Subversion Repositories shark

Rev

Rev 582 | Go to most recent revision | Blame | Compare with Previous | Last modification | View Log | RSS feed

/*
 *      Pentium 4/Xeon CPU on demand clock modulation/speed scaling
 *      (C) 2002 - 2003 Dominik Brodowski <linux@brodo.de>
 *      (C) 2002 Zwane Mwaikambo <zwane@commfireservices.com>
 *      (C) 2002 Arjan van de Ven <arjanv@redhat.com>
 *      (C) 2002 Tora T. Engstad
 *      All Rights Reserved
 *
 *      This program is free software; you can redistribute it and/or
 *      modify it under the terms of the GNU General Public License
 *      as published by the Free Software Foundation; either version
 *      2 of the License, or (at your option) any later version.
 *
 *      The author(s) of this software shall not be held liable for damages
 *      of any nature resulting due to the use of this software. This
 *      software is provided AS-IS with no warranties.
 *     
 *      Date            Errata                  Description
 *      20020525        N44, O17        12.5% or 25% DC causes lockup
 *
 */


#include <linuxcomp.h>

#include <linux/config.h>
#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/init.h>
#include <linux/smp.h>
#include <linux/cpufreq.h>
#include <linux/slab.h>
#include <linux/sched.h>

#include <asm/processor.h>
#include <asm/msr.h>
#include <asm/timex.h>

#define PFX     "cpufreq: "

/*
 * Duty Cycle (3bits), note DC_DISABLE is not specified in
 * intel docs i just use it to mean disable
 */

enum {
        DC_RESV, DC_DFLT, DC_25PT, DC_38PT, DC_50PT,
        DC_64PT, DC_75PT, DC_88PT, DC_DISABLE
};

#define DC_ENTRIES      8


static int has_N44_O17_errata[NR_CPUS];
static int stock_freq;


static int cpufreq_p4_setdc(unsigned int cpu, unsigned int newstate)
{
        u32 l, h;
        cpumask_t cpus_allowed, affected_cpu_map;
        struct cpufreq_freqs freqs;
        int hyperthreading = 0;
        int sibling = 0;

        if (!cpu_online(cpu) || (newstate > DC_DISABLE) ||
                (newstate == DC_RESV))
                return -EINVAL;

        /* switch to physical CPU where state is to be changed*/
        cpus_allowed = current->cpus_allowed;

        /* only run on CPU to be set, or on its sibling */
       affected_cpu_map = cpumask_of_cpu(cpu);
#ifdef CONFIG_X86_HT
        hyperthreading = ((cpu_has_ht) && (smp_num_siblings == 2));
        if (hyperthreading) {
                sibling = cpu_sibling_map[cpu];
                cpu_set(sibling, affected_cpu_map);
        }
#endif
        set_cpus_allowed(current, affected_cpu_map);
        BUG_ON(!cpu_isset(smp_processor_id(), affected_cpu_map));

        /* get current state */
        rdmsr(MSR_IA32_THERM_CONTROL, l, h);
        if (l & 0x10) {
                l = l >> 1;
                l &= 0x7;
        } else
                l = DC_DISABLE;
       
        if (l == newstate) {
                set_cpus_allowed(current, cpus_allowed);
                return 0;
        } else if (l == DC_RESV) {
                printk(KERN_ERR PFX "BIG FAT WARNING: currently in invalid setting\n");
        }

        /* notifiers */
        freqs.old = stock_freq * l / 8;
        freqs.new = stock_freq * newstate / 8;
        freqs.cpu = cpu;
        cpufreq_notify_transition(&freqs, CPUFREQ_PRECHANGE);
        if (hyperthreading) {
                freqs.cpu = sibling;
                cpufreq_notify_transition(&freqs, CPUFREQ_PRECHANGE);
        }

        rdmsr(MSR_IA32_THERM_STATUS, l, h);
#if 0
        if (l & 0x01)
                printk(KERN_DEBUG PFX "CPU#%d currently thermal throttled\n", cpu);
#endif
        if (has_N44_O17_errata[cpu] && (newstate == DC_25PT || newstate == DC_DFLT))
                newstate = DC_38PT;

        rdmsr(MSR_IA32_THERM_CONTROL, l, h);
        if (newstate == DC_DISABLE) {
                /* printk(KERN_INFO PFX "CPU#%d disabling modulation\n", cpu); */
                wrmsr(MSR_IA32_THERM_CONTROL, l & ~(1<<4), h);
        } else {
                /* printk(KERN_INFO PFX "CPU#%d setting duty cycle to %d%%\n",
                        cpu, ((125 * newstate) / 10)); */

                /* bits 63 - 5  : reserved
                 * bit  4       : enable/disable
                 * bits 3-1     : duty cycle
                 * bit  0       : reserved
                 */

                l = (l & ~14);
                l = l | (1<<4) | ((newstate & 0x7)<<1);
                wrmsr(MSR_IA32_THERM_CONTROL, l, h);
        }

        set_cpus_allowed(current, cpus_allowed);

        /* notifiers */
        cpufreq_notify_transition(&freqs, CPUFREQ_POSTCHANGE);
        if (hyperthreading) {
                freqs.cpu = cpu;
                cpufreq_notify_transition(&freqs, CPUFREQ_POSTCHANGE);
        }

        return 0;
}


static struct cpufreq_frequency_table p4clockmod_table[] = {
        {DC_RESV, CPUFREQ_ENTRY_INVALID},
        {DC_DFLT, 0},
        {DC_25PT, 0},
        {DC_38PT, 0},
        {DC_50PT, 0},
        {DC_64PT, 0},
        {DC_75PT, 0},
        {DC_88PT, 0},
        {DC_DISABLE, 0},
        {DC_RESV, CPUFREQ_TABLE_END},
};


static int cpufreq_p4_target(struct cpufreq_policy *policy,
                             unsigned int target_freq,
                             unsigned int relation)
{
        unsigned int    newstate = DC_RESV;

        if (cpufreq_frequency_table_target(policy, &p4clockmod_table[0], target_freq, relation, &newstate))
                return -EINVAL;

        cpufreq_p4_setdc(policy->cpu, p4clockmod_table[newstate].index);

        return 0;
}


static int cpufreq_p4_verify(struct cpufreq_policy *policy)
{
        return cpufreq_frequency_table_verify(policy, &p4clockmod_table[0]);
}


static int cpufreq_p4_cpu_init(struct cpufreq_policy *policy)
{
        struct cpuinfo_x86 *c = &cpu_data[policy->cpu];
        int cpuid = 0;
        unsigned int i;

        /* Errata workaround */
        cpuid = (c->x86 << 8) | (c->x86_model << 4) | c->x86_mask;
        switch (cpuid) {
        case 0x0f07:
        case 0x0f0a:
        case 0x0f11:
        case 0x0f12:
                has_N44_O17_errata[policy->cpu] = 1;
        }
       
        /* get frequency */
        if (!stock_freq) {
                if (cpu_khz)
                        stock_freq = cpu_khz;
                else {
                        printk(KERN_INFO PFX "unknown core frequency - please use module parameter 'stock_freq'\n");
                        return -EINVAL;
                }
        }

        /* table init */
        for (i=1; (p4clockmod_table[i].frequency != CPUFREQ_TABLE_END); i++) {
                if ((i<2) && (has_N44_O17_errata[policy->cpu]))
                        p4clockmod_table[i].frequency = CPUFREQ_ENTRY_INVALID;
                else
                        p4clockmod_table[i].frequency = (stock_freq * i)/8;
        }
        cpufreq_frequency_table_get_attr(p4clockmod_table, policy->cpu);
       
        /* cpuinfo and default policy values */
        policy->governor = 0; //!!!CPUFREQ_DEFAULT_GOVERNOR;
        policy->cpuinfo.transition_latency = 1000;
        policy->cur = stock_freq;

        return cpufreq_frequency_table_cpuinfo(policy, &p4clockmod_table[0]);
}


static int cpufreq_p4_cpu_exit(struct cpufreq_policy *policy)
{
        cpufreq_frequency_table_put_attr(policy->cpu);    
        return cpufreq_p4_setdc(policy->cpu, DC_DISABLE);
}

static struct freq_attr* p4clockmod_attr[] = {
        &cpufreq_freq_attr_scaling_available_freqs,
        NULL,
};

static struct cpufreq_driver p4clockmod_driver = {
        .verify         = cpufreq_p4_verify,
        .target         = cpufreq_p4_target,
        .init           = cpufreq_p4_cpu_init,
        .exit           = cpufreq_p4_cpu_exit,
        .name           = "p4-clockmod",
        .owner          = THIS_MODULE,
        .attr           = p4clockmod_attr,
};


/*static*/ int __init cpufreq_p4_init(void)
{      
        struct cpuinfo_x86 *c = cpu_data;

        /*
         * THERM_CONTROL is architectural for IA32 now, so
         * we can rely on the capability checks
         */

        if (c->x86_vendor != X86_VENDOR_INTEL)
                return -ENODEV;

        if (!test_bit(X86_FEATURE_ACPI, c->x86_capability) ||
                !test_bit(X86_FEATURE_ACC, c->x86_capability))
                return -ENODEV;

        printk(KERN_INFO PFX "P4/Xeon(TM) CPU On-Demand Clock Modulation available\n");

        return cpufreq_register_driver(&p4clockmod_driver);
}


/*static*/ void __exit cpufreq_p4_exit(void)
{
        cpufreq_unregister_driver(&p4clockmod_driver);
}


MODULE_PARM(stock_freq, "i");

MODULE_AUTHOR ("Zwane Mwaikambo <zwane@commfireservices.com>");
MODULE_DESCRIPTION ("cpufreq driver for Pentium(TM) 4/Xeon(TM)");
MODULE_LICENSE ("GPL");

module_init(cpufreq_p4_init);
module_exit(cpufreq_p4_exit);