readable, and turn the check into the below to succeed only when
> queue is empty:
> 
>         (n != -1 || (errno != EAGAIN && errno != EWOULDBLOCK))

Well, looks like adding MSG_PEEK exposed a bug in the test. I'll fix that.

Thanks,
Michal

Re: [PATCH bpf-next 1/5] selftests/bpf: sockmap_redir: Simplify try_recv()

[Jakub Sitnicki]

On Tue, Sep 09, 2025 at 11:25 PM +02, Michal Luczaj wrote:
> On 9/9/25 12:15, Jakub Sitnicki wrote:
>> On Tue, Sep 09, 2025 at 11:51 AM +02, Jakub Sitnicki wrote:
>>> On Fri, Sep 05, 2025 at 01:11 PM +02, Michal Luczaj wrote:
>>>> try_recv() was meant to support both @expect_success cases, but all the
>>>> callers use @expect_success=false anyway. Drop the unused logic and fold in
>>>> MSG_DONTWAIT. Adapt callers.
>>>>
>>>> Subtle change here: recv() return value of 0 will also be considered (an
>>>> unexpected) success.
>>>>
>>>> Signed-off-by: Michal Luczaj <mhal@rbox.co>
>>>> ---
>>>>  .../selftests/bpf/prog_tests/sockmap_redir.c       | 25 +++++++++-------------
>>>>  1 file changed, 10 insertions(+), 15 deletions(-)
>>>>
>>>> diff --git a/tools/testing/selftests/bpf/prog_tests/sockmap_redir.c b/tools/testing/selftests/bpf/prog_tests/sockmap_redir.c
>>>> index 9c461d93113db20de65ac353f92dfdbe32ffbd3b..c1bf1076e8152b7d83c3e07e2dce746b5a39cf7e 100644
>>>> --- a/tools/testing/selftests/bpf/prog_tests/sockmap_redir.c
>>>> +++ b/tools/testing/selftests/bpf/prog_tests/sockmap_redir.c
>>>> @@ -144,17 +144,14 @@ static void get_redir_params(struct redir_spec *redir,
>>>>  		*redirect_flags = 0;
>>>>  }
>>>>  
>>>> -static void try_recv(const char *prefix, int fd, int flags, bool expect_success)
>>>> +static void fail_recv(const char *prefix, int fd, int more_flags)
>>>>  {
>>>>  	ssize_t n;
>>>>  	char buf;
>>>>  
>>>> -	errno = 0;
>>>> -	n = recv(fd, &buf, 1, flags);
>>>> -	if (n < 0 && expect_success)
>>>> -		FAIL_ERRNO("%s: unexpected failure: retval=%zd", prefix, n);
>>>> -	if (!n && !expect_success)
>>>> -		FAIL("%s: expected failure: retval=%zd", prefix, n);
>>>> +	n = recv(fd, &buf, 1, MSG_DONTWAIT | more_flags);
>>>> +	if (n >= 0)
>>>> +		FAIL("%s: unexpected success: retval=%zd", prefix, n);
>>>>  }
>>>
>>> This bit, which you highlighted in the description, I don't get.
>>>
>>> If we're expecting to receive exactly one byte, why treat a short read
>>> as a succcess? Why not make it a strict "n != 1" check?
>>>
>>> [...]
>> 
>> Nevermind. It makes sense now. We do want to report a failure for 0-len
>> msg recv as well. You're effectively checking if the rcv queue is empty.
>> 
>> I'd add MSG_PEEK, to signal that we're _just checking_ if the socket is
>> readable, and turn the check into the below to succeed only when
>> queue is empty:
>> 
>>         (n != -1 || (errno != EAGAIN && errno != EWOULDBLOCK))
>
> Well, looks like adding MSG_PEEK exposed a bug in the test. I'll fix that.

The gift that keeps on giving xD

Other alternatives that should also work, but who knows:

- select/poll/epoll readability check
- ioctl(SIOCINQ) but no way to tell if 0-len msg is pending

Re: [PATCH bpf-next 2/5] selftests/bpf: sockmap_redir: Fix OOB handling

[Jakub Sitnicki]

Sorry for the super-long time-to-feedback.

On Fri, Sep 05, 2025 at 01:11 PM +02, Michal Luczaj wrote:
> In some test cases, OOB packets might have been left unread. Flush them out
> and introduce additional checks.
>
> Signed-off-by: Michal Luczaj <mhal@rbox.co>
> ---
>  tools/testing/selftests/bpf/prog_tests/sockmap_redir.c | 16 ++++++++++++++++
>  1 file changed, 16 insertions(+)
>
> diff --git a/tools/testing/selftests/bpf/prog_tests/sockmap_redir.c b/tools/testing/selftests/bpf/prog_tests/sockmap_redir.c
> index c1bf1076e8152b7d83c3e07e2dce746b5a39cf7e..4997e72c14345b274367f3f2f4115c39d1ae48c9 100644
> --- a/tools/testing/selftests/bpf/prog_tests/sockmap_redir.c
> +++ b/tools/testing/selftests/bpf/prog_tests/sockmap_redir.c
> @@ -184,6 +184,19 @@ static void handle_unsupported(int sd_send, int sd_peer, int sd_in, int sd_out,
>  			FAIL_ERRNO("unsupported: packet missing, retval=%zd", n);
>  	}
>  
> +	/* af_unix send("ab", MSG_OOB) spits out 2 packets, but only the latter
> +	 * ("b") is designated OOB. If the peer is in a sockmap, the OOB packet
> +	 * will be silently dropped. Otherwise OOB stays in the queue and should
> +	 * be taken care of.
> +	 */
> +	if ((send_flags & MSG_OOB) && !pass && !drop) {

Nit: There's a similar check a few lines before that:

	if (pass == 0 && drop == 0 && (status & UNSUPPORTED_RACY_VERD)) {

For readability it might make sense to introduce a helper flag:

        bool no_verdict = !pass && !drop; /* prog didn't run */

> +		errno = 0;
> +		n = recv_timeout(sd_peer, &recv_buf, 1, MSG_OOB, IO_TIMEOUT_SEC);
> +		/* Ignore unsupported sk_msg error */
> +		if (n != 1 && errno != EOPNOTSUPP)
> +			FAIL_ERRNO("recv(OOB): retval=%zd", n);
> +	}
> +
>  	/* Ensure queues are empty */
>  	fail_recv("bpf.recv(sd_send)", sd_send, 0);
>  	if (sd_in != sd_send)
> @@ -192,6 +205,9 @@ static void handle_unsupported(int sd_send, int sd_peer, int sd_in, int sd_out,
>  	fail_recv("bpf.recv(sd_out)", sd_out, 0);
>  	if (sd_recv != sd_out)
>  		fail_recv("bpf.recv(sd_recv)", sd_recv, 0);
> +
> +	fail_recv("recv(sd_peer, OOB)", sd_peer, MSG_OOB);
> +	fail_recv("recv(sd_out, OOB)", sd_out, MSG_OOB);
>  }
>  
>  static void test_send_redir_recv(int sd_send, int send_flags, int sd_peer,

Reviewed-by: Jakub Sitnicki <jakub@cloudflare.com>

Re: [PATCH bpf-next 3/5] selftests/bpf: sockmap_redir: Rename functions

[Jakub Sitnicki]

On Fri, Sep 05, 2025 at 01:11 PM +02, Michal Luczaj wrote:
> Preparatory patch before adding SK_PASS/SK_DROP support:
>
> test_redir() => test_sockets()
> test_socket() => test_redir()
> test_send_redir_recv() => test_send_recv()
>
> After the change (and the following patch) the call stack will be:
>
> serial_test_sockmap_redir
>   test_map
>     test_sockets
>       test_redir
>         test_send_recv
>       (test_verdict)
>         (test_send_recv)
>
> Signed-off-by: Michal Luczaj <mhal@rbox.co>
> ---

Reviewed-by: Jakub Sitnicki <jakub@cloudflare.com>

Re: [PATCH bpf-next 4/5] selftests/bpf: sockmap_redir: Let test specify skel's redirect_type

[Jakub Sitnicki]

On Fri, Sep 05, 2025 at 01:11 PM +02, Michal Luczaj wrote:
> Preparatory patch before adding SK_PASS/SK_DROP support: allow to
> dynamically switch BPF program's redirect_type. This way, after setting up
> for a redirection, test can make the BPF program skip the actual
> bpf_{sk,msg}_redirect_{map,hash} part and return a specified verdict.
>
> Signed-off-by: Michal Luczaj <mhal@rbox.co>
> ---

Reviewed-by: Jakub Sitnicki <jakub@cloudflare.com>

Re: [PATCH bpf-next 5/5] selftests/bpf: sockmap_redir: Support no-redirect SK_DROP/SK_PASS

[Jakub Sitnicki]

On Fri, Sep 05, 2025 at 01:11 PM +02, Michal Luczaj wrote:
> Add tests that make the BPF programs skip the actual redirect and
> immediately return SK_DROP/SK_PASS.
>
> Suggested-by: Jiayuan Chen <mrpre@163.com>
> Signed-off-by: Michal Luczaj <mhal@rbox.co>
> ---

Reviewed-by: Jakub Sitnicki <jakub@cloudflare.com>

Re: [PATCH bpf-next 2/5] selftests/bpf: sockmap_redir: Fix OOB handling

[Jakub Sitnicki]

On Fri, Sep 05, 2025 at 01:19 PM +02, Michal Luczaj wrote:
> On 9/5/25 13:11, Michal Luczaj wrote:
>> In some test cases, OOB packets might have been left unread. Flush them out
>> and introduce additional checks.
>> 
>> Signed-off-by: Michal Luczaj <mhal@rbox.co>
>
> Sorry, this should also have:
>
> Fixes: f0709263a07e ("selftests/bpf: Add selftest for sockmap/hashmap
> redirection")

I'm not sure if we're backporting selftest changes to stable.