summaryrefslogtreecommitdiffstats
path: root/tools/perf/tests/openat-syscall-all-cpus.c
blob: a572f87e9c8d8e25665bf75e1c45b1f86ff8a6e9 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
#include "evsel.h"
#include "tests.h"
#include "thread_map.h"
#include "cpumap.h"
#include "debug.h"
#include "stat.h"

int test__openat_syscall_event_on_all_cpus(void)
{
	int err = -1, fd, cpu;
	struct cpu_map *cpus;
	struct perf_evsel *evsel;
	unsigned int nr_openat_calls = 111, i;
	cpu_set_t cpu_set;
	struct thread_map *threads = thread_map__new(-1, getpid(), UINT_MAX);
	char sbuf[STRERR_BUFSIZE];

	if (threads == NULL) {
		pr_debug("thread_map__new\n");
		return -1;
	}

	cpus = cpu_map__new(NULL);
	if (cpus == NULL) {
		pr_debug("cpu_map__new\n");
		goto out_thread_map_delete;
	}

	CPU_ZERO(&cpu_set);

	evsel = perf_evsel__newtp("syscalls", "sys_enter_openat");
	if (evsel == NULL) {
		if (tracefs_configured())
			pr_debug("is tracefs mounted on /sys/kernel/tracing?\n");
		else if (debugfs_configured())
			pr_debug("is debugfs mounted on /sys/kernel/debug?\n");
		else
			pr_debug("Neither tracefs or debugfs is enabled in this kernel\n");
		goto out_thread_map_delete;
	}

	if (perf_evsel__open(evsel, cpus, threads) < 0) {
		pr_debug("failed to open counter: %s, "
			 "tweak /proc/sys/kernel/perf_event_paranoid?\n",
			 strerror_r(errno, sbuf, sizeof(sbuf)));
		goto out_evsel_delete;
	}

	for (cpu = 0; cpu < cpus->nr; ++cpu) {
		unsigned int ncalls = nr_openat_calls + cpu;
		/*
		 * XXX eventually lift this restriction in a way that
		 * keeps perf building on older glibc installations
		 * without CPU_ALLOC. 1024 cpus in 2010 still seems
		 * a reasonable upper limit tho :-)
		 */
		if (cpus->map[cpu] >= CPU_SETSIZE) {
			pr_debug("Ignoring CPU %d\n", cpus->map[cpu]);
			continue;
		}

		CPU_SET(cpus->map[cpu], &cpu_set);
		if (sched_setaffinity(0, sizeof(cpu_set), &cpu_set) < 0) {
			pr_debug("sched_setaffinity() failed on CPU %d: %s ",
				 cpus->map[cpu],
				 strerror_r(errno, sbuf, sizeof(sbuf)));
			goto out_close_fd;
		}
		for (i = 0; i < ncalls; ++i) {
			fd = openat(0, "/etc/passwd", O_RDONLY);
			close(fd);
		}
		CPU_CLR(cpus->map[cpu], &cpu_set);
	}

	/*
	 * Here we need to explicitely preallocate the counts, as if
	 * we use the auto allocation it will allocate just for 1 cpu,
	 * as we start by cpu 0.
	 */
	if (perf_evsel__alloc_counts(evsel, cpus->nr, 1) < 0) {
		pr_debug("perf_evsel__alloc_counts(ncpus=%d)\n", cpus->nr);
		goto out_close_fd;
	}

	err = 0;

	for (cpu = 0; cpu < cpus->nr; ++cpu) {
		unsigned int expected;

		if (cpus->map[cpu] >= CPU_SETSIZE)
			continue;

		if (perf_evsel__read_on_cpu(evsel, cpu, 0) < 0) {
			pr_debug("perf_evsel__read_on_cpu\n");
			err = -1;
			break;
		}

		expected = nr_openat_calls + cpu;
		if (perf_counts(evsel->counts, cpu, 0)->val != expected) {
			pr_debug("perf_evsel__read_on_cpu: expected to intercept %d calls on cpu %d, got %" PRIu64 "\n",
				 expected, cpus->map[cpu], perf_counts(evsel->counts, cpu, 0)->val);
			err = -1;
		}
	}

	perf_evsel__free_counts(evsel);
out_close_fd:
	perf_evsel__close_fd(evsel, 1, threads->nr);
out_evsel_delete:
	perf_evsel__delete(evsel);
out_thread_map_delete:
	thread_map__put(threads);
	return err;
}