[PATCH 5/5] perf ftrace: Implement cpu and task filters in BPF

[Namhyung Kim <namhyung@kernel.org>]

Honor cpu and task options to set up filters (by pid or tid) in the
BPF program.  For example, the following command will show latency of
the mutex_lock for process 2570.

  # perf ftrace latency -b -T mutex_lock -p 2570 sleep 3
  #   DURATION     |      COUNT | GRAPH                          |
       0 - 1    us |        675 | ############################## |
       1 - 2    us |          9 |                                |
       2 - 4    us |          0 |                                |
       4 - 8    us |          0 |                                |
       8 - 16   us |          0 |                                |
      16 - 32   us |          0 |                                |
      32 - 64   us |          0 |                                |
      64 - 128  us |          0 |                                |
     128 - 256  us |          0 |                                |
     256 - 512  us |          0 |                                |
     512 - 1024 us |          0 |                                |
       1 - 2    ms |          0 |                                |
       2 - 4    ms |          0 |                                |
       4 - 8    ms |          0 |                                |
       8 - 16   ms |          0 |                                |
      16 - 32   ms |          0 |                                |
      32 - 64   ms |          0 |                                |
      64 - 128  ms |          0 |                                |
     128 - 256  ms |          0 |                                |
     256 - 512  ms |          0 |                                |
     512 - 1024 ms |          0 |                                |
       1 - ...   s |          0 |                                |

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
---
 tools/perf/util/bpf_ftrace.c                | 41 +++++++++++++++++++++
 tools/perf/util/bpf_skel/func_latency.bpf.c | 21 +++++++++++
 2 files changed, 62 insertions(+)

diff --git a/tools/perf/util/bpf_ftrace.c b/tools/perf/util/bpf_ftrace.c
index 1975a6fe73c9..18d6f1558163 100644
--- a/tools/perf/util/bpf_ftrace.c
+++ b/tools/perf/util/bpf_ftrace.c
@@ -6,7 +6,10 @@
 #include <linux/err.h>
 
 #include "util/ftrace.h"
+#include "util/cpumap.h"
+#include "util/thread_map.h"
 #include "util/debug.h"
+#include "util/evlist.h"
 #include "util/bpf_counter.h"
 
 #include "util/bpf_skel/func_latency.skel.h"
@@ -16,6 +19,7 @@ static struct func_latency_bpf *skel;
 int perf_ftrace__latency_prepare_bpf(struct perf_ftrace *ftrace)
 {
 	int fd, err;
+	int i, ncpus = 1, ntasks = 1;
 	struct filter_entry *func;
 	struct bpf_link *begin_link, *end_link;
 
@@ -33,6 +37,17 @@ int perf_ftrace__latency_prepare_bpf(struct perf_ftrace *ftrace)
 		return -1;
 	}
 
+	/* don't need to set cpu filter for system-wide mode */
+	if (ftrace->target.cpu_list) {
+		ncpus = perf_cpu_map__nr(ftrace->evlist->core.cpus);
+		bpf_map__set_max_entries(skel->maps.cpu_filter, ncpus);
+	}
+
+	if (target__has_task(&ftrace->target) || target__none(&ftrace->target)) {
+		ntasks = perf_thread_map__nr(ftrace->evlist->core.threads);
+		bpf_map__set_max_entries(skel->maps.task_filter, ntasks);
+	}
+
 	set_max_rlimit();
 
 	err = func_latency_bpf__load(skel);
@@ -41,6 +56,32 @@ int perf_ftrace__latency_prepare_bpf(struct perf_ftrace *ftrace)
 		goto out;
 	}
 
+	if (ftrace->target.cpu_list) {
+		u32 cpu;
+		u8 val = 1;
+
+		skel->bss->has_cpu = 1;
+		fd = bpf_map__fd(skel->maps.cpu_filter);
+
+		for (i = 0; i < ncpus; i++) {
+			cpu = cpu_map__cpu(ftrace->evlist->core.cpus, i);
+			bpf_map_update_elem(fd, &cpu, &val, BPF_ANY);
+		}
+	}
+
+	if (target__has_task(&ftrace->target) || target__none(&ftrace->target)) {
+		u32 pid;
+		u8 val = 1;
+
+		skel->bss->has_task = 1;
+		fd = bpf_map__fd(skel->maps.task_filter);
+
+		for (i = 0; i < ntasks; i++) {
+			pid = perf_thread_map__pid(ftrace->evlist->core.threads, i);
+			bpf_map_update_elem(fd, &pid, &val, BPF_ANY);
+		}
+	}
+
 	begin_link = bpf_program__attach_kprobe(skel->progs.func_begin,
 						 false, func->name);
 	if (IS_ERR(begin_link)) {
diff --git a/tools/perf/util/bpf_skel/func_latency.bpf.c b/tools/perf/util/bpf_skel/func_latency.bpf.c
index d7d31cfeabf8..17a2a53b4f16 100644
--- a/tools/perf/util/bpf_skel/func_latency.bpf.c
+++ b/tools/perf/util/bpf_skel/func_latency.bpf.c
@@ -36,6 +36,8 @@ struct {
 
 
 int enabled = 0;
+int has_cpu = 0;
+int has_task = 0;
 
 SEC("kprobe/func")
 int BPF_PROG(func_begin)
@@ -46,6 +48,25 @@ int BPF_PROG(func_begin)
 		return 0;
 
 	key = bpf_get_current_pid_tgid();
+
+	if (has_cpu) {
+		__u32 cpu = bpf_get_smp_processor_id();
+		__u8 *ok;
+
+		ok = bpf_map_lookup_elem(&cpu_filter, &cpu);
+		if (!ok)
+			return 0;
+	}
+
+	if (has_task) {
+		__u32 pid = key & 0xffffffff;
+		__u8 *ok;
+
+		ok = bpf_map_lookup_elem(&task_filter, &pid);
+		if (!ok)
+			return 0;
+	}
+
 	now = bpf_ktime_get_ns();
 
 	// overwrite timestamp for nested functions
-- 
2.34.0.rc2.393.gf8c9666880-goog