linux/drivers/xen/xen-acpi-processor.c

599 lines
17 KiB
C
Raw Normal View History

xen/acpi-processor: C and P-state driver that uploads said data to hypervisor. This driver solves three problems: 1). Parse and upload ACPI0007 (or PROCESSOR_TYPE) information to the hypervisor - aka P-states (cpufreq data). 2). Upload the the Cx state information (cpuidle data). 3). Inhibit CPU frequency scaling drivers from loading. The reason for wanting to solve 1) and 2) is such that the Xen hypervisor is the only one that knows the CPU usage of different guests and can make the proper decision of when to put CPUs and packages in proper states. Unfortunately the hypervisor has no support to parse ACPI DSDT tables, hence it needs help from the initial domain to provide this information. The reason for 3) is that we do not want the initial domain to change P-states while the hypervisor is doing it as well - it causes rather some funny cases of P-states transitions. For this to work, the driver parses the Power Management data and uploads said information to the Xen hypervisor. It also calls acpi_processor_notify_smm() to inhibit the other CPU frequency scaling drivers from being loaded. Everything revolves around the 'struct acpi_processor' structure which gets updated during the bootup cycle in different stages. At the startup, when the ACPI parser starts, the C-state information is processed (processor_idle) and saved in said structure as 'power' element. Later on, the CPU frequency scaling driver (powernow-k8 or acpi_cpufreq), would call the the acpi_processor_* (processor_perflib functions) to parse P-states information and populate in the said structure the 'performance' element. Since we do not want the CPU frequency scaling drivers from loading we have to call the acpi_processor_* functions to parse the P-states and call "acpi_processor_notify_smm" to stop them from loading. There is also one oddity in this driver which is that under Xen, the physical online CPU count can be different from the virtual online CPU count. Meaning that the macros 'for_[online|possible]_cpu' would process only up to virtual online CPU count. We on the other hand want to process the full amount of physical CPUs. For that, the driver checks if the ACPI IDs count is different from the APIC ID count - which can happen if the user choose to use dom0_max_vcpu argument. In such a case a backup of the PM structure is used and uploaded to the hypervisor. [v1-v2: Initial RFC implementations that were posted] [v3: Changed the name to passthru suggested by Pasi Kärkkäinen <pasik@iki.fi>] [v4: Added vCPU != pCPU support - aka dom0_max_vcpus support] [v5: Cleaned up the driver, fix bug under Athlon XP] [v6: Changed the driver to a CPU frequency governor] [v7: Jan Beulich <jbeulich@suse.com> suggestion to make it a cpufreq scaling driver made me rework it as driver that inhibits cpufreq scaling driver] [v8: Per Jan's review comments, fixed up the driver] [v9: Allow to continue even if acpi_processor_preregister_perf.. fails] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
2012-02-03 22:03:20 +01:00
/*
* Copyright 2012 by Oracle Inc
* Author: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
*
* This code borrows ideas from https://lkml.org/lkml/2011/11/30/249
* so many thanks go to Kevin Tian <kevin.tian@intel.com>
* and Yu Ke <ke.yu@intel.com>.
*
* This program is free software; you can redistribute it and/or modify it
* under the terms and conditions of the GNU General Public License,
* version 2, as published by the Free Software Foundation.
*
* This program is distributed in the hope it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
* more details.
*
*/
#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
xen/acpi-processor: C and P-state driver that uploads said data to hypervisor. This driver solves three problems: 1). Parse and upload ACPI0007 (or PROCESSOR_TYPE) information to the hypervisor - aka P-states (cpufreq data). 2). Upload the the Cx state information (cpuidle data). 3). Inhibit CPU frequency scaling drivers from loading. The reason for wanting to solve 1) and 2) is such that the Xen hypervisor is the only one that knows the CPU usage of different guests and can make the proper decision of when to put CPUs and packages in proper states. Unfortunately the hypervisor has no support to parse ACPI DSDT tables, hence it needs help from the initial domain to provide this information. The reason for 3) is that we do not want the initial domain to change P-states while the hypervisor is doing it as well - it causes rather some funny cases of P-states transitions. For this to work, the driver parses the Power Management data and uploads said information to the Xen hypervisor. It also calls acpi_processor_notify_smm() to inhibit the other CPU frequency scaling drivers from being loaded. Everything revolves around the 'struct acpi_processor' structure which gets updated during the bootup cycle in different stages. At the startup, when the ACPI parser starts, the C-state information is processed (processor_idle) and saved in said structure as 'power' element. Later on, the CPU frequency scaling driver (powernow-k8 or acpi_cpufreq), would call the the acpi_processor_* (processor_perflib functions) to parse P-states information and populate in the said structure the 'performance' element. Since we do not want the CPU frequency scaling drivers from loading we have to call the acpi_processor_* functions to parse the P-states and call "acpi_processor_notify_smm" to stop them from loading. There is also one oddity in this driver which is that under Xen, the physical online CPU count can be different from the virtual online CPU count. Meaning that the macros 'for_[online|possible]_cpu' would process only up to virtual online CPU count. We on the other hand want to process the full amount of physical CPUs. For that, the driver checks if the ACPI IDs count is different from the APIC ID count - which can happen if the user choose to use dom0_max_vcpu argument. In such a case a backup of the PM structure is used and uploaded to the hypervisor. [v1-v2: Initial RFC implementations that were posted] [v3: Changed the name to passthru suggested by Pasi Kärkkäinen <pasik@iki.fi>] [v4: Added vCPU != pCPU support - aka dom0_max_vcpus support] [v5: Cleaned up the driver, fix bug under Athlon XP] [v6: Changed the driver to a CPU frequency governor] [v7: Jan Beulich <jbeulich@suse.com> suggestion to make it a cpufreq scaling driver made me rework it as driver that inhibits cpufreq scaling driver] [v8: Per Jan's review comments, fixed up the driver] [v9: Allow to continue even if acpi_processor_preregister_perf.. fails] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
2012-02-03 22:03:20 +01:00
#include <linux/cpumask.h>
#include <linux/cpufreq.h>
#include <linux/freezer.h>
#include <linux/kernel.h>
#include <linux/kthread.h>
#include <linux/init.h>
#include <linux/module.h>
#include <linux/types.h>
#include <linux/syscore_ops.h>
xen/acpi-processor: C and P-state driver that uploads said data to hypervisor. This driver solves three problems: 1). Parse and upload ACPI0007 (or PROCESSOR_TYPE) information to the hypervisor - aka P-states (cpufreq data). 2). Upload the the Cx state information (cpuidle data). 3). Inhibit CPU frequency scaling drivers from loading. The reason for wanting to solve 1) and 2) is such that the Xen hypervisor is the only one that knows the CPU usage of different guests and can make the proper decision of when to put CPUs and packages in proper states. Unfortunately the hypervisor has no support to parse ACPI DSDT tables, hence it needs help from the initial domain to provide this information. The reason for 3) is that we do not want the initial domain to change P-states while the hypervisor is doing it as well - it causes rather some funny cases of P-states transitions. For this to work, the driver parses the Power Management data and uploads said information to the Xen hypervisor. It also calls acpi_processor_notify_smm() to inhibit the other CPU frequency scaling drivers from being loaded. Everything revolves around the 'struct acpi_processor' structure which gets updated during the bootup cycle in different stages. At the startup, when the ACPI parser starts, the C-state information is processed (processor_idle) and saved in said structure as 'power' element. Later on, the CPU frequency scaling driver (powernow-k8 or acpi_cpufreq), would call the the acpi_processor_* (processor_perflib functions) to parse P-states information and populate in the said structure the 'performance' element. Since we do not want the CPU frequency scaling drivers from loading we have to call the acpi_processor_* functions to parse the P-states and call "acpi_processor_notify_smm" to stop them from loading. There is also one oddity in this driver which is that under Xen, the physical online CPU count can be different from the virtual online CPU count. Meaning that the macros 'for_[online|possible]_cpu' would process only up to virtual online CPU count. We on the other hand want to process the full amount of physical CPUs. For that, the driver checks if the ACPI IDs count is different from the APIC ID count - which can happen if the user choose to use dom0_max_vcpu argument. In such a case a backup of the PM structure is used and uploaded to the hypervisor. [v1-v2: Initial RFC implementations that were posted] [v3: Changed the name to passthru suggested by Pasi Kärkkäinen <pasik@iki.fi>] [v4: Added vCPU != pCPU support - aka dom0_max_vcpus support] [v5: Cleaned up the driver, fix bug under Athlon XP] [v6: Changed the driver to a CPU frequency governor] [v7: Jan Beulich <jbeulich@suse.com> suggestion to make it a cpufreq scaling driver made me rework it as driver that inhibits cpufreq scaling driver] [v8: Per Jan's review comments, fixed up the driver] [v9: Allow to continue even if acpi_processor_preregister_perf.. fails] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
2012-02-03 22:03:20 +01:00
#include <acpi/acpi_bus.h>
#include <acpi/acpi_drivers.h>
#include <acpi/processor.h>
#include <xen/xen.h>
xen/acpi-processor: C and P-state driver that uploads said data to hypervisor. This driver solves three problems: 1). Parse and upload ACPI0007 (or PROCESSOR_TYPE) information to the hypervisor - aka P-states (cpufreq data). 2). Upload the the Cx state information (cpuidle data). 3). Inhibit CPU frequency scaling drivers from loading. The reason for wanting to solve 1) and 2) is such that the Xen hypervisor is the only one that knows the CPU usage of different guests and can make the proper decision of when to put CPUs and packages in proper states. Unfortunately the hypervisor has no support to parse ACPI DSDT tables, hence it needs help from the initial domain to provide this information. The reason for 3) is that we do not want the initial domain to change P-states while the hypervisor is doing it as well - it causes rather some funny cases of P-states transitions. For this to work, the driver parses the Power Management data and uploads said information to the Xen hypervisor. It also calls acpi_processor_notify_smm() to inhibit the other CPU frequency scaling drivers from being loaded. Everything revolves around the 'struct acpi_processor' structure which gets updated during the bootup cycle in different stages. At the startup, when the ACPI parser starts, the C-state information is processed (processor_idle) and saved in said structure as 'power' element. Later on, the CPU frequency scaling driver (powernow-k8 or acpi_cpufreq), would call the the acpi_processor_* (processor_perflib functions) to parse P-states information and populate in the said structure the 'performance' element. Since we do not want the CPU frequency scaling drivers from loading we have to call the acpi_processor_* functions to parse the P-states and call "acpi_processor_notify_smm" to stop them from loading. There is also one oddity in this driver which is that under Xen, the physical online CPU count can be different from the virtual online CPU count. Meaning that the macros 'for_[online|possible]_cpu' would process only up to virtual online CPU count. We on the other hand want to process the full amount of physical CPUs. For that, the driver checks if the ACPI IDs count is different from the APIC ID count - which can happen if the user choose to use dom0_max_vcpu argument. In such a case a backup of the PM structure is used and uploaded to the hypervisor. [v1-v2: Initial RFC implementations that were posted] [v3: Changed the name to passthru suggested by Pasi Kärkkäinen <pasik@iki.fi>] [v4: Added vCPU != pCPU support - aka dom0_max_vcpus support] [v5: Cleaned up the driver, fix bug under Athlon XP] [v6: Changed the driver to a CPU frequency governor] [v7: Jan Beulich <jbeulich@suse.com> suggestion to make it a cpufreq scaling driver made me rework it as driver that inhibits cpufreq scaling driver] [v8: Per Jan's review comments, fixed up the driver] [v9: Allow to continue even if acpi_processor_preregister_perf.. fails] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
2012-02-03 22:03:20 +01:00
#include <xen/interface/platform.h>
#include <asm/xen/hypercall.h>
static int no_hypercall;
MODULE_PARM_DESC(off, "Inhibit the hypercall.");
module_param_named(off, no_hypercall, int, 0400);
/*
* Note: Do not convert the acpi_id* below to cpumask_var_t or use cpumask_bit
* - as those shrink to nr_cpu_bits (which is dependent on possible_cpu), which
* can be less than what we want to put in. Instead use the 'nr_acpi_bits'
* which is dynamically computed based on the MADT or x2APIC table.
*/
static unsigned int nr_acpi_bits;
/* Mutex to protect the acpi_ids_done - for CPU hotplug use. */
static DEFINE_MUTEX(acpi_ids_mutex);
/* Which ACPI ID we have processed from 'struct acpi_processor'. */
static unsigned long *acpi_ids_done;
/* Which ACPI ID exist in the SSDT/DSDT processor definitions. */
static unsigned long *acpi_id_present;
xen/acpi-processor: C and P-state driver that uploads said data to hypervisor. This driver solves three problems: 1). Parse and upload ACPI0007 (or PROCESSOR_TYPE) information to the hypervisor - aka P-states (cpufreq data). 2). Upload the the Cx state information (cpuidle data). 3). Inhibit CPU frequency scaling drivers from loading. The reason for wanting to solve 1) and 2) is such that the Xen hypervisor is the only one that knows the CPU usage of different guests and can make the proper decision of when to put CPUs and packages in proper states. Unfortunately the hypervisor has no support to parse ACPI DSDT tables, hence it needs help from the initial domain to provide this information. The reason for 3) is that we do not want the initial domain to change P-states while the hypervisor is doing it as well - it causes rather some funny cases of P-states transitions. For this to work, the driver parses the Power Management data and uploads said information to the Xen hypervisor. It also calls acpi_processor_notify_smm() to inhibit the other CPU frequency scaling drivers from being loaded. Everything revolves around the 'struct acpi_processor' structure which gets updated during the bootup cycle in different stages. At the startup, when the ACPI parser starts, the C-state information is processed (processor_idle) and saved in said structure as 'power' element. Later on, the CPU frequency scaling driver (powernow-k8 or acpi_cpufreq), would call the the acpi_processor_* (processor_perflib functions) to parse P-states information and populate in the said structure the 'performance' element. Since we do not want the CPU frequency scaling drivers from loading we have to call the acpi_processor_* functions to parse the P-states and call "acpi_processor_notify_smm" to stop them from loading. There is also one oddity in this driver which is that under Xen, the physical online CPU count can be different from the virtual online CPU count. Meaning that the macros 'for_[online|possible]_cpu' would process only up to virtual online CPU count. We on the other hand want to process the full amount of physical CPUs. For that, the driver checks if the ACPI IDs count is different from the APIC ID count - which can happen if the user choose to use dom0_max_vcpu argument. In such a case a backup of the PM structure is used and uploaded to the hypervisor. [v1-v2: Initial RFC implementations that were posted] [v3: Changed the name to passthru suggested by Pasi Kärkkäinen <pasik@iki.fi>] [v4: Added vCPU != pCPU support - aka dom0_max_vcpus support] [v5: Cleaned up the driver, fix bug under Athlon XP] [v6: Changed the driver to a CPU frequency governor] [v7: Jan Beulich <jbeulich@suse.com> suggestion to make it a cpufreq scaling driver made me rework it as driver that inhibits cpufreq scaling driver] [v8: Per Jan's review comments, fixed up the driver] [v9: Allow to continue even if acpi_processor_preregister_perf.. fails] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
2012-02-03 22:03:20 +01:00
/* And if there is an _CST definition (or a PBLK) for the ACPI IDs */
static unsigned long *acpi_id_cst_present;
xen/acpi-processor: C and P-state driver that uploads said data to hypervisor. This driver solves three problems: 1). Parse and upload ACPI0007 (or PROCESSOR_TYPE) information to the hypervisor - aka P-states (cpufreq data). 2). Upload the the Cx state information (cpuidle data). 3). Inhibit CPU frequency scaling drivers from loading. The reason for wanting to solve 1) and 2) is such that the Xen hypervisor is the only one that knows the CPU usage of different guests and can make the proper decision of when to put CPUs and packages in proper states. Unfortunately the hypervisor has no support to parse ACPI DSDT tables, hence it needs help from the initial domain to provide this information. The reason for 3) is that we do not want the initial domain to change P-states while the hypervisor is doing it as well - it causes rather some funny cases of P-states transitions. For this to work, the driver parses the Power Management data and uploads said information to the Xen hypervisor. It also calls acpi_processor_notify_smm() to inhibit the other CPU frequency scaling drivers from being loaded. Everything revolves around the 'struct acpi_processor' structure which gets updated during the bootup cycle in different stages. At the startup, when the ACPI parser starts, the C-state information is processed (processor_idle) and saved in said structure as 'power' element. Later on, the CPU frequency scaling driver (powernow-k8 or acpi_cpufreq), would call the the acpi_processor_* (processor_perflib functions) to parse P-states information and populate in the said structure the 'performance' element. Since we do not want the CPU frequency scaling drivers from loading we have to call the acpi_processor_* functions to parse the P-states and call "acpi_processor_notify_smm" to stop them from loading. There is also one oddity in this driver which is that under Xen, the physical online CPU count can be different from the virtual online CPU count. Meaning that the macros 'for_[online|possible]_cpu' would process only up to virtual online CPU count. We on the other hand want to process the full amount of physical CPUs. For that, the driver checks if the ACPI IDs count is different from the APIC ID count - which can happen if the user choose to use dom0_max_vcpu argument. In such a case a backup of the PM structure is used and uploaded to the hypervisor. [v1-v2: Initial RFC implementations that were posted] [v3: Changed the name to passthru suggested by Pasi Kärkkäinen <pasik@iki.fi>] [v4: Added vCPU != pCPU support - aka dom0_max_vcpus support] [v5: Cleaned up the driver, fix bug under Athlon XP] [v6: Changed the driver to a CPU frequency governor] [v7: Jan Beulich <jbeulich@suse.com> suggestion to make it a cpufreq scaling driver made me rework it as driver that inhibits cpufreq scaling driver] [v8: Per Jan's review comments, fixed up the driver] [v9: Allow to continue even if acpi_processor_preregister_perf.. fails] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
2012-02-03 22:03:20 +01:00
static int push_cxx_to_hypervisor(struct acpi_processor *_pr)
{
struct xen_platform_op op = {
.cmd = XENPF_set_processor_pminfo,
.interface_version = XENPF_INTERFACE_VERSION,
.u.set_pminfo.id = _pr->acpi_id,
.u.set_pminfo.type = XEN_PM_CX,
};
struct xen_processor_cx *dst_cx, *dst_cx_states = NULL;
struct acpi_processor_cx *cx;
unsigned int i, ok;
int ret = 0;
dst_cx_states = kcalloc(_pr->power.count,
sizeof(struct xen_processor_cx), GFP_KERNEL);
if (!dst_cx_states)
return -ENOMEM;
for (ok = 0, i = 1; i <= _pr->power.count; i++) {
cx = &_pr->power.states[i];
if (!cx->valid)
continue;
dst_cx = &(dst_cx_states[ok++]);
dst_cx->reg.space_id = ACPI_ADR_SPACE_SYSTEM_IO;
if (cx->entry_method == ACPI_CSTATE_SYSTEMIO) {
dst_cx->reg.bit_width = 8;
dst_cx->reg.bit_offset = 0;
dst_cx->reg.access_size = 1;
} else {
dst_cx->reg.space_id = ACPI_ADR_SPACE_FIXED_HARDWARE;
if (cx->entry_method == ACPI_CSTATE_FFH) {
/* NATIVE_CSTATE_BEYOND_HALT */
dst_cx->reg.bit_offset = 2;
dst_cx->reg.bit_width = 1; /* VENDOR_INTEL */
}
dst_cx->reg.access_size = 0;
}
dst_cx->reg.address = cx->address;
dst_cx->type = cx->type;
dst_cx->latency = cx->latency;
dst_cx->dpcnt = 0;
set_xen_guest_handle(dst_cx->dp, NULL);
}
if (!ok) {
pr_debug("No _Cx for ACPI CPU %u\n", _pr->acpi_id);
xen/acpi-processor: C and P-state driver that uploads said data to hypervisor. This driver solves three problems: 1). Parse and upload ACPI0007 (or PROCESSOR_TYPE) information to the hypervisor - aka P-states (cpufreq data). 2). Upload the the Cx state information (cpuidle data). 3). Inhibit CPU frequency scaling drivers from loading. The reason for wanting to solve 1) and 2) is such that the Xen hypervisor is the only one that knows the CPU usage of different guests and can make the proper decision of when to put CPUs and packages in proper states. Unfortunately the hypervisor has no support to parse ACPI DSDT tables, hence it needs help from the initial domain to provide this information. The reason for 3) is that we do not want the initial domain to change P-states while the hypervisor is doing it as well - it causes rather some funny cases of P-states transitions. For this to work, the driver parses the Power Management data and uploads said information to the Xen hypervisor. It also calls acpi_processor_notify_smm() to inhibit the other CPU frequency scaling drivers from being loaded. Everything revolves around the 'struct acpi_processor' structure which gets updated during the bootup cycle in different stages. At the startup, when the ACPI parser starts, the C-state information is processed (processor_idle) and saved in said structure as 'power' element. Later on, the CPU frequency scaling driver (powernow-k8 or acpi_cpufreq), would call the the acpi_processor_* (processor_perflib functions) to parse P-states information and populate in the said structure the 'performance' element. Since we do not want the CPU frequency scaling drivers from loading we have to call the acpi_processor_* functions to parse the P-states and call "acpi_processor_notify_smm" to stop them from loading. There is also one oddity in this driver which is that under Xen, the physical online CPU count can be different from the virtual online CPU count. Meaning that the macros 'for_[online|possible]_cpu' would process only up to virtual online CPU count. We on the other hand want to process the full amount of physical CPUs. For that, the driver checks if the ACPI IDs count is different from the APIC ID count - which can happen if the user choose to use dom0_max_vcpu argument. In such a case a backup of the PM structure is used and uploaded to the hypervisor. [v1-v2: Initial RFC implementations that were posted] [v3: Changed the name to passthru suggested by Pasi Kärkkäinen <pasik@iki.fi>] [v4: Added vCPU != pCPU support - aka dom0_max_vcpus support] [v5: Cleaned up the driver, fix bug under Athlon XP] [v6: Changed the driver to a CPU frequency governor] [v7: Jan Beulich <jbeulich@suse.com> suggestion to make it a cpufreq scaling driver made me rework it as driver that inhibits cpufreq scaling driver] [v8: Per Jan's review comments, fixed up the driver] [v9: Allow to continue even if acpi_processor_preregister_perf.. fails] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
2012-02-03 22:03:20 +01:00
kfree(dst_cx_states);
return -EINVAL;
}
op.u.set_pminfo.power.count = ok;
op.u.set_pminfo.power.flags.bm_control = _pr->flags.bm_control;
op.u.set_pminfo.power.flags.bm_check = _pr->flags.bm_check;
op.u.set_pminfo.power.flags.has_cst = _pr->flags.has_cst;
op.u.set_pminfo.power.flags.power_setup_done =
_pr->flags.power_setup_done;
set_xen_guest_handle(op.u.set_pminfo.power.states, dst_cx_states);
if (!no_hypercall)
ret = HYPERVISOR_dom0_op(&op);
if (!ret) {
pr_debug("ACPI CPU%u - C-states uploaded.\n", _pr->acpi_id);
for (i = 1; i <= _pr->power.count; i++) {
cx = &_pr->power.states[i];
if (!cx->valid)
continue;
pr_debug(" C%d: %s %d uS\n",
cx->type, cx->desc, (u32)cx->latency);
}
} else if (ret != -EINVAL)
/* EINVAL means the ACPI ID is incorrect - meaning the ACPI
* table is referencing a non-existing CPU - which can happen
* with broken ACPI tables. */
pr_err("(CX): Hypervisor error (%d) for ACPI CPU%u\n",
xen/acpi-processor: C and P-state driver that uploads said data to hypervisor. This driver solves three problems: 1). Parse and upload ACPI0007 (or PROCESSOR_TYPE) information to the hypervisor - aka P-states (cpufreq data). 2). Upload the the Cx state information (cpuidle data). 3). Inhibit CPU frequency scaling drivers from loading. The reason for wanting to solve 1) and 2) is such that the Xen hypervisor is the only one that knows the CPU usage of different guests and can make the proper decision of when to put CPUs and packages in proper states. Unfortunately the hypervisor has no support to parse ACPI DSDT tables, hence it needs help from the initial domain to provide this information. The reason for 3) is that we do not want the initial domain to change P-states while the hypervisor is doing it as well - it causes rather some funny cases of P-states transitions. For this to work, the driver parses the Power Management data and uploads said information to the Xen hypervisor. It also calls acpi_processor_notify_smm() to inhibit the other CPU frequency scaling drivers from being loaded. Everything revolves around the 'struct acpi_processor' structure which gets updated during the bootup cycle in different stages. At the startup, when the ACPI parser starts, the C-state information is processed (processor_idle) and saved in said structure as 'power' element. Later on, the CPU frequency scaling driver (powernow-k8 or acpi_cpufreq), would call the the acpi_processor_* (processor_perflib functions) to parse P-states information and populate in the said structure the 'performance' element. Since we do not want the CPU frequency scaling drivers from loading we have to call the acpi_processor_* functions to parse the P-states and call "acpi_processor_notify_smm" to stop them from loading. There is also one oddity in this driver which is that under Xen, the physical online CPU count can be different from the virtual online CPU count. Meaning that the macros 'for_[online|possible]_cpu' would process only up to virtual online CPU count. We on the other hand want to process the full amount of physical CPUs. For that, the driver checks if the ACPI IDs count is different from the APIC ID count - which can happen if the user choose to use dom0_max_vcpu argument. In such a case a backup of the PM structure is used and uploaded to the hypervisor. [v1-v2: Initial RFC implementations that were posted] [v3: Changed the name to passthru suggested by Pasi Kärkkäinen <pasik@iki.fi>] [v4: Added vCPU != pCPU support - aka dom0_max_vcpus support] [v5: Cleaned up the driver, fix bug under Athlon XP] [v6: Changed the driver to a CPU frequency governor] [v7: Jan Beulich <jbeulich@suse.com> suggestion to make it a cpufreq scaling driver made me rework it as driver that inhibits cpufreq scaling driver] [v8: Per Jan's review comments, fixed up the driver] [v9: Allow to continue even if acpi_processor_preregister_perf.. fails] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
2012-02-03 22:03:20 +01:00
ret, _pr->acpi_id);
kfree(dst_cx_states);
return ret;
}
static struct xen_processor_px *
xen_copy_pss_data(struct acpi_processor *_pr,
struct xen_processor_performance *dst_perf)
{
struct xen_processor_px *dst_states = NULL;
unsigned int i;
BUILD_BUG_ON(sizeof(struct xen_processor_px) !=
sizeof(struct acpi_processor_px));
dst_states = kcalloc(_pr->performance->state_count,
sizeof(struct xen_processor_px), GFP_KERNEL);
if (!dst_states)
return ERR_PTR(-ENOMEM);
dst_perf->state_count = _pr->performance->state_count;
for (i = 0; i < _pr->performance->state_count; i++) {
/* Fortunatly for us, they are both the same size */
memcpy(&(dst_states[i]), &(_pr->performance->states[i]),
sizeof(struct acpi_processor_px));
}
return dst_states;
}
static int xen_copy_psd_data(struct acpi_processor *_pr,
struct xen_processor_performance *dst)
{
struct acpi_psd_package *pdomain;
BUILD_BUG_ON(sizeof(struct xen_psd_package) !=
sizeof(struct acpi_psd_package));
/* This information is enumerated only if acpi_processor_preregister_performance
* has been called.
*/
dst->shared_type = _pr->performance->shared_type;
pdomain = &(_pr->performance->domain_info);
/* 'acpi_processor_preregister_performance' does not parse if the
* num_processors <= 1, but Xen still requires it. Do it manually here.
*/
if (pdomain->num_processors <= 1) {
if (pdomain->coord_type == DOMAIN_COORD_TYPE_SW_ALL)
dst->shared_type = CPUFREQ_SHARED_TYPE_ALL;
else if (pdomain->coord_type == DOMAIN_COORD_TYPE_HW_ALL)
dst->shared_type = CPUFREQ_SHARED_TYPE_HW;
else if (pdomain->coord_type == DOMAIN_COORD_TYPE_SW_ANY)
dst->shared_type = CPUFREQ_SHARED_TYPE_ANY;
}
memcpy(&(dst->domain_info), pdomain, sizeof(struct acpi_psd_package));
return 0;
}
static int xen_copy_pct_data(struct acpi_pct_register *pct,
struct xen_pct_register *dst_pct)
{
/* It would be nice if you could just do 'memcpy(pct, dst_pct') but
* sadly the Xen structure did not have the proper padding so the
* descriptor field takes two (dst_pct) bytes instead of one (pct).
*/
dst_pct->descriptor = pct->descriptor;
dst_pct->length = pct->length;
dst_pct->space_id = pct->space_id;
dst_pct->bit_width = pct->bit_width;
dst_pct->bit_offset = pct->bit_offset;
dst_pct->reserved = pct->reserved;
dst_pct->address = pct->address;
return 0;
}
static int push_pxx_to_hypervisor(struct acpi_processor *_pr)
{
int ret = 0;
struct xen_platform_op op = {
.cmd = XENPF_set_processor_pminfo,
.interface_version = XENPF_INTERFACE_VERSION,
.u.set_pminfo.id = _pr->acpi_id,
.u.set_pminfo.type = XEN_PM_PX,
};
struct xen_processor_performance *dst_perf;
struct xen_processor_px *dst_states = NULL;
dst_perf = &op.u.set_pminfo.perf;
dst_perf->platform_limit = _pr->performance_platform_limit;
dst_perf->flags |= XEN_PX_PPC;
xen_copy_pct_data(&(_pr->performance->control_register),
&dst_perf->control_register);
xen_copy_pct_data(&(_pr->performance->status_register),
&dst_perf->status_register);
dst_perf->flags |= XEN_PX_PCT;
dst_states = xen_copy_pss_data(_pr, dst_perf);
if (!IS_ERR_OR_NULL(dst_states)) {
set_xen_guest_handle(dst_perf->states, dst_states);
dst_perf->flags |= XEN_PX_PSS;
}
if (!xen_copy_psd_data(_pr, dst_perf))
dst_perf->flags |= XEN_PX_PSD;
if (dst_perf->flags != (XEN_PX_PSD | XEN_PX_PSS | XEN_PX_PCT | XEN_PX_PPC)) {
pr_warn("ACPI CPU%u missing some P-state data (%x), skipping\n",
xen/acpi-processor: C and P-state driver that uploads said data to hypervisor. This driver solves three problems: 1). Parse and upload ACPI0007 (or PROCESSOR_TYPE) information to the hypervisor - aka P-states (cpufreq data). 2). Upload the the Cx state information (cpuidle data). 3). Inhibit CPU frequency scaling drivers from loading. The reason for wanting to solve 1) and 2) is such that the Xen hypervisor is the only one that knows the CPU usage of different guests and can make the proper decision of when to put CPUs and packages in proper states. Unfortunately the hypervisor has no support to parse ACPI DSDT tables, hence it needs help from the initial domain to provide this information. The reason for 3) is that we do not want the initial domain to change P-states while the hypervisor is doing it as well - it causes rather some funny cases of P-states transitions. For this to work, the driver parses the Power Management data and uploads said information to the Xen hypervisor. It also calls acpi_processor_notify_smm() to inhibit the other CPU frequency scaling drivers from being loaded. Everything revolves around the 'struct acpi_processor' structure which gets updated during the bootup cycle in different stages. At the startup, when the ACPI parser starts, the C-state information is processed (processor_idle) and saved in said structure as 'power' element. Later on, the CPU frequency scaling driver (powernow-k8 or acpi_cpufreq), would call the the acpi_processor_* (processor_perflib functions) to parse P-states information and populate in the said structure the 'performance' element. Since we do not want the CPU frequency scaling drivers from loading we have to call the acpi_processor_* functions to parse the P-states and call "acpi_processor_notify_smm" to stop them from loading. There is also one oddity in this driver which is that under Xen, the physical online CPU count can be different from the virtual online CPU count. Meaning that the macros 'for_[online|possible]_cpu' would process only up to virtual online CPU count. We on the other hand want to process the full amount of physical CPUs. For that, the driver checks if the ACPI IDs count is different from the APIC ID count - which can happen if the user choose to use dom0_max_vcpu argument. In such a case a backup of the PM structure is used and uploaded to the hypervisor. [v1-v2: Initial RFC implementations that were posted] [v3: Changed the name to passthru suggested by Pasi Kärkkäinen <pasik@iki.fi>] [v4: Added vCPU != pCPU support - aka dom0_max_vcpus support] [v5: Cleaned up the driver, fix bug under Athlon XP] [v6: Changed the driver to a CPU frequency governor] [v7: Jan Beulich <jbeulich@suse.com> suggestion to make it a cpufreq scaling driver made me rework it as driver that inhibits cpufreq scaling driver] [v8: Per Jan's review comments, fixed up the driver] [v9: Allow to continue even if acpi_processor_preregister_perf.. fails] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
2012-02-03 22:03:20 +01:00
_pr->acpi_id, dst_perf->flags);
ret = -ENODEV;
goto err_free;
}
if (!no_hypercall)
ret = HYPERVISOR_dom0_op(&op);
if (!ret) {
struct acpi_processor_performance *perf;
unsigned int i;
perf = _pr->performance;
pr_debug("ACPI CPU%u - P-states uploaded.\n", _pr->acpi_id);
for (i = 0; i < perf->state_count; i++) {
pr_debug(" %cP%d: %d MHz, %d mW, %d uS\n",
(i == perf->state ? '*' : ' '), i,
(u32) perf->states[i].core_frequency,
(u32) perf->states[i].power,
(u32) perf->states[i].transition_latency);
}
} else if (ret != -EINVAL)
/* EINVAL means the ACPI ID is incorrect - meaning the ACPI
* table is referencing a non-existing CPU - which can happen
* with broken ACPI tables. */
pr_warn("(_PXX): Hypervisor error (%d) for ACPI CPU%u\n",
ret, _pr->acpi_id);
xen/acpi-processor: C and P-state driver that uploads said data to hypervisor. This driver solves three problems: 1). Parse and upload ACPI0007 (or PROCESSOR_TYPE) information to the hypervisor - aka P-states (cpufreq data). 2). Upload the the Cx state information (cpuidle data). 3). Inhibit CPU frequency scaling drivers from loading. The reason for wanting to solve 1) and 2) is such that the Xen hypervisor is the only one that knows the CPU usage of different guests and can make the proper decision of when to put CPUs and packages in proper states. Unfortunately the hypervisor has no support to parse ACPI DSDT tables, hence it needs help from the initial domain to provide this information. The reason for 3) is that we do not want the initial domain to change P-states while the hypervisor is doing it as well - it causes rather some funny cases of P-states transitions. For this to work, the driver parses the Power Management data and uploads said information to the Xen hypervisor. It also calls acpi_processor_notify_smm() to inhibit the other CPU frequency scaling drivers from being loaded. Everything revolves around the 'struct acpi_processor' structure which gets updated during the bootup cycle in different stages. At the startup, when the ACPI parser starts, the C-state information is processed (processor_idle) and saved in said structure as 'power' element. Later on, the CPU frequency scaling driver (powernow-k8 or acpi_cpufreq), would call the the acpi_processor_* (processor_perflib functions) to parse P-states information and populate in the said structure the 'performance' element. Since we do not want the CPU frequency scaling drivers from loading we have to call the acpi_processor_* functions to parse the P-states and call "acpi_processor_notify_smm" to stop them from loading. There is also one oddity in this driver which is that under Xen, the physical online CPU count can be different from the virtual online CPU count. Meaning that the macros 'for_[online|possible]_cpu' would process only up to virtual online CPU count. We on the other hand want to process the full amount of physical CPUs. For that, the driver checks if the ACPI IDs count is different from the APIC ID count - which can happen if the user choose to use dom0_max_vcpu argument. In such a case a backup of the PM structure is used and uploaded to the hypervisor. [v1-v2: Initial RFC implementations that were posted] [v3: Changed the name to passthru suggested by Pasi Kärkkäinen <pasik@iki.fi>] [v4: Added vCPU != pCPU support - aka dom0_max_vcpus support] [v5: Cleaned up the driver, fix bug under Athlon XP] [v6: Changed the driver to a CPU frequency governor] [v7: Jan Beulich <jbeulich@suse.com> suggestion to make it a cpufreq scaling driver made me rework it as driver that inhibits cpufreq scaling driver] [v8: Per Jan's review comments, fixed up the driver] [v9: Allow to continue even if acpi_processor_preregister_perf.. fails] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
2012-02-03 22:03:20 +01:00
err_free:
if (!IS_ERR_OR_NULL(dst_states))
kfree(dst_states);
return ret;
}
static int upload_pm_data(struct acpi_processor *_pr)
{
int err = 0;
mutex_lock(&acpi_ids_mutex);
if (__test_and_set_bit(_pr->acpi_id, acpi_ids_done)) {
mutex_unlock(&acpi_ids_mutex);
return -EBUSY;
}
if (_pr->flags.power)
err = push_cxx_to_hypervisor(_pr);
if (_pr->performance && _pr->performance->states)
err |= push_pxx_to_hypervisor(_pr);
mutex_unlock(&acpi_ids_mutex);
return err;
}
static unsigned int __init get_max_acpi_id(void)
{
struct xenpf_pcpuinfo *info;
struct xen_platform_op op = {
.cmd = XENPF_get_cpuinfo,
.interface_version = XENPF_INTERFACE_VERSION,
};
int ret = 0;
unsigned int i, last_cpu, max_acpi_id = 0;
info = &op.u.pcpu_info;
info->xen_cpuid = 0;
ret = HYPERVISOR_dom0_op(&op);
if (ret)
return NR_CPUS;
/* The max_present is the same irregardless of the xen_cpuid */
last_cpu = op.u.pcpu_info.max_present;
for (i = 0; i <= last_cpu; i++) {
info->xen_cpuid = i;
ret = HYPERVISOR_dom0_op(&op);
if (ret)
continue;
max_acpi_id = max(info->acpi_id, max_acpi_id);
}
max_acpi_id *= 2; /* Slack for CPU hotplug support. */
pr_debug("Max ACPI ID: %u\n", max_acpi_id);
xen/acpi-processor: C and P-state driver that uploads said data to hypervisor. This driver solves three problems: 1). Parse and upload ACPI0007 (or PROCESSOR_TYPE) information to the hypervisor - aka P-states (cpufreq data). 2). Upload the the Cx state information (cpuidle data). 3). Inhibit CPU frequency scaling drivers from loading. The reason for wanting to solve 1) and 2) is such that the Xen hypervisor is the only one that knows the CPU usage of different guests and can make the proper decision of when to put CPUs and packages in proper states. Unfortunately the hypervisor has no support to parse ACPI DSDT tables, hence it needs help from the initial domain to provide this information. The reason for 3) is that we do not want the initial domain to change P-states while the hypervisor is doing it as well - it causes rather some funny cases of P-states transitions. For this to work, the driver parses the Power Management data and uploads said information to the Xen hypervisor. It also calls acpi_processor_notify_smm() to inhibit the other CPU frequency scaling drivers from being loaded. Everything revolves around the 'struct acpi_processor' structure which gets updated during the bootup cycle in different stages. At the startup, when the ACPI parser starts, the C-state information is processed (processor_idle) and saved in said structure as 'power' element. Later on, the CPU frequency scaling driver (powernow-k8 or acpi_cpufreq), would call the the acpi_processor_* (processor_perflib functions) to parse P-states information and populate in the said structure the 'performance' element. Since we do not want the CPU frequency scaling drivers from loading we have to call the acpi_processor_* functions to parse the P-states and call "acpi_processor_notify_smm" to stop them from loading. There is also one oddity in this driver which is that under Xen, the physical online CPU count can be different from the virtual online CPU count. Meaning that the macros 'for_[online|possible]_cpu' would process only up to virtual online CPU count. We on the other hand want to process the full amount of physical CPUs. For that, the driver checks if the ACPI IDs count is different from the APIC ID count - which can happen if the user choose to use dom0_max_vcpu argument. In such a case a backup of the PM structure is used and uploaded to the hypervisor. [v1-v2: Initial RFC implementations that were posted] [v3: Changed the name to passthru suggested by Pasi Kärkkäinen <pasik@iki.fi>] [v4: Added vCPU != pCPU support - aka dom0_max_vcpus support] [v5: Cleaned up the driver, fix bug under Athlon XP] [v6: Changed the driver to a CPU frequency governor] [v7: Jan Beulich <jbeulich@suse.com> suggestion to make it a cpufreq scaling driver made me rework it as driver that inhibits cpufreq scaling driver] [v8: Per Jan's review comments, fixed up the driver] [v9: Allow to continue even if acpi_processor_preregister_perf.. fails] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
2012-02-03 22:03:20 +01:00
return max_acpi_id;
}
/*
* The read_acpi_id and check_acpi_ids are there to support the Xen
* oddity of virtual CPUs != physical CPUs in the initial domain.
* The user can supply 'xen_max_vcpus=X' on the Xen hypervisor line
* which will band the amount of CPUs the initial domain can see.
* In general that is OK, except it plays havoc with any of the
* for_each_[present|online]_cpu macros which are banded to the virtual
* CPU amount.
*/
static acpi_status
xen/acpi-processor: C and P-state driver that uploads said data to hypervisor. This driver solves three problems: 1). Parse and upload ACPI0007 (or PROCESSOR_TYPE) information to the hypervisor - aka P-states (cpufreq data). 2). Upload the the Cx state information (cpuidle data). 3). Inhibit CPU frequency scaling drivers from loading. The reason for wanting to solve 1) and 2) is such that the Xen hypervisor is the only one that knows the CPU usage of different guests and can make the proper decision of when to put CPUs and packages in proper states. Unfortunately the hypervisor has no support to parse ACPI DSDT tables, hence it needs help from the initial domain to provide this information. The reason for 3) is that we do not want the initial domain to change P-states while the hypervisor is doing it as well - it causes rather some funny cases of P-states transitions. For this to work, the driver parses the Power Management data and uploads said information to the Xen hypervisor. It also calls acpi_processor_notify_smm() to inhibit the other CPU frequency scaling drivers from being loaded. Everything revolves around the 'struct acpi_processor' structure which gets updated during the bootup cycle in different stages. At the startup, when the ACPI parser starts, the C-state information is processed (processor_idle) and saved in said structure as 'power' element. Later on, the CPU frequency scaling driver (powernow-k8 or acpi_cpufreq), would call the the acpi_processor_* (processor_perflib functions) to parse P-states information and populate in the said structure the 'performance' element. Since we do not want the CPU frequency scaling drivers from loading we have to call the acpi_processor_* functions to parse the P-states and call "acpi_processor_notify_smm" to stop them from loading. There is also one oddity in this driver which is that under Xen, the physical online CPU count can be different from the virtual online CPU count. Meaning that the macros 'for_[online|possible]_cpu' would process only up to virtual online CPU count. We on the other hand want to process the full amount of physical CPUs. For that, the driver checks if the ACPI IDs count is different from the APIC ID count - which can happen if the user choose to use dom0_max_vcpu argument. In such a case a backup of the PM structure is used and uploaded to the hypervisor. [v1-v2: Initial RFC implementations that were posted] [v3: Changed the name to passthru suggested by Pasi Kärkkäinen <pasik@iki.fi>] [v4: Added vCPU != pCPU support - aka dom0_max_vcpus support] [v5: Cleaned up the driver, fix bug under Athlon XP] [v6: Changed the driver to a CPU frequency governor] [v7: Jan Beulich <jbeulich@suse.com> suggestion to make it a cpufreq scaling driver made me rework it as driver that inhibits cpufreq scaling driver] [v8: Per Jan's review comments, fixed up the driver] [v9: Allow to continue even if acpi_processor_preregister_perf.. fails] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
2012-02-03 22:03:20 +01:00
read_acpi_id(acpi_handle handle, u32 lvl, void *context, void **rv)
{
u32 acpi_id;
acpi_status status;
acpi_object_type acpi_type;
unsigned long long tmp;
union acpi_object object = { 0 };
struct acpi_buffer buffer = { sizeof(union acpi_object), &object };
acpi_io_address pblk = 0;
status = acpi_get_type(handle, &acpi_type);
if (ACPI_FAILURE(status))
return AE_OK;
switch (acpi_type) {
case ACPI_TYPE_PROCESSOR:
status = acpi_evaluate_object(handle, NULL, NULL, &buffer);
if (ACPI_FAILURE(status))
return AE_OK;
acpi_id = object.processor.proc_id;
pblk = object.processor.pblk_address;
break;
case ACPI_TYPE_DEVICE:
status = acpi_evaluate_integer(handle, "_UID", NULL, &tmp);
if (ACPI_FAILURE(status))
return AE_OK;
acpi_id = tmp;
break;
default:
return AE_OK;
}
/* There are more ACPI Processor objects than in x2APIC or MADT.
* This can happen with incorrect ACPI SSDT declerations. */
if (acpi_id > nr_acpi_bits) {
pr_debug("We only have %u, trying to set %u\n",
xen/acpi-processor: C and P-state driver that uploads said data to hypervisor. This driver solves three problems: 1). Parse and upload ACPI0007 (or PROCESSOR_TYPE) information to the hypervisor - aka P-states (cpufreq data). 2). Upload the the Cx state information (cpuidle data). 3). Inhibit CPU frequency scaling drivers from loading. The reason for wanting to solve 1) and 2) is such that the Xen hypervisor is the only one that knows the CPU usage of different guests and can make the proper decision of when to put CPUs and packages in proper states. Unfortunately the hypervisor has no support to parse ACPI DSDT tables, hence it needs help from the initial domain to provide this information. The reason for 3) is that we do not want the initial domain to change P-states while the hypervisor is doing it as well - it causes rather some funny cases of P-states transitions. For this to work, the driver parses the Power Management data and uploads said information to the Xen hypervisor. It also calls acpi_processor_notify_smm() to inhibit the other CPU frequency scaling drivers from being loaded. Everything revolves around the 'struct acpi_processor' structure which gets updated during the bootup cycle in different stages. At the startup, when the ACPI parser starts, the C-state information is processed (processor_idle) and saved in said structure as 'power' element. Later on, the CPU frequency scaling driver (powernow-k8 or acpi_cpufreq), would call the the acpi_processor_* (processor_perflib functions) to parse P-states information and populate in the said structure the 'performance' element. Since we do not want the CPU frequency scaling drivers from loading we have to call the acpi_processor_* functions to parse the P-states and call "acpi_processor_notify_smm" to stop them from loading. There is also one oddity in this driver which is that under Xen, the physical online CPU count can be different from the virtual online CPU count. Meaning that the macros 'for_[online|possible]_cpu' would process only up to virtual online CPU count. We on the other hand want to process the full amount of physical CPUs. For that, the driver checks if the ACPI IDs count is different from the APIC ID count - which can happen if the user choose to use dom0_max_vcpu argument. In such a case a backup of the PM structure is used and uploaded to the hypervisor. [v1-v2: Initial RFC implementations that were posted] [v3: Changed the name to passthru suggested by Pasi Kärkkäinen <pasik@iki.fi>] [v4: Added vCPU != pCPU support - aka dom0_max_vcpus support] [v5: Cleaned up the driver, fix bug under Athlon XP] [v6: Changed the driver to a CPU frequency governor] [v7: Jan Beulich <jbeulich@suse.com> suggestion to make it a cpufreq scaling driver made me rework it as driver that inhibits cpufreq scaling driver] [v8: Per Jan's review comments, fixed up the driver] [v9: Allow to continue even if acpi_processor_preregister_perf.. fails] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
2012-02-03 22:03:20 +01:00
nr_acpi_bits, acpi_id);
return AE_OK;
}
/* OK, There is a ACPI Processor object */
__set_bit(acpi_id, acpi_id_present);
pr_debug("ACPI CPU%u w/ PBLK:0x%lx\n", acpi_id, (unsigned long)pblk);
xen/acpi-processor: C and P-state driver that uploads said data to hypervisor. This driver solves three problems: 1). Parse and upload ACPI0007 (or PROCESSOR_TYPE) information to the hypervisor - aka P-states (cpufreq data). 2). Upload the the Cx state information (cpuidle data). 3). Inhibit CPU frequency scaling drivers from loading. The reason for wanting to solve 1) and 2) is such that the Xen hypervisor is the only one that knows the CPU usage of different guests and can make the proper decision of when to put CPUs and packages in proper states. Unfortunately the hypervisor has no support to parse ACPI DSDT tables, hence it needs help from the initial domain to provide this information. The reason for 3) is that we do not want the initial domain to change P-states while the hypervisor is doing it as well - it causes rather some funny cases of P-states transitions. For this to work, the driver parses the Power Management data and uploads said information to the Xen hypervisor. It also calls acpi_processor_notify_smm() to inhibit the other CPU frequency scaling drivers from being loaded. Everything revolves around the 'struct acpi_processor' structure which gets updated during the bootup cycle in different stages. At the startup, when the ACPI parser starts, the C-state information is processed (processor_idle) and saved in said structure as 'power' element. Later on, the CPU frequency scaling driver (powernow-k8 or acpi_cpufreq), would call the the acpi_processor_* (processor_perflib functions) to parse P-states information and populate in the said structure the 'performance' element. Since we do not want the CPU frequency scaling drivers from loading we have to call the acpi_processor_* functions to parse the P-states and call "acpi_processor_notify_smm" to stop them from loading. There is also one oddity in this driver which is that under Xen, the physical online CPU count can be different from the virtual online CPU count. Meaning that the macros 'for_[online|possible]_cpu' would process only up to virtual online CPU count. We on the other hand want to process the full amount of physical CPUs. For that, the driver checks if the ACPI IDs count is different from the APIC ID count - which can happen if the user choose to use dom0_max_vcpu argument. In such a case a backup of the PM structure is used and uploaded to the hypervisor. [v1-v2: Initial RFC implementations that were posted] [v3: Changed the name to passthru suggested by Pasi Kärkkäinen <pasik@iki.fi>] [v4: Added vCPU != pCPU support - aka dom0_max_vcpus support] [v5: Cleaned up the driver, fix bug under Athlon XP] [v6: Changed the driver to a CPU frequency governor] [v7: Jan Beulich <jbeulich@suse.com> suggestion to make it a cpufreq scaling driver made me rework it as driver that inhibits cpufreq scaling driver] [v8: Per Jan's review comments, fixed up the driver] [v9: Allow to continue even if acpi_processor_preregister_perf.. fails] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
2012-02-03 22:03:20 +01:00
status = acpi_evaluate_object(handle, "_CST", NULL, &buffer);
if (ACPI_FAILURE(status)) {
if (!pblk)
return AE_OK;
}
/* .. and it has a C-state */
__set_bit(acpi_id, acpi_id_cst_present);
return AE_OK;
}
static int check_acpi_ids(struct acpi_processor *pr_backup)
xen/acpi-processor: C and P-state driver that uploads said data to hypervisor. This driver solves three problems: 1). Parse and upload ACPI0007 (or PROCESSOR_TYPE) information to the hypervisor - aka P-states (cpufreq data). 2). Upload the the Cx state information (cpuidle data). 3). Inhibit CPU frequency scaling drivers from loading. The reason for wanting to solve 1) and 2) is such that the Xen hypervisor is the only one that knows the CPU usage of different guests and can make the proper decision of when to put CPUs and packages in proper states. Unfortunately the hypervisor has no support to parse ACPI DSDT tables, hence it needs help from the initial domain to provide this information. The reason for 3) is that we do not want the initial domain to change P-states while the hypervisor is doing it as well - it causes rather some funny cases of P-states transitions. For this to work, the driver parses the Power Management data and uploads said information to the Xen hypervisor. It also calls acpi_processor_notify_smm() to inhibit the other CPU frequency scaling drivers from being loaded. Everything revolves around the 'struct acpi_processor' structure which gets updated during the bootup cycle in different stages. At the startup, when the ACPI parser starts, the C-state information is processed (processor_idle) and saved in said structure as 'power' element. Later on, the CPU frequency scaling driver (powernow-k8 or acpi_cpufreq), would call the the acpi_processor_* (processor_perflib functions) to parse P-states information and populate in the said structure the 'performance' element. Since we do not want the CPU frequency scaling drivers from loading we have to call the acpi_processor_* functions to parse the P-states and call "acpi_processor_notify_smm" to stop them from loading. There is also one oddity in this driver which is that under Xen, the physical online CPU count can be different from the virtual online CPU count. Meaning that the macros 'for_[online|possible]_cpu' would process only up to virtual online CPU count. We on the other hand want to process the full amount of physical CPUs. For that, the driver checks if the ACPI IDs count is different from the APIC ID count - which can happen if the user choose to use dom0_max_vcpu argument. In such a case a backup of the PM structure is used and uploaded to the hypervisor. [v1-v2: Initial RFC implementations that were posted] [v3: Changed the name to passthru suggested by Pasi Kärkkäinen <pasik@iki.fi>] [v4: Added vCPU != pCPU support - aka dom0_max_vcpus support] [v5: Cleaned up the driver, fix bug under Athlon XP] [v6: Changed the driver to a CPU frequency governor] [v7: Jan Beulich <jbeulich@suse.com> suggestion to make it a cpufreq scaling driver made me rework it as driver that inhibits cpufreq scaling driver] [v8: Per Jan's review comments, fixed up the driver] [v9: Allow to continue even if acpi_processor_preregister_perf.. fails] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
2012-02-03 22:03:20 +01:00
{
if (!pr_backup)
return -ENODEV;
if (acpi_id_present && acpi_id_cst_present)
/* OK, done this once .. skip to uploading */
goto upload;
xen/acpi-processor: C and P-state driver that uploads said data to hypervisor. This driver solves three problems: 1). Parse and upload ACPI0007 (or PROCESSOR_TYPE) information to the hypervisor - aka P-states (cpufreq data). 2). Upload the the Cx state information (cpuidle data). 3). Inhibit CPU frequency scaling drivers from loading. The reason for wanting to solve 1) and 2) is such that the Xen hypervisor is the only one that knows the CPU usage of different guests and can make the proper decision of when to put CPUs and packages in proper states. Unfortunately the hypervisor has no support to parse ACPI DSDT tables, hence it needs help from the initial domain to provide this information. The reason for 3) is that we do not want the initial domain to change P-states while the hypervisor is doing it as well - it causes rather some funny cases of P-states transitions. For this to work, the driver parses the Power Management data and uploads said information to the Xen hypervisor. It also calls acpi_processor_notify_smm() to inhibit the other CPU frequency scaling drivers from being loaded. Everything revolves around the 'struct acpi_processor' structure which gets updated during the bootup cycle in different stages. At the startup, when the ACPI parser starts, the C-state information is processed (processor_idle) and saved in said structure as 'power' element. Later on, the CPU frequency scaling driver (powernow-k8 or acpi_cpufreq), would call the the acpi_processor_* (processor_perflib functions) to parse P-states information and populate in the said structure the 'performance' element. Since we do not want the CPU frequency scaling drivers from loading we have to call the acpi_processor_* functions to parse the P-states and call "acpi_processor_notify_smm" to stop them from loading. There is also one oddity in this driver which is that under Xen, the physical online CPU count can be different from the virtual online CPU count. Meaning that the macros 'for_[online|possible]_cpu' would process only up to virtual online CPU count. We on the other hand want to process the full amount of physical CPUs. For that, the driver checks if the ACPI IDs count is different from the APIC ID count - which can happen if the user choose to use dom0_max_vcpu argument. In such a case a backup of the PM structure is used and uploaded to the hypervisor. [v1-v2: Initial RFC implementations that were posted] [v3: Changed the name to passthru suggested by Pasi Kärkkäinen <pasik@iki.fi>] [v4: Added vCPU != pCPU support - aka dom0_max_vcpus support] [v5: Cleaned up the driver, fix bug under Athlon XP] [v6: Changed the driver to a CPU frequency governor] [v7: Jan Beulich <jbeulich@suse.com> suggestion to make it a cpufreq scaling driver made me rework it as driver that inhibits cpufreq scaling driver] [v8: Per Jan's review comments, fixed up the driver] [v9: Allow to continue even if acpi_processor_preregister_perf.. fails] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
2012-02-03 22:03:20 +01:00
/* All online CPUs have been processed at this stage. Now verify
* whether in fact "online CPUs" == physical CPUs.
*/
acpi_id_present = kcalloc(BITS_TO_LONGS(nr_acpi_bits), sizeof(unsigned long), GFP_KERNEL);
if (!acpi_id_present)
return -ENOMEM;
acpi_id_cst_present = kcalloc(BITS_TO_LONGS(nr_acpi_bits), sizeof(unsigned long), GFP_KERNEL);
if (!acpi_id_cst_present) {
kfree(acpi_id_present);
return -ENOMEM;
}
acpi_walk_namespace(ACPI_TYPE_PROCESSOR, ACPI_ROOT_OBJECT,
ACPI_UINT32_MAX,
read_acpi_id, NULL, NULL, NULL);
acpi_get_devices("ACPI0007", read_acpi_id, NULL, NULL);
upload:
xen/acpi-processor: C and P-state driver that uploads said data to hypervisor. This driver solves three problems: 1). Parse and upload ACPI0007 (or PROCESSOR_TYPE) information to the hypervisor - aka P-states (cpufreq data). 2). Upload the the Cx state information (cpuidle data). 3). Inhibit CPU frequency scaling drivers from loading. The reason for wanting to solve 1) and 2) is such that the Xen hypervisor is the only one that knows the CPU usage of different guests and can make the proper decision of when to put CPUs and packages in proper states. Unfortunately the hypervisor has no support to parse ACPI DSDT tables, hence it needs help from the initial domain to provide this information. The reason for 3) is that we do not want the initial domain to change P-states while the hypervisor is doing it as well - it causes rather some funny cases of P-states transitions. For this to work, the driver parses the Power Management data and uploads said information to the Xen hypervisor. It also calls acpi_processor_notify_smm() to inhibit the other CPU frequency scaling drivers from being loaded. Everything revolves around the 'struct acpi_processor' structure which gets updated during the bootup cycle in different stages. At the startup, when the ACPI parser starts, the C-state information is processed (processor_idle) and saved in said structure as 'power' element. Later on, the CPU frequency scaling driver (powernow-k8 or acpi_cpufreq), would call the the acpi_processor_* (processor_perflib functions) to parse P-states information and populate in the said structure the 'performance' element. Since we do not want the CPU frequency scaling drivers from loading we have to call the acpi_processor_* functions to parse the P-states and call "acpi_processor_notify_smm" to stop them from loading. There is also one oddity in this driver which is that under Xen, the physical online CPU count can be different from the virtual online CPU count. Meaning that the macros 'for_[online|possible]_cpu' would process only up to virtual online CPU count. We on the other hand want to process the full amount of physical CPUs. For that, the driver checks if the ACPI IDs count is different from the APIC ID count - which can happen if the user choose to use dom0_max_vcpu argument. In such a case a backup of the PM structure is used and uploaded to the hypervisor. [v1-v2: Initial RFC implementations that were posted] [v3: Changed the name to passthru suggested by Pasi Kärkkäinen <pasik@iki.fi>] [v4: Added vCPU != pCPU support - aka dom0_max_vcpus support] [v5: Cleaned up the driver, fix bug under Athlon XP] [v6: Changed the driver to a CPU frequency governor] [v7: Jan Beulich <jbeulich@suse.com> suggestion to make it a cpufreq scaling driver made me rework it as driver that inhibits cpufreq scaling driver] [v8: Per Jan's review comments, fixed up the driver] [v9: Allow to continue even if acpi_processor_preregister_perf.. fails] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
2012-02-03 22:03:20 +01:00
if (!bitmap_equal(acpi_id_present, acpi_ids_done, nr_acpi_bits)) {
unsigned int i;
for_each_set_bit(i, acpi_id_present, nr_acpi_bits) {
pr_backup->acpi_id = i;
/* Mask out C-states if there are no _CST or PBLK */
pr_backup->flags.power = test_bit(i, acpi_id_cst_present);
(void)upload_pm_data(pr_backup);
}
}
xen/acpi-processor: C and P-state driver that uploads said data to hypervisor. This driver solves three problems: 1). Parse and upload ACPI0007 (or PROCESSOR_TYPE) information to the hypervisor - aka P-states (cpufreq data). 2). Upload the the Cx state information (cpuidle data). 3). Inhibit CPU frequency scaling drivers from loading. The reason for wanting to solve 1) and 2) is such that the Xen hypervisor is the only one that knows the CPU usage of different guests and can make the proper decision of when to put CPUs and packages in proper states. Unfortunately the hypervisor has no support to parse ACPI DSDT tables, hence it needs help from the initial domain to provide this information. The reason for 3) is that we do not want the initial domain to change P-states while the hypervisor is doing it as well - it causes rather some funny cases of P-states transitions. For this to work, the driver parses the Power Management data and uploads said information to the Xen hypervisor. It also calls acpi_processor_notify_smm() to inhibit the other CPU frequency scaling drivers from being loaded. Everything revolves around the 'struct acpi_processor' structure which gets updated during the bootup cycle in different stages. At the startup, when the ACPI parser starts, the C-state information is processed (processor_idle) and saved in said structure as 'power' element. Later on, the CPU frequency scaling driver (powernow-k8 or acpi_cpufreq), would call the the acpi_processor_* (processor_perflib functions) to parse P-states information and populate in the said structure the 'performance' element. Since we do not want the CPU frequency scaling drivers from loading we have to call the acpi_processor_* functions to parse the P-states and call "acpi_processor_notify_smm" to stop them from loading. There is also one oddity in this driver which is that under Xen, the physical online CPU count can be different from the virtual online CPU count. Meaning that the macros 'for_[online|possible]_cpu' would process only up to virtual online CPU count. We on the other hand want to process the full amount of physical CPUs. For that, the driver checks if the ACPI IDs count is different from the APIC ID count - which can happen if the user choose to use dom0_max_vcpu argument. In such a case a backup of the PM structure is used and uploaded to the hypervisor. [v1-v2: Initial RFC implementations that were posted] [v3: Changed the name to passthru suggested by Pasi Kärkkäinen <pasik@iki.fi>] [v4: Added vCPU != pCPU support - aka dom0_max_vcpus support] [v5: Cleaned up the driver, fix bug under Athlon XP] [v6: Changed the driver to a CPU frequency governor] [v7: Jan Beulich <jbeulich@suse.com> suggestion to make it a cpufreq scaling driver made me rework it as driver that inhibits cpufreq scaling driver] [v8: Per Jan's review comments, fixed up the driver] [v9: Allow to continue even if acpi_processor_preregister_perf.. fails] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
2012-02-03 22:03:20 +01:00
return 0;
}
static int __init check_prereq(void)
{
struct cpuinfo_x86 *c = &cpu_data(0);
if (!xen_initial_domain())
return -ENODEV;
if (!acpi_gbl_FADT.smi_command)
return -ENODEV;
if (c->x86_vendor == X86_VENDOR_INTEL) {
if (!cpu_has(c, X86_FEATURE_EST))
return -ENODEV;
return 0;
}
if (c->x86_vendor == X86_VENDOR_AMD) {
/* Copied from powernow-k8.h, can't include ../cpufreq/powernow
* as we get compile warnings for the static functions.
*/
#define CPUID_FREQ_VOLT_CAPABILITIES 0x80000007
#define USE_HW_PSTATE 0x00000080
u32 eax, ebx, ecx, edx;
cpuid(CPUID_FREQ_VOLT_CAPABILITIES, &eax, &ebx, &ecx, &edx);
if ((edx & USE_HW_PSTATE) != USE_HW_PSTATE)
return -ENODEV;
return 0;
}
return -ENODEV;
}
/* acpi_perf_data is a pointer to percpu data. */
static struct acpi_processor_performance __percpu *acpi_perf_data;
static void free_acpi_perf_data(void)
{
unsigned int i;
/* Freeing a NULL pointer is OK, and alloc_percpu zeroes. */
for_each_possible_cpu(i)
free_cpumask_var(per_cpu_ptr(acpi_perf_data, i)
->shared_cpu_map);
free_percpu(acpi_perf_data);
}
static int xen_upload_processor_pm_data(void)
xen/acpi-processor: C and P-state driver that uploads said data to hypervisor. This driver solves three problems: 1). Parse and upload ACPI0007 (or PROCESSOR_TYPE) information to the hypervisor - aka P-states (cpufreq data). 2). Upload the the Cx state information (cpuidle data). 3). Inhibit CPU frequency scaling drivers from loading. The reason for wanting to solve 1) and 2) is such that the Xen hypervisor is the only one that knows the CPU usage of different guests and can make the proper decision of when to put CPUs and packages in proper states. Unfortunately the hypervisor has no support to parse ACPI DSDT tables, hence it needs help from the initial domain to provide this information. The reason for 3) is that we do not want the initial domain to change P-states while the hypervisor is doing it as well - it causes rather some funny cases of P-states transitions. For this to work, the driver parses the Power Management data and uploads said information to the Xen hypervisor. It also calls acpi_processor_notify_smm() to inhibit the other CPU frequency scaling drivers from being loaded. Everything revolves around the 'struct acpi_processor' structure which gets updated during the bootup cycle in different stages. At the startup, when the ACPI parser starts, the C-state information is processed (processor_idle) and saved in said structure as 'power' element. Later on, the CPU frequency scaling driver (powernow-k8 or acpi_cpufreq), would call the the acpi_processor_* (processor_perflib functions) to parse P-states information and populate in the said structure the 'performance' element. Since we do not want the CPU frequency scaling drivers from loading we have to call the acpi_processor_* functions to parse the P-states and call "acpi_processor_notify_smm" to stop them from loading. There is also one oddity in this driver which is that under Xen, the physical online CPU count can be different from the virtual online CPU count. Meaning that the macros 'for_[online|possible]_cpu' would process only up to virtual online CPU count. We on the other hand want to process the full amount of physical CPUs. For that, the driver checks if the ACPI IDs count is different from the APIC ID count - which can happen if the user choose to use dom0_max_vcpu argument. In such a case a backup of the PM structure is used and uploaded to the hypervisor. [v1-v2: Initial RFC implementations that were posted] [v3: Changed the name to passthru suggested by Pasi Kärkkäinen <pasik@iki.fi>] [v4: Added vCPU != pCPU support - aka dom0_max_vcpus support] [v5: Cleaned up the driver, fix bug under Athlon XP] [v6: Changed the driver to a CPU frequency governor] [v7: Jan Beulich <jbeulich@suse.com> suggestion to make it a cpufreq scaling driver made me rework it as driver that inhibits cpufreq scaling driver] [v8: Per Jan's review comments, fixed up the driver] [v9: Allow to continue even if acpi_processor_preregister_perf.. fails] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
2012-02-03 22:03:20 +01:00
{
struct acpi_processor *pr_backup = NULL;
unsigned int i;
int rc = 0;
pr_info("Uploading Xen processor PM info\n");
for_each_possible_cpu(i) {
struct acpi_processor *_pr;
_pr = per_cpu(processors, i /* APIC ID */);
if (!_pr)
continue;
if (!pr_backup) {
pr_backup = kzalloc(sizeof(struct acpi_processor), GFP_KERNEL);
if (pr_backup)
memcpy(pr_backup, _pr, sizeof(struct acpi_processor));
}
(void)upload_pm_data(_pr);
}
rc = check_acpi_ids(pr_backup);
kfree(pr_backup);
return rc;
}
static void xen_acpi_processor_resume(void)
{
bitmap_zero(acpi_ids_done, nr_acpi_bits);
xen_upload_processor_pm_data();
}
static struct syscore_ops xap_syscore_ops = {
.resume = xen_acpi_processor_resume,
};
static int __init xen_acpi_processor_init(void)
{
xen/acpi-processor: C and P-state driver that uploads said data to hypervisor. This driver solves three problems: 1). Parse and upload ACPI0007 (or PROCESSOR_TYPE) information to the hypervisor - aka P-states (cpufreq data). 2). Upload the the Cx state information (cpuidle data). 3). Inhibit CPU frequency scaling drivers from loading. The reason for wanting to solve 1) and 2) is such that the Xen hypervisor is the only one that knows the CPU usage of different guests and can make the proper decision of when to put CPUs and packages in proper states. Unfortunately the hypervisor has no support to parse ACPI DSDT tables, hence it needs help from the initial domain to provide this information. The reason for 3) is that we do not want the initial domain to change P-states while the hypervisor is doing it as well - it causes rather some funny cases of P-states transitions. For this to work, the driver parses the Power Management data and uploads said information to the Xen hypervisor. It also calls acpi_processor_notify_smm() to inhibit the other CPU frequency scaling drivers from being loaded. Everything revolves around the 'struct acpi_processor' structure which gets updated during the bootup cycle in different stages. At the startup, when the ACPI parser starts, the C-state information is processed (processor_idle) and saved in said structure as 'power' element. Later on, the CPU frequency scaling driver (powernow-k8 or acpi_cpufreq), would call the the acpi_processor_* (processor_perflib functions) to parse P-states information and populate in the said structure the 'performance' element. Since we do not want the CPU frequency scaling drivers from loading we have to call the acpi_processor_* functions to parse the P-states and call "acpi_processor_notify_smm" to stop them from loading. There is also one oddity in this driver which is that under Xen, the physical online CPU count can be different from the virtual online CPU count. Meaning that the macros 'for_[online|possible]_cpu' would process only up to virtual online CPU count. We on the other hand want to process the full amount of physical CPUs. For that, the driver checks if the ACPI IDs count is different from the APIC ID count - which can happen if the user choose to use dom0_max_vcpu argument. In such a case a backup of the PM structure is used and uploaded to the hypervisor. [v1-v2: Initial RFC implementations that were posted] [v3: Changed the name to passthru suggested by Pasi Kärkkäinen <pasik@iki.fi>] [v4: Added vCPU != pCPU support - aka dom0_max_vcpus support] [v5: Cleaned up the driver, fix bug under Athlon XP] [v6: Changed the driver to a CPU frequency governor] [v7: Jan Beulich <jbeulich@suse.com> suggestion to make it a cpufreq scaling driver made me rework it as driver that inhibits cpufreq scaling driver] [v8: Per Jan's review comments, fixed up the driver] [v9: Allow to continue even if acpi_processor_preregister_perf.. fails] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
2012-02-03 22:03:20 +01:00
unsigned int i;
int rc = check_prereq();
if (rc)
return rc;
nr_acpi_bits = get_max_acpi_id() + 1;
acpi_ids_done = kcalloc(BITS_TO_LONGS(nr_acpi_bits), sizeof(unsigned long), GFP_KERNEL);
if (!acpi_ids_done)
return -ENOMEM;
acpi_perf_data = alloc_percpu(struct acpi_processor_performance);
if (!acpi_perf_data) {
pr_debug("Memory allocation error for acpi_perf_data\n");
xen/acpi-processor: C and P-state driver that uploads said data to hypervisor. This driver solves three problems: 1). Parse and upload ACPI0007 (or PROCESSOR_TYPE) information to the hypervisor - aka P-states (cpufreq data). 2). Upload the the Cx state information (cpuidle data). 3). Inhibit CPU frequency scaling drivers from loading. The reason for wanting to solve 1) and 2) is such that the Xen hypervisor is the only one that knows the CPU usage of different guests and can make the proper decision of when to put CPUs and packages in proper states. Unfortunately the hypervisor has no support to parse ACPI DSDT tables, hence it needs help from the initial domain to provide this information. The reason for 3) is that we do not want the initial domain to change P-states while the hypervisor is doing it as well - it causes rather some funny cases of P-states transitions. For this to work, the driver parses the Power Management data and uploads said information to the Xen hypervisor. It also calls acpi_processor_notify_smm() to inhibit the other CPU frequency scaling drivers from being loaded. Everything revolves around the 'struct acpi_processor' structure which gets updated during the bootup cycle in different stages. At the startup, when the ACPI parser starts, the C-state information is processed (processor_idle) and saved in said structure as 'power' element. Later on, the CPU frequency scaling driver (powernow-k8 or acpi_cpufreq), would call the the acpi_processor_* (processor_perflib functions) to parse P-states information and populate in the said structure the 'performance' element. Since we do not want the CPU frequency scaling drivers from loading we have to call the acpi_processor_* functions to parse the P-states and call "acpi_processor_notify_smm" to stop them from loading. There is also one oddity in this driver which is that under Xen, the physical online CPU count can be different from the virtual online CPU count. Meaning that the macros 'for_[online|possible]_cpu' would process only up to virtual online CPU count. We on the other hand want to process the full amount of physical CPUs. For that, the driver checks if the ACPI IDs count is different from the APIC ID count - which can happen if the user choose to use dom0_max_vcpu argument. In such a case a backup of the PM structure is used and uploaded to the hypervisor. [v1-v2: Initial RFC implementations that were posted] [v3: Changed the name to passthru suggested by Pasi Kärkkäinen <pasik@iki.fi>] [v4: Added vCPU != pCPU support - aka dom0_max_vcpus support] [v5: Cleaned up the driver, fix bug under Athlon XP] [v6: Changed the driver to a CPU frequency governor] [v7: Jan Beulich <jbeulich@suse.com> suggestion to make it a cpufreq scaling driver made me rework it as driver that inhibits cpufreq scaling driver] [v8: Per Jan's review comments, fixed up the driver] [v9: Allow to continue even if acpi_processor_preregister_perf.. fails] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
2012-02-03 22:03:20 +01:00
kfree(acpi_ids_done);
return -ENOMEM;
}
for_each_possible_cpu(i) {
if (!zalloc_cpumask_var_node(
&per_cpu_ptr(acpi_perf_data, i)->shared_cpu_map,
GFP_KERNEL, cpu_to_node(i))) {
rc = -ENOMEM;
goto err_out;
}
}
/* Do initialization in ACPI core. It is OK to fail here. */
(void)acpi_processor_preregister_performance(acpi_perf_data);
for_each_possible_cpu(i) {
acpi: Export the acpi_processor_get_performance_info The git commit d5aaffa9dd531c978c6f3fea06a2972653bd7fc8 (cpufreq: handle cpufreq being disabled for all exported function) tightens the cpufreq API by returning errors when disable_cpufreq() had been called. The problem we are hitting is that the module xen-acpi-processor which uses the ACPI's functions: acpi_processor_register_performance, acpi_processor_preregister_performance, and acpi_processor_notify_smm fails at acpi_processor_register_performance with -22. Note that earlier during bootup in arch/x86/xen/setup.c there is also an call to cpufreq's API: disable_cpufreq(). This is b/c we want the Linux kernel to parse the ACPI data, but leave the cpufreq decisions to the hypervisor. In v3.9 all the checks that d5aaffa9dd531c978c6f3fea06a2972653bd7fc8 added are now hit and the calls to cpufreq_register_notifier will now fail. This means that acpi_processor_ppc_init ends up printing: "Warning: Processor Platform Limit not supported" and the acpi_processor_ppc_status is not set. The repercussions of that is that the call to acpi_processor_register_performance fails right away at: if (!(acpi_processor_ppc_status & PPC_REGISTERED)) and we don't progress any further on parsing and extracting the _P* objects. The only reason the Xen code called that function was b/c it was exported and the only way to gather the P-states. But we can also just make acpi_processor_get_performance_info be exported and not use acpi_processor_register_performance. This patch does so. Acked-by: Rafael J. Wysocki <rjw@sisk.pl> Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
2013-03-05 19:42:54 +01:00
struct acpi_processor *pr;
xen/acpi-processor: C and P-state driver that uploads said data to hypervisor. This driver solves three problems: 1). Parse and upload ACPI0007 (or PROCESSOR_TYPE) information to the hypervisor - aka P-states (cpufreq data). 2). Upload the the Cx state information (cpuidle data). 3). Inhibit CPU frequency scaling drivers from loading. The reason for wanting to solve 1) and 2) is such that the Xen hypervisor is the only one that knows the CPU usage of different guests and can make the proper decision of when to put CPUs and packages in proper states. Unfortunately the hypervisor has no support to parse ACPI DSDT tables, hence it needs help from the initial domain to provide this information. The reason for 3) is that we do not want the initial domain to change P-states while the hypervisor is doing it as well - it causes rather some funny cases of P-states transitions. For this to work, the driver parses the Power Management data and uploads said information to the Xen hypervisor. It also calls acpi_processor_notify_smm() to inhibit the other CPU frequency scaling drivers from being loaded. Everything revolves around the 'struct acpi_processor' structure which gets updated during the bootup cycle in different stages. At the startup, when the ACPI parser starts, the C-state information is processed (processor_idle) and saved in said structure as 'power' element. Later on, the CPU frequency scaling driver (powernow-k8 or acpi_cpufreq), would call the the acpi_processor_* (processor_perflib functions) to parse P-states information and populate in the said structure the 'performance' element. Since we do not want the CPU frequency scaling drivers from loading we have to call the acpi_processor_* functions to parse the P-states and call "acpi_processor_notify_smm" to stop them from loading. There is also one oddity in this driver which is that under Xen, the physical online CPU count can be different from the virtual online CPU count. Meaning that the macros 'for_[online|possible]_cpu' would process only up to virtual online CPU count. We on the other hand want to process the full amount of physical CPUs. For that, the driver checks if the ACPI IDs count is different from the APIC ID count - which can happen if the user choose to use dom0_max_vcpu argument. In such a case a backup of the PM structure is used and uploaded to the hypervisor. [v1-v2: Initial RFC implementations that were posted] [v3: Changed the name to passthru suggested by Pasi Kärkkäinen <pasik@iki.fi>] [v4: Added vCPU != pCPU support - aka dom0_max_vcpus support] [v5: Cleaned up the driver, fix bug under Athlon XP] [v6: Changed the driver to a CPU frequency governor] [v7: Jan Beulich <jbeulich@suse.com> suggestion to make it a cpufreq scaling driver made me rework it as driver that inhibits cpufreq scaling driver] [v8: Per Jan's review comments, fixed up the driver] [v9: Allow to continue even if acpi_processor_preregister_perf.. fails] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
2012-02-03 22:03:20 +01:00
struct acpi_processor_performance *perf;
acpi: Export the acpi_processor_get_performance_info The git commit d5aaffa9dd531c978c6f3fea06a2972653bd7fc8 (cpufreq: handle cpufreq being disabled for all exported function) tightens the cpufreq API by returning errors when disable_cpufreq() had been called. The problem we are hitting is that the module xen-acpi-processor which uses the ACPI's functions: acpi_processor_register_performance, acpi_processor_preregister_performance, and acpi_processor_notify_smm fails at acpi_processor_register_performance with -22. Note that earlier during bootup in arch/x86/xen/setup.c there is also an call to cpufreq's API: disable_cpufreq(). This is b/c we want the Linux kernel to parse the ACPI data, but leave the cpufreq decisions to the hypervisor. In v3.9 all the checks that d5aaffa9dd531c978c6f3fea06a2972653bd7fc8 added are now hit and the calls to cpufreq_register_notifier will now fail. This means that acpi_processor_ppc_init ends up printing: "Warning: Processor Platform Limit not supported" and the acpi_processor_ppc_status is not set. The repercussions of that is that the call to acpi_processor_register_performance fails right away at: if (!(acpi_processor_ppc_status & PPC_REGISTERED)) and we don't progress any further on parsing and extracting the _P* objects. The only reason the Xen code called that function was b/c it was exported and the only way to gather the P-states. But we can also just make acpi_processor_get_performance_info be exported and not use acpi_processor_register_performance. This patch does so. Acked-by: Rafael J. Wysocki <rjw@sisk.pl> Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
2013-03-05 19:42:54 +01:00
pr = per_cpu(processors, i);
xen/acpi-processor: C and P-state driver that uploads said data to hypervisor. This driver solves three problems: 1). Parse and upload ACPI0007 (or PROCESSOR_TYPE) information to the hypervisor - aka P-states (cpufreq data). 2). Upload the the Cx state information (cpuidle data). 3). Inhibit CPU frequency scaling drivers from loading. The reason for wanting to solve 1) and 2) is such that the Xen hypervisor is the only one that knows the CPU usage of different guests and can make the proper decision of when to put CPUs and packages in proper states. Unfortunately the hypervisor has no support to parse ACPI DSDT tables, hence it needs help from the initial domain to provide this information. The reason for 3) is that we do not want the initial domain to change P-states while the hypervisor is doing it as well - it causes rather some funny cases of P-states transitions. For this to work, the driver parses the Power Management data and uploads said information to the Xen hypervisor. It also calls acpi_processor_notify_smm() to inhibit the other CPU frequency scaling drivers from being loaded. Everything revolves around the 'struct acpi_processor' structure which gets updated during the bootup cycle in different stages. At the startup, when the ACPI parser starts, the C-state information is processed (processor_idle) and saved in said structure as 'power' element. Later on, the CPU frequency scaling driver (powernow-k8 or acpi_cpufreq), would call the the acpi_processor_* (processor_perflib functions) to parse P-states information and populate in the said structure the 'performance' element. Since we do not want the CPU frequency scaling drivers from loading we have to call the acpi_processor_* functions to parse the P-states and call "acpi_processor_notify_smm" to stop them from loading. There is also one oddity in this driver which is that under Xen, the physical online CPU count can be different from the virtual online CPU count. Meaning that the macros 'for_[online|possible]_cpu' would process only up to virtual online CPU count. We on the other hand want to process the full amount of physical CPUs. For that, the driver checks if the ACPI IDs count is different from the APIC ID count - which can happen if the user choose to use dom0_max_vcpu argument. In such a case a backup of the PM structure is used and uploaded to the hypervisor. [v1-v2: Initial RFC implementations that were posted] [v3: Changed the name to passthru suggested by Pasi Kärkkäinen <pasik@iki.fi>] [v4: Added vCPU != pCPU support - aka dom0_max_vcpus support] [v5: Cleaned up the driver, fix bug under Athlon XP] [v6: Changed the driver to a CPU frequency governor] [v7: Jan Beulich <jbeulich@suse.com> suggestion to make it a cpufreq scaling driver made me rework it as driver that inhibits cpufreq scaling driver] [v8: Per Jan's review comments, fixed up the driver] [v9: Allow to continue even if acpi_processor_preregister_perf.. fails] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
2012-02-03 22:03:20 +01:00
perf = per_cpu_ptr(acpi_perf_data, i);
xen/acpi-processor: Don't dereference struct acpi_processor on all CPUs. With git commit c705c78c0d0835a4aa5d0d9a3422e3218462030c "acpi: Export the acpi_processor_get_performance_info" we are now using a different mechanism to access the P-states. The acpi_processor per-cpu structure is set and filtered by the core ACPI code which shrinks the per_cpu contents to only online CPUs. In the past we would call acpi_processor_register_performance() which would have not tried to dereference offline cpus. With the new patch and the fact that the loop we take is for for_all_possible_cpus we end up crashing on some machines. We could modify the loop to be for online_cpus - but all the other loops in the code use possible_cpus (for a good reason) - so lets leave it as so and just check if per_cpu(processor) is NULL. With this patch we will bypass the !online but possible CPUs. This fixes: IP: [<ffffffffa00d13b5>] xen_acpi_processor_init+0x1b6/0xe01 [xen_acpi_processor] PGD 4126e6067 PUD 4126e3067 PMD 0 Oops: 0002 [#1] SMP Pid: 432, comm: modprobe Not tainted 3.9.0-rc3+ #28 To be filled by O.E.M. To be filled by O.E.M./M5A97 RIP: e030:[<ffffffffa00d13b5>] [<ffffffffa00d13b5>] xen_acpi_processor_init+0x1b6/0xe01 [xen_acpi_processor] RSP: e02b:ffff88040c8a3ce8 EFLAGS: 00010282 .. snip.. Call Trace: [<ffffffffa00d11ff>] ? read_acpi_id+0x12b/0x12b [xen_acpi_processor] [<ffffffff8100215a>] do_one_initcall+0x12a/0x180 [<ffffffff810c42c3>] load_module+0x1cd3/0x2870 [<ffffffff81319b70>] ? ddebug_proc_open+0xc0/0xc0 [<ffffffff810c4f37>] sys_init_module+0xd7/0x120 [<ffffffff8166ce19>] system_call_fastpath+0x16/0x1b on some machines. Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
2013-03-19 19:35:30 +01:00
if (!pr)
continue;
acpi: Export the acpi_processor_get_performance_info The git commit d5aaffa9dd531c978c6f3fea06a2972653bd7fc8 (cpufreq: handle cpufreq being disabled for all exported function) tightens the cpufreq API by returning errors when disable_cpufreq() had been called. The problem we are hitting is that the module xen-acpi-processor which uses the ACPI's functions: acpi_processor_register_performance, acpi_processor_preregister_performance, and acpi_processor_notify_smm fails at acpi_processor_register_performance with -22. Note that earlier during bootup in arch/x86/xen/setup.c there is also an call to cpufreq's API: disable_cpufreq(). This is b/c we want the Linux kernel to parse the ACPI data, but leave the cpufreq decisions to the hypervisor. In v3.9 all the checks that d5aaffa9dd531c978c6f3fea06a2972653bd7fc8 added are now hit and the calls to cpufreq_register_notifier will now fail. This means that acpi_processor_ppc_init ends up printing: "Warning: Processor Platform Limit not supported" and the acpi_processor_ppc_status is not set. The repercussions of that is that the call to acpi_processor_register_performance fails right away at: if (!(acpi_processor_ppc_status & PPC_REGISTERED)) and we don't progress any further on parsing and extracting the _P* objects. The only reason the Xen code called that function was b/c it was exported and the only way to gather the P-states. But we can also just make acpi_processor_get_performance_info be exported and not use acpi_processor_register_performance. This patch does so. Acked-by: Rafael J. Wysocki <rjw@sisk.pl> Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
2013-03-05 19:42:54 +01:00
pr->performance = perf;
rc = acpi_processor_get_performance_info(pr);
if (rc)
xen/acpi-processor: C and P-state driver that uploads said data to hypervisor. This driver solves three problems: 1). Parse and upload ACPI0007 (or PROCESSOR_TYPE) information to the hypervisor - aka P-states (cpufreq data). 2). Upload the the Cx state information (cpuidle data). 3). Inhibit CPU frequency scaling drivers from loading. The reason for wanting to solve 1) and 2) is such that the Xen hypervisor is the only one that knows the CPU usage of different guests and can make the proper decision of when to put CPUs and packages in proper states. Unfortunately the hypervisor has no support to parse ACPI DSDT tables, hence it needs help from the initial domain to provide this information. The reason for 3) is that we do not want the initial domain to change P-states while the hypervisor is doing it as well - it causes rather some funny cases of P-states transitions. For this to work, the driver parses the Power Management data and uploads said information to the Xen hypervisor. It also calls acpi_processor_notify_smm() to inhibit the other CPU frequency scaling drivers from being loaded. Everything revolves around the 'struct acpi_processor' structure which gets updated during the bootup cycle in different stages. At the startup, when the ACPI parser starts, the C-state information is processed (processor_idle) and saved in said structure as 'power' element. Later on, the CPU frequency scaling driver (powernow-k8 or acpi_cpufreq), would call the the acpi_processor_* (processor_perflib functions) to parse P-states information and populate in the said structure the 'performance' element. Since we do not want the CPU frequency scaling drivers from loading we have to call the acpi_processor_* functions to parse the P-states and call "acpi_processor_notify_smm" to stop them from loading. There is also one oddity in this driver which is that under Xen, the physical online CPU count can be different from the virtual online CPU count. Meaning that the macros 'for_[online|possible]_cpu' would process only up to virtual online CPU count. We on the other hand want to process the full amount of physical CPUs. For that, the driver checks if the ACPI IDs count is different from the APIC ID count - which can happen if the user choose to use dom0_max_vcpu argument. In such a case a backup of the PM structure is used and uploaded to the hypervisor. [v1-v2: Initial RFC implementations that were posted] [v3: Changed the name to passthru suggested by Pasi Kärkkäinen <pasik@iki.fi>] [v4: Added vCPU != pCPU support - aka dom0_max_vcpus support] [v5: Cleaned up the driver, fix bug under Athlon XP] [v6: Changed the driver to a CPU frequency governor] [v7: Jan Beulich <jbeulich@suse.com> suggestion to make it a cpufreq scaling driver made me rework it as driver that inhibits cpufreq scaling driver] [v8: Per Jan's review comments, fixed up the driver] [v9: Allow to continue even if acpi_processor_preregister_perf.. fails] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
2012-02-03 22:03:20 +01:00
goto err_out;
}
rc = xen_upload_processor_pm_data();
if (rc)
goto err_unregister;
xen/acpi-processor: C and P-state driver that uploads said data to hypervisor. This driver solves three problems: 1). Parse and upload ACPI0007 (or PROCESSOR_TYPE) information to the hypervisor - aka P-states (cpufreq data). 2). Upload the the Cx state information (cpuidle data). 3). Inhibit CPU frequency scaling drivers from loading. The reason for wanting to solve 1) and 2) is such that the Xen hypervisor is the only one that knows the CPU usage of different guests and can make the proper decision of when to put CPUs and packages in proper states. Unfortunately the hypervisor has no support to parse ACPI DSDT tables, hence it needs help from the initial domain to provide this information. The reason for 3) is that we do not want the initial domain to change P-states while the hypervisor is doing it as well - it causes rather some funny cases of P-states transitions. For this to work, the driver parses the Power Management data and uploads said information to the Xen hypervisor. It also calls acpi_processor_notify_smm() to inhibit the other CPU frequency scaling drivers from being loaded. Everything revolves around the 'struct acpi_processor' structure which gets updated during the bootup cycle in different stages. At the startup, when the ACPI parser starts, the C-state information is processed (processor_idle) and saved in said structure as 'power' element. Later on, the CPU frequency scaling driver (powernow-k8 or acpi_cpufreq), would call the the acpi_processor_* (processor_perflib functions) to parse P-states information and populate in the said structure the 'performance' element. Since we do not want the CPU frequency scaling drivers from loading we have to call the acpi_processor_* functions to parse the P-states and call "acpi_processor_notify_smm" to stop them from loading. There is also one oddity in this driver which is that under Xen, the physical online CPU count can be different from the virtual online CPU count. Meaning that the macros 'for_[online|possible]_cpu' would process only up to virtual online CPU count. We on the other hand want to process the full amount of physical CPUs. For that, the driver checks if the ACPI IDs count is different from the APIC ID count - which can happen if the user choose to use dom0_max_vcpu argument. In such a case a backup of the PM structure is used and uploaded to the hypervisor. [v1-v2: Initial RFC implementations that were posted] [v3: Changed the name to passthru suggested by Pasi Kärkkäinen <pasik@iki.fi>] [v4: Added vCPU != pCPU support - aka dom0_max_vcpus support] [v5: Cleaned up the driver, fix bug under Athlon XP] [v6: Changed the driver to a CPU frequency governor] [v7: Jan Beulich <jbeulich@suse.com> suggestion to make it a cpufreq scaling driver made me rework it as driver that inhibits cpufreq scaling driver] [v8: Per Jan's review comments, fixed up the driver] [v9: Allow to continue even if acpi_processor_preregister_perf.. fails] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
2012-02-03 22:03:20 +01:00
register_syscore_ops(&xap_syscore_ops);
xen/acpi-processor: C and P-state driver that uploads said data to hypervisor. This driver solves three problems: 1). Parse and upload ACPI0007 (or PROCESSOR_TYPE) information to the hypervisor - aka P-states (cpufreq data). 2). Upload the the Cx state information (cpuidle data). 3). Inhibit CPU frequency scaling drivers from loading. The reason for wanting to solve 1) and 2) is such that the Xen hypervisor is the only one that knows the CPU usage of different guests and can make the proper decision of when to put CPUs and packages in proper states. Unfortunately the hypervisor has no support to parse ACPI DSDT tables, hence it needs help from the initial domain to provide this information. The reason for 3) is that we do not want the initial domain to change P-states while the hypervisor is doing it as well - it causes rather some funny cases of P-states transitions. For this to work, the driver parses the Power Management data and uploads said information to the Xen hypervisor. It also calls acpi_processor_notify_smm() to inhibit the other CPU frequency scaling drivers from being loaded. Everything revolves around the 'struct acpi_processor' structure which gets updated during the bootup cycle in different stages. At the startup, when the ACPI parser starts, the C-state information is processed (processor_idle) and saved in said structure as 'power' element. Later on, the CPU frequency scaling driver (powernow-k8 or acpi_cpufreq), would call the the acpi_processor_* (processor_perflib functions) to parse P-states information and populate in the said structure the 'performance' element. Since we do not want the CPU frequency scaling drivers from loading we have to call the acpi_processor_* functions to parse the P-states and call "acpi_processor_notify_smm" to stop them from loading. There is also one oddity in this driver which is that under Xen, the physical online CPU count can be different from the virtual online CPU count. Meaning that the macros 'for_[online|possible]_cpu' would process only up to virtual online CPU count. We on the other hand want to process the full amount of physical CPUs. For that, the driver checks if the ACPI IDs count is different from the APIC ID count - which can happen if the user choose to use dom0_max_vcpu argument. In such a case a backup of the PM structure is used and uploaded to the hypervisor. [v1-v2: Initial RFC implementations that were posted] [v3: Changed the name to passthru suggested by Pasi Kärkkäinen <pasik@iki.fi>] [v4: Added vCPU != pCPU support - aka dom0_max_vcpus support] [v5: Cleaned up the driver, fix bug under Athlon XP] [v6: Changed the driver to a CPU frequency governor] [v7: Jan Beulich <jbeulich@suse.com> suggestion to make it a cpufreq scaling driver made me rework it as driver that inhibits cpufreq scaling driver] [v8: Per Jan's review comments, fixed up the driver] [v9: Allow to continue even if acpi_processor_preregister_perf.. fails] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
2012-02-03 22:03:20 +01:00
return 0;
err_unregister:
for_each_possible_cpu(i) {
struct acpi_processor_performance *perf;
perf = per_cpu_ptr(acpi_perf_data, i);
acpi_processor_unregister_performance(perf, i);
}
err_out:
/* Freeing a NULL pointer is OK: alloc_percpu zeroes. */
free_acpi_perf_data();
kfree(acpi_ids_done);
return rc;
}
static void __exit xen_acpi_processor_exit(void)
{
int i;
unregister_syscore_ops(&xap_syscore_ops);
xen/acpi-processor: C and P-state driver that uploads said data to hypervisor. This driver solves three problems: 1). Parse and upload ACPI0007 (or PROCESSOR_TYPE) information to the hypervisor - aka P-states (cpufreq data). 2). Upload the the Cx state information (cpuidle data). 3). Inhibit CPU frequency scaling drivers from loading. The reason for wanting to solve 1) and 2) is such that the Xen hypervisor is the only one that knows the CPU usage of different guests and can make the proper decision of when to put CPUs and packages in proper states. Unfortunately the hypervisor has no support to parse ACPI DSDT tables, hence it needs help from the initial domain to provide this information. The reason for 3) is that we do not want the initial domain to change P-states while the hypervisor is doing it as well - it causes rather some funny cases of P-states transitions. For this to work, the driver parses the Power Management data and uploads said information to the Xen hypervisor. It also calls acpi_processor_notify_smm() to inhibit the other CPU frequency scaling drivers from being loaded. Everything revolves around the 'struct acpi_processor' structure which gets updated during the bootup cycle in different stages. At the startup, when the ACPI parser starts, the C-state information is processed (processor_idle) and saved in said structure as 'power' element. Later on, the CPU frequency scaling driver (powernow-k8 or acpi_cpufreq), would call the the acpi_processor_* (processor_perflib functions) to parse P-states information and populate in the said structure the 'performance' element. Since we do not want the CPU frequency scaling drivers from loading we have to call the acpi_processor_* functions to parse the P-states and call "acpi_processor_notify_smm" to stop them from loading. There is also one oddity in this driver which is that under Xen, the physical online CPU count can be different from the virtual online CPU count. Meaning that the macros 'for_[online|possible]_cpu' would process only up to virtual online CPU count. We on the other hand want to process the full amount of physical CPUs. For that, the driver checks if the ACPI IDs count is different from the APIC ID count - which can happen if the user choose to use dom0_max_vcpu argument. In such a case a backup of the PM structure is used and uploaded to the hypervisor. [v1-v2: Initial RFC implementations that were posted] [v3: Changed the name to passthru suggested by Pasi Kärkkäinen <pasik@iki.fi>] [v4: Added vCPU != pCPU support - aka dom0_max_vcpus support] [v5: Cleaned up the driver, fix bug under Athlon XP] [v6: Changed the driver to a CPU frequency governor] [v7: Jan Beulich <jbeulich@suse.com> suggestion to make it a cpufreq scaling driver made me rework it as driver that inhibits cpufreq scaling driver] [v8: Per Jan's review comments, fixed up the driver] [v9: Allow to continue even if acpi_processor_preregister_perf.. fails] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
2012-02-03 22:03:20 +01:00
kfree(acpi_ids_done);
kfree(acpi_id_present);
kfree(acpi_id_cst_present);
xen/acpi-processor: C and P-state driver that uploads said data to hypervisor. This driver solves three problems: 1). Parse and upload ACPI0007 (or PROCESSOR_TYPE) information to the hypervisor - aka P-states (cpufreq data). 2). Upload the the Cx state information (cpuidle data). 3). Inhibit CPU frequency scaling drivers from loading. The reason for wanting to solve 1) and 2) is such that the Xen hypervisor is the only one that knows the CPU usage of different guests and can make the proper decision of when to put CPUs and packages in proper states. Unfortunately the hypervisor has no support to parse ACPI DSDT tables, hence it needs help from the initial domain to provide this information. The reason for 3) is that we do not want the initial domain to change P-states while the hypervisor is doing it as well - it causes rather some funny cases of P-states transitions. For this to work, the driver parses the Power Management data and uploads said information to the Xen hypervisor. It also calls acpi_processor_notify_smm() to inhibit the other CPU frequency scaling drivers from being loaded. Everything revolves around the 'struct acpi_processor' structure which gets updated during the bootup cycle in different stages. At the startup, when the ACPI parser starts, the C-state information is processed (processor_idle) and saved in said structure as 'power' element. Later on, the CPU frequency scaling driver (powernow-k8 or acpi_cpufreq), would call the the acpi_processor_* (processor_perflib functions) to parse P-states information and populate in the said structure the 'performance' element. Since we do not want the CPU frequency scaling drivers from loading we have to call the acpi_processor_* functions to parse the P-states and call "acpi_processor_notify_smm" to stop them from loading. There is also one oddity in this driver which is that under Xen, the physical online CPU count can be different from the virtual online CPU count. Meaning that the macros 'for_[online|possible]_cpu' would process only up to virtual online CPU count. We on the other hand want to process the full amount of physical CPUs. For that, the driver checks if the ACPI IDs count is different from the APIC ID count - which can happen if the user choose to use dom0_max_vcpu argument. In such a case a backup of the PM structure is used and uploaded to the hypervisor. [v1-v2: Initial RFC implementations that were posted] [v3: Changed the name to passthru suggested by Pasi Kärkkäinen <pasik@iki.fi>] [v4: Added vCPU != pCPU support - aka dom0_max_vcpus support] [v5: Cleaned up the driver, fix bug under Athlon XP] [v6: Changed the driver to a CPU frequency governor] [v7: Jan Beulich <jbeulich@suse.com> suggestion to make it a cpufreq scaling driver made me rework it as driver that inhibits cpufreq scaling driver] [v8: Per Jan's review comments, fixed up the driver] [v9: Allow to continue even if acpi_processor_preregister_perf.. fails] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
2012-02-03 22:03:20 +01:00
for_each_possible_cpu(i) {
struct acpi_processor_performance *perf;
perf = per_cpu_ptr(acpi_perf_data, i);
acpi_processor_unregister_performance(perf, i);
}
free_acpi_perf_data();
}
MODULE_AUTHOR("Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>");
MODULE_DESCRIPTION("Xen ACPI Processor P-states (and Cx) driver which uploads PM data to Xen hypervisor");
MODULE_LICENSE("GPL");
/* We want to be loaded before the CPU freq scaling drivers are loaded.
* They are loaded in late_initcall. */
device_initcall(xen_acpi_processor_init);
module_exit(xen_acpi_processor_exit);