perf header: Support CPU DOMAIN relation info

The '/proc/schedstat' file gives info about load balancing statistics
within a given domain.

It also contains the cpu_mask giving information about the sibling cpus
and domain names after schedstat version 17.

Storing this information in perf header will help tools like `perf sched
stats` for better analysis.

Signed-off-by: Swapnil Sapkal <swapnil.sapkal@amd.com>
Tested-by: Chen Yu <yu.c.chen@intel.com>
Acked-by: Ian Rogers <irogers@google.com>
Acked-by: Namhyung Kim <namhyung@kernel.org>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Anubhav Shelat <ashelat@redhat.com>
Cc: Ben Gainey <ben.gainey@arm.com>
Cc: Blake Jones <blakejones@google.com>
Cc: Chun-Tse Shao <ctshao@google.com>
Cc: David Vernet <void@manifault.com>
Cc: Dmitriy Vyukov <dvyukov@google.com>
Cc: Dr. David Alan Gilbert <linux@treblig.org>
Cc: Gautham Shenoy <gautham.shenoy@amd.com>
Cc: Graham Woodward <graham.woodward@arm.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: James Clark <james.clark@arm.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Juri Lelli <juri.lelli@redhat.com>
Cc: K Prateek Nayak <kprateek.nayak@amd.com>
Cc: Kan Liang <kan.liang@linux.intel.com>
Cc: Leo Yan <leo.yan@arm.com>
Cc: Madadi Vineeth Reddy <vineethr@linux.ibm.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Ravi Bangoria <ravi.bangoria@amd.com>
Cc: Sandipan Das <sandipan.das@amd.com>
Cc: Santosh Shukla <santosh.shukla@amd.com>
Cc: Shrikanth Hegde <sshegde@linux.ibm.com>
Cc: Steven Rostedt (VMware) <rostedt@goodmis.org>
Cc: Tejun Heo <tj@kernel.org>
Cc: Thomas Falcon <thomas.falcon@intel.com>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Vincent Guittot <vincent.guittot@linaro.org>
Cc: Yang Jihong <yangjihong@bytedance.com>
Cc: Yujie Liu <yujie.liu@intel.com>
Cc: Zhongqiu Han <quic_zhonhan@quicinc.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

tools/perf/Documentation/perf.data-file-format.txt

@@ -447,6 +447,23 @@ struct {
 	} [nr_pmu];
 };
 
+	HEADER_CPU_DOMAIN_INFO = 32,
+
+List of cpu-domain relation info. The format of the data is as below.
+
+struct domain_info {
+	int domain;
+	char dname[];
+	char cpumask[];
+	char cpulist[];
+};
+
+struct cpu_domain_info {
+	int cpu;
+	int nr_domains;
+	struct domain_info domains[];
+};
+
 	other bits are reserved and should ignored for now
 	HEADER_FEAT_BITS	= 256,
 

tools/perf/builtin-inject.c

@@ -2133,6 +2133,7 @@ static bool keep_feat(struct perf_inject *inject, int feat)
 	case HEADER_CLOCK_DATA:
 	case HEADER_HYBRID_TOPOLOGY:
 	case HEADER_PMU_CAPS:
+	case HEADER_CPU_DOMAIN_INFO:
 		return true;
 	/* Information that can be updated */
 	case HEADER_BUILD_ID:

tools/perf/util/env.c

@@ -216,6 +216,34 @@ static void perf_env__purge_bpf(struct perf_env *env __maybe_unused)
 }
 #endif // HAVE_LIBBPF_SUPPORT
 
+void free_cpu_domain_info(struct cpu_domain_map **cd_map, u32 schedstat_version, u32 nr)
+{
+	if (!cd_map)
+		return;
+
+	for (u32 i = 0; i < nr; i++) {
+		if (!cd_map[i])
+			continue;
+
+		for (u32 j = 0; j < cd_map[i]->nr_domains; j++) {
+			struct domain_info *d_info = cd_map[i]->domains[j];
+
+			if (!d_info)
+				continue;
+
+			if (schedstat_version >= 17)
+				zfree(&d_info->dname);
+
+			zfree(&d_info->cpumask);
+			zfree(&d_info->cpulist);
+			zfree(&d_info);
+		}
+		zfree(&cd_map[i]->domains);
+		zfree(&cd_map[i]);
+	}
+	zfree(&cd_map);
+}
+
 void perf_env__exit(struct perf_env *env)
 {
 	int i, j;
@@ -265,6 +293,7 @@ void perf_env__exit(struct perf_env *env)
 		zfree(&env->pmu_caps[i].pmu_name);
 	}
 	zfree(&env->pmu_caps);
+	free_cpu_domain_info(env->cpu_domain, env->schedstat_version, env->nr_cpus_avail);
 }
 
 void perf_env__init(struct perf_env *env)

tools/perf/util/env.h

@@ -54,6 +54,19 @@ struct pmu_caps {
 	char            *pmu_name;
 };
 
+struct domain_info {
+	u32	domain;
+	char	*dname;
+	char	*cpumask;
+	char	*cpulist;
+};
+
+struct cpu_domain_map {
+	u32			cpu;
+	u32			nr_domains;
+	struct domain_info	**domains;
+};
+
 typedef const char *(arch_syscalls__strerrno_t)(int err);
 
 struct perf_env {
@@ -70,6 +83,8 @@ struct perf_env {
 	unsigned int		max_branches;
 	unsigned int		br_cntr_nr;
 	unsigned int		br_cntr_width;
+	unsigned int		schedstat_version;
+	unsigned int		max_sched_domains;
 	int			kernel_is_64_bit;
 
 	int			nr_cmdline;
@@ -92,6 +107,7 @@ struct perf_env {
 	char			**cpu_pmu_caps;
 	struct cpu_topology_map	*cpu;
 	struct cpu_cache_level	*caches;
+	struct cpu_domain_map	**cpu_domain;
 	int			 caches_cnt;
 	u32			comp_ratio;
 	u32			comp_ver;
@@ -151,6 +167,7 @@ struct bpf_prog_info_node;
 struct btf_node;
 
 int perf_env__read_core_pmu_caps(struct perf_env *env);
+void free_cpu_domain_info(struct cpu_domain_map **cd_map, u32 schedstat_version, u32 nr);
 void perf_env__exit(struct perf_env *env);
 
 int perf_env__kernel_is_64_bit(struct perf_env *env);

tools/perf/util/header.c

@@ -1614,6 +1614,162 @@ static int write_pmu_caps(struct feat_fd *ff,
 	return 0;
 }
 
+static struct cpu_domain_map **build_cpu_domain_map(u32 *schedstat_version, u32 *max_sched_domains,
+						    u32 nr)
+{
+	struct domain_info *domain_info;
+	struct cpu_domain_map **cd_map;
+	char dname[16], cpumask[256];
+	char cpulist[1024];
+	char *line = NULL;
+	u32 cpu, domain;
+	u32 dcount = 0;
+	size_t len;
+	FILE *fp;
+
+	fp = fopen("/proc/schedstat", "r");
+	if (!fp) {
+		pr_err("Failed to open /proc/schedstat\n");
+		return NULL;
+	}
+
+	cd_map = zalloc(sizeof(*cd_map) * nr);
+	if (!cd_map)
+		goto out;
+
+	while (getline(&line, &len, fp) > 0) {
+		int retval;
+
+		if (strncmp(line, "version", 7) == 0) {
+			retval = sscanf(line, "version %d\n", schedstat_version);
+			if (retval != 1)
+				continue;
+
+		} else if (strncmp(line, "cpu", 3) == 0) {
+			retval = sscanf(line, "cpu%u %*s", &cpu);
+			if (retval == 1) {
+				cd_map[cpu] = zalloc(sizeof(*cd_map[cpu]));
+				if (!cd_map[cpu])
+					goto out_free_line;
+				cd_map[cpu]->cpu = cpu;
+			} else
+				continue;
+
+			dcount = 0;
+		} else if (strncmp(line, "domain", 6) == 0) {
+			struct domain_info **temp_domains;
+
+			dcount++;
+			temp_domains = realloc(cd_map[cpu]->domains, dcount * sizeof(domain_info));
+			if (!temp_domains)
+				goto out_free_line;
+			else
+				cd_map[cpu]->domains = temp_domains;
+
+			domain_info = zalloc(sizeof(*domain_info));
+			if (!domain_info)
+				goto out_free_line;
+
+			cd_map[cpu]->domains[dcount - 1] = domain_info;
+
+			if (*schedstat_version >= 17) {
+				retval = sscanf(line, "domain%u %s %s %*s", &domain, dname,
+						cpumask);
+				if (retval != 3)
+					continue;
+
+				domain_info->dname = strdup(dname);
+				if (!domain_info->dname)
+					goto out_free_line;
+			} else {
+				retval = sscanf(line, "domain%u %s %*s", &domain, cpumask);
+				if (retval != 2)
+					continue;
+			}
+
+			domain_info->domain = domain;
+			if (domain > *max_sched_domains)
+				*max_sched_domains = domain;
+
+			domain_info->cpumask = strdup(cpumask);
+			if (!domain_info->cpumask)
+				goto out_free_line;
+
+			cpumask_to_cpulist(cpumask, cpulist);
+			domain_info->cpulist = strdup(cpulist);
+			if (!domain_info->cpulist)
+				goto out_free_line;
+
+			cd_map[cpu]->nr_domains = dcount;
+		}
+	}
+
+out_free_line:
+	free(line);
+out:
+	fclose(fp);
+	return cd_map;
+}
+
+static int write_cpu_domain_info(struct feat_fd *ff,
+				 struct evlist *evlist __maybe_unused)
+{
+	u32 max_sched_domains = 0, schedstat_version = 0;
+	struct cpu_domain_map **cd_map;
+	u32 i, j, nr, ret;
+
+	nr = cpu__max_present_cpu().cpu;
+
+	cd_map = build_cpu_domain_map(&schedstat_version, &max_sched_domains, nr);
+	if (!cd_map)
+		return -1;
+
+	ret = do_write(ff, &schedstat_version, sizeof(u32));
+	if (ret < 0)
+		goto out;
+
+	max_sched_domains += 1;
+	ret = do_write(ff, &max_sched_domains, sizeof(u32));
+	if (ret < 0)
+		goto out;
+
+	for (i = 0; i < nr; i++) {
+		if (!cd_map[i])
+			continue;
+
+		ret = do_write(ff, &cd_map[i]->cpu, sizeof(u32));
+		if (ret < 0)
+			goto out;
+
+		ret = do_write(ff, &cd_map[i]->nr_domains, sizeof(u32));
+		if (ret < 0)
+			goto out;
+
+		for (j = 0; j < cd_map[i]->nr_domains; j++) {
+			ret = do_write(ff, &cd_map[i]->domains[j]->domain, sizeof(u32));
+			if (ret < 0)
+				goto out;
+			if (schedstat_version >= 17) {
+				ret = do_write_string(ff, cd_map[i]->domains[j]->dname);
+				if (ret < 0)
+					goto out;
+			}
+
+			ret = do_write_string(ff, cd_map[i]->domains[j]->cpumask);
+			if (ret < 0)
+				goto out;
+
+			ret = do_write_string(ff, cd_map[i]->domains[j]->cpulist);
+			if (ret < 0)
+				goto out;
+		}
+	}
+
+out:
+	free_cpu_domain_info(cd_map, schedstat_version, nr);
+	return ret;
+}
+
 static void print_hostname(struct feat_fd *ff, FILE *fp)
 {
 	fprintf(fp, "# hostname : %s\n", ff->ph->env.hostname);
@@ -2247,6 +2403,39 @@ static void print_mem_topology(struct feat_fd *ff, FILE *fp)
 	}
 }
 
+static void print_cpu_domain_info(struct feat_fd *ff, FILE *fp)
+{
+	struct cpu_domain_map **cd_map = ff->ph->env.cpu_domain;
+	u32 nr = ff->ph->env.nr_cpus_avail;
+	struct domain_info *d_info;
+	u32 i, j;
+
+	fprintf(fp, "# schedstat version	: %u\n", ff->ph->env.schedstat_version);
+	fprintf(fp, "# Maximum sched domains	: %u\n", ff->ph->env.max_sched_domains);
+
+	for (i = 0; i < nr; i++) {
+		if (!cd_map[i])
+			continue;
+
+		fprintf(fp, "# cpu		: %u\n", cd_map[i]->cpu);
+		fprintf(fp, "# nr_domains	: %u\n", cd_map[i]->nr_domains);
+
+		for (j = 0; j < cd_map[i]->nr_domains; j++) {
+			d_info = cd_map[i]->domains[j];
+			if (!d_info)
+				continue;
+
+			fprintf(fp, "# Domain		: %u\n", d_info->domain);
+
+			if (ff->ph->env.schedstat_version >= 17)
+				fprintf(fp, "# Domain name      : %s\n", d_info->dname);
+
+			fprintf(fp, "# Domain cpu map   : %s\n", d_info->cpumask);
+			fprintf(fp, "# Domain cpu list  : %s\n", d_info->cpulist);
+		}
+	}
+}
+
 static int __event_process_build_id(struct perf_record_header_build_id *bev,
 				    char *filename,
 				    struct perf_session *session)
@@ -3388,6 +3577,102 @@ err:
 	return ret;
 }
 
+static int process_cpu_domain_info(struct feat_fd *ff, void *data __maybe_unused)
+{
+	u32 schedstat_version, max_sched_domains, cpu, domain, nr_domains;
+	struct perf_env *env = &ff->ph->env;
+	char *dname, *cpumask, *cpulist;
+	struct cpu_domain_map **cd_map;
+	struct domain_info *d_info;
+	u32 nra, nr, i, j;
+	int ret;
+
+	nra = env->nr_cpus_avail;
+	nr = env->nr_cpus_online;
+
+	cd_map = zalloc(sizeof(*cd_map) * nra);
+	if (!cd_map)
+		return -1;
+
+	env->cpu_domain = cd_map;
+
+	ret = do_read_u32(ff, &schedstat_version);
+	if (ret)
+		return ret;
+
+	env->schedstat_version = schedstat_version;
+
+	ret = do_read_u32(ff, &max_sched_domains);
+	if (ret)
+		return ret;
+
+	env->max_sched_domains = max_sched_domains;
+
+	for (i = 0; i < nr; i++) {
+		if (do_read_u32(ff, &cpu))
+			return -1;
+
+		cd_map[cpu] = zalloc(sizeof(*cd_map[cpu]));
+		if (!cd_map[cpu])
+			return -1;
+
+		cd_map[cpu]->cpu = cpu;
+
+		if (do_read_u32(ff, &nr_domains))
+			return -1;
+
+		cd_map[cpu]->nr_domains = nr_domains;
+
+		cd_map[cpu]->domains = zalloc(sizeof(*d_info) * max_sched_domains);
+		if (!cd_map[cpu]->domains)
+			return -1;
+
+		for (j = 0; j < nr_domains; j++) {
+			if (do_read_u32(ff, &domain))
+				return -1;
+
+			d_info = zalloc(sizeof(*d_info));
+			if (!d_info)
+				return -1;
+
+			cd_map[cpu]->domains[domain] = d_info;
+			d_info->domain = domain;
+
+			if (schedstat_version >= 17) {
+				dname = do_read_string(ff);
+				if (!dname)
+					return -1;
+
+				d_info->dname = zalloc(strlen(dname) + 1);
+				if (!d_info->dname)
+					return -1;
+
+				d_info->dname = strdup(dname);
+			}
+
+			cpumask = do_read_string(ff);
+			if (!cpumask)
+				return -1;
+
+			d_info->cpumask = zalloc(strlen(cpumask) + 1);
+			if (!d_info->cpumask)
+				return -1;
+			d_info->cpumask = strdup(cpumask);
+
+			cpulist = do_read_string(ff);
+			if (!cpulist)
+				return -1;
+
+			d_info->cpulist = zalloc(strlen(cpulist) + 1);
+			if (!d_info->cpulist)
+				return -1;
+			d_info->cpulist = strdup(cpulist);
+		}
+	}
+
+	return ret;
+}
+
 #define FEAT_OPR(n, func, __full_only) \
 	[HEADER_##n] = {					\
 		.name	    = __stringify(n),			\
@@ -3453,6 +3738,7 @@ 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(CPU_DOMAIN_INFO,	cpu_domain_info,	true),
 };
 
 struct header_print_data {

tools/perf/util/header.h

@@ -53,6 +53,7 @@ enum {
 	HEADER_CLOCK_DATA,
 	HEADER_HYBRID_TOPOLOGY,
 	HEADER_PMU_CAPS,
+	HEADER_CPU_DOMAIN_INFO,
 	HEADER_LAST_FEATURE,
 	HEADER_FEAT_BITS	= 256,
 };

tools/perf/util/util.c

@@ -257,6 +257,48 @@ static int rm_rf_kcore_dir(const char *path)
 	return 0;
 }
 
+void cpumask_to_cpulist(char *cpumask, char *cpulist)
+{
+	int i, j, bm_size, nbits;
+	int len = strlen(cpumask);
+	unsigned long *bm;
+	char cpus[1024];
+
+	for (i = 0; i < len; i++) {
+		if (cpumask[i] == ',') {
+			for (j = i; j < len; j++)
+				cpumask[j] = cpumask[j + 1];
+		}
+	}
+
+	len = strlen(cpumask);
+	bm_size = (len + 15) / 16;
+	nbits = bm_size * 64;
+	if (nbits <= 0)
+		return;
+
+	bm = calloc(bm_size, sizeof(unsigned long));
+	if (!cpumask)
+		goto free_bm;
+
+	for (i = 0; i < bm_size; i++) {
+		char blk[17];
+		int blklen = len > 16 ? 16 : len;
+
+		strncpy(blk, cpumask + len - blklen, blklen);
+		blk[blklen] = '\0';
+		bm[i] = strtoul(blk, NULL, 16);
+		cpumask[len - blklen] = '\0';
+		len = strlen(cpumask);
+	}
+
+	bitmap_scnprintf(bm, nbits, cpus, sizeof(cpus));
+	strcpy(cpulist, cpus);
+
+free_bm:
+	free(bm);
+}
+
 int rm_rf_perf_data(const char *path)
 {
 	const char *pat[] = {

tools/perf/util/util.h

@@ -11,6 +11,7 @@
 #include <stdbool.h>
 #include <stddef.h>
 #include <linux/compiler.h>
+#include <linux/bitmap.h>
 #include <sys/types.h>
 #ifndef __cplusplus
 #include <internal/cpumap.h>
@@ -48,6 +49,8 @@ bool sysctl__nmi_watchdog_enabled(void);
 
 int perf_tip(char **strp, const char *dirpath);
 
+void cpumask_to_cpulist(char *cpumask, char *cpulist);
+
 #ifndef HAVE_SCHED_GETCPU_SUPPORT
 int sched_getcpu(void);
 #endif