tools/power turbostat: Add RAPL psys as a built-in counter

Introduce the counter as a part of global, platform counters structure.
We open the counter for only one cpu, but otherwise treat it as an
ordinary RAPL counter, allowing for grouped perf read.

The counter is disabled by default, because it's interpretation may
require additional, platform specific information, making it unsuitable
for general use.

Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>

Author: Patryk Wlazlyn

tools/power/x86/turbostat/turbostat.8

@@ -190,6 +190,8 @@ The system configuration dump (if --quiet is not used) is followed by statistics
 .PP
 \fBRAMWatt\fP Watts consumed by the DRAM DIMMS -- available only on server processors.
 .PP
+\fBSysWatt\fP Watts consumed by the whole platform (RAPL PSYS). Disabled by default. May require platform specific information to interpret the data, making it not suitable for general use.
+.PP
 \fBPKG_%\fP percent of the interval that RAPL throttling was active on the Package.  Note that the system summary is the sum of the package throttling time, and thus may be higher than 100% on a multi-package system.  Note that the meaning of this field is model specific.  For example, some hardware increments this counter when RAPL responds to thermal limits, but does not increment this counter when RAPL responds to power limits.  Comparing PkgWatt and PkgTmp to system limits is necessary.
 .PP
 \fBRAM_%\fP percent of the interval that RAPL throttling was active on DRAM.

tools/power/x86/turbostat/turbostat.c

@@ -200,6 +200,8 @@ struct msr_counter bic[] = {
 	{ 0x0, "SAMMHz", NULL, 0, 0, 0, NULL, 0 },
 	{ 0x0, "SAMAMHz", NULL, 0, 0, 0, NULL, 0 },
 	{ 0x0, "Die%c6", NULL, 0, 0, 0, NULL, 0 },
+	{ 0x0, "SysWatt", NULL, 0, 0, 0, NULL, 0 },
+	{ 0x0, "Sys_J", NULL, 0, 0, 0, NULL, 0 },
 };
 
 #define MAX_BIC (sizeof(bic) / sizeof(struct msr_counter))
@@ -262,14 +264,16 @@ struct msr_counter bic[] = {
 #define	BIC_SAMMHz		(1ULL << 56)
 #define	BIC_SAMACTMHz		(1ULL << 57)
 #define	BIC_Diec6		(1ULL << 58)
+#define	BIC_SysWatt		(1ULL << 59)
+#define	BIC_Sys_J		(1ULL << 60)
 
 #define BIC_TOPOLOGY (BIC_Package | BIC_Node | BIC_CoreCnt | BIC_PkgCnt | BIC_Core | BIC_CPU | BIC_Die )
 #define BIC_THERMAL_PWR ( BIC_CoreTmp | BIC_PkgTmp | BIC_PkgWatt | BIC_CorWatt | BIC_GFXWatt | BIC_RAMWatt | BIC_PKG__ | BIC_RAM__)
 #define BIC_FREQUENCY (BIC_Avg_MHz | BIC_Busy | BIC_Bzy_MHz | BIC_TSC_MHz | BIC_GFXMHz | BIC_GFXACTMHz | BIC_SAMMHz | BIC_SAMACTMHz | BIC_UNCORE_MHZ)
 #define BIC_IDLE (BIC_sysfs | BIC_CPU_c1 | BIC_CPU_c3 | BIC_CPU_c6 | BIC_CPU_c7 | BIC_GFX_rc6 | BIC_Pkgpc2 | BIC_Pkgpc3 | BIC_Pkgpc6 | BIC_Pkgpc7 | BIC_Pkgpc8 | BIC_Pkgpc9 | BIC_Pkgpc10 | BIC_CPU_LPI | BIC_SYS_LPI | BIC_Mod_c6 | BIC_Totl_c0 | BIC_Any_c0 | BIC_GFX_c0 | BIC_CPUGFX | BIC_SAM_mc6 | BIC_Diec6)
 #define BIC_OTHER ( BIC_IRQ | BIC_SMI | BIC_ThreadC | BIC_CoreTmp | BIC_IPC)
 
-#define BIC_DISABLED_BY_DEFAULT	(BIC_USEC | BIC_TOD | BIC_APIC | BIC_X2APIC)
+#define BIC_DISABLED_BY_DEFAULT	(BIC_USEC | BIC_TOD | BIC_APIC | BIC_X2APIC | BIC_SysWatt | BIC_Sys_J)
 
 unsigned long long bic_enabled = (0xFFFFFFFFFFFFFFFFULL & ~BIC_DISABLED_BY_DEFAULT);
 unsigned long long bic_present = BIC_USEC | BIC_TOD | BIC_sysfs | BIC_APIC | BIC_X2APIC;
@@ -507,12 +511,15 @@ enum rapl_msrs {
 	RAPL_AMD_PWR_UNIT = BIT(14),	/* 0xc0010299 MSR_AMD_RAPL_POWER_UNIT */
 	RAPL_AMD_CORE_ENERGY_STAT = BIT(15),	/* 0xc001029a MSR_AMD_CORE_ENERGY_STATUS */
 	RAPL_AMD_PKG_ENERGY_STAT = BIT(16),	/* 0xc001029b MSR_AMD_PKG_ENERGY_STATUS */
+	RAPL_PLATFORM_ENERGY_LIMIT = BIT(17),	/* 0x64c MSR_PLATFORM_ENERGY_LIMIT */
+	RAPL_PLATFORM_ENERGY_STATUS = BIT(18),	/* 0x64d MSR_PLATFORM_ENERGY_STATUS */
 };
 
 #define RAPL_PKG	(RAPL_PKG_ENERGY_STATUS | RAPL_PKG_POWER_LIMIT)
 #define RAPL_DRAM	(RAPL_DRAM_ENERGY_STATUS | RAPL_DRAM_POWER_LIMIT)
 #define RAPL_CORE	(RAPL_CORE_ENERGY_STATUS | RAPL_CORE_POWER_LIMIT)
 #define RAPL_GFX	(RAPL_GFX_POWER_LIMIT | RAPL_GFX_ENERGY_STATUS)
+#define RAPL_PSYS	(RAPL_PLATFORM_ENERGY_STATUS | RAPL_PLATFORM_ENERGY_LIMIT)
 
 #define RAPL_PKG_ALL	(RAPL_PKG | RAPL_PKG_PERF_STATUS | RAPL_PKG_POWER_INFO)
 #define RAPL_DRAM_ALL	(RAPL_DRAM | RAPL_DRAM_PERF_STATUS | RAPL_DRAM_POWER_INFO)
@@ -713,7 +720,7 @@ static const struct platform_features skl_features = {
 	.has_ext_cst_msrs = 1,
 	.trl_msrs = TRL_BASE,
 	.tcc_offset_bits = 6,
-	.rapl_msrs = RAPL_PKG_ALL | RAPL_CORE_ALL | RAPL_DRAM | RAPL_DRAM_PERF_STATUS | RAPL_GFX,
+	.rapl_msrs = RAPL_PKG_ALL | RAPL_CORE_ALL | RAPL_DRAM | RAPL_DRAM_PERF_STATUS | RAPL_GFX | RAPL_PSYS,
 	.enable_tsc_tweak = 1,
 };
 
@@ -730,7 +737,7 @@ static const struct platform_features cnl_features = {
 	.has_ext_cst_msrs = 1,
 	.trl_msrs = TRL_BASE,
 	.tcc_offset_bits = 6,
-	.rapl_msrs = RAPL_PKG_ALL | RAPL_CORE_ALL | RAPL_DRAM | RAPL_DRAM_PERF_STATUS | RAPL_GFX,
+	.rapl_msrs = RAPL_PKG_ALL | RAPL_CORE_ALL | RAPL_DRAM | RAPL_DRAM_PERF_STATUS | RAPL_GFX | RAPL_PSYS,
 	.enable_tsc_tweak = 1,
 };
 
@@ -797,7 +804,7 @@ static const struct platform_features icx_features = {
 	.has_irtl_msrs = 1,
 	.has_cst_prewake_bit = 1,
 	.trl_msrs = TRL_BASE | TRL_CORECOUNT,
-	.rapl_msrs = RAPL_PKG_ALL | RAPL_DRAM_ALL,
+	.rapl_msrs = RAPL_PKG_ALL | RAPL_DRAM_ALL | RAPL_PSYS,
 	.has_fixed_rapl_unit = 1,
 };
 
@@ -813,7 +820,7 @@ static const struct platform_features spr_features = {
 	.has_irtl_msrs = 1,
 	.has_cst_prewake_bit = 1,
 	.trl_msrs = TRL_BASE | TRL_CORECOUNT,
-	.rapl_msrs = RAPL_PKG_ALL | RAPL_DRAM_ALL,
+	.rapl_msrs = RAPL_PKG_ALL | RAPL_DRAM_ALL | RAPL_PSYS,
 };
 
 static const struct platform_features srf_features = {
@@ -829,7 +836,7 @@ static const struct platform_features srf_features = {
 	.has_irtl_msrs = 1,
 	.has_cst_prewake_bit = 1,
 	.trl_msrs = TRL_BASE | TRL_CORECOUNT,
-	.rapl_msrs = RAPL_PKG_ALL | RAPL_DRAM_ALL,
+	.rapl_msrs = RAPL_PKG_ALL | RAPL_DRAM_ALL | RAPL_PSYS,
 };
 
 static const struct platform_features grr_features = {
@@ -845,7 +852,7 @@ static const struct platform_features grr_features = {
 	.has_irtl_msrs = 1,
 	.has_cst_prewake_bit = 1,
 	.trl_msrs = TRL_BASE | TRL_CORECOUNT,
-	.rapl_msrs = RAPL_PKG_ALL | RAPL_DRAM_ALL,
+	.rapl_msrs = RAPL_PKG_ALL | RAPL_DRAM_ALL | RAPL_PSYS,
 };
 
 static const struct platform_features slv_features = {
@@ -1108,6 +1115,7 @@ enum rapl_rci_index {
 	RAPL_RCI_INDEX_PKG_PERF_STATUS = 4,
 	RAPL_RCI_INDEX_DRAM_PERF_STATUS = 5,
 	RAPL_RCI_INDEX_CORE_ENERGY = 6,
+	RAPL_RCI_INDEX_ENERGY_PLATFORM = 7,
 	NUM_RAPL_COUNTERS,
 };
 
@@ -1134,6 +1142,7 @@ struct rapl_counter_info_t {
 struct rapl_counter_info_t *rapl_counter_info_perdomain;
 unsigned int rapl_counter_info_perdomain_size;
 
+#define RAPL_COUNTER_FLAG_PLATFORM_COUNTER (1u << 0)
 #define RAPL_COUNTER_FLAG_USE_MSR_SUM (1u << 1)
 
 struct rapl_counter_arch_info {
@@ -1255,6 +1264,19 @@ static const struct rapl_counter_arch_info rapl_counter_arch_infos[] = {
 	 .compat_scale = 1.0,
 	 .flags = 0,
 	  },
+	{
+	 .feature_mask = RAPL_PSYS,
+	 .perf_subsys = "power",
+	 .perf_name = "energy-psys",
+	 .msr = MSR_PLATFORM_ENERGY_STATUS,
+	 .msr_mask = 0x00000000FFFFFFFF,
+	 .msr_shift = 0,
+	 .platform_rapl_msr_scale = &rapl_energy_units,
+	 .rci_index = RAPL_RCI_INDEX_ENERGY_PLATFORM,
+	 .bic = BIC_SysWatt | BIC_Sys_J,
+	 .compat_scale = 1.0,
+	 .flags = RAPL_COUNTER_FLAG_PLATFORM_COUNTER | RAPL_COUNTER_FLAG_USE_MSR_SUM,
+	  },
 };
 
 struct rapl_counter {
@@ -1682,6 +1704,7 @@ enum {
 	IDX_PP1_ENERGY,
 	IDX_PKG_PERF,
 	IDX_DRAM_PERF,
+	IDX_PSYS_ENERGY,
 	IDX_COUNT,
 };
 
@@ -1726,6 +1749,9 @@ off_t idx_to_offset(int idx)
 	case IDX_DRAM_PERF:
 		offset = MSR_DRAM_PERF_STATUS;
 		break;
+	case IDX_PSYS_ENERGY:
+		offset = MSR_PLATFORM_ENERGY_STATUS;
+		break;
 	default:
 		offset = -1;
 	}
@@ -1756,6 +1782,9 @@ int offset_to_idx(off_t offset)
 	case MSR_DRAM_PERF_STATUS:
 		idx = IDX_DRAM_PERF;
 		break;
+	case MSR_PLATFORM_ENERGY_STATUS:
+		idx = IDX_PSYS_ENERGY;
+		break;
 	default:
 		idx = -1;
 	}
@@ -1777,6 +1806,8 @@ int idx_valid(int idx)
 		return platform->rapl_msrs & RAPL_PKG_PERF_STATUS;
 	case IDX_DRAM_PERF:
 		return platform->rapl_msrs & RAPL_DRAM_PERF_STATUS;
+	case IDX_PSYS_ENERGY:
+		return platform->rapl_msrs & RAPL_PSYS;
 	default:
 		return 0;
 	}
@@ -1848,6 +1879,10 @@ struct system_summary {
 	struct pkg_data packages;
 } average;
 
+struct platform_counters {
+	struct rapl_counter energy_psys;	/* MSR_PLATFORM_ENERGY_STATUS */
+} platform_counters_odd, platform_counters_even;
+
 struct cpu_topology {
 	int physical_package_id;
 	int die_id;
@@ -2512,13 +2547,19 @@ void print_header(char *delim)
 		ppmt = ppmt->next;
 	}
 
+	if (DO_BIC(BIC_SysWatt))
+		outp += sprintf(outp, "%sSysWatt", (printed++ ? delim : ""));
+	if (DO_BIC(BIC_Sys_J))
+		outp += sprintf(outp, "%sSys_J", (printed++ ? delim : ""));
+
 	outp += sprintf(outp, "\n");
 }
 
 int dump_counters(struct thread_data *t, struct core_data *c, struct pkg_data *p)
 {
 	int i;
 	struct msr_counter *mp;
+	struct platform_counters *pplat_cnt = p == package_odd ? &platform_counters_odd : &platform_counters_even;
 
 	outp += sprintf(outp, "t %p, c %p, p %p\n", t, c, p);
 
@@ -2590,6 +2631,7 @@ int dump_counters(struct thread_data *t, struct core_data *c, struct pkg_data *p
 		outp += sprintf(outp, "Joules COR: %0llX\n", p->energy_cores.raw_value);
 		outp += sprintf(outp, "Joules GFX: %0llX\n", p->energy_gfx.raw_value);
 		outp += sprintf(outp, "Joules RAM: %0llX\n", p->energy_dram.raw_value);
+		outp += sprintf(outp, "Joules PSYS: %0llX\n", pplat_cnt->energy_psys.raw_value);
 		outp += sprintf(outp, "Throttle PKG: %0llX\n", p->rapl_pkg_perf_status.raw_value);
 		outp += sprintf(outp, "Throttle RAM: %0llX\n", p->rapl_dram_perf_status.raw_value);
 		outp += sprintf(outp, "PTM: %dC\n", p->pkg_temp_c);
@@ -2628,6 +2670,9 @@ double rapl_counter_get_value(const struct rapl_counter *c, enum rapl_unit desir
  */
 int format_counters(struct thread_data *t, struct core_data *c, struct pkg_data *p)
 {
+	static int count;
+
+	struct platform_counters *pplat_cnt = NULL;
 	double interval_float, tsc;
 	char *fmt8;
 	int i;
@@ -2637,6 +2682,11 @@ int format_counters(struct thread_data *t, struct core_data *c, struct pkg_data
 	char *delim = "\t";
 	int printed = 0;
 
+	if (t == &average.threads) {
+		pplat_cnt = count & 1 ? &platform_counters_odd : &platform_counters_even;
+		++count;
+	}
+
 	/* if showing only 1st thread in core and this isn't one, bail out */
 	if (show_core_only && !is_cpu_first_thread_in_core(t, c, p))
 		return 0;
@@ -3093,6 +3143,13 @@ int format_counters(struct thread_data *t, struct core_data *c, struct pkg_data
 		}
 	}
 
+	if (DO_BIC(BIC_SysWatt) && (t == &average.threads))
+		outp += sprintf(outp, fmt8, (printed++ ? delim : ""),
+				rapl_counter_get_value(&pplat_cnt->energy_psys, RAPL_UNIT_WATTS, interval_float));
+	if (DO_BIC(BIC_Sys_J) && (t == &average.threads))
+		outp += sprintf(outp, fmt8, (printed++ ? delim : ""),
+				rapl_counter_get_value(&pplat_cnt->energy_psys, RAPL_UNIT_JOULES, interval_float));
+
 done:
 	if (*(outp - 1) != '\n')
 		outp += sprintf(outp, "\n");
@@ -3400,6 +3457,11 @@ int delta_cpu(struct thread_data *t, struct core_data *c,
 	return retval;
 }
 
+void delta_platform(struct platform_counters *new, struct platform_counters *old)
+{
+	old->energy_psys.raw_value = new->energy_psys.raw_value - old->energy_psys.raw_value;
+}
+
 void rapl_counter_clear(struct rapl_counter *c)
 {
 	c->raw_value = 0;
@@ -4129,13 +4191,17 @@ static size_t cstate_counter_info_count_perf(const struct cstate_counter_info_t
 
 void write_rapl_counter(struct rapl_counter *rc, struct rapl_counter_info_t *rci, unsigned int idx)
 {
+	if (rci->source[idx] == COUNTER_SOURCE_NONE)
+		return;
+
 	rc->raw_value = rci->data[idx];
 	rc->unit = rci->unit[idx];
 	rc->scale = rci->scale[idx];
 }
 
 int get_rapl_counters(int cpu, unsigned int domain, struct core_data *c, struct pkg_data *p)
 {
+	struct platform_counters *pplat_cnt = p == package_odd ? &platform_counters_odd : &platform_counters_even;
 	unsigned long long perf_data[NUM_RAPL_COUNTERS + 1];
 	struct rapl_counter_info_t *rci;
 
@@ -4163,6 +4229,7 @@ int get_rapl_counters(int cpu, unsigned int domain, struct core_data *c, struct
 	for (unsigned int i = 0, pi = 1; i < NUM_RAPL_COUNTERS; ++i) {
 		switch (rci->source[i]) {
 		case COUNTER_SOURCE_NONE:
+			rci->data[i] = 0;
 			break;
 
 		case COUNTER_SOURCE_PERF:
@@ -4201,14 +4268,15 @@ int get_rapl_counters(int cpu, unsigned int domain, struct core_data *c, struct
 		}
 	}
 
-	BUILD_BUG_ON(NUM_RAPL_COUNTERS != 7);
+	BUILD_BUG_ON(NUM_RAPL_COUNTERS != 8);
 	write_rapl_counter(&p->energy_pkg, rci, RAPL_RCI_INDEX_ENERGY_PKG);
 	write_rapl_counter(&p->energy_cores, rci, RAPL_RCI_INDEX_ENERGY_CORES);
 	write_rapl_counter(&p->energy_dram, rci, RAPL_RCI_INDEX_DRAM);
 	write_rapl_counter(&p->energy_gfx, rci, RAPL_RCI_INDEX_GFX);
 	write_rapl_counter(&p->rapl_pkg_perf_status, rci, RAPL_RCI_INDEX_PKG_PERF_STATUS);
 	write_rapl_counter(&p->rapl_dram_perf_status, rci, RAPL_RCI_INDEX_DRAM_PERF_STATUS);
 	write_rapl_counter(&c->core_energy, rci, RAPL_RCI_INDEX_CORE_ENERGY);
+	write_rapl_counter(&pplat_cnt->energy_psys, rci, RAPL_RCI_INDEX_ENERGY_PLATFORM);
 
 	return 0;
 }
@@ -6144,6 +6212,7 @@ void turbostat_loop()
 			re_initialize();
 			goto restart;
 		}
+		delta_platform(&platform_counters_odd, &platform_counters_even);
 		compute_average(EVEN_COUNTERS);
 		format_all_counters(EVEN_COUNTERS);
 		flush_output_stdout();
@@ -6167,6 +6236,7 @@ void turbostat_loop()
 			re_initialize();
 			goto restart;
 		}
+		delta_platform(&platform_counters_even, &platform_counters_odd);
 		compute_average(ODD_COUNTERS);
 		format_all_counters(ODD_COUNTERS);
 		flush_output_stdout();
@@ -6945,8 +7015,8 @@ void rapl_probe_intel(void)
 	unsigned long long msr;
 	unsigned int time_unit;
 	double tdp;
-	const unsigned long long bic_watt_bits = BIC_PkgWatt | BIC_CorWatt | BIC_RAMWatt | BIC_GFXWatt;
-	const unsigned long long bic_joules_bits = BIC_Pkg_J | BIC_Cor_J | BIC_RAM_J | BIC_GFX_J;
+	const unsigned long long bic_watt_bits = BIC_SysWatt | BIC_PkgWatt | BIC_CorWatt | BIC_RAMWatt | BIC_GFXWatt;
+	const unsigned long long bic_joules_bits = BIC_Sys_J | BIC_Pkg_J | BIC_Cor_J | BIC_RAM_J | BIC_GFX_J;
 
 	if (rapl_joules)
 		bic_enabled &= ~bic_watt_bits;
@@ -7606,6 +7676,9 @@ void rapl_perf_init(void)
 
 			domain_visited[next_domain] = 1;
 
+			if ((cai->flags & RAPL_COUNTER_FLAG_PLATFORM_COUNTER) && (cpu != base_cpu))
+				continue;
+
 			struct rapl_counter_info_t *rci = &rapl_counter_info_perdomain[next_domain];
 
 			/* Check if the counter is enabled and accessible */