selftests/net: make so_txtime more robust to timer variance

The SO_TXTIME test depends on accurate timers. In some virtualized
environments the test has been reported to be flaky. This is easily
reproduced by disabling kvm acceleration in Qemu.

Allow greater variance in a run and retry to further reduce flakiness.

Observed errors are one of two kinds: either the packet arrives too
early or late at recv(), or it was dropped in the qdisc itself and the
recv() call times out.

In the latter case, the qdisc queues a notification to the error
queue of the send socket. Also explicitly report this cause.

Link: https://lore.kernel.org/netdev/[e-mail]
Reported-by: Naresh Kamboju <[e-mail]>
Signed-off-by: Willem de Bruijn <[e-mail]>
Signed-off-by: Jakub Kicinski <[e-mail]>

tools/testing/selftests/net/so_txtime.c

···
 #include <arpa/inet.h>
 #include <error.h>
 #include <errno.h>
+#include <inttypes.h>
 #include <linux/net_tstamp.h>
+#include <linux/errqueue.h>
+#include <linux/ipv6.h>
+#include <linux/tcp.h>
 #include <stdbool.h>
 #include <stdlib.h>
 #include <stdio.h>
···
 static bool	cfg_do_ipv4;
 static bool	cfg_do_ipv6;
 static uint16_t	cfg_port	= 8000;
-static int	cfg_variance_us	= 2000;
+static int	cfg_variance_us	= 4000;
 
 static uint64_t glob_tstart;
 
···
 static struct timed_send cfg_in[MAX_NUM_PKT];
 static struct timed_send cfg_out[MAX_NUM_PKT];
 static int cfg_num_pkt;
+
+static int cfg_errq_level;
+static int cfg_errq_type;
 
 static uint64_t gettime_ns(void)
 {
···
 
 }
 
-static void do_recv_one(int fdr, struct timed_send *ts)
+static bool do_recv_one(int fdr, struct timed_send *ts)
 {
 	int64_t tstop, texpect;
 	char rbuf[2];
 	int ret;
 
 	ret = recv(fdr, rbuf, sizeof(rbuf), 0);
+	if (ret == -1 && errno == EAGAIN)
+		return true;
 	if (ret == -1)
 		error(1, errno, "read");
 	if (ret != 1)
···
 
 	if (labs(tstop - texpect) > cfg_variance_us)
 		error(1, 0, "exceeds variance (%d us)", cfg_variance_us);
+
+	return false;
 }
 
 static void do_recv_verify_empty(int fdr)
···
 		error(1, 0, "recv: not empty as expected (%d, %d)", ret, errno);
 }
 
+static void do_recv_errqueue_timeout(int fdt)
+{
+	char control[CMSG_SPACE(sizeof(struct sock_extended_err)) +
+		     CMSG_SPACE(sizeof(struct sockaddr_in6))] = {0};
+	char data[sizeof(struct ipv6hdr) +
+		  sizeof(struct tcphdr) + 1];
+	struct sock_extended_err *err;
+	struct msghdr msg = {0};
+	struct iovec iov = {0};
+	struct cmsghdr *cm;
+	int64_t tstamp = 0;
+	int ret;
+
+	iov.iov_base = data;
+	iov.iov_len = sizeof(data);
+
+	msg.msg_iov = &iov;
+	msg.msg_iovlen = 1;
+
+	msg.msg_control = control;
+	msg.msg_controllen = sizeof(control);
+
+	while (1) {
+		ret = recvmsg(fdt, &msg, MSG_ERRQUEUE);
+		if (ret == -1 && errno == EAGAIN)
+			break;
+		if (ret == -1)
+			error(1, errno, "errqueue");
+		if (msg.msg_flags != MSG_ERRQUEUE)
+			error(1, 0, "errqueue: flags 0x%x\n", msg.msg_flags);
+
+		cm = CMSG_FIRSTHDR(&msg);
+		if (cm->cmsg_level != cfg_errq_level ||
+		    cm->cmsg_type != cfg_errq_type)
+			error(1, 0, "errqueue: type 0x%x.0x%x\n",
+				    cm->cmsg_level, cm->cmsg_type);
+
+		err = (struct sock_extended_err *)CMSG_DATA(cm);
+		if (err->ee_origin != SO_EE_ORIGIN_TXTIME)
+			error(1, 0, "errqueue: origin 0x%x\n", err->ee_origin);
+		if (err->ee_code != ECANCELED)
+			error(1, 0, "errqueue: code 0x%x\n", err->ee_code);
+
+		tstamp = ((int64_t) err->ee_data) << 32 | err->ee_info;
+		tstamp -= (int64_t) glob_tstart;
+		tstamp /= 1000 * 1000;
+		fprintf(stderr, "send: pkt %c at %" PRId64 "ms dropped\n",
+				data[ret - 1], tstamp);
+
+		msg.msg_flags = 0;
+		msg.msg_controllen = sizeof(control);
+	}
+
+	error(1, 0, "recv: timeout");
+}
+
 static void setsockopt_txtime(int fd)
 {
 	struct sock_txtime so_txtime_val = { .clockid = cfg_clockid };
 	struct sock_txtime so_txtime_val_read = { 0 };
 	socklen_t vallen = sizeof(so_txtime_val);
+
+	so_txtime_val.flags = SOF_TXTIME_REPORT_ERRORS;
 
 	if (setsockopt(fd, SOL_SOCKET, SO_TXTIME,
 		       &so_txtime_val, sizeof(so_txtime_val)))
···
 	for (i = 0; i < cfg_num_pkt; i++)
 		do_send_one(fdt, &cfg_in[i]);
 	for (i = 0; i < cfg_num_pkt; i++)
-		do_recv_one(fdr, &cfg_out[i]);
+		if (do_recv_one(fdr, &cfg_out[i]))
+			do_recv_errqueue_timeout(fdt);
 
 	do_recv_verify_empty(fdr);
 
···
 		addr6.sin6_family = AF_INET6;
 		addr6.sin6_port = htons(cfg_port);
 		addr6.sin6_addr = in6addr_loopback;
+
+		cfg_errq_level = SOL_IPV6;
+		cfg_errq_type = IPV6_RECVERR;
+
 		do_test((void *)&addr6, sizeof(addr6));
 	}
 
···
 		addr4.sin_family = AF_INET;
 		addr4.sin_port = htons(cfg_port);
 		addr4.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
+
+		cfg_errq_level = SOL_IP;
+		cfg_errq_type = IP_RECVERR;
+
 		do_test((void *)&addr4, sizeof(addr4));
 	}
 

tools/testing/selftests/net/so_txtime.sh

···
 
 # Run in network namespace
 if [[ $# -eq 0 ]]; then
-	./in_netns.sh $0 __subprocess
+	if ! ./in_netns.sh $0 __subprocess; then
+		# test is time sensitive, can be flaky
+		echo "test failed: retry once"
+		./in_netns.sh $0 __subprocess
+	fi
+
 	exit $?
 fi
 
···
 ./so_txtime -4 -6 -c mono a,10,b,20 a,10,b,20
 ./so_txtime -4 -6 -c mono a,20,b,10 b,20,a,20
 
-if tc qdisc replace dev lo root etf clockid CLOCK_TAI delta 200000; then
+if tc qdisc replace dev lo root etf clockid CLOCK_TAI delta 400000; then
 	! ./so_txtime -4 -6 -c tai a,-1 a,-1
 	! ./so_txtime -4 -6 -c tai a,0 a,0
 	./so_txtime -4 -6 -c tai a,10 a,10