diff --git a/tools/testing/selftests/bpf/prog_tests/perf_link.c b/tools/testing/selftests/bpf/prog_tests/perf_link.c
new file mode 100644
index 000000000000..b1abd0c46607
--- /dev/null
+++ b/tools/testing/selftests/bpf/prog_tests/perf_link.c
@@ -0,0 +1,89 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2021 Facebook */
+#define _GNU_SOURCE
+#include <pthread.h>
+#include <sched.h>
+#include <test_progs.h>
+#include "test_perf_link.skel.h"
+
+static void burn_cpu(void)
+{
+	volatile int j = 0;
+	cpu_set_t cpu_set;
+	int i, err;
+
+	/* generate some branches on cpu 0 */
+	CPU_ZERO(&cpu_set);
+	CPU_SET(0, &cpu_set);
+	err = pthread_setaffinity_np(pthread_self(), sizeof(cpu_set), &cpu_set);
+	ASSERT_OK(err, "set_thread_affinity");
+
+	/* spin the loop for a while (random high number) */
+	for (i = 0; i < 1000000; ++i)
+		++j;
+}
+
+void test_perf_link(void)
+{
+	struct test_perf_link *skel = NULL;
+	struct perf_event_attr attr;
+	int pfd = -1, link_fd = -1, err;
+	int run_cnt_before, run_cnt_after;
+	struct bpf_link_info info;
+	__u32 info_len = sizeof(info);
+
+	/* create perf event */
+	memset(&attr, 0, sizeof(attr));
+	attr.size = sizeof(attr);
+	attr.type = PERF_TYPE_SOFTWARE;
+	attr.config = PERF_COUNT_SW_CPU_CLOCK;
+	attr.freq = 1;
+	attr.sample_freq = 4000;
+	pfd = syscall(__NR_perf_event_open, &attr, -1, 0, -1, PERF_FLAG_FD_CLOEXEC);
+	if (!ASSERT_GE(pfd, 0, "perf_fd"))
+		goto cleanup;
+
+	skel = test_perf_link__open_and_load();
+	if (!ASSERT_OK_PTR(skel, "skel_load"))
+		goto cleanup;
+
+	link_fd = bpf_link_create(bpf_program__fd(skel->progs.handler), pfd,
+				  BPF_PERF_EVENT, NULL);
+	if (!ASSERT_GE(link_fd, 0, "link_fd"))
+		goto cleanup;
+
+	memset(&info, 0, sizeof(info));
+	err = bpf_obj_get_info_by_fd(link_fd, &info, &info_len);
+	if (!ASSERT_OK(err, "link_get_info"))
+		goto cleanup;
+
+	ASSERT_EQ(info.type, BPF_LINK_TYPE_PERF_EVENT, "link_type");
+	ASSERT_GT(info.id, 0, "link_id");
+	ASSERT_GT(info.prog_id, 0, "link_prog_id");
+
+	/* ensure we get at least one perf_event prog execution */
+	burn_cpu();
+	ASSERT_GT(skel->bss->run_cnt, 0, "run_cnt");
+
+	/* perf_event is still active, but we close link and BPF program
+	 * shouldn't be executed anymore
+	 */
+	close(link_fd);
+	link_fd = -1;
+
+	/* make sure there are no stragglers */
+	kern_sync_rcu();
+
+	run_cnt_before = skel->bss->run_cnt;
+	burn_cpu();
+	run_cnt_after = skel->bss->run_cnt;
+
+	ASSERT_EQ(run_cnt_before, run_cnt_after, "run_cnt_before_after");
+
+cleanup:
+	if (link_fd >= 0)
+		close(link_fd);
+	if (pfd >= 0)
+		close(pfd);
+	test_perf_link__destroy(skel);
+}
diff --git a/tools/testing/selftests/bpf/progs/test_perf_link.c b/tools/testing/selftests/bpf/progs/test_perf_link.c
new file mode 100644
index 000000000000..c1db9fd98d0c
--- /dev/null
+++ b/tools/testing/selftests/bpf/progs/test_perf_link.c
@@ -0,0 +1,16 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2021 Facebook */
+#include "vmlinux.h"
+#include <bpf/bpf_helpers.h>
+#include <bpf/bpf_tracing.h>
+
+int run_cnt = 0;
+
+SEC("perf_event")
+int handler(struct pt_regs *ctx)
+{
+	__sync_fetch_and_add(&run_cnt, 1);
+	return 0;
+}
+
+char _license[] SEC("license") = "GPL";