[PATCH] perf utilities: cln_size header

[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@proton.me>
Signed-off-by: Ricky Ringler <ricky.ringler@proton.me>
---
 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

Re: [PATCH] perf utilities: cln_size header

[Namhyung Kim]

Hello,

On Sat, Feb 14, 2026 at 04:07:25AM +0000, Ricky Ringler wrote:
> 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

Sorry for the delay.  Can you please rebase this on top of the current
perf-tools-next?  We added a couple more header features recently.

Thanks,
Namhyung

> 
> Tested-by: Ricky Ringler <ricky.ringler@proton.me>
> Signed-off-by: Ricky Ringler <ricky.ringler@proton.me>
> ---
>  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
> 
> 

[PATCH v2] 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@proton.me>
Signed-off-by: Ricky Ringler <ricky.ringler@proton.me>
---
 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

Re: [PATCH v2] perf utilities: cln_size header

[Ricky Ringler]

Apologies all. I just realized I rebased on the kernel instead of perf-tools-next. Please hold off reviewing. I will submit a v3 patch rebased on perf-tools-next.

On Thursday, March 5th, 2026 at 5:57 PM, Ricky Ringler <ricky.ringler@proton.me> wrote:

> 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@proton.me>
> Signed-off-by: Ricky Ringler <ricky.ringler@proton.me>
> ---
>  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
> 
>

[PATCH v3] 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@proton.me>
Signed-off-by: Ricky Ringler <ricky.ringler@proton.me>
---
 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 5b29f4296861..2675d32f88cf 100644
--- a/tools/perf/builtin-inject.c
+++ b/tools/perf/builtin-inject.c
@@ -2134,6 +2134,7 @@ static bool keep_feat(struct perf_inject *inject, int feat)
 	case HEADER_HYBRID_TOPOLOGY:
 	case HEADER_PMU_CAPS:
 	case HEADER_CPU_DOMAIN_INFO:
+	case HEADER_CLNq_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 a4501cbca375..c7052ac1f856 100644
--- a/tools/perf/util/env.h
+++ b/tools/perf/util/env.h
@@ -112,6 +112,7 @@ struct perf_env {
 	struct cpu_cache_level	*caches;
 	struct cpu_domain_map	**cpu_domain;
 	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 9142a8ba4019..4f65faafe75d 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>
@@ -1304,6 +1305,18 @@ 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)
 {
@@ -2261,6 +2274,11 @@ 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",
@@ -3154,6 +3172,16 @@ 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;
@@ -3763,6 +3791,7 @@ const struct perf_header_feature_ops feat_ops[HEADER_LAST_FEATURE] = {
 	FEAT_OPR(PMU_CAPS,	pmu_caps,	false),
 	FEAT_OPR(CPU_DOMAIN_INFO,	cpu_domain_info,	true),
 	FEAT_OPR(E_MACHINE,	e_machine,	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 cc40ac796f52..be315040727f 100644
--- a/tools/perf/util/header.h
+++ b/tools/perf/util/header.h
@@ -55,6 +55,7 @@ enum {
 	HEADER_PMU_CAPS,
 	HEADER_CPU_DOMAIN_INFO,
 	HEADER_E_MACHINE,
+	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 42d5cd7ef4e2..13287ade784e 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

Re: [PATCH v3] perf utilities: cln_size header

[Ian Rogers]

On Sun, Mar 8, 2026 at 10:21 AM Ricky Ringler <ricky.ringler@proton.me> wrote:
>
> 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@proton.me>
> Signed-off-by: Ricky Ringler <ricky.ringler@proton.me>
> ---
>  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 5b29f4296861..2675d32f88cf 100644
> --- a/tools/perf/builtin-inject.c
> +++ b/tools/perf/builtin-inject.c
> @@ -2134,6 +2134,7 @@ static bool keep_feat(struct perf_inject *inject, int feat)
>         case HEADER_HYBRID_TOPOLOGY:
>         case HEADER_PMU_CAPS:
>         case HEADER_CPU_DOMAIN_INFO:
> +       case HEADER_CLNq_SIZE:

Looks like a typo, s/HEADER_CLNq_SIZE/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 a4501cbca375..c7052ac1f856 100644
> --- a/tools/perf/util/env.h
> +++ b/tools/perf/util/env.h
> @@ -112,6 +112,7 @@ struct perf_env {
>         struct cpu_cache_level  *caches;
>         struct cpu_domain_map   **cpu_domain;
>         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 9142a8ba4019..4f65faafe75d 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>
> @@ -1304,6 +1305,18 @@ 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;

Did you mean to assign "cln_size = DEFAULT_CACHELINE_SIZE" here? It
seems strange to something with value 0 the value 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)
>  {
> @@ -2261,6 +2274,11 @@ 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)

No need for the __maybe_unused on fp here.

> +{
> +       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",
> @@ -3154,6 +3172,16 @@ 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;
> @@ -3763,6 +3791,7 @@ const struct perf_header_feature_ops feat_ops[HEADER_LAST_FEATURE] = {
>         FEAT_OPR(PMU_CAPS,      pmu_caps,       false),
>         FEAT_OPR(CPU_DOMAIN_INFO,       cpu_domain_info,        true),
>         FEAT_OPR(E_MACHINE,     e_machine,      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 cc40ac796f52..be315040727f 100644
> --- a/tools/perf/util/header.h
> +++ b/tools/perf/util/header.h
> @@ -55,6 +55,7 @@ enum {
>         HEADER_PMU_CAPS,
>         HEADER_CPU_DOMAIN_INFO,
>         HEADER_E_MACHINE,
> +       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 42d5cd7ef4e2..13287ade784e 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;

I think on these 2 lines, prefer evsel__session(evsel) rather than the
direct access.

Thanks,
Ian

> +       }
> +
> +       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
>
>
>

[PATCH v4] 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.

V4: Ian feedback
V3: Rebase off perf-tools-next round two
V2: Rebase off perf-tools-next

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@proton.me>
Signed-off-by: Ricky Ringler <ricky.ringler@proton.me>
---
 tools/perf/builtin-inject.c |  2 +-
 tools/perf/util/header.c    | 11 ++++++++---
 tools/perf/util/sort.c      |  8 ++++++--
 3 files changed, 15 insertions(+), 6 deletions(-)

diff --git a/tools/perf/builtin-inject.c b/tools/perf/builtin-inject.c
index 2675d32f88cf..11ac7c8c4be3 100644
--- a/tools/perf/builtin-inject.c
+++ b/tools/perf/builtin-inject.c
@@ -2134,7 +2134,7 @@ static bool keep_feat(struct perf_inject *inject, int feat)
 	case HEADER_HYBRID_TOPOLOGY:
 	case HEADER_PMU_CAPS:
 	case HEADER_CPU_DOMAIN_INFO:
-	case HEADER_CLNq_SIZE:
+	case HEADER_CLN_SIZE:
 		return true;
 	/* Information that can be updated */
 	case HEADER_BUILD_ID:
diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c
index 4f65faafe75d..2d39da470267 100644
--- a/tools/perf/util/header.c
+++ b/tools/perf/util/header.c
@@ -1310,8 +1310,13 @@ static int write_cln_size(struct feat_fd *ff,
 {
 	int cln_size = cacheline_size();
 
-	if (!cln_size)
-		cln_size = 0;
+
+	if (!cln_size) {
+		int default_cacheline_size = 64;
+
+		cln_size = default_cacheline_size;
+	}
+
 	ff->ph->env.cln_size = cln_size;
 
 	return do_write(ff, &cln_size, sizeof(cln_size));
@@ -2274,7 +2279,7 @@ 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)
+static void print_cln_size(struct feat_fd *ff, FILE *fp)
 {
 	fprintf(fp, "# cacheline size: %u\n", ff->ph->env.cln_size);
 }
diff --git a/tools/perf/util/sort.c b/tools/perf/util/sort.c
index 13287ade784e..5f617cf03d5d 100644
--- a/tools/perf/util/sort.c
+++ b/tools/perf/util/sort.c
@@ -2483,8 +2483,12 @@ hist_entry__cln_size(struct hist_entry *he)
 	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 (evsel) {
+			struct perf_session *session = evsel__session(evsel);
+
+			ret = session->header.env.cln_size;
+		}
 	}
 
 	if (!ret || ret < 1) {
-- 
2.53.0

Re: [PATCH v4] perf utilities: cln_size header

[Namhyung Kim]

Hello,

On Sat, Mar 21, 2026 at 08:41:55PM +0000, Ricky Ringler wrote:
> Store cacheline size during perf record in header, so
> that cacheline size can be used for other features, like
> sort.
> 
> V4: Ian feedback
> V3: Rebase off perf-tools-next round two
> V2: Rebase off perf-tools-next
> 
> Follow-up patch from message ID "aYZiQk6Uftzlb_JV@x1"

Please squash the patches into a single change and send it again.

> 
> Testing:
> - Built perf
> - Ran record + report with feat enabled
> - Ran record + report with feat disabled
> 
> Tested-by: Ricky Ringler <ricky.ringler@proton.me>

Tested-by tag from the author may not be meaningful.  We always expect
authors to test their patches. :)

> Signed-off-by: Ricky Ringler <ricky.ringler@proton.me>
> ---
>  tools/perf/builtin-inject.c |  2 +-
>  tools/perf/util/header.c    | 11 ++++++++---
>  tools/perf/util/sort.c      |  8 ++++++--
>  3 files changed, 15 insertions(+), 6 deletions(-)
> 
> diff --git a/tools/perf/builtin-inject.c b/tools/perf/builtin-inject.c
> index 2675d32f88cf..11ac7c8c4be3 100644
> --- a/tools/perf/builtin-inject.c
> +++ b/tools/perf/builtin-inject.c
> @@ -2134,7 +2134,7 @@ static bool keep_feat(struct perf_inject *inject, int feat)
>  	case HEADER_HYBRID_TOPOLOGY:
>  	case HEADER_PMU_CAPS:
>  	case HEADER_CPU_DOMAIN_INFO:
> -	case HEADER_CLNq_SIZE:
> +	case HEADER_CLN_SIZE:
>  		return true;
>  	/* Information that can be updated */
>  	case HEADER_BUILD_ID:
> diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c
> index 4f65faafe75d..2d39da470267 100644
> --- a/tools/perf/util/header.c
> +++ b/tools/perf/util/header.c
> @@ -1310,8 +1310,13 @@ static int write_cln_size(struct feat_fd *ff,
>  {
>  	int cln_size = cacheline_size();
>  
> -	if (!cln_size)
> -		cln_size = 0;
> +
> +	if (!cln_size) {
> +		int default_cacheline_size = 64;

Let's get rid of the local variable.

Thanks,
Namhyung

> +
> +		cln_size = default_cacheline_size;
> +	}
> +
>  	ff->ph->env.cln_size = cln_size;
>  
>  	return do_write(ff, &cln_size, sizeof(cln_size));
> @@ -2274,7 +2279,7 @@ 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)
> +static void print_cln_size(struct feat_fd *ff, FILE *fp)
>  {
>  	fprintf(fp, "# cacheline size: %u\n", ff->ph->env.cln_size);
>  }
> diff --git a/tools/perf/util/sort.c b/tools/perf/util/sort.c
> index 13287ade784e..5f617cf03d5d 100644
> --- a/tools/perf/util/sort.c
> +++ b/tools/perf/util/sort.c
> @@ -2483,8 +2483,12 @@ hist_entry__cln_size(struct hist_entry *he)
>  	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 (evsel) {
> +			struct perf_session *session = evsel__session(evsel);
> +
> +			ret = session->header.env.cln_size;
> +		}
>  	}
>  
>  	if (!ret || ret < 1) {
> -- 
> 2.53.0
> 
>