[PATCH liburing v2 0/2] add tx timestamp tests

[PATCH liburing v2 0/2] add tx timestamp tests

[Pavel Begunkov]

Add definitions / tests test for tx timestamping io_uring API. See

https://lore.kernel.org/all/cover.1750065793.git.asml.silence@gmail.com/

Pavel Begunkov (2):
  Sync io_uring.h with tx timestamp api
  tests: add a tx timestamp test

 src/include/liburing/io_uring.h |  16 ++
 test/Makefile                   |   1 +
 test/timestamp.c                | 373 ++++++++++++++++++++++++++++++++
 3 files changed, 390 insertions(+)
 create mode 100644 test/timestamp.c

-- 
2.49.0

[PATCH liburing v2 1/2] Sync io_uring.h with tx timestamp api

[Pavel Begunkov]

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
---
 src/include/liburing/io_uring.h | 16 ++++++++++++++++
 1 file changed, 16 insertions(+)

diff --git a/src/include/liburing/io_uring.h b/src/include/liburing/io_uring.h
index 73d29976..94de038d 100644
--- a/src/include/liburing/io_uring.h
+++ b/src/include/liburing/io_uring.h
@@ -915,6 +915,22 @@ enum io_uring_socket_op {
 	SOCKET_URING_OP_SIOCOUTQ,
 	SOCKET_URING_OP_GETSOCKOPT,
 	SOCKET_URING_OP_SETSOCKOPT,
+	SOCKET_URING_OP_TX_TIMESTAMP,
+};
+
+/*
+ * SOCKET_URING_OP_TX_TIMESTAMP definitions
+ */
+
+#define IORING_TIMESTAMP_HW_SHIFT	16
+/* The cqe->flags bit from which the timestamp type is stored */
+#define IORING_TIMESTAMP_TYPE_SHIFT	(IORING_TIMESTAMP_HW_SHIFT + 1)
+/* The cqe->flags flag signifying whether it's a hardware timestamp */
+#define IORING_CQE_F_TSTAMP_HW		((__u32)1 << IORING_TIMESTAMP_HW_SHIFT);
+
+struct io_timespec {
+	__u64		tv_sec;
+	__u64		tv_nsec;
 };
 
 /* Zero copy receive refill queue entry */
-- 
2.49.0

[PATCH liburing v2 2/2] tests: add a tx timestamp test

[Pavel Begunkov]

Add a test for tx timestamping io_uring API.

https://lore.kernel.org/all/cover.1750065793.git.asml.silence@gmail.com/

The test itself is just an adapted version of txtimestamp.c
kernel selftest.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
---
 test/Makefile    |   1 +
 test/timestamp.c | 373 +++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 374 insertions(+)
 create mode 100644 test/timestamp.c

diff --git a/test/Makefile b/test/Makefile
index ee09f5a8..99b272d7 100644
--- a/test/Makefile
+++ b/test/Makefile
@@ -249,6 +249,7 @@ test_srcs := \
 	xattr.c \
 	zcrx.c \
 	vec-regbuf.c \
+	timestamp.c \
 	# EOL
 
 # Please keep this list sorted alphabetically.
diff --git a/test/timestamp.c b/test/timestamp.c
new file mode 100644
index 00000000..98180557
--- /dev/null
+++ b/test/timestamp.c
@@ -0,0 +1,373 @@
+#include <arpa/inet.h>
+#include <error.h>
+#include <errno.h>
+#include <inttypes.h>
+#include <linux/errqueue.h>
+#include <linux/ipv6.h>
+#include <linux/net_tstamp.h>
+#include <netinet/in.h>
+#include <netinet/ip.h>
+#include <netinet/udp.h>
+#include <netinet/tcp.h>
+#include <stdarg.h>
+#include <stdbool.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <sys/socket.h>
+#include <time.h>
+#include <unistd.h>
+#include <assert.h>
+
+#include "liburing.h"
+#include "helpers.h"
+
+#ifndef SCM_TS_OPT_ID
+/* It's generally 81 except for a few selected archs. Jens requested it
+ * to be set here, please report if the test fails.
+ */
+#define SCM_TS_OPT_ID 81
+#endif
+
+static const int cfg_payload_len = 10;
+static uint16_t dest_port = 9000;
+static uint32_t ts_opt_id = 81;
+static bool cfg_use_cmsg_opt_id = false;
+static char buffer[128];
+static const bool verbose = false;
+
+static struct sockaddr_in6 daddr6;
+
+static int saved_tskey = -1;
+static int saved_tskey_type = -1;
+
+struct ctx {
+	int family;
+	int proto;
+	int report_opt;
+	int num_pkts;
+};
+
+static int validate_key(int tskey, int tstype, struct ctx *ctx)
+{
+	int stepsize;
+
+	/* compare key for each subsequent request
+	 * must only test for one type, the first one requested
+	 */
+	if (saved_tskey == -1 || cfg_use_cmsg_opt_id)
+		saved_tskey_type = tstype;
+	else if (saved_tskey_type != tstype)
+		return 0;
+
+	stepsize = ctx->proto == SOCK_STREAM ? cfg_payload_len : 1;
+	stepsize = cfg_use_cmsg_opt_id ? 0 : stepsize;
+	if (tskey != saved_tskey + stepsize) {
+		fprintf(stderr, "ERROR: key %d, expected %d\n",
+				tskey, saved_tskey + stepsize);
+		return -EINVAL;
+	}
+
+	saved_tskey = tskey;
+	return 0;
+}
+
+static int test_prep_sock(int family, int proto, unsigned report_opt)
+{
+	int ipproto = proto == SOCK_STREAM ? IPPROTO_TCP : IPPROTO_UDP;
+	unsigned int sock_opt;
+	int fd, val = 1;
+
+	fd = socket(family, proto, ipproto);
+	if (fd < 0)
+		error(1, errno, "socket");
+
+	if (proto == SOCK_STREAM) {
+		if (setsockopt(fd, IPPROTO_TCP, TCP_NODELAY,
+			       (char*) &val, sizeof(val)))
+			error(1, 0, "setsockopt no nagle");
+
+		if (connect(fd, (void *) &daddr6, sizeof(daddr6)))
+			error(1, errno, "connect ipv6");
+	}
+
+	sock_opt = SOF_TIMESTAMPING_SOFTWARE |
+		   SOF_TIMESTAMPING_OPT_CMSG |
+		   SOF_TIMESTAMPING_OPT_ID;
+	sock_opt |= report_opt;
+	sock_opt |= SOF_TIMESTAMPING_OPT_TSONLY;
+
+	if (setsockopt(fd, SOL_SOCKET, SO_TIMESTAMPING,
+		       (char *) &sock_opt, sizeof(sock_opt)))
+		error(1, 0, "setsockopt timestamping");
+
+	return fd;
+}
+
+#define MAX_PACKETS 32
+
+struct send_req {
+	struct msghdr msg;
+	struct iovec iov;
+	char control[CMSG_SPACE(sizeof(uint32_t))];
+};
+
+static void queue_ts_cmd(struct io_uring *ring, int fd)
+{
+	struct io_uring_sqe *sqe;
+
+	sqe = io_uring_get_sqe(ring);
+	io_uring_prep_rw(IORING_OP_URING_CMD, sqe, fd, NULL, 0, 0);
+	sqe->cmd_op = SOCKET_URING_OP_TX_TIMESTAMP;
+	sqe->user_data = 43;
+}
+
+static void queue_send(struct io_uring *ring, int fd, void *buf, struct send_req *r,
+			int proto)
+{
+	struct io_uring_sqe *sqe;
+
+	r->iov.iov_base = buf;
+	r->iov.iov_len = cfg_payload_len;
+
+	memset(&r->msg, 0, sizeof(r->msg));
+	r->msg.msg_iov = &r->iov;
+	r->msg.msg_iovlen = 1;
+	if (proto == SOCK_STREAM) {
+		r->msg.msg_name = (void *)&daddr6;
+		r->msg.msg_namelen = sizeof(daddr6);
+	}
+
+	if (cfg_use_cmsg_opt_id) {
+		struct cmsghdr *cmsg;
+
+		memset(r->control, 0, sizeof(r->control));
+		r->msg.msg_control = r->control;
+		r->msg.msg_controllen = CMSG_SPACE(sizeof(uint32_t));
+
+		cmsg = CMSG_FIRSTHDR(&r->msg);
+		cmsg->cmsg_level = SOL_SOCKET;
+		cmsg->cmsg_type = SCM_TS_OPT_ID;
+		cmsg->cmsg_len = CMSG_LEN(sizeof(uint32_t));
+
+		*((uint32_t *)CMSG_DATA(cmsg)) = ts_opt_id;
+		saved_tskey = ts_opt_id;
+	}
+
+	sqe = io_uring_get_sqe(ring);
+	io_uring_prep_sendmsg(sqe, fd, &r->msg, 0);
+	sqe->user_data = 0;
+}
+
+static const char *get_tstype_name(int tstype)
+{
+	if (tstype == SCM_TSTAMP_SCHED)
+		return "ENQ";
+	if (tstype == SCM_TSTAMP_SND)
+		return "SND";
+	if (tstype == SCM_TSTAMP_ACK)
+		return "ACK";
+	return "unknown";
+}
+
+static int do_test(struct ctx *ctx)
+{
+	struct send_req reqs[MAX_PACKETS];
+	struct io_uring_cqe *cqe;
+	struct io_uring ring;
+	unsigned long head;
+	int cqes_seen = 0;
+	int i, fd, ret;
+	int ts_expected = 0, ts_got = 0;
+
+	ts_expected += !!(ctx->report_opt & SOF_TIMESTAMPING_TX_SCHED);
+	ts_expected += !!(ctx->report_opt & SOF_TIMESTAMPING_TX_SOFTWARE);
+	ts_expected += !!(ctx->report_opt & SOF_TIMESTAMPING_TX_ACK);
+
+	ret = t_create_ring(32, &ring, IORING_SETUP_CQE32);
+	if (ret == T_SETUP_SKIP)
+		return T_EXIT_SKIP;
+	else if (ret)
+		t_error(1, ret, "queue init\n");
+
+	assert(ctx->num_pkts <= MAX_PACKETS);
+
+	fd = test_prep_sock(ctx->family, ctx->proto, ctx->report_opt);
+	if (fd < 0)
+		t_error(1, fd, "can't create socket\n");
+
+	memset(buffer, 'a', cfg_payload_len);
+	saved_tskey = -1;
+
+	if (cfg_use_cmsg_opt_id)
+		saved_tskey = ts_opt_id;
+
+	for (i =  0; i < ctx->num_pkts; i++) {
+		queue_send(&ring, fd, buffer, &reqs[i], ctx->proto);
+		ret = io_uring_submit(&ring);
+		if (ret != 1)
+			t_error(1, ret, "submit failed");
+
+		ret = io_uring_wait_cqe(&ring, &cqe);
+		if (ret || cqe->res != cfg_payload_len) {
+			fprintf(stderr, "wait send cqe, %d %d, expected %d\n",
+				ret, cqe->res, cfg_payload_len);
+			return T_EXIT_FAIL;
+		}
+		io_uring_cqe_seen(&ring, cqe);
+	}
+
+	usleep(200000);
+
+	queue_ts_cmd(&ring, fd);
+	ret = io_uring_submit(&ring);
+	if (ret != 1)
+		t_error(1, ret, "submit failed");
+
+	ret = io_uring_wait_cqe(&ring, &cqe);
+	if (ret) {
+		fprintf(stderr, "wait_cqe failed %d\n", ret);
+		return T_EXIT_FAIL;
+	}
+
+	io_uring_for_each_cqe(&ring, head, cqe) {
+		struct io_timespec *ts;
+		int tskey, tstype;
+		bool hwts;
+
+		cqes_seen++;
+
+		if (!(cqe->flags & IORING_CQE_F_MORE)) {
+			if (cqe->res == -EINVAL || cqe->res == -EOPNOTSUPP)
+				return T_EXIT_SKIP;
+			if (cqe->res)
+				t_error(1, 0, "failed cqe %i", cqe->res);
+			break;
+		}
+
+		ts = (void *)(cqe + 1);
+		tstype = cqe->flags >> IORING_TIMESTAMP_TYPE_SHIFT;
+		tskey = cqe->res;
+		hwts = cqe->flags & IORING_CQE_F_TSTAMP_HW;
+
+		ts_got++;
+		if (verbose)
+			fprintf(stderr, "ts: key %x, type %i (%s), is hw %i, sec %lu, nsec %lu\n",
+				tskey, tstype, get_tstype_name(tstype), hwts,
+				(unsigned long)ts->tv_sec,
+				(unsigned long)ts->tv_nsec);
+
+		ret = validate_key(tskey, tstype, ctx);
+		if (ret)
+			return T_EXIT_FAIL;
+	}
+
+	if (ts_got != ts_expected) {
+		fprintf(stderr, "expected %i timestamps, got %i\n",
+			ts_expected, ts_got);
+		return -EINVAL;
+	}
+
+	close(fd);
+	io_uring_cq_advance(&ring, cqes_seen);
+	io_uring_queue_exit(&ring);
+	return T_EXIT_PASS;
+}
+
+static void resolve_hostname(const char *name, int port)
+{
+	memset(&daddr6, 0, sizeof(daddr6));
+	daddr6.sin6_family = AF_INET6;
+	daddr6.sin6_port = htons(port);
+	if (inet_pton(AF_INET6, name, &daddr6.sin6_addr) != 1)
+		t_error(1, 0, "ipv6 parse error: %s", name);
+}
+
+static void do_listen(int family, int type, void *addr, int alen)
+{
+	int fd;
+
+	fd = socket(family, type, 0);
+	if (fd == -1)
+		error(1, errno, "socket rx");
+
+	if (bind(fd, addr, alen))
+		error(1, errno, "bind rx");
+
+	if (type == SOCK_STREAM && listen(fd, 10))
+		error(1, errno, "listen rx");
+
+	/* leave fd open, will be closed on process exit.
+	 * this enables connect() to succeed and avoids icmp replies
+	 */
+}
+
+static int do_main(int family, int proto)
+{
+	struct ctx ctx;
+	int ret;
+
+	ctx.num_pkts = 1;
+	ctx.family = family;
+	ctx.proto = proto;
+
+	if (verbose)
+		fprintf(stderr, "test SND\n");
+	ctx.report_opt = SOF_TIMESTAMPING_TX_SOFTWARE;
+	ret = do_test(&ctx);
+	if (ret) {
+		if (ret == T_EXIT_SKIP)
+			fprintf(stderr, "no timestamp cmd, skip\n");
+		return ret;
+	}
+
+	if (verbose)
+		fprintf(stderr, "test ENQ\n");
+	ctx.report_opt = SOF_TIMESTAMPING_TX_SCHED;
+	ret = do_test(&ctx);
+	if (ret)
+		return T_EXIT_FAIL;
+
+	if (verbose)
+		fprintf(stderr, "test ENQ + SND\n");
+	ctx.report_opt = SOF_TIMESTAMPING_TX_SCHED | SOF_TIMESTAMPING_TX_SOFTWARE;
+	ret = do_test(&ctx);
+	if (ret)
+		return T_EXIT_FAIL;
+
+	if (proto == SOCK_STREAM) {
+		if (verbose)
+			fprintf(stderr, "test ACK\n");
+		ctx.report_opt = SOF_TIMESTAMPING_TX_ACK;
+		ret = do_test(&ctx);
+		if (ret)
+			return T_EXIT_FAIL;
+
+		if (verbose)
+			fprintf(stderr, "test SND + ACK\n");
+		ctx.report_opt = SOF_TIMESTAMPING_TX_SOFTWARE |
+				  SOF_TIMESTAMPING_TX_ACK;
+		ret = do_test(&ctx);
+		if (ret)
+			return T_EXIT_FAIL;
+
+		if (verbose)
+			fprintf(stderr, "test ENQ + SND + ACK\n");
+		ctx.report_opt = SOF_TIMESTAMPING_TX_SCHED |
+				  SOF_TIMESTAMPING_TX_SOFTWARE |
+				  SOF_TIMESTAMPING_TX_ACK;
+		ret = do_test(&ctx);
+		if (ret)
+			return T_EXIT_FAIL;
+	}
+	return 0;
+}
+
+int main(int argc, char **argv)
+{
+	const char *hostname = "::1";
+
+	resolve_hostname(hostname, dest_port);
+	do_listen(PF_INET6, SOCK_STREAM, &daddr6, sizeof(daddr6));
+	return do_main(PF_INET6, SOCK_STREAM);
+}
-- 
2.49.0

Re: [PATCH liburing v2 0/2] add tx timestamp tests

[Jens Axboe]

On Mon, 30 Jun 2025 18:10:29 +0100, Pavel Begunkov wrote:
> Add definitions / tests test for tx timestamping io_uring API. See
> 
> https://lore.kernel.org/all/cover.1750065793.git.asml.silence@gmail.com/
> 
> Pavel Begunkov (2):
>   Sync io_uring.h with tx timestamp api
>   tests: add a tx timestamp test
> 
> [...]

Applied, thanks!

[1/2] Sync io_uring.h with tx timestamp api
      commit: 2b11d4753496335cec916d3c0a46c181ea79c6b6
[2/2] tests: add a tx timestamp test
      commit: 21224848af24d379d54fbf1bd43a60861fe19f9b

Best regards,
-- 
Jens Axboe