soreuseport: BPF selection functional test for TCP

Unfortunately the existing test relied on packet payload in order to
map incoming packets to sockets.  In order to get this to work with TCP,
TCP_FASTOPEN needed to be used.

Since the fast open path is slightly different than the standard TCP path,
I created a second test which sends to reuseport group members based
on receiving cpu core id.  This will probably serve as a better
real-world example use as well.

Signed-off-by: Craig Gallek <kraig@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

4 files changed, 370 insertions, 8 deletions

diff --git a/tools/testing/selftests/net/.gitignore b/tools/testing/selftests/net/.gitignore
index 6fb23366b258..69bb3fc38fb2 100644
--- a/tools/testing/selftests/net/.gitignore
+++ b/tools/testing/selftests/net/.gitignore
@@ -2,3 +2,4 @@ socket
 psock_fanout
 psock_tpacket
 reuseport_bpf
+reuseport_bpf_cpu
diff --git a/tools/testing/selftests/net/Makefile b/tools/testing/selftests/net/Makefile
index 41449b5ad0a9..c658792d47b4 100644
--- a/tools/testing/selftests/net/Makefile
+++ b/tools/testing/selftests/net/Makefile
@@ -4,7 +4,7 @@ CFLAGS = -Wall -O2 -g
 
 CFLAGS += -I../../../../usr/include/
 
-NET_PROGS = socket psock_fanout psock_tpacket reuseport_bpf
+NET_PROGS = socket psock_fanout psock_tpacket reuseport_bpf reuseport_bpf_cpu
 
 all: $(NET_PROGS)
 %: %.c
diff --git a/tools/testing/selftests/net/reuseport_bpf.c b/tools/testing/selftests/net/reuseport_bpf.c
index bec1b5dd2530..96ba386b1b7b 100644
--- a/tools/testing/selftests/net/reuseport_bpf.c
+++ b/tools/testing/selftests/net/reuseport_bpf.c
@@ -9,10 +9,12 @@
 
 #include <errno.h>
 #include <error.h>
+#include <fcntl.h>
 #include <linux/bpf.h>
 #include <linux/filter.h>
 #include <linux/unistd.h>
 #include <netinet/in.h>
+#include <netinet/tcp.h>
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
@@ -169,9 +171,15 @@ static void build_recv_group(const struct test_params p, int fd[], uint16_t mod,
 		if (bind(fd[i], addr, sockaddr_size()))
 			error(1, errno, "failed to bind recv socket %d", i);
 
-		if (p.protocol == SOCK_STREAM)
+		if (p.protocol == SOCK_STREAM) {
+			opt = 4;
+			if (setsockopt(fd[i], SOL_TCP, TCP_FASTOPEN, &opt,
+				       sizeof(opt)))
+				error(1, errno,
+				      "failed to set TCP_FASTOPEN on %d", i);
 			if (listen(fd[i], p.recv_socks * 10))
 				error(1, errno, "failed to listen on socket");
+		}
 	}
 	free(addr);
 }
@@ -189,10 +197,8 @@ static void send_from(struct test_params p, uint16_t sport, char *buf,
 
 	if (bind(fd, saddr, sockaddr_size()))
 		error(1, errno, "failed to bind send socket");
-	if (connect(fd, daddr, sockaddr_size()))
-		error(1, errno, "failed to connect");
 
-	if (send(fd, buf, len, 0) < 0)
+	if (sendto(fd, buf, len, MSG_FASTOPEN, daddr, sockaddr_size()) < 0)
 		error(1, errno, "failed to send message");
 
 	close(fd);
@@ -260,7 +266,7 @@ static void test_recv_order(const struct test_params p, int fd[], int mod)
 	}
 }
 
-static void test_reuseport_ebpf(const struct test_params p)
+static void test_reuseport_ebpf(struct test_params p)
 {
 	int i, fd[p.recv_socks];
 
@@ -268,6 +274,7 @@ static void test_reuseport_ebpf(const struct test_params p)
 	build_recv_group(p, fd, p.recv_socks, attach_ebpf);
 	test_recv_order(p, fd, p.recv_socks);
 
+	p.send_port_min += p.recv_socks * 2;
 	fprintf(stderr, "Reprograming, testing mod %zd...\n", p.recv_socks / 2);
 	attach_ebpf(fd[0], p.recv_socks / 2);
 	test_recv_order(p, fd, p.recv_socks / 2);
@@ -276,7 +283,7 @@ static void test_reuseport_ebpf(const struct test_params p)
 		close(fd[i]);
 }
 
-static void test_reuseport_cbpf(const struct test_params p)
+static void test_reuseport_cbpf(struct test_params p)
 {
 	int i, fd[p.recv_socks];
 
@@ -284,6 +291,7 @@ static void test_reuseport_cbpf(const struct test_params p)
 	build_recv_group(p, fd, p.recv_socks, attach_cbpf);
 	test_recv_order(p, fd, p.recv_socks);
 
+	p.send_port_min += p.recv_socks * 2;
 	fprintf(stderr, "Reprograming, testing mod %zd...\n", p.recv_socks / 2);
 	attach_cbpf(fd[0], p.recv_socks / 2);
 	test_recv_order(p, fd, p.recv_socks / 2);
@@ -377,7 +385,7 @@ static void test_filter_no_reuseport(const struct test_params p)
 
 static void test_filter_without_bind(void)
 {
-	int fd1, fd2;
+	int fd1, fd2, opt = 1;
 
 	fprintf(stderr, "Testing filter add without bind...\n");
 	fd1 = socket(AF_INET, SOCK_DGRAM, 0);
@@ -386,6 +394,10 @@ static void test_filter_without_bind(void)
 	fd2 = socket(AF_INET, SOCK_DGRAM, 0);
 	if (fd2 < 0)
 		error(1, errno, "failed to create socket 2");
+	if (setsockopt(fd1, SOL_SOCKET, SO_REUSEPORT, &opt, sizeof(opt)))
+		error(1, errno, "failed to set SO_REUSEPORT on socket 1");
+	if (setsockopt(fd2, SOL_SOCKET, SO_REUSEPORT, &opt, sizeof(opt)))
+		error(1, errno, "failed to set SO_REUSEPORT on socket 2");
 
 	attach_ebpf(fd1, 10);
 	attach_cbpf(fd2, 10);
@@ -394,6 +406,32 @@ static void test_filter_without_bind(void)
 	close(fd2);
 }
 
+void enable_fastopen(void)
+{
+	int fd = open("/proc/sys/net/ipv4/tcp_fastopen", 0);
+	int rw_mask = 3;  /* bit 1: client side; bit-2 server side */
+	int val, size;
+	char buf[16];
+
+	if (fd < 0)
+		error(1, errno, "Unable to open tcp_fastopen sysctl");
+	if (read(fd, buf, sizeof(buf)) <= 0)
+		error(1, errno, "Unable to read tcp_fastopen sysctl");
+	val = atoi(buf);
+	close(fd);
+
+	if ((val & rw_mask) != rw_mask) {
+		fd = open("/proc/sys/net/ipv4/tcp_fastopen", O_RDWR);
+		if (fd < 0)
+			error(1, errno,
+			      "Unable to open tcp_fastopen sysctl for writing");
+		val |= rw_mask;
+		size = snprintf(buf, 16, "%d", val);
+		if (write(fd, buf, size) <= 0)
+			error(1, errno, "Unable to write tcp_fastopen sysctl");
+		close(fd);
+	}
+}
 
 int main(void)
 {
@@ -506,6 +544,71 @@ int main(void)
 		.recv_port = 8007,
 		.send_port_min = 9100});
 
+	/* TCP fastopen is required for the TCP tests */
+	enable_fastopen();
+	fprintf(stderr, "---- IPv4 TCP ----\n");
+	test_reuseport_ebpf((struct test_params) {
+		.recv_family = AF_INET,
+		.send_family = AF_INET,
+		.protocol = SOCK_STREAM,
+		.recv_socks = 10,
+		.recv_port = 8008,
+		.send_port_min = 9120});
+	test_reuseport_cbpf((struct test_params) {
+		.recv_family = AF_INET,
+		.send_family = AF_INET,
+		.protocol = SOCK_STREAM,
+		.recv_socks = 10,
+		.recv_port = 8009,
+		.send_port_min = 9160});
+	test_extra_filter((struct test_params) {
+		.recv_family = AF_INET,
+		.protocol = SOCK_STREAM,
+		.recv_port = 8010});
+	test_filter_no_reuseport((struct test_params) {
+		.recv_family = AF_INET,
+		.protocol = SOCK_STREAM,
+		.recv_port = 8011});
+
+	fprintf(stderr, "---- IPv6 TCP ----\n");
+	test_reuseport_ebpf((struct test_params) {
+		.recv_family = AF_INET6,
+		.send_family = AF_INET6,
+		.protocol = SOCK_STREAM,
+		.recv_socks = 10,
+		.recv_port = 8012,
+		.send_port_min = 9200});
+	test_reuseport_cbpf((struct test_params) {
+		.recv_family = AF_INET6,
+		.send_family = AF_INET6,
+		.protocol = SOCK_STREAM,
+		.recv_socks = 10,
+		.recv_port = 8013,
+		.send_port_min = 9240});
+	test_extra_filter((struct test_params) {
+		.recv_family = AF_INET6,
+		.protocol = SOCK_STREAM,
+		.recv_port = 8014});
+	test_filter_no_reuseport((struct test_params) {
+		.recv_family = AF_INET6,
+		.protocol = SOCK_STREAM,
+		.recv_port = 8015});
+
+	fprintf(stderr, "---- IPv6 TCP w/ mapped IPv4 ----\n");
+	test_reuseport_ebpf((struct test_params) {
+		.recv_family = AF_INET6,
+		.send_family = AF_INET,
+		.protocol = SOCK_STREAM,
+		.recv_socks = 10,
+		.recv_port = 8016,
+		.send_port_min = 9320});
+	test_reuseport_cbpf((struct test_params) {
+		.recv_family = AF_INET6,
+		.send_family = AF_INET,
+		.protocol = SOCK_STREAM,
+		.recv_socks = 10,
+		.recv_port = 8017,
+		.send_port_min = 9360});
 
 	test_filter_without_bind();
 
diff --git a/tools/testing/selftests/net/reuseport_bpf_cpu.c b/tools/testing/selftests/net/reuseport_bpf_cpu.c
new file mode 100644
index 000000000000..b23d6f54de7b
--- /dev/null
+++ b/tools/testing/selftests/net/reuseport_bpf_cpu.c
@@ -0,0 +1,258 @@
+/*
+ * Test functionality of BPF filters with SO_REUSEPORT.  This program creates
+ * an SO_REUSEPORT receiver group containing one socket per CPU core. It then
+ * creates a BPF program that will select a socket from this group based
+ * on the core id that receives the packet.  The sending code artificially
+ * moves itself to run on different core ids and sends one message from
+ * each core.  Since these packets are delivered over loopback, they should
+ * arrive on the same core that sent them.  The receiving code then ensures
+ * that the packet was received on the socket for the corresponding core id.
+ * This entire process is done for several different core id permutations
+ * and for each IPv4/IPv6 and TCP/UDP combination.
+ */
+
+#define _GNU_SOURCE
+
+#include <arpa/inet.h>
+#include <errno.h>
+#include <error.h>
+#include <linux/filter.h>
+#include <linux/in.h>
+#include <linux/unistd.h>
+#include <sched.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <sys/epoll.h>
+#include <sys/types.h>
+#include <sys/socket.h>
+#include <unistd.h>
+
+static const int PORT = 8888;
+
+static void build_rcv_group(int *rcv_fd, size_t len, int family, int proto)
+{
+	struct sockaddr_storage addr;
+	struct sockaddr_in  *addr4;
+	struct sockaddr_in6 *addr6;
+	size_t i;
+	int opt;
+
+	switch (family) {
+	case AF_INET:
+		addr4 = (struct sockaddr_in *)&addr;
+		addr4->sin_family = AF_INET;
+		addr4->sin_addr.s_addr = htonl(INADDR_ANY);
+		addr4->sin_port = htons(PORT);
+		break;
+	case AF_INET6:
+		addr6 = (struct sockaddr_in6 *)&addr;
+		addr6->sin6_family = AF_INET6;
+		addr6->sin6_addr = in6addr_any;
+		addr6->sin6_port = htons(PORT);
+		break;
+	default:
+		error(1, 0, "Unsupported family %d", family);
+	}
+
+	for (i = 0; i < len; ++i) {
+		rcv_fd[i] = socket(family, proto, 0);
+		if (rcv_fd[i] < 0)
+			error(1, errno, "failed to create receive socket");
+
+		opt = 1;
+		if (setsockopt(rcv_fd[i], SOL_SOCKET, SO_REUSEPORT, &opt,
+			       sizeof(opt)))
+			error(1, errno, "failed to set SO_REUSEPORT");
+
+		if (bind(rcv_fd[i], (struct sockaddr *)&addr, sizeof(addr)))
+			error(1, errno, "failed to bind receive socket");
+
+		if (proto == SOCK_STREAM && listen(rcv_fd[i], len * 10))
+			error(1, errno, "failed to listen on receive port");
+	}
+}
+
+static void attach_bpf(int fd)
+{
+	struct sock_filter code[] = {
+		/* A = raw_smp_processor_id() */
+		{ BPF_LD  | BPF_W | BPF_ABS, 0, 0, SKF_AD_OFF + SKF_AD_CPU },
+		/* return A */
+		{ BPF_RET | BPF_A, 0, 0, 0 },
+	};
+	struct sock_fprog p = {
+		.len = 2,
+		.filter = code,
+	};
+
+	if (setsockopt(fd, SOL_SOCKET, SO_ATTACH_REUSEPORT_CBPF, &p, sizeof(p)))
+		error(1, errno, "failed to set SO_ATTACH_REUSEPORT_CBPF");
+}
+
+static void send_from_cpu(int cpu_id, int family, int proto)
+{
+	struct sockaddr_storage saddr, daddr;
+	struct sockaddr_in  *saddr4, *daddr4;
+	struct sockaddr_in6 *saddr6, *daddr6;
+	cpu_set_t cpu_set;
+	int fd;
+
+	switch (family) {
+	case AF_INET:
+		saddr4 = (struct sockaddr_in *)&saddr;
+		saddr4->sin_family = AF_INET;
+		saddr4->sin_addr.s_addr = htonl(INADDR_ANY);
+		saddr4->sin_port = 0;
+
+		daddr4 = (struct sockaddr_in *)&daddr;
+		daddr4->sin_family = AF_INET;
+		daddr4->sin_addr.s_addr = htonl(INADDR_LOOPBACK);
+		daddr4->sin_port = htons(PORT);
+		break;
+	case AF_INET6:
+		saddr6 = (struct sockaddr_in6 *)&saddr;
+		saddr6->sin6_family = AF_INET6;
+		saddr6->sin6_addr = in6addr_any;
+		saddr6->sin6_port = 0;
+
+		daddr6 = (struct sockaddr_in6 *)&daddr;
+		daddr6->sin6_family = AF_INET6;
+		daddr6->sin6_addr = in6addr_loopback;
+		daddr6->sin6_port = htons(PORT);
+		break;
+	default:
+		error(1, 0, "Unsupported family %d", family);
+	}
+
+	memset(&cpu_set, 0, sizeof(cpu_set));
+	CPU_SET(cpu_id, &cpu_set);
+	if (sched_setaffinity(0, sizeof(cpu_set), &cpu_set) < 0)
+		error(1, errno, "failed to pin to cpu");
+
+	fd = socket(family, proto, 0);
+	if (fd < 0)
+		error(1, errno, "failed to create send socket");
+
+	if (bind(fd, (struct sockaddr *)&saddr, sizeof(saddr)))
+		error(1, errno, "failed to bind send socket");
+
+	if (connect(fd, (struct sockaddr *)&daddr, sizeof(daddr)))
+		error(1, errno, "failed to connect send socket");
+
+	if (send(fd, "a", 1, 0) < 0)
+		error(1, errno, "failed to send message");
+
+	close(fd);
+}
+
+static
+void receive_on_cpu(int *rcv_fd, int len, int epfd, int cpu_id, int proto)
+{
+	struct epoll_event ev;
+	int i, fd;
+	char buf[8];
+
+	i = epoll_wait(epfd, &ev, 1, -1);
+	if (i < 0)
+		error(1, errno, "epoll_wait failed");
+
+	if (proto == SOCK_STREAM) {
+		fd = accept(ev.data.fd, NULL, NULL);
+		if (fd < 0)
+			error(1, errno, "failed to accept");
+		i = recv(fd, buf, sizeof(buf), 0);
+		close(fd);
+	} else {
+		i = recv(ev.data.fd, buf, sizeof(buf), 0);
+	}
+
+	if (i < 0)
+		error(1, errno, "failed to recv");
+
+	for (i = 0; i < len; ++i)
+		if (ev.data.fd == rcv_fd[i])
+			break;
+	if (i == len)
+		error(1, 0, "failed to find socket");
+	fprintf(stderr, "send cpu %d, receive socket %d\n", cpu_id, i);
+	if (cpu_id != i)
+		error(1, 0, "cpu id/receive socket mismatch");
+}
+
+static void test(int *rcv_fd, int len, int family, int proto)
+{
+	struct epoll_event ev;
+	int epfd, cpu;
+
+	build_rcv_group(rcv_fd, len, family, proto);
+	attach_bpf(rcv_fd[0]);
+
+	epfd = epoll_create(1);
+	if (epfd < 0)
+		error(1, errno, "failed to create epoll");
+	for (cpu = 0; cpu < len; ++cpu) {
+		ev.events = EPOLLIN;
+		ev.data.fd = rcv_fd[cpu];
+		if (epoll_ctl(epfd, EPOLL_CTL_ADD, rcv_fd[cpu], &ev))
+			error(1, errno, "failed to register sock epoll");
+	}
+
+	/* Forward iterate */
+	for (cpu = 0; cpu < len; ++cpu) {
+		send_from_cpu(cpu, family, proto);
+		receive_on_cpu(rcv_fd, len, epfd, cpu, proto);
+	}
+
+	/* Reverse iterate */
+	for (cpu = len - 1; cpu >= 0; --cpu) {
+		send_from_cpu(cpu, family, proto);
+		receive_on_cpu(rcv_fd, len, epfd, cpu, proto);
+	}
+
+	/* Even cores */
+	for (cpu = 0; cpu < len; cpu += 2) {
+		send_from_cpu(cpu, family, proto);
+		receive_on_cpu(rcv_fd, len, epfd, cpu, proto);
+	}
+
+	/* Odd cores */
+	for (cpu = 1; cpu < len; cpu += 2) {
+		send_from_cpu(cpu, family, proto);
+		receive_on_cpu(rcv_fd, len, epfd, cpu, proto);
+	}
+
+	close(epfd);
+	for (cpu = 0; cpu < len; ++cpu)
+		close(rcv_fd[cpu]);
+}
+
+int main(void)
+{
+	int *rcv_fd, cpus;
+
+	cpus = sysconf(_SC_NPROCESSORS_ONLN);
+	if (cpus <= 0)
+		error(1, errno, "failed counting cpus");
+
+	rcv_fd = calloc(cpus, sizeof(int));
+	if (!rcv_fd)
+		error(1, 0, "failed to allocate array");
+
+	fprintf(stderr, "---- IPv4 UDP ----\n");
+	test(rcv_fd, cpus, AF_INET, SOCK_DGRAM);
+
+	fprintf(stderr, "---- IPv6 UDP ----\n");
+	test(rcv_fd, cpus, AF_INET6, SOCK_DGRAM);
+
+	fprintf(stderr, "---- IPv4 TCP ----\n");
+	test(rcv_fd, cpus, AF_INET, SOCK_STREAM);
+
+	fprintf(stderr, "---- IPv6 TCP ----\n");
+	test(rcv_fd, cpus, AF_INET6, SOCK_STREAM);
+
+	free(rcv_fd);
+
+	fprintf(stderr, "SUCCESS\n");
+	return 0;
+}