From 6300b034143cab98d472276054de2781eb3ce97a Mon Sep 17 00:00:00 2001 From: Timothy Pearson Date: Mon, 26 Jan 2015 17:53:22 -0600 Subject: amd/amdfam10: Add runtime ACPI _PSS generation Skeleton and ACPI generator interface taken from model_fxx powernow_acpi.c Small portions of FIDVID MSR code taken from model_10xxx fidvid.c Nearly completely rewritten for the P-state-based K10 CPU TEST: KFSN4-DRE with dual Opteron 8356 CPUs Verified CPU per-core dynamic state change with system load Verified reported P-state count and frequencies Stress-tested each CPU (all cores simultaneously) to verify proper P0 transition and configuration. Change-Id: Icf620ec96a3f163b62d96b5988184996641dd439 Signed-off-by: Timothy Pearson Reviewed-on: http://review.coreboot.org/8284 Tested-by: build bot (Jenkins) Reviewed-by: Alexandru Gagniuc --- src/cpu/amd/model_10xxx/powernow_acpi.c | 228 +++++++++++++++++++++++++++++++- 1 file changed, 225 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/cpu/amd/model_10xxx/powernow_acpi.c b/src/cpu/amd/model_10xxx/powernow_acpi.c index 97f2d16a6c..e4e0289146 100644 --- a/src/cpu/amd/model_10xxx/powernow_acpi.c +++ b/src/cpu/amd/model_10xxx/powernow_acpi.c @@ -1,8 +1,9 @@ /* * This file is part of the coreboot project. * - * Copyright (C) 2008 Advanced Micro Devices, Inc. + * Copyright (C) 2007-2008 Advanced Micro Devices, Inc. * Copyright (C) 2009 Rudolf Marek + * Copyright (C) 2015 Timothy Pearson , Raptor Engineering * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by @@ -23,9 +24,230 @@ #include #include #include +#include +#include +#include +#include +#include +#include +#include + +static void write_pstates_for_core(u8 pstate_num, u16 *pstate_feq, u32 *pstate_power, + u32 *pstate_latency, u32 *pstate_control, + u32 *pstate_status, int coreID, + u32 pcontrol_blk, u8 plen, u8 onlyBSP) +{ + int i; + + if ((onlyBSP) && (coreID != 0)) { + plen = 0; + pcontrol_blk = 0; + } + + acpigen_write_processor(coreID, pcontrol_blk, plen); + acpigen_write_empty_PCT(); + acpigen_write_name("_PSS"); + + /* add later to total sum */ + acpigen_write_package(pstate_num); + + for (i = 0;i < pstate_num; i++) + acpigen_write_PSS_package(pstate_feq[i], + pstate_power[i], + pstate_latency[i], + pstate_latency[i], + pstate_control[i], + pstate_status[i]); + + /* update the package size */ + acpigen_pop_len(); + + acpigen_write_PPC(pstate_num); + /* patch the whole Processor token length */ + acpigen_pop_len(); +} + +/* +* For details of this algorithm, please refer to the BDKG 3.62 page 69 +*/ +static void pstates_algorithm(u32 pcontrol_blk, u8 plen, u8 onlyBSP) +{ + u8 processor_brand[49]; + u32 *v; + struct cpuid_result cpuid1; + + u16 Pstate_feq[10]; + u32 Pstate_power[10]; + u32 Pstate_latency[10]; + u32 Pstate_control[10]; + u32 Pstate_status[10]; + u8 Pstate_num; + u8 cmp_cap; + u8 index; + msr_t msr; + + /* Get the Processor Brand String using cpuid(0x8000000x) command x=2,3,4 */ + cpuid1 = cpuid(0x80000002); + v = (u32 *) processor_brand; + v[0] = cpuid1.eax; + v[1] = cpuid1.ebx; + v[2] = cpuid1.ecx; + v[3] = cpuid1.edx; + cpuid1 = cpuid(0x80000003); + v[4] = cpuid1.eax; + v[5] = cpuid1.ebx; + v[6] = cpuid1.ecx; + v[7] = cpuid1.edx; + cpuid1 = cpuid(0x80000004); + v[8] = cpuid1.eax; + v[9] = cpuid1.ebx; + v[10] = cpuid1.ecx; + v[11] = cpuid1.edx; + processor_brand[48] = 0; + printk(BIOS_INFO, "processor_brand=%s\n", processor_brand); + + /* + * Based on the CPU socket type,cmp_cap and pwr_lmt , get the power limit. + * socket_type : 0x10 SocketF; 0x11 AM2/ASB1 ; 0x12 S1G1 + * cmp_cap : 0x0 SingleCore ; 0x1 DualCore ; 0x2 TripleCore ; 0x3 QuadCore ; 0x5 QuintupleCore ; 0x6 HexCore + */ + printk(BIOS_INFO, "Pstates Algorithm ...\n"); + cmp_cap = + (pci_read_config16(dev_find_slot(0, PCI_DEVFN(0x18, 3)), 0xE8) & + 0x7000) >> 12; + + Pstate_num = 0; + + /* See if the CPUID(0x80000007) returned EDX[7]==1b */ + cpuid1 = cpuid(0x80000007); + if ((cpuid1.edx & 0x80) != 0x80) { + printk(BIOS_INFO, "No valid set of P-states\n"); + goto write_pstates; + } + + uint32_t dtemp; + uint8_t pviModeFlag; + uint8_t Pstate_max; + uint8_t cpufid; + uint8_t cpudid; + uint8_t cpuvid; + uint8_t cpuidd; + uint8_t cpuidv; + uint8_t power_step_up; + uint8_t power_step_down; + uint8_t pll_lock_time; + uint32_t expanded_cpuidv; + uint32_t core_frequency; + uint32_t core_power; + uint32_t core_latency; + uint32_t core_voltage; /* multiplied by 10000 */ + + /* Determine if this is a PVI or SVI system */ + dtemp = pci_read_config32(dev_find_slot(0, PCI_DEVFN(0x18, 3)), 0xA0); + + if (dtemp & PVI_MODE) + pviModeFlag = 1; + else + pviModeFlag = 0; + + /* Get PSmax's index */ + msr = rdmsr(0xC0010061); + Pstate_max = (uint8_t) ((msr.lo >> PS_MAX_VAL_SHFT) & BIT_MASK_3); + + /* Determine if all enabled Pstates have the same fidvid */ + uint8_t i; + uint8_t cpufid_prev = (rdmsr(0xC0010064).lo & 0x3f); + uint8_t all_enabled_cores_have_same_cpufid = 1; + for (i = 1; i < Pstate_max; i++) { + cpufid = rdmsr(0xC0010064 + i).lo & 0x3f; + if (cpufid != cpufid_prev) { + all_enabled_cores_have_same_cpufid = 0; + break; + } + } + + /* Populate tables with all Pstate information */ + for (Pstate_num = 0; Pstate_num < Pstate_max; Pstate_num++) { + /* Get power state information */ + msr = rdmsr(0xC0010064 + Pstate_num); + cpufid = (msr.lo & 0x3f); + cpudid = (msr.lo & 0x1c0) >> 6; + cpuvid = (msr.lo & 0xfe00) >> 9; + cpuidd = (msr.hi & 0xff); + cpuidv = (msr.hi & 0x300) >> 8; + core_frequency = (100 * (cpufid + 0x10)) / (0x01 << cpudid); + if (pviModeFlag) { + if (cpuvid >= 0x20) { + core_voltage = 7625 - (((cpuvid - 0x20) * 10000) / 80); + } + else { + core_voltage = 15500 - ((cpuvid * 10000) / 40); + } + } + else { + cpuvid = cpuvid & 0x7f; + if (cpuvid >= 0x7c) + core_voltage = 0; + else + core_voltage = 15500 - ((cpuvid * 10000) / 80); + } + switch (cpuidv) { + case 0x0: + expanded_cpuidv = 1; + break; + case 0x1: + expanded_cpuidv = 10; + break; + case 0x2: + expanded_cpuidv = 100; + break; + case 0x3: + expanded_cpuidv = 1000; + break; + } + core_power = (core_voltage * cpuidd) / (expanded_cpuidv * 10); + + /* Calculate transition latency */ + dtemp = pci_read_config32(dev_find_slot(0, PCI_DEVFN(0x18, 3)), 0xD4); + power_step_up = (dtemp & 0xf000000) >> 24; + power_step_down = (dtemp & 0xf00000) >> 20; + dtemp = pci_read_config32(dev_find_slot(0, PCI_DEVFN(0x18, 3)), 0xA0); + pll_lock_time = (pll_lock_time & 0x3800) >> 11; + if (all_enabled_cores_have_same_cpufid) + core_latency = ((12 * power_step_down) + power_step_up) / 1000; + else + core_latency = (12 * (power_step_down + power_step_up) / 1000) + + pll_lock_time; + + Pstate_feq[Pstate_num] = core_frequency; + Pstate_power[Pstate_num] = core_power; + Pstate_latency[Pstate_num] = core_latency; + Pstate_control[Pstate_num] = Pstate_num; + Pstate_status[Pstate_num] = Pstate_num; + } + + /* Print Pstate frequency, power, and latency */ + for (index = 0; index < Pstate_num; index++) { + printk(BIOS_INFO, "Pstate_freq[%d] = %dMHz\t", index, + Pstate_feq[index]); + printk(BIOS_INFO, "Pstate_power[%d] = %dmw\n", index, + Pstate_power[index]); + printk(BIOS_INFO, "Pstate_latency[%d] = %dus\n", index, + Pstate_latency[index]); + } + +write_pstates: + for (index = 0; index < (cmp_cap + 1); index++) + write_pstates_for_core(Pstate_num, Pstate_feq, Pstate_power, + Pstate_latency, Pstate_control, Pstate_status, + index, pcontrol_blk, plen, onlyBSP); +} -/* FIXME: this should be implemented but right now all boards hardcode it. */ void amd_generate_powernow(u32 pcontrol_blk, u8 plen, u8 onlyBSP) { - return; + char pscope[] = "\\_PR"; + + acpigen_write_scope(pscope); + pstates_algorithm(pcontrol_blk, plen, onlyBSP); + acpigen_pop_len(); } -- cgit v1.2.3