perf record: Add cgroup support for off-cpu profiling

This covers two different use cases.  The first one is cgroup
filtering given by -G/--cgroup option which controls the off-cpu
profiling for tasks in the given cgroups only.

The other use case is cgroup sampling which is enabled by
--all-cgroups option and it adds PERF_SAMPLE_CGROUP to the sample_type
to set the cgroup id of the task in the sample data.

Example output.

  $ sudo perf record -a --off-cpu --all-cgroups sleep 1

  $ sudo perf report --stdio -s comm,cgroup --call-graph=no
  ...
  # Samples: 144  of event 'offcpu-time'
  # Event count (approx.): 48452045427
  #
  # Children      Self  Command          Cgroup
  # ........  ........  ...............  ..........................................
  #
      61.57%     5.60%  Chrome_ChildIOT  /user.slice/user-657345.slice/user@657345.service/app.slice/...
      29.51%     7.38%  Web Content      /user.slice/user-657345.slice/user@657345.service/app.slice/...
      17.48%     1.59%  Chrome_IOThread  /user.slice/user-657345.slice/user@657345.service/app.slice/...
      16.48%     4.12%  pipewire-pulse   /user.slice/user-657345.slice/user@657345.service/session.slice/...
      14.48%     2.07%  perf             /user.slice/user-657345.slice/user@657345.service/app.slice/...
      14.30%     7.15%  CompositorTileW  /user.slice/user-657345.slice/user@657345.service/app.slice/...
      13.33%     6.67%  Timer            /user.slice/user-657345.slice/user@657345.service/app.slice/...
  ...
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Acked-by: Ian Rogers <irogers@google.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Blake Jones <blakejones@google.com>
Cc: Hao Luo <haoluo@google.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Milian Wolff <milian.wolff@kdab.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Song Liu <songliubraving@fb.com>
Cc: bpf@vger.kernel.org
Link: https://lore.kernel.org/r/20220518224725.742882-6-namhyung@kernel.orgSigned-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

tools/perf/builtin-record.c

@@ -892,7 +892,7 @@ static int record__config_text_poke(struct evlist *evlist)
 
 static int record__config_off_cpu(struct record *rec)
 {
-	return off_cpu_prepare(rec->evlist, &rec->opts.target);
+	return off_cpu_prepare(rec->evlist, &rec->opts.target, &rec->opts);
 }
 
 static bool record__kcore_readable(struct machine *machine)

tools/perf/util/bpf_off_cpu.c

@@ -5,10 +5,12 @@
 #include "util/evlist.h"
 #include "util/off_cpu.h"
 #include "util/perf-hooks.h"
+#include "util/record.h"
 #include "util/session.h"
 #include "util/target.h"
 #include "util/cpumap.h"
 #include "util/thread_map.h"
+#include "util/cgroup.h"
 #include <bpf/bpf.h>
 
 #include "bpf_skel/off_cpu.skel.h"
@@ -24,6 +26,7 @@ struct off_cpu_key {
 	u32 tgid;
 	u32 stack_id;
 	u32 state;
+	u64 cgroup_id;
 };
 
 union off_cpu_data {
@@ -116,10 +119,11 @@ static void check_sched_switch_args(void)
 	}
 }
 
-int off_cpu_prepare(struct evlist *evlist, struct target *target)
+int off_cpu_prepare(struct evlist *evlist, struct target *target,
+		    struct record_opts *opts)
 {
 	int err, fd, i;
-	int ncpus = 1, ntasks = 1;
+	int ncpus = 1, ntasks = 1, ncgrps = 1;
 
 	if (off_cpu_config(evlist) < 0) {
 		pr_err("Failed to config off-cpu BPF event\n");
@@ -143,6 +147,21 @@ int off_cpu_prepare(struct evlist *evlist, struct target *target)
 		bpf_map__set_max_entries(skel->maps.task_filter, ntasks);
 	}
 
+	if (evlist__first(evlist)->cgrp) {
+		ncgrps = evlist->core.nr_entries - 1; /* excluding a dummy */
+		bpf_map__set_max_entries(skel->maps.cgroup_filter, ncgrps);
+
+		if (!cgroup_is_v2("perf_event"))
+			skel->rodata->uses_cgroup_v1 = true;
+	}
+
+	if (opts->record_cgroup) {
+		skel->rodata->needs_cgroup = true;
+
+		if (!cgroup_is_v2("perf_event"))
+			skel->rodata->uses_cgroup_v1 = true;
+	}
+
 	set_max_rlimit();
 	check_sched_switch_args();
 
@@ -178,6 +197,29 @@ int off_cpu_prepare(struct evlist *evlist, struct target *target)
 		}
 	}
 
+	if (evlist__first(evlist)->cgrp) {
+		struct evsel *evsel;
+		u8 val = 1;
+
+		skel->bss->has_cgroup = 1;
+		fd = bpf_map__fd(skel->maps.cgroup_filter);
+
+		evlist__for_each_entry(evlist, evsel) {
+			struct cgroup *cgrp = evsel->cgrp;
+
+			if (cgrp == NULL)
+				continue;
+
+			if (!cgrp->id && read_cgroup_id(cgrp) < 0) {
+				pr_err("Failed to read cgroup id of %s\n",
+				       cgrp->name);
+				goto out;
+			}
+
+			bpf_map_update_elem(fd, &cgrp->id, &val, BPF_ANY);
+		}
+	}
+
 	err = off_cpu_bpf__attach(skel);
 	if (err) {
 		pr_err("Failed to attach off-cpu BPF skeleton\n");
@@ -275,6 +317,8 @@ int off_cpu_write(struct perf_session *session)
 			/* calculate sample callchain data array length */
 			n += len + 2;
 		}
+		if (sample_type & PERF_SAMPLE_CGROUP)
+			data.array[n++] = key.cgroup_id;
 		/* TODO: handle more sample types */
 
 		size = n * sizeof(u64);

tools/perf/util/bpf_skel/off_cpu.bpf.c

@@ -26,6 +26,7 @@ struct offcpu_key {
 	__u32 tgid;
 	__u32 stack_id;
 	__u32 state;
+	__u64 cgroup_id;
 };
 
 struct {
@@ -63,6 +64,13 @@ struct {
 	__uint(max_entries, 1);
 } task_filter SEC(".maps");
 
+struct {
+	__uint(type, BPF_MAP_TYPE_HASH);
+	__uint(key_size, sizeof(__u64));
+	__uint(value_size, sizeof(__u8));
+	__uint(max_entries, 1);
+} cgroup_filter SEC(".maps");
+
 /* old kernel task_struct definition */
 struct task_struct___old {
 	long state;
@@ -71,8 +79,11 @@ struct task_struct___old {
 int enabled = 0;
 int has_cpu = 0;
 int has_task = 0;
+int has_cgroup = 0;
 
 const volatile bool has_prev_state = false;
+const volatile bool needs_cgroup = false;
+const volatile bool uses_cgroup_v1 = false;
 
 /*
  * Old kernel used to call it task_struct->state and now it's '__state'.
@@ -92,6 +103,18 @@ static inline int get_task_state(struct task_struct *t)
 	return BPF_CORE_READ(t_old, state);
 }
 
+static inline __u64 get_cgroup_id(struct task_struct *t)
+{
+	struct cgroup *cgrp;
+
+	if (uses_cgroup_v1)
+		cgrp = BPF_CORE_READ(t, cgroups, subsys[perf_event_cgrp_id], cgroup);
+	else
+		cgrp = BPF_CORE_READ(t, cgroups, dfl_cgrp);
+
+	return BPF_CORE_READ(cgrp, kn, id);
+}
+
 static inline int can_record(struct task_struct *t, int state)
 {
 	/* kernel threads don't have user stack */
@@ -120,6 +143,15 @@ static inline int can_record(struct task_struct *t, int state)
 			return 0;
 	}
 
+	if (has_cgroup) {
+		__u8 *ok;
+		__u64 cgrp_id = get_cgroup_id(t);
+
+		ok = bpf_map_lookup_elem(&cgroup_filter, &cgrp_id);
+		if (!ok)
+			return 0;
+	}
+
 	return 1;
 }
 
@@ -156,6 +188,7 @@ static int off_cpu_stat(u64 *ctx, struct task_struct *prev,
 			.tgid = next->tgid,
 			.stack_id = pelem->stack_id,
 			.state = pelem->state,
+			.cgroup_id = needs_cgroup ? get_cgroup_id(next) : 0,
 		};
 		__u64 delta = ts - pelem->timestamp;
 		__u64 *total;

tools/perf/util/off_cpu.h

@@ -4,15 +4,18 @@
 struct evlist;
 struct target;
 struct perf_session;
+struct record_opts;
 
 #define OFFCPU_EVENT  "offcpu-time"
 
 #ifdef HAVE_BPF_SKEL
-int off_cpu_prepare(struct evlist *evlist, struct target *target);
+int off_cpu_prepare(struct evlist *evlist, struct target *target,
+		    struct record_opts *opts);
 int off_cpu_write(struct perf_session *session);
 #else
 static inline int off_cpu_prepare(struct evlist *evlist __maybe_unused,
-				  struct target *target __maybe_unused)
+				  struct target *target __maybe_unused,
+				  struct record_opts *opts __maybe_unused)
 {
 	return -1;
 }