[PATCH] perf utilities: cln_size header

[Ricky Ringler]

Store cacheline size during perf record in header, so
that cacheline size can be used for other features, like
sort.

Follow-up patch from message ID "aYZiQk6Uftzlb_JV@x1"

 Testing:
- Built perf
- Ran record + report with feat enabled
- Ran record + report with feat disabled

Tested-by: Ricky Ringler <ricky.ringler@xxxxxxxxx>
Signed-off-by: Ricky Ringler <ricky.ringler@xxxxxxxxx>
---
 tools/perf/builtin-inject.c |  1 +
 tools/perf/util/env.h       |  1 +
 tools/perf/util/header.c    | 29 +++++++++++++++++++++++++++++
 tools/perf/util/header.h    |  1 +
 tools/perf/util/sort.c      | 37 ++++++++++++++++++++++++++-----------
 5 files changed, 58 insertions(+), 11 deletions(-)

diff --git a/tools/perf/builtin-inject.c b/tools/perf/builtin-inject.c
index aa7be4fb5838..9639154459d9 100644
--- a/tools/perf/builtin-inject.c
+++ b/tools/perf/builtin-inject.c
@@ -2047,6 +2047,7 @@ static bool keep_feat(int feat)
 	case HEADER_CLOCK_DATA:
 	case HEADER_HYBRID_TOPOLOGY:
 	case HEADER_PMU_CAPS:
+	case HEADER_CLN_SIZE:
 		return true;
 	/* Information that can be updated */
 	case HEADER_BUILD_ID:
diff --git a/tools/perf/util/env.h b/tools/perf/util/env.h
index 9977b85523a8..04580c64847b 100644
--- a/tools/perf/util/env.h
+++ b/tools/perf/util/env.h
@@ -93,6 +93,7 @@ struct perf_env {
 	struct cpu_topology_map	*cpu;
 	struct cpu_cache_level	*caches;
 	int			 caches_cnt;
+	unsigned int		cln_size;
 	u32			comp_ratio;
 	u32			comp_ver;
 	u32			comp_type;
diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c
index f5cad377c99e..ad15829acc69 100644
--- a/tools/perf/util/header.c
+++ b/tools/perf/util/header.c
@@ -54,6 +54,7 @@
 #include "bpf-event.h"
 #include "bpf-utils.h"
 #include "clockid.h"
+#include "cacheline.h"
 
 #include <linux/ctype.h>
 #include <internal/lib.h>
@@ -1288,6 +1289,18 @@ static int write_cache(struct feat_fd *ff,
 	return ret;
 }
 
+static int write_cln_size(struct feat_fd *ff,
+		       struct evlist *evlist __maybe_unused)
+{
+	int cln_size = cacheline_size();
+
+	if (!cln_size)
+		cln_size = 0;
+	ff->ph->env.cln_size = cln_size;
+
+	return do_write(ff, &cln_size, sizeof(cln_size));
+}
+
 static int write_stat(struct feat_fd *ff __maybe_unused,
 		      struct evlist *evlist __maybe_unused)
 {
@@ -2084,6 +2097,11 @@ static void print_cache(struct feat_fd *ff, FILE *fp __maybe_unused)
 	}
 }
 
+static void print_cln_size(struct feat_fd *ff, FILE *fp __maybe_unused)
+{
+	fprintf(fp, "# cacheline size: %u\n", ff->ph->env.cln_size);
+}
+
 static void print_compressed(struct feat_fd *ff, FILE *fp)
 {
 	fprintf(fp, "# compressed : %s, level = %d, ratio = %d\n",
@@ -2933,6 +2951,16 @@ static int process_cache(struct feat_fd *ff, void *data __maybe_unused)
 	return -1;
 }
 
+static int process_cln_size(struct feat_fd *ff, void *data __maybe_unused)
+{
+	struct perf_env *env = &ff->ph->env;
+
+	if (do_read_u32(ff, &env->cln_size))
+		return -1;
+
+	return 0;
+}
+
 static int process_sample_time(struct feat_fd *ff, void *data __maybe_unused)
 {
 	struct perf_session *session;
@@ -3453,6 +3481,7 @@ const struct perf_header_feature_ops feat_ops[HEADER_LAST_FEATURE] = {
 	FEAT_OPR(CLOCK_DATA,	clock_data,	false),
 	FEAT_OPN(HYBRID_TOPOLOGY,	hybrid_topology,	true),
 	FEAT_OPR(PMU_CAPS,	pmu_caps,	false),
+	FEAT_OPR(CLN_SIZE,	cln_size,	false),
 };
 
 struct header_print_data {
diff --git a/tools/perf/util/header.h b/tools/perf/util/header.h
index c058021c3150..04394331630f 100644
--- a/tools/perf/util/header.h
+++ b/tools/perf/util/header.h
@@ -53,6 +53,7 @@ enum {
 	HEADER_CLOCK_DATA,
 	HEADER_HYBRID_TOPOLOGY,
 	HEADER_PMU_CAPS,
+	HEADER_CLN_SIZE,
 	HEADER_LAST_FEATURE,
 	HEADER_FEAT_BITS	= 256,
 };
diff --git a/tools/perf/util/sort.c b/tools/perf/util/sort.c
index aa79eb6476dd..e636b9f88e5c 100644
--- a/tools/perf/util/sort.c
+++ b/tools/perf/util/sort.c
@@ -30,6 +30,7 @@
 #include "time-utils.h"
 #include "cgroup.h"
 #include "machine.h"
+#include "session.h"
 #include "trace-event.h"
 #include <linux/kernel.h>
 #include <linux/string.h>
@@ -2474,7 +2475,26 @@ struct sort_entry sort_type_offset = {
 
 /* --sort typecln */
 
-#define DEFAULT_CACHELINE_SIZE 64
+static int
+hist_entry__cln_size(struct hist_entry *he)
+{
+	int ret = 0;
+
+	if (he && he->hists) {
+		struct evsel *evsel = hists_to_evsel(he->hists);
+
+		if (evsel && evsel->evlist->session && evsel->evlist->session)
+			ret = evsel->evlist->session->header.env.cln_size;
+	}
+
+	if (!ret || ret < 1) {
+		int default_cacheline_size = 64; // avoid div/0 later
+
+		ret = default_cacheline_size;
+	}
+
+	return ret;
+}
 
 static int64_t
 sort__typecln_sort(struct hist_entry *left, struct hist_entry *right)
@@ -2482,11 +2502,9 @@ sort__typecln_sort(struct hist_entry *left, struct hist_entry *right)
 	struct annotated_data_type *left_type = left->mem_type;
 	struct annotated_data_type *right_type = right->mem_type;
 	int64_t left_cln, right_cln;
+	int64_t cln_size_left = hist_entry__cln_size(left);
+	int64_t cln_size_right = hist_entry__cln_size(right);
 	int64_t ret;
-	int cln_size = cacheline_size();
-
-	if (cln_size == 0)
-		cln_size = DEFAULT_CACHELINE_SIZE;
 
 	if (!left_type) {
 		sort__type_init(left);
@@ -2502,8 +2520,8 @@ sort__typecln_sort(struct hist_entry *left, struct hist_entry *right)
 	if (ret)
 		return ret;
 
-	left_cln = left->mem_type_off / cln_size;
-	right_cln = right->mem_type_off / cln_size;
+	left_cln = left->mem_type_off / cln_size_left;
+	right_cln = right->mem_type_off / cln_size_right;
 	return left_cln - right_cln;
 }
 
@@ -2511,10 +2529,7 @@ static int hist_entry__typecln_snprintf(struct hist_entry *he, char *bf,
 				     size_t size, unsigned int width __maybe_unused)
 {
 	struct annotated_data_type *he_type = he->mem_type;
-	int cln_size = cacheline_size();
-
-	if (cln_size == 0)
-		cln_size = DEFAULT_CACHELINE_SIZE;
+	int cln_size = hist_entry__cln_size(he);
 
 	return repsep_snprintf(bf, size, "%s: cache-line %d", he_type->self.type_name,
 			       he->mem_type_off / cln_size);
-- 
2.53.0