[net-next, v3 0/2] Adding SO_PEEK_OFF for TCPv6

[net-next, v3 0/2] Adding SO_PEEK_OFF for TCPv6

[jmaloy]

From: Jon Maloy <jmaloy@redhat.com>

Adding SO_PEEK_OFF for TCPv6 and selftest for both TCPv4 and TCPv6

Jon Maloy (2):
  tcp: add SO_PEEK_OFF socket option tor TCPv6
  selftests: add selftest for tcp SO_PEEK_OFF support

 net/ipv6/af_inet6.c                           |   1 +
 tools/testing/selftests/net/Makefile          |   1 +
 tools/testing/selftests/net/tcp_so_peek_off.c | 183 ++++++++++++++++++
 3 files changed, 185 insertions(+)
 create mode 100644 tools/testing/selftests/net/tcp_so_peek_off.c

-- 
2.45.2

[net-next, v3 1/2] tcp: add SO_PEEK_OFF socket option tor TCPv6

[jmaloy]

From: Jon Maloy <jmaloy@redhat.com>

When doing further testing of the recently added SO_PEEK_OFF feature
for TCP I realized I had omitted to add it for TCP/IPv6.

I do that here.

Fixes: 05ea491641d3 ("tcp: add support for SO_PEEK_OFF socket option")
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Stefano Brivio <sbrivio@redhat.com>
Tested-by: Stefano Brivio <sbrivio@redhat.com>
Signed-off-by: Jon Maloy <jmaloy@redhat.com>

---
v2: Removed redundant whitespace between "Fixes" and
    "Signed-off-by" lines. Based on feedback from J. Kicinski.
v3: Rephrased commit log to make it clear how the TCP/IPv6 omission
    was discovered. Based on feedback from J. Kicinski.
---
 net/ipv6/af_inet6.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/net/ipv6/af_inet6.c b/net/ipv6/af_inet6.c
index 90d2c7e3f5e9..ba69b86f1c7d 100644
--- a/net/ipv6/af_inet6.c
+++ b/net/ipv6/af_inet6.c
@@ -708,6 +708,7 @@ const struct proto_ops inet6_stream_ops = {
 	.splice_eof	   = inet_splice_eof,
 	.sendmsg_locked    = tcp_sendmsg_locked,
 	.splice_read	   = tcp_splice_read,
+	.set_peek_off      = sk_set_peek_off,
 	.read_sock	   = tcp_read_sock,
 	.read_skb	   = tcp_read_skb,
 	.peek_len	   = tcp_peek_len,
-- 
2.45.2

[net-next, v3 2/2] selftests: add selftest for tcp SO_PEEK_OFF support

[jmaloy]

From: Jon Maloy <jmaloy@redhat.com>

We add a selftest to check that the new feature added in
commit 05ea491641d3 ("tcp: add support for SO_PEEK_OFF socket option")
works correctly.

Reviewed-by: Jason Xing <kerneljasonxing@gmail.com>
Reviewed-by: Stefano Brivio <sbrivio@redhat.com>
Tested-by: Stefano Brivio <sbrivio@redhat.com>
Signed-off-by: Jon Maloy <jmaloy@redhat.com>

---
v2: Made correction based on Jason Xing's comment.
v3: Removed redundant whitespace at end of file.
    Based on comment from Jakub Kicinski.
---
 tools/testing/selftests/net/Makefile          |   1 +
 tools/testing/selftests/net/tcp_so_peek_off.c | 183 ++++++++++++++++++
 2 files changed, 184 insertions(+)
 create mode 100644 tools/testing/selftests/net/tcp_so_peek_off.c

diff --git a/tools/testing/selftests/net/Makefile b/tools/testing/selftests/net/Makefile
index 8eaffd7a641c..1179e3261bef 100644
--- a/tools/testing/selftests/net/Makefile
+++ b/tools/testing/selftests/net/Makefile
@@ -80,6 +80,7 @@ TEST_PROGS += io_uring_zerocopy_tx.sh
 TEST_GEN_FILES += bind_bhash
 TEST_GEN_PROGS += sk_bind_sendto_listen
 TEST_GEN_PROGS += sk_connect_zero_addr
+TEST_GEN_PROGS += tcp_so_peek_off
 TEST_PROGS += test_ingress_egress_chaining.sh
 TEST_GEN_PROGS += so_incoming_cpu
 TEST_PROGS += sctp_vrf.sh
diff --git a/tools/testing/selftests/net/tcp_so_peek_off.c b/tools/testing/selftests/net/tcp_so_peek_off.c
new file mode 100644
index 000000000000..df8a39d9d3c3
--- /dev/null
+++ b/tools/testing/selftests/net/tcp_so_peek_off.c
@@ -0,0 +1,183 @@
+// SPDX-License-Identifier: GPL-2.0
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <unistd.h>
+#include <errno.h>
+#include <sys/types.h>
+#include <netinet/in.h>
+#include <arpa/inet.h>
+#include "../kselftest.h"
+
+static char *afstr(int af)
+{
+	return af == AF_INET ? "TCP/IPv4" : "TCP/IPv6";
+}
+
+int tcp_peek_offset_probe(sa_family_t af)
+{
+	int optv = 0;
+	int ret = 0;
+	int s;
+
+	s = socket(af, SOCK_STREAM | SOCK_CLOEXEC, IPPROTO_TCP);
+	if (s < 0) {
+		ksft_perror("Temporary TCP socket creation failed");
+	} else {
+		if (!setsockopt(s, SOL_SOCKET, SO_PEEK_OFF, &optv, sizeof(int)))
+			ret = 1;
+		else
+			printf("%s does not support SO_PEEK_OFF\n", afstr(af));
+		close(s);
+	}
+	return ret;
+}
+
+static void tcp_peek_offset_set(int s, int offset)
+{
+	if (setsockopt(s, SOL_SOCKET, SO_PEEK_OFF, &offset, sizeof(offset)))
+		ksft_perror("Failed to set SO_PEEK_OFF value\n");
+}
+
+static int tcp_peek_offset_get(int s)
+{
+	int offset;
+	socklen_t len = sizeof(offset);
+
+	if (getsockopt(s, SOL_SOCKET, SO_PEEK_OFF, &offset, &len))
+		ksft_perror("Failed to get SO_PEEK_OFF value\n");
+	return offset;
+}
+
+static int tcp_peek_offset_test(sa_family_t af)
+{
+	union {
+		struct sockaddr sa;
+		struct sockaddr_in a4;
+		struct sockaddr_in6 a6;
+	} a;
+	int res = 0;
+	int s[2] = {0, 0};
+	int recv_sock = 0;
+	int offset = 0;
+	ssize_t len;
+	char buf;
+
+	memset(&a, 0, sizeof(a));
+	a.sa.sa_family = af;
+
+	s[0] = socket(af, SOCK_STREAM, IPPROTO_TCP);
+	s[1] = socket(af, SOCK_STREAM | SOCK_NONBLOCK, IPPROTO_TCP);
+
+	if (s[0] < 0 || s[1] < 0) {
+		ksft_perror("Temporary socket creation failed\n");
+		goto out;
+	}
+	if (bind(s[0], &a.sa, sizeof(a)) < 0) {
+		ksft_perror("Temporary socket bind() failed\n");
+		goto out;
+	}
+	if (getsockname(s[0], &a.sa, &((socklen_t) { sizeof(a) })) < 0) {
+		ksft_perror("Temporary socket getsockname() failed\n");
+		goto out;
+	}
+	if (listen(s[0], 0) < 0) {
+		ksft_perror("Temporary socket listen() failed\n");
+		goto out;
+	}
+	if (connect(s[1], &a.sa, sizeof(a)) >= 0 || errno != EINPROGRESS) {
+		ksft_perror("Temporary socket connect() failed\n");
+		goto out;
+	}
+	recv_sock = accept(s[0], NULL, NULL);
+	if (recv_sock <= 0) {
+		ksft_perror("Temporary socket accept() failed\n");
+		goto out;
+	}
+
+	/* Some basic tests of getting/setting offset */
+	offset = tcp_peek_offset_get(recv_sock);
+	if (offset != -1) {
+		ksft_perror("Initial value of socket offset not -1\n");
+		goto out;
+	}
+	tcp_peek_offset_set(recv_sock, 0);
+	offset = tcp_peek_offset_get(recv_sock);
+	if (offset != 0) {
+		ksft_perror("Failed to set socket offset to 0\n");
+		goto out;
+	}
+
+	/* Transfer a message */
+	if (send(s[1], (char *)("ab"), 2, 0) <= 0 || errno != EINPROGRESS) {
+		ksft_perror("Temporary probe socket send() failed\n");
+		goto out;
+	}
+	/* Read first byte */
+	len = recv(recv_sock, &buf, 1, MSG_PEEK);
+	if (len != 1 || buf != 'a') {
+		ksft_perror("Failed to read first byte of message\n");
+		goto out;
+	}
+	offset = tcp_peek_offset_get(recv_sock);
+	if (offset != 1) {
+		ksft_perror("Offset not forwarded correctly at first byte\n");
+		goto out;
+	}
+	/* Try to read beyond last byte */
+	len = recv(recv_sock, &buf, 2, MSG_PEEK);
+	if (len != 1 || buf != 'b') {
+		ksft_perror("Failed to read last byte of message\n");
+		goto out;
+	}
+	offset = tcp_peek_offset_get(recv_sock);
+	if (offset != 2) {
+		ksft_perror("Offset not forwarded correctly at last byte\n");
+		goto out;
+	}
+	/* Flush message */
+	len = recv(recv_sock, NULL, 2, MSG_TRUNC);
+	if (len != 2) {
+		ksft_perror("Failed to flush message\n");
+		goto out;
+	}
+	offset = tcp_peek_offset_get(recv_sock);
+	if (offset != 0) {
+		ksft_perror("Offset not reverted correctly after flush\n");
+		goto out;
+	}
+
+	printf("%s with MSG_PEEK_OFF works correctly\n", afstr(af));
+	res = 1;
+out:
+	if (recv_sock >= 0)
+		close(recv_sock);
+	if (s[1] >= 0)
+		close(s[1]);
+	if (s[0] >= 0)
+		close(s[0]);
+	return res;
+}
+
+int main(void)
+{
+	int res4, res6;
+
+	res4 = tcp_peek_offset_probe(AF_INET);
+	res6 = tcp_peek_offset_probe(AF_INET6);
+
+	if (!res4 && !res6)
+		return KSFT_SKIP;
+
+	if (res4)
+		res4 = tcp_peek_offset_test(AF_INET);
+
+	if (res6)
+		res6 = tcp_peek_offset_test(AF_INET6);
+
+	if (!res4 || !res6)
+		return KSFT_FAIL;
+
+	return KSFT_PASS;
+}
-- 
2.45.2

Re: [net-next, v3 1/2] tcp: add SO_PEEK_OFF socket option tor TCPv6

[Jason Xing]

On Thu, Aug 29, 2024 at 2:38 AM <jmaloy@redhat.com> wrote:
>
> From: Jon Maloy <jmaloy@redhat.com>
>
> When doing further testing of the recently added SO_PEEK_OFF feature
> for TCP I realized I had omitted to add it for TCP/IPv6.
>
> I do that here.
>
> Fixes: 05ea491641d3 ("tcp: add support for SO_PEEK_OFF socket option")
> Reviewed-by: Eric Dumazet <edumazet@google.com>
> Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
> Reviewed-by: Stefano Brivio <sbrivio@redhat.com>
> Tested-by: Stefano Brivio <sbrivio@redhat.com>
> Signed-off-by: Jon Maloy <jmaloy@redhat.com>

Reviewed-by: Jason Xing <kerneljasonxing@gmail.com>

Re: [net-next, v3 0/2] Adding SO_PEEK_OFF for TCPv6

[Jakub Kicinski]

On Wed, 28 Aug 2024 14:37:50 -0400 jmaloy@redhat.com wrote:
> Adding SO_PEEK_OFF for TCPv6 and selftest for both TCPv4 and TCPv6

Applied, thanks.