4 files changed, 287 insertions, 0 deletions

diff --git a/tools/perf/util/Build b/tools/perf/util/Build
index 2e5bfbb69960..294b12430d73 100644
--- a/tools/perf/util/Build
+++ b/tools/perf/util/Build
@@ -144,6 +144,7 @@ perf-$(CONFIG_LIBBPF) += bpf-loader.o
 perf-$(CONFIG_LIBBPF) += bpf_map.o
 perf-$(CONFIG_PERF_BPF_SKEL) += bpf_counter.o
 perf-$(CONFIG_PERF_BPF_SKEL) += bpf_counter_cgroup.o
+perf-$(CONFIG_PERF_BPF_SKEL) += bpf_ftrace.o
 perf-$(CONFIG_BPF_PROLOGUE) += bpf-prologue.o
 perf-$(CONFIG_LIBELF) += symbol-elf.o
 perf-$(CONFIG_LIBELF) += probe-file.o
diff --git a/tools/perf/util/bpf_ftrace.c b/tools/perf/util/bpf_ftrace.c
new file mode 100644
index 000000000000..ec4e2f5a2fc4
--- /dev/null
+++ b/tools/perf/util/bpf_ftrace.c
@@ -0,0 +1,112 @@
+#include <stdio.h>
+#include <fcntl.h>
+#include <stdint.h>
+#include <stdlib.h>
+
+#include <linux/err.h>
+
+#include "util/ftrace.h"
+#include "util/cpumap.h"
+#include "util/debug.h"
+#include "util/bpf_counter.h"
+
+#include "util/bpf_skel/func_latency.skel.h"
+
+static struct func_latency_bpf *skel;
+
+int perf_ftrace__latency_prepare_bpf(struct perf_ftrace *ftrace)
+{
+	int err;
+	struct filter_entry *func;
+
+	if (!list_is_singular(&ftrace->filters)) {
+		pr_err("ERROR: %s target function(s).\n",
+		       list_empty(&ftrace->filters) ? "No" : "Too many");
+		return -1;
+	}
+
+	func = list_first_entry(&ftrace->filters, struct filter_entry, list);
+
+	skel = func_latency_bpf__open();
+	if (!skel) {
+		pr_err("Failed to open func latency skeleton\n");
+		return -1;
+	}
+
+	set_max_rlimit();
+
+	err = func_latency_bpf__load(skel);
+	if (err) {
+		pr_err("Failed to load func latency skeleton\n");
+		goto out;
+	}
+
+	skel->links.func_begin = bpf_program__attach_kprobe(skel->progs.func_begin,
+							    false, func->name);
+	if (IS_ERR(skel->links.func_begin)) {
+		pr_err("Failed to attach fentry program\n");
+		err = PTR_ERR(skel->links.func_begin);
+		goto out;
+	}
+
+	skel->links.func_end = bpf_program__attach_kprobe(skel->progs.func_end,
+							  true, func->name);
+	if (IS_ERR(skel->links.func_end)) {
+		pr_err("Failed to attach fexit program\n");
+		err = PTR_ERR(skel->links.func_end);
+		goto out;
+	}
+
+	/* XXX: we don't actually use this fd - just for poll() */
+	return open("/dev/null", O_RDONLY);
+
+out:
+	return err;
+}
+
+int perf_ftrace__latency_start_bpf(struct perf_ftrace *ftrace __maybe_unused)
+{
+	skel->bss->enabled = 1;
+	return 0;
+}
+
+int perf_ftrace__latency_stop_bpf(struct perf_ftrace *ftrace __maybe_unused)
+{
+	skel->bss->enabled = 0;
+	return 0;
+}
+
+int perf_ftrace__latency_read_bpf(struct perf_ftrace *ftrace __maybe_unused,
+				  int buckets[])
+{
+	int i, fd, err;
+	u32 idx;
+	u64 *hist;
+	int ncpus = cpu__max_cpu();
+
+	fd = bpf_map__fd(skel->maps.latency);
+
+	hist = calloc(ncpus, sizeof(*hist));
+	if (hist == NULL)
+		return -ENOMEM;
+
+	for (idx = 0; idx < NUM_BUCKET; idx++) {
+		err = bpf_map_lookup_elem(fd, &idx, hist);
+		if (err) {
+			buckets[idx] = 0;
+			continue;
+		}
+
+		for (i = 0; i < ncpus; i++)
+			buckets[idx] += hist[i];
+	}
+
+	free(hist);
+	return 0;
+}
+
+int perf_ftrace__latency_cleanup_bpf(struct perf_ftrace *ftrace __maybe_unused)
+{
+	func_latency_bpf__destroy(skel);
+	return 0;
+}
diff --git a/tools/perf/util/bpf_skel/func_latency.bpf.c b/tools/perf/util/bpf_skel/func_latency.bpf.c
new file mode 100644
index 000000000000..ccd96b09fc42
--- /dev/null
+++ b/tools/perf/util/bpf_skel/func_latency.bpf.c
@@ -0,0 +1,93 @@
+// SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+// Copyright (c) 2021 Google
+#include "vmlinux.h"
+#include <bpf/bpf_helpers.h>
+#include <bpf/bpf_tracing.h>
+
+// This should be in sync with "util/ftrace.h"
+#define NUM_BUCKET  22
+
+struct {
+	__uint(type, BPF_MAP_TYPE_HASH);
+	__uint(key_size, sizeof(__u64));
+	__uint(value_size, sizeof(__u64));
+	__uint(max_entries, 10000);
+} functime SEC(".maps");
+
+struct {
+	__uint(type, BPF_MAP_TYPE_HASH);
+	__uint(key_size, sizeof(__u32));
+	__uint(value_size, sizeof(__u8));
+	__uint(max_entries, 1);
+} cpu_filter SEC(".maps");
+
+struct {
+	__uint(type, BPF_MAP_TYPE_HASH);
+	__uint(key_size, sizeof(__u32));
+	__uint(value_size, sizeof(__u8));
+	__uint(max_entries, 1);
+} task_filter SEC(".maps");
+
+struct {
+	__uint(type, BPF_MAP_TYPE_PERCPU_ARRAY);
+	__uint(key_size, sizeof(__u32));
+	__uint(value_size, sizeof(__u64));
+	__uint(max_entries, NUM_BUCKET);
+} latency SEC(".maps");
+
+
+int enabled = 0;
+
+SEC("kprobe/func")
+int BPF_PROG(func_begin)
+{
+	__u64 key, now;
+
+	if (!enabled)
+		return 0;
+
+	key = bpf_get_current_pid_tgid();
+	now = bpf_ktime_get_ns();
+
+	// overwrite timestamp for nested functions
+	bpf_map_update_elem(&functime, &key, &now, BPF_ANY);
+	return 0;
+}
+
+SEC("kretprobe/func")
+int BPF_PROG(func_end)
+{
+	__u64 tid;
+	__u64 *start;
+
+	if (!enabled)
+		return 0;
+
+	tid = bpf_get_current_pid_tgid();
+
+	start = bpf_map_lookup_elem(&functime, &tid);
+	if (start) {
+		__s64 delta = bpf_ktime_get_ns() - *start;
+		__u32 key;
+		__u64 *hist;
+
+		bpf_map_delete_elem(&functime, &tid);
+
+		if (delta < 0)
+			return 0;
+
+		// calculate index using delta in usec
+		for (key = 0; key < (NUM_BUCKET - 1); key++) {
+			if (delta < ((1000UL) << key))
+				break;
+		}
+
+		hist = bpf_map_lookup_elem(&latency, &key);
+		if (!hist)
+			return 0;
+
+		*hist += 1;
+	}
+
+	return 0;
+}
diff --git a/tools/perf/util/ftrace.h b/tools/perf/util/ftrace.h
new file mode 100644
index 000000000000..887f68a185f7
--- /dev/null
+++ b/tools/perf/util/ftrace.h
@@ -0,0 +1,81 @@
+#ifndef __PERF_FTRACE_H__
+#define __PERF_FTRACE_H__
+
+#include <linux/list.h>
+
+#include "target.h"
+
+struct evlist;
+
+struct perf_ftrace {
+	struct evlist		*evlist;
+	struct target		target;
+	const char		*tracer;
+	struct list_head	filters;
+	struct list_head	notrace;
+	struct list_head	graph_funcs;
+	struct list_head	nograph_funcs;
+	unsigned long		percpu_buffer_size;
+	bool			inherit;
+	int			graph_depth;
+	int			func_stack_trace;
+	int			func_irq_info;
+	int			graph_nosleep_time;
+	int			graph_noirqs;
+	int			graph_verbose;
+	int			graph_thresh;
+	unsigned int		initial_delay;
+};
+
+struct filter_entry {
+	struct list_head	list;
+	char			name[];
+};
+
+#define NUM_BUCKET  22  /* 20 + 2 (for outliers in both direction) */
+
+#ifdef HAVE_BPF_SKEL
+
+int perf_ftrace__latency_prepare_bpf(struct perf_ftrace *ftrace);
+int perf_ftrace__latency_start_bpf(struct perf_ftrace *ftrace);
+int perf_ftrace__latency_stop_bpf(struct perf_ftrace *ftrace);
+int perf_ftrace__latency_read_bpf(struct perf_ftrace *ftrace,
+				  int buckets[]);
+int perf_ftrace__latency_cleanup_bpf(struct perf_ftrace *ftrace);
+
+#else  /* !HAVE_BPF_SKEL */
+
+static inline int
+perf_ftrace__latency_prepare_bpf(struct perf_ftrace *ftrace __maybe_unused)
+{
+	return -1;
+}
+
+static inline int
+perf_ftrace__latency_start_bpf(struct perf_ftrace *ftrace __maybe_unused)
+{
+	return -1;
+}
+
+static inline int
+perf_ftrace__latency_stop_bpf(struct perf_ftrace *ftrace __maybe_unused)
+{
+	return -1;
+}
+
+static inline int
+perf_ftrace__latency_read_bpf(struct perf_ftrace *ftrace __maybe_unused,
+			      int buckets[] __maybe_unused)
+{
+	return -1;
+}
+
+static inline int
+perf_ftrace__latency_cleanup_bpf(struct perf_ftrace *ftrace __maybe_unused)
+{
+	return -1;
+}
+
+#endif  /* HAVE_BPF_SKEL */
+
+#endif  /* __PERF_FTRACE_H__ */