perf kwork top: Implements BPF-based cpu usage statistics

Use BPF to collect statistics on the CPU usage based on perf BPF skeletons.

Example usage:

  # perf kwork top -h

   Usage: perf kwork top [<options>]

      -b, --use-bpf         Use BPF to measure task cpu usage
      -C, --cpu <cpu>       list of cpus to profile
      -i, --input <file>    input file name
      -n, --name <name>     event name to profile
      -s, --sort <key[,key2...]>
                            sort by key(s): rate, runtime, tid
          --time <str>      Time span for analysis (start,stop)

  #
  # perf kwork -k sched top -b
  Starting trace, Hit <Ctrl+C> to stop and report
  ^C
  Total  : 160702.425 ms, 8 cpus
  %Cpu(s):  36.00% id,   0.00% hi,   0.00% si
  %Cpu0   [||||||||||||||||||              61.66%]
  %Cpu1   [||||||||||||||||||              61.27%]
  %Cpu2   [|||||||||||||||||||             66.40%]
  %Cpu3   [||||||||||||||||||              61.28%]
  %Cpu4   [||||||||||||||||||              61.82%]
  %Cpu5   [|||||||||||||||||||||||         77.41%]
  %Cpu6   [||||||||||||||||||              61.73%]
  %Cpu7   [||||||||||||||||||              63.25%]

        PID     SPID    %CPU           RUNTIME  COMMMAND
    -------------------------------------------------------------
          0        0   38.72       8089.463 ms  [swapper/1]
          0        0   38.71       8084.547 ms  [swapper/3]
          0        0   38.33       8007.532 ms  [swapper/0]
          0        0   38.26       7992.985 ms  [swapper/6]
          0        0   38.17       7971.865 ms  [swapper/4]
          0        0   36.74       7447.765 ms  [swapper/7]
          0        0   33.59       6486.942 ms  [swapper/2]
          0        0   22.58       3771.268 ms  [swapper/5]
       9545     9351    2.48        447.136 ms  sched-messaging
       9574     9351    2.09        418.583 ms  sched-messaging
       9724     9351    2.05        372.407 ms  sched-messaging
       9531     9351    2.01        368.804 ms  sched-messaging
       9512     9351    2.00        362.250 ms  sched-messaging
       9514     9351    1.95        357.767 ms  sched-messaging
       9538     9351    1.86        384.476 ms  sched-messaging
       9712     9351    1.84        386.490 ms  sched-messaging
       9723     9351    1.83        380.021 ms  sched-messaging
       9722     9351    1.82        382.738 ms  sched-messaging
       9517     9351    1.81        354.794 ms  sched-messaging
       9559     9351    1.79        344.305 ms  sched-messaging
       9725     9351    1.77        365.315 ms  sched-messaging
  <SNIP>

  # perf kwork -k sched top -b -n perf
  Starting trace, Hit <Ctrl+C> to stop and report
  ^C
  Total  : 151563.332 ms, 8 cpus
  %Cpu(s):  26.49% id,   0.00% hi,   0.00% si
  %Cpu0   [                                 0.01%]
  %Cpu1   [                                 0.00%]
  %Cpu2   [                                 0.00%]
  %Cpu3   [                                 0.00%]
  %Cpu4   [                                 0.00%]
  %Cpu5   [                                 0.00%]
  %Cpu6   [                                 0.00%]
  %Cpu7   [                                 0.00%]

        PID     SPID    %CPU           RUNTIME  COMMMAND
    -------------------------------------------------------------
       9754     9754    0.01          2.303 ms  perf

  #
  # perf kwork -k sched top -b -C 2,3,4
  Starting trace, Hit <Ctrl+C> to stop and report
  ^C
  Total  :  48016.721 ms, 3 cpus
  %Cpu(s):  27.82% id,   0.00% hi,   0.00% si
  %Cpu2   [||||||||||||||||||||||          74.68%]
  %Cpu3   [|||||||||||||||||||||           71.06%]
  %Cpu4   [|||||||||||||||||||||           70.91%]

        PID     SPID    %CPU           RUNTIME  COMMMAND
    -------------------------------------------------------------
          0        0   29.08       4734.998 ms  [swapper/4]
          0        0   28.93       4710.029 ms  [swapper/3]
          0        0   25.31       3912.363 ms  [swapper/2]
      10248    10158    1.62        264.931 ms  sched-messaging
      10253    10158    1.62        265.136 ms  sched-messaging
      10158    10158    1.60        263.013 ms  bash
      10360    10158    1.49        243.639 ms  sched-messaging
      10413    10158    1.48        238.604 ms  sched-messaging
      10531    10158    1.47        234.067 ms  sched-messaging
      10400    10158    1.47        240.631 ms  sched-messaging
      10355    10158    1.47        230.586 ms  sched-messaging
      10377    10158    1.43        234.835 ms  sched-messaging
      10526    10158    1.42        232.045 ms  sched-messaging
      10298    10158    1.41        222.396 ms  sched-messaging
      10410    10158    1.38        221.853 ms  sched-messaging
      10364    10158    1.38        226.042 ms  sched-messaging
      10480    10158    1.36        213.633 ms  sched-messaging
      10370    10158    1.36        223.620 ms  sched-messaging
      10553    10158    1.34        217.169 ms  sched-messaging
      10291    10158    1.34        211.516 ms  sched-messaging
      10251    10158    1.34        218.813 ms  sched-messaging
      10522    10158    1.33        218.498 ms  sched-messaging
      10288    10158    1.33        216.787 ms  sched-messaging
  <SNIP>

Reviewed-by: Ian Rogers <irogers@google.com>
Signed-off-by: Yang Jihong <yangjihong1@huawei.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Kan Liang <kan.liang@linux.intel.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Ravi Bangoria <ravi.bangoria@amd.com>
Cc: Sandipan Das <sandipan.das@amd.com>
Link: https://lore.kernel.org/r/20230812084917.169338-15-yangjihong1@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

7 files changed, 565 insertions, 3 deletions

diff --git a/tools/perf/Documentation/perf-kwork.txt b/tools/perf/Documentation/perf-kwork.txt
index 34d6c285e527..109ace1d5e90 100644
--- a/tools/perf/Documentation/perf-kwork.txt
+++ b/tools/perf/Documentation/perf-kwork.txt
@@ -33,6 +33,7 @@ There are several variants of 'perf kwork':
         perf kwork latency -b
         perf kwork timehist
         perf kwork top
+        perf kwork top -b
 
    By default it shows the individual work events such as irq, workqeueu,
    including the run time and delay (time between raise and actually entry):
@@ -181,6 +182,10 @@ OPTIONS for 'perf kwork timehist'
 OPTIONS for 'perf kwork top'
 ---------------------------------
 
+-b::
+--use-bpf::
+	Use BPF to measure task cpu usage.
+
 -C::
 --cpu::
 	Only show events for the given CPU(s) (comma separated list).
diff --git a/tools/perf/Makefile.perf b/tools/perf/Makefile.perf
index 37af6df7b978..cb0422debdb9 100644
--- a/tools/perf/Makefile.perf
+++ b/tools/perf/Makefile.perf
@@ -1032,6 +1032,7 @@ SKELETONS += $(SKEL_OUT)/bperf_leader.skel.h $(SKEL_OUT)/bperf_follower.skel.h
 SKELETONS += $(SKEL_OUT)/bperf_cgroup.skel.h $(SKEL_OUT)/func_latency.skel.h
 SKELETONS += $(SKEL_OUT)/off_cpu.skel.h $(SKEL_OUT)/lock_contention.skel.h
 SKELETONS += $(SKEL_OUT)/kwork_trace.skel.h $(SKEL_OUT)/sample_filter.skel.h
+SKELETONS += $(SKEL_OUT)/kwork_top.skel.h
 SKELETONS += $(SKEL_OUT)/bench_uprobe.skel.h
 SKELETONS += $(SKEL_OUT)/augmented_raw_syscalls.skel.h
 
diff --git a/tools/perf/builtin-kwork.c b/tools/perf/builtin-kwork.c
index d5949ff4bd15..04b966801643 100644
--- a/tools/perf/builtin-kwork.c
+++ b/tools/perf/builtin-kwork.c
@@ -1619,8 +1619,14 @@ static void top_print_header(struct perf_kwork *kwork __maybe_unused)
 	int ret;
 
 	printf("\n ");
-	ret = printf(" %*s  %*s  %*s  %-*s",
+	ret = printf(" %*s %s%*s%s %*s  %*s  %-*s",
 		     PRINT_PID_WIDTH, "PID",
+
+		     kwork->use_bpf ? " " : "",
+		     kwork->use_bpf ? PRINT_PID_WIDTH : 0,
+		     kwork->use_bpf ? "SPID" : "",
+		     kwork->use_bpf ? " " : "",
+
 		     PRINT_CPU_USAGE_WIDTH, "%CPU",
 		     PRINT_RUNTIME_HEADER_WIDTH + RPINT_DECIMAL_WIDTH, "RUNTIME",
 		     PRINT_TASK_NAME_WIDTH, "COMMMAND");
@@ -1640,6 +1646,12 @@ static int top_print_work(struct perf_kwork *kwork __maybe_unused, struct kwork_
 	ret += printf(" %*ld ", PRINT_PID_WIDTH, work->id);
 
 	/*
+	 * tgid
+	 */
+	if (kwork->use_bpf)
+		ret += printf(" %*d ", PRINT_PID_WIDTH, work->tgid);
+
+	/*
 	 * cpu usage
 	 */
 	ret += printf(" %*.*f ",
@@ -1656,7 +1668,13 @@ static int top_print_work(struct perf_kwork *kwork __maybe_unused, struct kwork_
 	/*
 	 * command
 	 */
-	ret += printf(" %-*s", PRINT_TASK_NAME_WIDTH, work->name);
+	if (kwork->use_bpf)
+		ret += printf(" %s%s%s",
+			      work->is_kthread ? "[" : "",
+			      work->name,
+			      work->is_kthread ? "]" : "");
+	else
+		ret += printf(" %-*s", PRINT_TASK_NAME_WIDTH, work->name);
 
 	printf("\n");
 	return ret;
@@ -2153,6 +2171,36 @@ next:
 	printf("\n");
 }
 
+static int perf_kwork__top_bpf(struct perf_kwork *kwork)
+{
+	int ret;
+
+	signal(SIGINT, sig_handler);
+	signal(SIGTERM, sig_handler);
+
+	ret = perf_kwork__top_prepare_bpf(kwork);
+	if (ret)
+		return -1;
+
+	printf("Starting trace, Hit <Ctrl+C> to stop and report\n");
+
+	perf_kwork__top_start();
+
+	/*
+	 * a simple pause, wait here for stop signal
+	 */
+	pause();
+
+	perf_kwork__top_finish();
+
+	perf_kwork__top_read_bpf(kwork);
+
+	perf_kwork__top_cleanup_bpf();
+
+	return 0;
+
+}
+
 static int perf_kwork__top(struct perf_kwork *kwork)
 {
 	struct __top_cpus_runtime *cpus_runtime;
@@ -2165,7 +2213,11 @@ static int perf_kwork__top(struct perf_kwork *kwork)
 	kwork->top_stat.cpus_runtime = cpus_runtime;
 	bitmap_zero(kwork->top_stat.all_cpus_bitmap, MAX_NR_CPUS);
 
-	ret = perf_kwork__read_events(kwork);
+	if (kwork->use_bpf)
+		ret = perf_kwork__top_bpf(kwork);
+	else
+		ret = perf_kwork__read_events(kwork);
+
 	if (ret)
 		goto out;
 
@@ -2380,6 +2432,10 @@ int cmd_kwork(int argc, const char **argv)
 		   "Time span for analysis (start,stop)"),
 	OPT_STRING('i', "input", &input_name, "file",
 		   "input file name"),
+#ifdef HAVE_BPF_SKEL
+	OPT_BOOLEAN('b', "use-bpf", &kwork.use_bpf,
+		    "Use BPF to measure task cpu usage"),
+#endif
 	OPT_PARENT(kwork_options)
 	};
 	const char *kwork_usage[] = {
diff --git a/tools/perf/util/Build b/tools/perf/util/Build
index 6d657c9927f7..0ea5a9d368d4 100644
--- a/tools/perf/util/Build
+++ b/tools/perf/util/Build
@@ -165,6 +165,7 @@ endif
 
 ifeq ($(CONFIG_LIBTRACEEVENT),y)
   perf-$(CONFIG_PERF_BPF_SKEL) += bpf_kwork.o
+  perf-$(CONFIG_PERF_BPF_SKEL) += bpf_kwork_top.o
 endif
 
 perf-$(CONFIG_LIBELF) += symbol-elf.o
diff --git a/tools/perf/util/bpf_kwork_top.c b/tools/perf/util/bpf_kwork_top.c
new file mode 100644
index 000000000000..42897ea22c61
--- /dev/null
+++ b/tools/perf/util/bpf_kwork_top.c
@@ -0,0 +1,286 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * bpf_kwork_top.c
+ *
+ * Copyright (c) 2022  Huawei Inc,  Yang Jihong <yangjihong1@huawei.com>
+ */
+
+#include <time.h>
+#include <fcntl.h>
+#include <signal.h>
+#include <stdio.h>
+#include <unistd.h>
+
+#include <linux/time64.h>
+
+#include "util/debug.h"
+#include "util/evsel.h"
+#include "util/kwork.h"
+
+#include <bpf/bpf.h>
+#include <perf/cpumap.h>
+
+#include "util/bpf_skel/kwork_top.skel.h"
+
+/*
+ * This should be in sync with "util/kwork_top.bpf.c"
+ */
+#define MAX_COMMAND_LEN 16
+
+struct time_data {
+	__u64 timestamp;
+};
+
+struct work_data {
+	__u64 runtime;
+};
+
+struct task_data {
+	__u32 tgid;
+	__u32 is_kthread;
+	char comm[MAX_COMMAND_LEN];
+};
+
+struct work_key {
+	__u32 type;
+	__u32 pid;
+	__u64 task_p;
+};
+
+struct task_key {
+	__u32 pid;
+	__u32 cpu;
+};
+
+struct kwork_class_bpf {
+	struct kwork_class *class;
+	void (*load_prepare)(void);
+};
+
+static struct kwork_top_bpf *skel;
+
+void perf_kwork__top_start(void)
+{
+	struct timespec ts;
+
+	clock_gettime(CLOCK_MONOTONIC, &ts);
+	skel->bss->from_timestamp = (u64)ts.tv_sec * NSEC_PER_SEC + ts.tv_nsec;
+	skel->bss->enabled = 1;
+	pr_debug("perf kwork top start at: %lld\n", skel->bss->from_timestamp);
+}
+
+void perf_kwork__top_finish(void)
+{
+	struct timespec ts;
+
+	skel->bss->enabled = 0;
+	clock_gettime(CLOCK_MONOTONIC, &ts);
+	skel->bss->to_timestamp = (u64)ts.tv_sec * NSEC_PER_SEC + ts.tv_nsec;
+	pr_debug("perf kwork top finish at: %lld\n", skel->bss->to_timestamp);
+}
+
+static void sched_load_prepare(void)
+{
+	bpf_program__set_autoload(skel->progs.on_switch, true);
+}
+
+static struct kwork_class_bpf kwork_sched_bpf = {
+	.load_prepare = sched_load_prepare,
+};
+
+static struct kwork_class_bpf *
+kwork_class_bpf_supported_list[KWORK_CLASS_MAX] = {
+	[KWORK_CLASS_SCHED]	= &kwork_sched_bpf,
+};
+
+static bool valid_kwork_class_type(enum kwork_class_type type)
+{
+	return type >= 0 && type < KWORK_CLASS_MAX ? true : false;
+}
+
+static int setup_filters(struct perf_kwork *kwork)
+{
+	u8 val = 1;
+	int i, nr_cpus, fd;
+	struct perf_cpu_map *map;
+
+	if (kwork->cpu_list) {
+		fd = bpf_map__fd(skel->maps.kwork_top_cpu_filter);
+		if (fd < 0) {
+			pr_debug("Invalid cpu filter fd\n");
+			return -1;
+		}
+
+		map = perf_cpu_map__new(kwork->cpu_list);
+		if (!map) {
+			pr_debug("Invalid cpu_list\n");
+			return -1;
+		}
+
+		nr_cpus = libbpf_num_possible_cpus();
+		for (i = 0; i < perf_cpu_map__nr(map); i++) {
+			struct perf_cpu cpu = perf_cpu_map__cpu(map, i);
+
+			if (cpu.cpu >= nr_cpus) {
+				perf_cpu_map__put(map);
+				pr_err("Requested cpu %d too large\n", cpu.cpu);
+				return -1;
+			}
+			bpf_map_update_elem(fd, &cpu.cpu, &val, BPF_ANY);
+		}
+		perf_cpu_map__put(map);
+
+		skel->bss->has_cpu_filter = 1;
+	}
+
+	return 0;
+}
+
+int perf_kwork__top_prepare_bpf(struct perf_kwork *kwork __maybe_unused)
+{
+	struct bpf_program *prog;
+	struct kwork_class *class;
+	struct kwork_class_bpf *class_bpf;
+	enum kwork_class_type type;
+
+	skel = kwork_top_bpf__open();
+	if (!skel) {
+		pr_debug("Failed to open kwork top skeleton\n");
+		return -1;
+	}
+
+	/*
+	 * set all progs to non-autoload,
+	 * then set corresponding progs according to config
+	 */
+	bpf_object__for_each_program(prog, skel->obj)
+		bpf_program__set_autoload(prog, false);
+
+	list_for_each_entry(class, &kwork->class_list, list) {
+		type = class->type;
+		if (!valid_kwork_class_type(type) ||
+		    !kwork_class_bpf_supported_list[type]) {
+			pr_err("Unsupported bpf trace class %s\n", class->name);
+			goto out;
+		}
+
+		class_bpf = kwork_class_bpf_supported_list[type];
+		class_bpf->class = class;
+
+		if (class_bpf->load_prepare)
+			class_bpf->load_prepare();
+	}
+
+	if (kwork_top_bpf__load(skel)) {
+		pr_debug("Failed to load kwork top skeleton\n");
+		goto out;
+	}
+
+	if (setup_filters(kwork))
+		goto out;
+
+	if (kwork_top_bpf__attach(skel)) {
+		pr_debug("Failed to attach kwork top skeleton\n");
+		goto out;
+	}
+
+	return 0;
+
+out:
+	kwork_top_bpf__destroy(skel);
+	return -1;
+}
+
+static void read_task_info(struct kwork_work *work)
+{
+	int fd;
+	struct task_data data;
+	struct task_key key = {
+		.pid = work->id,
+		.cpu = work->cpu,
+	};
+
+	fd = bpf_map__fd(skel->maps.kwork_top_tasks);
+	if (fd < 0) {
+		pr_debug("Invalid top tasks map fd\n");
+		return;
+	}
+
+	if (!bpf_map_lookup_elem(fd, &key, &data)) {
+		work->tgid = data.tgid;
+		work->is_kthread = data.is_kthread;
+		work->name = strdup(data.comm);
+	}
+}
+static int add_work(struct perf_kwork *kwork, struct work_key *key,
+		    struct work_data *data, int cpu)
+{
+	struct kwork_class_bpf *bpf_trace;
+	struct kwork_work *work;
+	struct kwork_work tmp = {
+		.id = key->pid,
+		.cpu = cpu,
+		.name = NULL,
+	};
+	enum kwork_class_type type = key->type;
+
+	if (!valid_kwork_class_type(type)) {
+		pr_debug("Invalid class type %d to add work\n", type);
+		return -1;
+	}
+
+	bpf_trace = kwork_class_bpf_supported_list[type];
+	tmp.class = bpf_trace->class;
+
+	work = perf_kwork_add_work(kwork, tmp.class, &tmp);
+	if (!work)
+		return -1;
+
+	work->total_runtime = data->runtime;
+	read_task_info(work);
+
+	return 0;
+}
+
+int perf_kwork__top_read_bpf(struct perf_kwork *kwork)
+{
+	int i, fd, nr_cpus;
+	struct work_data *data;
+	struct work_key key, prev;
+
+	fd = bpf_map__fd(skel->maps.kwork_top_works);
+	if (fd < 0) {
+		pr_debug("Invalid top runtime fd\n");
+		return -1;
+	}
+
+	nr_cpus = libbpf_num_possible_cpus();
+	data = calloc(nr_cpus, sizeof(struct work_data));
+	if (!data)
+		return -1;
+
+	memset(&prev, 0, sizeof(prev));
+	while (!bpf_map_get_next_key(fd, &prev, &key)) {
+		if ((bpf_map_lookup_elem(fd, &key, data)) != 0) {
+			pr_debug("Failed to lookup top elem\n");
+			return -1;
+		}
+
+		for (i = 0; i < nr_cpus; i++) {
+			if (data[i].runtime == 0)
+				continue;
+
+			if (add_work(kwork, &key, &data[i], i))
+				return -1;
+		}
+		prev = key;
+	}
+	free(data);
+
+	return 0;
+}
+
+void perf_kwork__top_cleanup_bpf(void)
+{
+	kwork_top_bpf__destroy(skel);
+}
diff --git a/tools/perf/util/bpf_skel/kwork_top.bpf.c b/tools/perf/util/bpf_skel/kwork_top.bpf.c
new file mode 100644
index 000000000000..47ad61608ec7
--- /dev/null
+++ b/tools/perf/util/bpf_skel/kwork_top.bpf.c
@@ -0,0 +1,187 @@
+// SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+// Copyright (c) 2022, Huawei
+
+#include "vmlinux.h"
+#include <bpf/bpf_helpers.h>
+#include <bpf/bpf_tracing.h>
+#include <bpf/bpf_core_read.h>
+
+/*
+ * This should be in sync with "util/kwork.h"
+ */
+enum kwork_class_type {
+	KWORK_CLASS_IRQ,
+	KWORK_CLASS_SOFTIRQ,
+	KWORK_CLASS_WORKQUEUE,
+	KWORK_CLASS_SCHED,
+	KWORK_CLASS_MAX,
+};
+
+#define MAX_ENTRIES     102400
+#define MAX_NR_CPUS     2048
+#define PF_KTHREAD      0x00200000
+#define MAX_COMMAND_LEN 16
+
+struct time_data {
+	__u64 timestamp;
+};
+
+struct work_data {
+	__u64 runtime;
+};
+
+struct task_data {
+	__u32 tgid;
+	__u32 is_kthread;
+	char comm[MAX_COMMAND_LEN];
+};
+
+struct work_key {
+	__u32 type;
+	__u32 pid;
+	__u64 task_p;
+};
+
+struct task_key {
+	__u32 pid;
+	__u32 cpu;
+};
+
+struct {
+	__uint(type, BPF_MAP_TYPE_TASK_STORAGE);
+	__uint(map_flags, BPF_F_NO_PREALLOC);
+	__type(key, int);
+	__type(value, struct time_data);
+} kwork_top_task_time SEC(".maps");
+
+struct {
+	__uint(type, BPF_MAP_TYPE_HASH);
+	__uint(key_size, sizeof(struct task_key));
+	__uint(value_size, sizeof(struct task_data));
+	__uint(max_entries, MAX_ENTRIES);
+} kwork_top_tasks SEC(".maps");
+
+struct {
+	__uint(type, BPF_MAP_TYPE_PERCPU_HASH);
+	__uint(key_size, sizeof(struct work_key));
+	__uint(value_size, sizeof(struct work_data));
+	__uint(max_entries, MAX_ENTRIES);
+} kwork_top_works SEC(".maps");
+
+struct {
+	__uint(type, BPF_MAP_TYPE_HASH);
+	__uint(key_size, sizeof(u32));
+	__uint(value_size, sizeof(u8));
+	__uint(max_entries, MAX_NR_CPUS);
+} kwork_top_cpu_filter SEC(".maps");
+
+int enabled = 0;
+
+int has_cpu_filter = 0;
+
+__u64 from_timestamp = 0;
+__u64 to_timestamp = 0;
+
+static __always_inline int cpu_is_filtered(__u32 cpu)
+{
+	__u8 *cpu_val;
+
+	if (has_cpu_filter) {
+		cpu_val = bpf_map_lookup_elem(&kwork_top_cpu_filter, &cpu);
+		if (!cpu_val)
+			return 1;
+	}
+
+	return 0;
+}
+
+static __always_inline void update_task_info(struct task_struct *task, __u32 cpu)
+{
+	struct task_key key = {
+		.pid = task->pid,
+		.cpu = cpu,
+	};
+
+	if (!bpf_map_lookup_elem(&kwork_top_tasks, &key)) {
+		struct task_data data = {
+			.tgid = task->tgid,
+			.is_kthread = task->flags & PF_KTHREAD ? 1 : 0,
+		};
+		BPF_CORE_READ_STR_INTO(&data.comm, task, comm);
+
+		bpf_map_update_elem(&kwork_top_tasks, &key, &data, BPF_ANY);
+	}
+}
+
+static __always_inline void update_work(struct work_key *key, __u64 delta)
+{
+	struct work_data *data;
+
+	data = bpf_map_lookup_elem(&kwork_top_works, key);
+	if (data) {
+		data->runtime += delta;
+	} else {
+		struct work_data new_data = {
+			.runtime = delta,
+		};
+
+		bpf_map_update_elem(&kwork_top_works, key, &new_data, BPF_ANY);
+	}
+}
+
+static void on_sched_out(struct task_struct *task, __u64 ts, __u32 cpu)
+{
+	__u64 delta;
+	struct time_data *pelem;
+
+	pelem = bpf_task_storage_get(&kwork_top_task_time, task, NULL, 0);
+	if (pelem)
+		delta = ts - pelem->timestamp;
+	else
+		delta = ts - from_timestamp;
+
+	struct work_key key = {
+		.type = KWORK_CLASS_SCHED,
+		.pid = task->pid,
+		.task_p = (__u64)task,
+	};
+
+	update_work(&key, delta);
+	update_task_info(task, cpu);
+}
+
+static void on_sched_in(struct task_struct *task, __u64 ts)
+{
+	struct time_data *pelem;
+
+	pelem = bpf_task_storage_get(&kwork_top_task_time, task, NULL,
+				     BPF_LOCAL_STORAGE_GET_F_CREATE);
+	if (pelem)
+		pelem->timestamp = ts;
+}
+
+SEC("tp_btf/sched_switch")
+int on_switch(u64 *ctx)
+{
+	struct task_struct *prev, *next;
+
+	prev = (struct task_struct *)ctx[1];
+	next = (struct task_struct *)ctx[2];
+
+	if (!enabled)
+		return 0;
+
+	__u32 cpu = bpf_get_smp_processor_id();
+
+	if (cpu_is_filtered(cpu))
+		return 0;
+
+	__u64 ts = bpf_ktime_get_ns();
+
+	on_sched_out(prev, ts, cpu);
+	on_sched_in(next, ts);
+
+	return 0;
+}
+
+char LICENSE[] SEC("license") = "Dual BSD/GPL";
diff --git a/tools/perf/util/kwork.h b/tools/perf/util/kwork.h
index 723b34385df6..76fe2a821bcf 100644
--- a/tools/perf/util/kwork.h
+++ b/tools/perf/util/kwork.h
@@ -135,6 +135,8 @@ struct kwork_work {
 	 * top report
 	 */
 	u32 cpu_usage;
+	u32 tgid;
+	bool is_kthread;
 };
 
 struct kwork_class {
@@ -264,6 +266,13 @@ void perf_kwork__report_cleanup_bpf(void);
 void perf_kwork__trace_start(void);
 void perf_kwork__trace_finish(void);
 
+int perf_kwork__top_prepare_bpf(struct perf_kwork *kwork);
+int perf_kwork__top_read_bpf(struct perf_kwork *kwork);
+void perf_kwork__top_cleanup_bpf(void);
+
+void perf_kwork__top_start(void);
+void perf_kwork__top_finish(void);
+
 #else  /* !HAVE_BPF_SKEL */
 
 static inline int
@@ -283,6 +292,23 @@ static inline void perf_kwork__report_cleanup_bpf(void) {}
 static inline void perf_kwork__trace_start(void) {}
 static inline void perf_kwork__trace_finish(void) {}
 
+static inline int
+perf_kwork__top_prepare_bpf(struct perf_kwork *kwork __maybe_unused)
+{
+	return -1;
+}
+
+static inline int
+perf_kwork__top_read_bpf(struct perf_kwork *kwork __maybe_unused)
+{
+	return -1;
+}
+
+static inline void perf_kwork__top_cleanup_bpf(void) {}
+
+static inline void perf_kwork__top_start(void) {}
+static inline void perf_kwork__top_finish(void) {}
+
 #endif  /* HAVE_BPF_SKEL */
 
 #endif  /* PERF_UTIL_KWORK_H */