[PATCH v2 29/57] x86: acpi: Add common Intel ACPI tables

Simon Glass sjg at chromium.org
Sun Aug 30 01:31:48 CEST 2020


Add various tables that are common to Intel CPUs. These functions can be
used by arch-specific CPU code.

Signed-off-by: Simon Glass <sjg at chromium.org>
---

(no changes since v1)

 arch/x86/cpu/intel_common/Makefile |   2 +
 arch/x86/cpu/intel_common/acpi.c   | 377 +++++++++++++++++++++++++++++
 arch/x86/cpu/intel_common/cpu.c    |  14 ++
 arch/x86/include/asm/acpi_table.h  |  22 ++
 arch/x86/include/asm/cpu_common.h  |   7 +
 arch/x86/include/asm/intel_acpi.h  |  52 ++++
 drivers/core/Kconfig               |   9 +
 7 files changed, 483 insertions(+)
 create mode 100644 arch/x86/cpu/intel_common/acpi.c
 create mode 100644 arch/x86/include/asm/intel_acpi.h

diff --git a/arch/x86/cpu/intel_common/Makefile b/arch/x86/cpu/intel_common/Makefile
index f1d1513a981..4a5cf17e41d 100644
--- a/arch/x86/cpu/intel_common/Makefile
+++ b/arch/x86/cpu/intel_common/Makefile
@@ -2,6 +2,8 @@
 #
 # Copyright (c) 2016 Google, Inc
 
+obj-$(CONFIG_INTEL_ACPIGEN) += acpi.o
+
 ifdef CONFIG_HAVE_MRC
 obj-$(CONFIG_$(SPL_TPL_)X86_16BIT_INIT) += car.o
 obj-$(CONFIG_$(SPL_TPL_)X86_32BIT_INIT) += me_status.o
diff --git a/arch/x86/cpu/intel_common/acpi.c b/arch/x86/cpu/intel_common/acpi.c
new file mode 100644
index 00000000000..a4d5fbd38a7
--- /dev/null
+++ b/arch/x86/cpu/intel_common/acpi.c
@@ -0,0 +1,377 @@
+// SPDX-License-Identifier: GPL-2.0+
+/*
+ * Generic Intel ACPI table generation
+ *
+ * Copyright (C) 2017 Intel Corp.
+ * Copyright 2019 Google LLC
+ *
+ * Modified from coreboot src/soc/intel/common/block/acpi.c
+ */
+
+#include <common.h>
+#include <bloblist.h>
+#include <cpu.h>
+#include <dm.h>
+#include <acpi/acpigen.h>
+#include <asm/acpigen.h>
+#include <asm/acpi_table.h>
+#include <asm/cpu.h>
+#include <asm/cpu_common.h>
+#include <asm/intel_acpi.h>
+#include <asm/ioapic.h>
+#include <asm/mpspec.h>
+#include <asm/smm.h>
+#include <asm/turbo.h>
+#include <asm/intel_gnvs.h>
+#include <asm/arch/iomap.h>
+#include <asm/arch/pm.h>
+#include <asm/arch/systemagent.h>
+#include <dm/acpi.h>
+#include <linux/err.h>
+#include <power/acpi_pmc.h>
+
+u32 acpi_fill_mcfg(u32 current)
+{
+	/* PCI Segment Group 0, Start Bus Number 0, End Bus Number is 255 */
+	current += acpi_create_mcfg_mmconfig((void *)current,
+					     CONFIG_MMCONF_BASE_ADDRESS, 0, 0,
+					     (CONFIG_SA_PCIEX_LENGTH >> 20)
+					     - 1);
+	return current;
+}
+
+static int acpi_sci_irq(void)
+{
+	int sci_irq = 9;
+	uint scis;
+	int ret;
+
+	ret = arch_read_sci_irq_select();
+	if (IS_ERR_VALUE(ret))
+		return log_msg_ret("sci_irq", ret);
+	scis = ret;
+	scis &= SCI_IRQ_MASK;
+	scis >>= SCI_IRQ_SHIFT;
+
+	/* Determine how SCI is routed. */
+	switch (scis) {
+	case SCIS_IRQ9:
+	case SCIS_IRQ10:
+	case SCIS_IRQ11:
+		sci_irq = scis - SCIS_IRQ9 + 9;
+		break;
+	case SCIS_IRQ20:
+	case SCIS_IRQ21:
+	case SCIS_IRQ22:
+	case SCIS_IRQ23:
+		sci_irq = scis - SCIS_IRQ20 + 20;
+		break;
+	default:
+		log_warning("Invalid SCI route! Defaulting to IRQ9\n");
+		sci_irq = 9;
+		break;
+	}
+
+	log_debug("SCI is IRQ%d\n", sci_irq);
+
+	return sci_irq;
+}
+
+static unsigned long acpi_madt_irq_overrides(unsigned long current)
+{
+	int sci = acpi_sci_irq();
+	u16 flags = MP_IRQ_TRIGGER_LEVEL;
+
+	if (sci < 0)
+		return log_msg_ret("sci irq", sci);
+
+	/* INT_SRC_OVR */
+	current += acpi_create_madt_irqoverride((void *)current, 0, 0, 2, 0);
+
+	flags |= arch_madt_sci_irq_polarity(sci);
+
+	/* SCI */
+	current +=
+	    acpi_create_madt_irqoverride((void *)current, 0, sci, sci, flags);
+
+	return current;
+}
+
+u32 acpi_fill_madt(u32 current)
+{
+	/* Local APICs */
+	current += acpi_create_madt_lapics(current);
+
+	/* IOAPIC */
+	current += acpi_create_madt_ioapic((void *)current, 2, IO_APIC_ADDR, 0);
+
+	return acpi_madt_irq_overrides(current);
+}
+
+void intel_acpi_fill_fadt(struct acpi_fadt *fadt)
+{
+	const u16 pmbase = IOMAP_ACPI_BASE;
+
+	/* Use ACPI 3.0 revision. */
+	fadt->header.revision = acpi_get_table_revision(ACPITAB_FADT);
+
+	fadt->sci_int = acpi_sci_irq();
+	fadt->smi_cmd = APM_CNT;
+	fadt->acpi_enable = APM_CNT_ACPI_ENABLE;
+	fadt->acpi_disable = APM_CNT_ACPI_DISABLE;
+	fadt->s4bios_req = 0x0;
+	fadt->pstate_cnt = 0;
+
+	fadt->pm1a_evt_blk = pmbase + PM1_STS;
+	fadt->pm1b_evt_blk = 0x0;
+	fadt->pm1a_cnt_blk = pmbase + PM1_CNT;
+	fadt->pm1b_cnt_blk = 0x0;
+
+	fadt->gpe0_blk = pmbase + GPE0_STS;
+
+	fadt->pm1_evt_len = 4;
+	fadt->pm1_cnt_len = 2;
+
+	/* GPE0 STS/EN pairs each 32 bits wide. */
+	fadt->gpe0_blk_len = 2 * GPE0_REG_MAX * sizeof(uint32_t);
+
+	fadt->flush_size = 0x400;	/* twice of cache size */
+	fadt->flush_stride = 0x10;	/* Cache line width  */
+	fadt->duty_offset = 1;
+	fadt->day_alrm = 0xd;
+
+	fadt->flags = ACPI_FADT_WBINVD | ACPI_FADT_C1_SUPPORTED |
+	    ACPI_FADT_C2_MP_SUPPORTED | ACPI_FADT_SLEEP_BUTTON |
+	    ACPI_FADT_RESET_REGISTER | ACPI_FADT_SEALED_CASE |
+	    ACPI_FADT_S4_RTC_WAKE | ACPI_FADT_PLATFORM_CLOCK;
+
+	fadt->reset_reg.space_id = 1;
+	fadt->reset_reg.bit_width = 8;
+	fadt->reset_reg.addrl = IO_PORT_RESET;
+	fadt->reset_value = RST_CPU | SYS_RST;
+
+	fadt->x_pm1a_evt_blk.space_id = 1;
+	fadt->x_pm1a_evt_blk.bit_width = fadt->pm1_evt_len * 8;
+	fadt->x_pm1a_evt_blk.addrl = pmbase + PM1_STS;
+
+	fadt->x_pm1b_evt_blk.space_id = 1;
+
+	fadt->x_pm1a_cnt_blk.space_id = 1;
+	fadt->x_pm1a_cnt_blk.bit_width = fadt->pm1_cnt_len * 8;
+	fadt->x_pm1a_cnt_blk.addrl = pmbase + PM1_CNT;
+
+	fadt->x_pm1b_cnt_blk.space_id = 1;
+
+	fadt->x_gpe1_blk.space_id = 1;
+}
+
+int intel_southbridge_write_acpi_tables(const struct udevice *dev,
+					struct acpi_ctx *ctx)
+{
+	int ret;
+
+	ret = acpi_write_dbg2_pci_uart(ctx, gd->cur_serial_dev,
+				       ACPI_ACCESS_SIZE_DWORD_ACCESS);
+	if (ret)
+		return log_msg_ret("dbg2", ret);
+
+	ret = acpi_write_hpet(ctx);
+	if (ret)
+		return log_msg_ret("hpet", ret);
+
+	return 0;
+}
+
+__weak u32 acpi_fill_soc_wake(u32 generic_pm1_en,
+			      const struct chipset_power_state *ps)
+{
+	return generic_pm1_en;
+}
+
+__weak int acpi_create_gnvs(struct acpi_global_nvs *gnvs)
+{
+	return 0;
+}
+
+int southbridge_inject_dsdt(const struct udevice *dev, struct acpi_ctx *ctx)
+{
+	struct acpi_global_nvs *gnvs;
+	int ret;
+
+	ret = bloblist_ensure_size(BLOBLISTT_ACPI_GNVS, sizeof(*gnvs),
+				   (void **)&gnvs);
+	if (ret)
+		return log_msg_ret("bloblist", ret);
+	memset(gnvs, '\0', sizeof(*gnvs));
+
+	ret = acpi_create_gnvs(gnvs);
+	if (ret)
+		return log_msg_ret("gnvs", ret);
+
+	/*
+	 * TODO(sjg at chromum.org): tell SMI about it
+	 * smm_setup_structures(gnvs, NULL, NULL);
+	 */
+
+	/* Add it to DSDT */
+	acpigen_write_scope(ctx, "\\");
+	acpigen_write_name_dword(ctx, "NVSA", (uintptr_t)gnvs);
+	acpigen_pop_len(ctx);
+
+	return 0;
+}
+
+static int calculate_power(int tdp, int p1_ratio, int ratio)
+{
+	u32 m;
+	u32 power;
+
+	/*
+	 * M = ((1.1 - ((p1_ratio - ratio) * 0.00625)) / 1.1) ^ 2
+	 *
+	 * Power = (ratio / p1_ratio) * m * tdp
+	 */
+
+	m = (110000 - ((p1_ratio - ratio) * 625)) / 11;
+	m = (m * m) / 1000;
+
+	power = ((ratio * 100000 / p1_ratio) / 100);
+	power *= (m / 100) * (tdp / 1000);
+	power /= 1000;
+
+	return power;
+}
+
+void generate_p_state_entries(struct acpi_ctx *ctx, int core,
+			      int cores_per_package)
+{
+	int ratio_min, ratio_max, ratio_turbo, ratio_step;
+	int coord_type, power_max, num_entries;
+	int ratio, power, clock, clock_max;
+	bool turbo;
+
+	coord_type = cpu_get_coord_type();
+	ratio_min = cpu_get_min_ratio();
+	ratio_max = cpu_get_max_ratio();
+	clock_max = (ratio_max * cpu_get_bus_clock_khz()) / 1000;
+	turbo = (turbo_get_state() == TURBO_ENABLED);
+
+	/* Calculate CPU TDP in mW */
+	power_max = cpu_get_power_max();
+
+	/* Write _PCT indicating use of FFixedHW */
+	acpigen_write_empty_pct(ctx);
+
+	/* Write _PPC with no limit on supported P-state */
+	acpigen_write_ppc_nvs(ctx);
+	/* Write PSD indicating configured coordination type */
+	acpigen_write_psd_package(ctx, core, 1, coord_type);
+
+	/* Add P-state entries in _PSS table */
+	acpigen_write_name(ctx, "_PSS");
+
+	/* Determine ratio points */
+	ratio_step = PSS_RATIO_STEP;
+	do {
+		num_entries = ((ratio_max - ratio_min) / ratio_step) + 1;
+		if (((ratio_max - ratio_min) % ratio_step) > 0)
+			num_entries += 1;
+		if (turbo)
+			num_entries += 1;
+		if (num_entries > PSS_MAX_ENTRIES)
+			ratio_step += 1;
+	} while (num_entries > PSS_MAX_ENTRIES);
+
+	/* _PSS package count depends on Turbo */
+	acpigen_write_package(ctx, num_entries);
+
+	/* P[T] is Turbo state if enabled */
+	if (turbo) {
+		ratio_turbo = cpu_get_max_turbo_ratio();
+
+		/* Add entry for Turbo ratio */
+		acpigen_write_pss_package(ctx, clock_max + 1,	/* MHz */
+					  power_max,		/* mW */
+					  PSS_LATENCY_TRANSITION,/* lat1 */
+					  PSS_LATENCY_BUSMASTER,/* lat2 */
+					  ratio_turbo << 8,	/* control */
+					  ratio_turbo << 8);	/* status */
+		num_entries -= 1;
+	}
+
+	/* First regular entry is max non-turbo ratio */
+	acpigen_write_pss_package(ctx, clock_max,	/* MHz */
+				  power_max,		/* mW */
+				  PSS_LATENCY_TRANSITION,/* lat1 */
+				  PSS_LATENCY_BUSMASTER,/* lat2 */
+				  ratio_max << 8,	/* control */
+				  ratio_max << 8);	/* status */
+	num_entries -= 1;
+
+	/* Generate the remaining entries */
+	for (ratio = ratio_min + ((num_entries - 1) * ratio_step);
+	     ratio >= ratio_min; ratio -= ratio_step) {
+		/* Calculate power at this ratio */
+		power = calculate_power(power_max, ratio_max, ratio);
+		clock = (ratio * cpu_get_bus_clock_khz()) / 1000;
+
+		acpigen_write_pss_package(ctx, clock,		/* MHz */
+					  power,		/* mW */
+					  PSS_LATENCY_TRANSITION,/* lat1 */
+					  PSS_LATENCY_BUSMASTER,/* lat2 */
+					  ratio << 8,		/* control */
+					  ratio << 8);		/* status */
+	}
+	/* Fix package length */
+	acpigen_pop_len(ctx);
+}
+
+void generate_t_state_entries(struct acpi_ctx *ctx, int core,
+			      int cores_per_package, struct acpi_tstate *entry,
+			      int nentries)
+{
+	if (!nentries)
+		return;
+
+	/* Indicate SW_ALL coordination for T-states */
+	acpigen_write_tsd_package(ctx, core, cores_per_package, SW_ALL);
+
+	/* Indicate FixedHW so OS will use MSR */
+	acpigen_write_empty_ptc(ctx);
+
+	/* Set NVS controlled T-state limit */
+	acpigen_write_tpc(ctx, "\\TLVL");
+
+	/* Write TSS table for MSR access */
+	acpigen_write_tss_package(ctx, entry, nentries);
+}
+
+int acpi_generate_cpu_header(struct acpi_ctx *ctx, int core_id,
+			     const struct acpi_cstate *c_state_map,
+			     int num_cstates)
+{
+	bool is_first = !core_id;
+
+	/* Generate processor \_PR.CPUx */
+	acpigen_write_processor(ctx, core_id, is_first ? ACPI_BASE_ADDRESS : 0,
+				is_first ? 6 : 0);
+
+	/* Generate C-state tables */
+	acpigen_write_cst_package(ctx, c_state_map, num_cstates);
+
+	return 0;
+}
+
+int acpi_generate_cpu_package_final(struct acpi_ctx *ctx, int cores_per_package)
+{
+	/*
+	 * PPKG is usually used for thermal management of the first and only
+	 * package
+	 */
+	acpigen_write_processor_package(ctx, "PPKG", 0, cores_per_package);
+
+	/* Add a method to notify processor nodes */
+	acpigen_write_processor_cnot(ctx, cores_per_package);
+
+	return 0;
+}
diff --git a/arch/x86/cpu/intel_common/cpu.c b/arch/x86/cpu/intel_common/cpu.c
index cb4ef84013a..d8a3d60ae72 100644
--- a/arch/x86/cpu/intel_common/cpu.c
+++ b/arch/x86/cpu/intel_common/cpu.c
@@ -291,3 +291,17 @@ int cpu_get_max_turbo_ratio(void)
 
 	return msr.lo & 0xff;
 }
+
+int cpu_get_cores_per_package(void)
+{
+	struct cpuid_result result;
+	int cores = 1;
+
+	if (gd->arch.x86_vendor != X86_VENDOR_INTEL)
+		return 1;
+
+	result = cpuid_ext(0xb, 1);
+	cores = result.ebx & 0xff;
+
+	return cores;
+}
diff --git a/arch/x86/include/asm/acpi_table.h b/arch/x86/include/asm/acpi_table.h
index 1b7ff509516..3245e447813 100644
--- a/arch/x86/include/asm/acpi_table.h
+++ b/arch/x86/include/asm/acpi_table.h
@@ -76,4 +76,26 @@ ulong write_acpi_tables(ulong start);
  */
 ulong acpi_get_rsdp_addr(void);
 
+/**
+ * arch_read_sci_irq_select() - Read the system-control interrupt number
+ *
+ * @returns value of IRQ register in the PMC
+ */
+int arch_read_sci_irq_select(void);
+
+/**
+ * arch_write_sci_irq_select() - Set the system-control interrupt number
+ *
+ * @scis: New value for IRQ register in the PMC
+ */
+int arch_write_sci_irq_select(uint scis);
+
+/**
+ * arch_madt_sci_irq_polarity() - Return the priority to use for the MADT
+ *
+ * @sci: System-control interrupt number
+ * @return priority to use (MP_IRQ_POLARITY_...)
+ */
+int arch_madt_sci_irq_polarity(int sci);
+
 #endif /* __ASM_ACPI_TABLE_H__ */
diff --git a/arch/x86/include/asm/cpu_common.h b/arch/x86/include/asm/cpu_common.h
index a7b7112d417..48f56c2aad9 100644
--- a/arch/x86/include/asm/cpu_common.h
+++ b/arch/x86/include/asm/cpu_common.h
@@ -177,4 +177,11 @@ int cpu_get_power_max(void);
  */
 int cpu_get_max_turbo_ratio(void);
 
+/**
+ * cpu_get_cores_per_package() - Get the number of CPU cores in each package
+ *
+ * @return number of cores
+ */
+int cpu_get_cores_per_package(void);
+
 #endif
diff --git a/arch/x86/include/asm/intel_acpi.h b/arch/x86/include/asm/intel_acpi.h
new file mode 100644
index 00000000000..a5781f1af45
--- /dev/null
+++ b/arch/x86/include/asm/intel_acpi.h
@@ -0,0 +1,52 @@
+/* SPDX-License-Identifier: GPL-2.0+ */
+/*
+ * Copyright (C) 2018, Bin Meng <bmeng.cn at gmail.com>
+ */
+
+#ifndef __ASM_INTEL_ACPI_H__
+#define __ASM_INTEL_ACPI_H__
+
+struct acpi_cstate;
+struct acpi_ctx;
+struct acpi_tstate;
+struct udevice;
+
+/**
+ * acpi_generate_cpu_header() - Start generating an ACPI CPU entry
+ *
+ * Generates the ACPI information for a CPU. After this, the caller should
+ * generate_p_state_entries(), generate_t_state_entries and then
+ * acpigen_pop_len() to close off this package.
+ *
+ * @ctx: ACPI context pointer
+ * @core_id: CPU core number, as numbered by the SoC
+ * @c_state_map: Information about each C state
+ * @num_cstates: Number of entries in @c_state_map
+ * @return 0 if OK, -ve on error
+ */
+int acpi_generate_cpu_header(struct acpi_ctx *ctx, int core_id,
+			     const struct acpi_cstate *c_state_map,
+			     int num_cstates);
+
+/**
+ * acpi_generate_cpu_package_final() - Write out the CPU PPKG entry
+ *
+ * This writes information about the CPUs in the package
+ *
+ * @ctx: ACPI context pointer
+ * @cores_per_package: Number of CPU cores in each package in the SoC
+ */
+int acpi_generate_cpu_package_final(struct acpi_ctx *ctx,
+				    int cores_per_package);
+
+void generate_p_state_entries(struct acpi_ctx *ctx, int core,
+			      int cores_per_package);
+void generate_t_state_entries(struct acpi_ctx *ctx, int core,
+			      int cores_per_package, struct acpi_tstate *entry,
+			      int nentries);
+int southbridge_inject_dsdt(const struct udevice *dev, struct acpi_ctx *ctx);
+
+int intel_southbridge_write_acpi_tables(const struct udevice *dev,
+					struct acpi_ctx *ctx);
+
+#endif /* __ASM_INTEL_ACPI_H__ */
diff --git a/drivers/core/Kconfig b/drivers/core/Kconfig
index 00d1d80dc38..1ca5d66141b 100644
--- a/drivers/core/Kconfig
+++ b/drivers/core/Kconfig
@@ -277,4 +277,13 @@ config ACPIGEN
 	  things like generating device-specific tables and returning the ACPI
 	  name of a device.
 
+config INTEL_ACPIGEN
+	bool "Support ACPI table generation for Intel SoCs"
+	depends on ACPIGEN
+	help
+	  This option adds some functions used for programatic generation of
+	  ACPI tables on Intel SoCs. This provides features for writing CPU
+	  information such as P states and T stages. Also included is a way
+	  to create a GNVS table and set it up.
+
 endmenu
-- 
2.28.0.402.g5ffc5be6b7-goog



More information about the U-Boot mailing list