These changes are the raw update to linux-4.4.6-rt14. Kernel sources
[kvmfornfv.git] / kernel / samples / bpf / tracex6_user.c
diff --git a/kernel/samples/bpf/tracex6_user.c b/kernel/samples/bpf/tracex6_user.c
new file mode 100644 (file)
index 0000000..8ea4976
--- /dev/null
@@ -0,0 +1,72 @@
+#include <stdio.h>
+#include <unistd.h>
+#include <stdlib.h>
+#include <stdbool.h>
+#include <string.h>
+#include <fcntl.h>
+#include <poll.h>
+#include <sys/ioctl.h>
+#include <linux/perf_event.h>
+#include <linux/bpf.h>
+#include "libbpf.h"
+#include "bpf_load.h"
+
+#define SAMPLE_PERIOD  0x7fffffffffffffffULL
+
+static void test_bpf_perf_event(void)
+{
+       int nr_cpus = sysconf(_SC_NPROCESSORS_CONF);
+       int *pmu_fd = malloc(nr_cpus * sizeof(int));
+       int status, i;
+
+       struct perf_event_attr attr_insn_pmu = {
+               .freq = 0,
+               .sample_period = SAMPLE_PERIOD,
+               .inherit = 0,
+               .type = PERF_TYPE_HARDWARE,
+               .read_format = 0,
+               .sample_type = 0,
+               .config = 0,/* PMU: cycles */
+       };
+
+       for (i = 0; i < nr_cpus; i++) {
+               pmu_fd[i] = perf_event_open(&attr_insn_pmu, -1/*pid*/, i/*cpu*/, -1/*group_fd*/, 0);
+               if (pmu_fd[i] < 0) {
+                       printf("event syscall failed\n");
+                       goto exit;
+               }
+
+               bpf_update_elem(map_fd[0], &i, &pmu_fd[i], BPF_ANY);
+               ioctl(pmu_fd[i], PERF_EVENT_IOC_ENABLE, 0);
+       }
+
+       status = system("ls > /dev/null");
+       if (status)
+               goto exit;
+       status = system("sleep 2");
+       if (status)
+               goto exit;
+
+exit:
+       for (i = 0; i < nr_cpus; i++)
+               close(pmu_fd[i]);
+       close(map_fd[0]);
+       free(pmu_fd);
+}
+
+int main(int argc, char **argv)
+{
+       char filename[256];
+
+       snprintf(filename, sizeof(filename), "%s_kern.o", argv[0]);
+
+       if (load_bpf_file(filename)) {
+               printf("%s", bpf_log_buf);
+               return 1;
+       }
+
+       test_bpf_perf_event();
+       read_trace_pipe();
+
+       return 0;
+}