/*
- * (c) 2003, 2004 Advanced Micro Devices, Inc.
+ * (c) 2003, 2004, 2005 Advanced Micro Devices, Inc.
* Your use of this code is subject to the terms and conditions of the
* GNU general public license version 2. See "COPYING" or
* http://www.gnu.org/licenses/gpl.html
*
- * Support : paul.devriendt@amd.com
+ * Support : mark.langsdorf@amd.com
*
* Based on the powernow-k7.c module written by Dave Jones.
* (C) 2003 Dave Jones <davej@codemonkey.org.uk> on behalf of SuSE Labs
*
* Valuable input gratefully received from Dave Jones, Pavel Machek,
* Dominik Brodowski, and others.
+ * Originally developed by Paul Devriendt.
* Processor information obtained from Chapter 9 (Power and Thermal Management)
* of the "BIOS and Kernel Developer's Guide for the AMD Athlon 64 and AMD
* Opteron Processors" available for download from www.amd.com
+ *
+ * Tables for specific CPUs can be infrerred from
+ * http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_docs/30430.pdf
*/
#include <linux/kernel.h>
#include <linux/cpufreq.h>
#include <linux/slab.h>
#include <linux/string.h>
+#include <linux/cpumask.h>
+#include <linux/sched.h> /* for current / set_cpus_allowed() */
#include <asm/msr.h>
#include <asm/io.h>
#ifdef CONFIG_X86_POWERNOW_K8_ACPI
#include <linux/acpi.h>
+#include <linux/mutex.h>
#include <acpi/processor.h>
#endif
#define PFX "powernow-k8: "
#define BFX PFX "BIOS error: "
-#define VERSION "version 1.00.09e"
+#define VERSION "version 1.60.2"
#include "powernow-k8.h"
/* serialize freq changes */
-static DECLARE_MUTEX(fidvid_sem);
+static DEFINE_MUTEX(fidvid_mutex);
static struct powernow_k8_data *powernow_data[NR_CPUS];
+#ifndef CONFIG_SMP
+static cpumask_t cpu_core_map[1];
+#endif
+
/* Return a frequency in MHz, given an input fid */
static u32 find_freq_from_fid(u32 fid)
{
return 1550-vid*25;
}
-/* Return the vco fid for an input fid */
+/* Return the vco fid for an input fid
+ *
+ * Each "low" fid has corresponding "high" fid, and you can get to "low" fids
+ * only from corresponding high fids. This returns "high" fid corresponding to
+ * "low" one.
+ */
static u32 convert_fid_to_vco_fid(u32 fid)
{
- if (fid < HI_FID_TABLE_BOTTOM) {
+ if (fid < HI_FID_TABLE_BOTTOM)
return 8 + (2 * fid);
- } else {
+ else
return fid;
- }
}
/*
u32 lo, hi;
u32 i = 0;
- lo = MSR_S_LO_CHANGE_PENDING;
- while (lo & MSR_S_LO_CHANGE_PENDING) {
- if (i++ > 0x1000000) {
- printk(KERN_ERR PFX "detected change pending stuck\n");
+ do {
+ if (i++ > 10000) {
+ dprintk("detected change pending stuck\n");
return 1;
}
rdmsr(MSR_FIDVID_STATUS, lo, hi);
- }
+ } while (lo & MSR_S_LO_CHANGE_PENDING);
data->currvid = hi & MSR_S_HI_CURRENT_VID;
data->currfid = lo & MSR_S_LO_CURRENT_FID;
{
u32 lo;
u32 savevid = data->currvid;
+ u32 i = 0;
if ((fid & INVALID_FID_MASK) || (data->currvid & INVALID_VID_MASK)) {
printk(KERN_ERR PFX "internal error - overflow on fid write\n");
dprintk("writing fid 0x%x, lo 0x%x, hi 0x%x\n",
fid, lo, data->plllock * PLL_LOCK_CONVERSION);
- wrmsr(MSR_FIDVID_CTL, lo, data->plllock * PLL_LOCK_CONVERSION);
-
- if (query_current_values_with_pending_wait(data))
- return 1;
+ do {
+ wrmsr(MSR_FIDVID_CTL, lo, data->plllock * PLL_LOCK_CONVERSION);
+ if (i++ > 100) {
+ printk(KERN_ERR PFX "internal error - pending bit very stuck - no further pstate changes possible\n");
+ return 1;
+ }
+ } while (query_current_values_with_pending_wait(data));
count_off_irt(data);
{
u32 lo;
u32 savefid = data->currfid;
+ int i = 0;
if ((data->currfid & INVALID_FID_MASK) || (vid & INVALID_VID_MASK)) {
printk(KERN_ERR PFX "internal error - overflow on vid write\n");
dprintk("writing vid 0x%x, lo 0x%x, hi 0x%x\n",
vid, lo, STOP_GRANT_5NS);
- wrmsr(MSR_FIDVID_CTL, lo, STOP_GRANT_5NS);
-
- if (query_current_values_with_pending_wait(data))
- return 1;
+ do {
+ wrmsr(MSR_FIDVID_CTL, lo, STOP_GRANT_5NS);
+ if (i++ > 100) {
+ printk(KERN_ERR PFX "internal error - pending bit very stuck - no further pstate changes possible\n");
+ return 1;
+ }
+ } while (query_current_values_with_pending_wait(data));
if (savefid != data->currfid) {
printk(KERN_ERR PFX "fid changed on vid trans, old 0x%x new 0x%x\n",
/*
* Reduce the vid by the max of step or reqvid.
* Decreasing vid codes represent increasing voltages:
- * vid of 0 is 1.550V, vid of 0x1e is 0.800V, vid of 0x1f is off.
+ * vid of 0 is 1.550V, vid of 0x1e is 0.800V, vid of VID_OFF is off.
*/
static int decrease_vid_code_by_step(struct powernow_k8_data *data, u32 reqvid, u32 step)
{
{
u32 rvosteps = data->rvo;
u32 savefid = data->currfid;
+ u32 maxvid, lo;
dprintk("ph1 (cpu%d): start, currfid 0x%x, currvid 0x%x, reqvid 0x%x, rvo 0x%x\n",
smp_processor_id(),
data->currfid, data->currvid, reqvid, data->rvo);
+ rdmsr(MSR_FIDVID_STATUS, lo, maxvid);
+ maxvid = 0x1f & (maxvid >> 16);
+ dprintk("ph1 maxvid=0x%x\n", maxvid);
+ if (reqvid < maxvid) /* lower numbers are higher voltages */
+ reqvid = maxvid;
+
while (data->currvid > reqvid) {
dprintk("ph1: curr 0x%x, req vid 0x%x\n",
data->currvid, reqvid);
return 1;
}
- while (rvosteps > 0) {
- if (data->currvid == 0) {
+ while ((rvosteps > 0) && ((data->rvo + data->currvid) > reqvid)) {
+ if (data->currvid == maxvid) {
rvosteps = 0;
} else {
dprintk("ph1: changing vid for rvo, req 0x%x\n",
/* Phase 2 - core frequency transition */
static int core_frequency_transition(struct powernow_k8_data *data, u32 reqfid)
{
- u32 vcoreqfid;
- u32 vcocurrfid;
- u32 vcofiddiff;
- u32 savevid = data->currvid;
+ u32 vcoreqfid, vcocurrfid, vcofiddiff, fid_interval, savevid = data->currvid;
if ((reqfid < HI_FID_TABLE_BOTTOM) && (data->currfid < HI_FID_TABLE_BOTTOM)) {
printk(KERN_ERR PFX "ph2: illegal lo-lo transition 0x%x 0x%x\n",
: vcoreqfid - vcocurrfid;
while (vcofiddiff > 2) {
+ (data->currfid & 1) ? (fid_interval = 1) : (fid_interval = 2);
+
if (reqfid > data->currfid) {
if (data->currfid > LO_FID_TABLE_TOP) {
- if (write_new_fid(data, data->currfid + 2)) {
+ if (write_new_fid(data, data->currfid + fid_interval)) {
return 1;
}
} else {
}
}
} else {
- if (write_new_fid(data, data->currfid - 2))
+ if (write_new_fid(data, data->currfid - fid_interval))
return 1;
}
oldmask = current->cpus_allowed;
set_cpus_allowed(current, cpumask_of_cpu(cpu));
- schedule();
if (smp_processor_id() != cpu) {
- printk(KERN_ERR "limiting to cpu %u failed\n", cpu);
+ printk(KERN_ERR PFX "limiting to cpu %u failed\n", cpu);
goto out;
}
goto out;
eax = cpuid_eax(CPUID_PROCESSOR_SIGNATURE);
+ if ((eax & CPUID_XFAM) != CPUID_XFAM_K8)
+ goto out;
+
if (((eax & CPUID_USE_XFAM_XMOD) != CPUID_USE_XFAM_XMOD) ||
- ((eax & CPUID_XFAM) != CPUID_XFAM_K8) ||
- ((eax & CPUID_XMOD) > CPUID_XMOD_REV_E)) {
+ ((eax & CPUID_XMOD) > CPUID_XMOD_REV_G)) {
printk(KERN_INFO PFX "Processor cpuid %x not supported\n", eax);
goto out;
}
out:
set_cpus_allowed(current, oldmask);
- schedule();
return rc;
-
}
static int check_pst_table(struct powernow_k8_data *data, struct pst_s *pst, u8 maxvid)
printk(KERN_ERR BFX "maxvid exceeded with pstate %d\n", j);
return -ENODEV;
}
- if ((pst[j].fid > MAX_FID)
- || (pst[j].fid & 1)
- || (j && (pst[j].fid < HI_FID_TABLE_BOTTOM))) {
+ if (pst[j].fid > MAX_FID) {
+ printk(KERN_ERR BFX "maxfid exceeded with pstate %d\n", j);
+ return -ENODEV;
+ }
+ if (j && (pst[j].fid < HI_FID_TABLE_BOTTOM)) {
/* Only first fid is allowed to be in "low" range */
- printk(KERN_ERR PFX "fid %d invalid : 0x%x\n", j, pst[j].fid);
+ printk(KERN_ERR BFX "two low fids - %d : 0x%x\n", j, pst[j].fid);
return -EINVAL;
}
if (pst[j].fid < lastfid)
lastfid = pst[j].fid;
}
if (lastfid & 1) {
- printk(KERN_ERR PFX "lastfid invalid\n");
+ printk(KERN_ERR BFX "lastfid invalid\n");
return -EINVAL;
}
if (lastfid > LO_FID_TABLE_TOP)
- printk(KERN_INFO PFX "first fid not from lo freq table\n");
+ printk(KERN_INFO BFX "first fid not from lo freq table\n");
return 0;
}
dprintk("table vers: 0x%x\n", psb->tableversion);
if (psb->tableversion != PSB_VERSION_1_4) {
- printk(KERN_INFO BFX "PSB table is not v1.4\n");
+ printk(KERN_ERR BFX "PSB table is not v1.4\n");
return -ENODEV;
}
return -ENODEV;
}
- data->vstable = psb->voltagestabilizationtime;
+ data->vstable = psb->vstable;
dprintk("voltage stabilization time: %d(*20us)\n", data->vstable);
dprintk("flags2: 0x%x\n", psb->flags2);
dprintk("isochronous relief time: %d\n", data->irt);
dprintk("maximum voltage step: %d - 0x%x\n", mvs, data->vidmvs);
- dprintk("numpst: 0x%x\n", psb->numpst);
- cpst = psb->numpst;
+ dprintk("numpst: 0x%x\n", psb->num_tables);
+ cpst = psb->num_tables;
if ((psb->cpuid == 0x00000fc0) || (psb->cpuid == 0x00000fe0) ){
thiscpuid = cpuid_eax(CPUID_PROCESSOR_SIGNATURE);
if ((thiscpuid == 0x00000fc0) || (thiscpuid == 0x00000fe0) ) {
dprintk("maxvid: 0x%x\n", psb->maxvid);
maxvid = psb->maxvid;
- data->numps = psb->numpstates;
+ data->numps = psb->numps;
dprintk("numpstates: 0x%x\n", data->numps);
return fill_powernow_table(data, (struct pst_s *)(psb+1), maxvid);
}
* BIOS and Kernel Developer's Guide, which is available on
* www.amd.com
*/
- printk(KERN_ERR PFX "BIOS error - no PSB\n");
+ printk(KERN_ERR PFX "BIOS error - no PSB or ACPI _PSS objects\n");
return -ENODEV;
}
data->irt = (data->acpi_data.states[index].control >> IRT_SHIFT) & IRT_MASK;
data->rvo = (data->acpi_data.states[index].control >> RVO_SHIFT) & RVO_MASK;
+ data->exttype = (data->acpi_data.states[index].control >> EXT_TYPE_SHIFT) & EXT_TYPE_MASK;
data->plllock = (data->acpi_data.states[index].control >> PLL_L_SHIFT) & PLL_L_MASK;
data->vidmvs = 1 << ((data->acpi_data.states[index].control >> MVS_SHIFT) & MVS_MASK);
data->vstable = (data->acpi_data.states[index].control >> VST_SHIFT) & VST_MASK;
struct cpufreq_frequency_table *powernow_table;
if (acpi_processor_register_performance(&data->acpi_data, data->cpu)) {
- dprintk("register performance failed\n");
+ dprintk("register performance failed: bad ACPI data\n");
return -EIO;
}
}
for (i = 0; i < data->acpi_data.state_count; i++) {
- u32 fid = data->acpi_data.states[i].control & FID_MASK;
- u32 vid = (data->acpi_data.states[i].control >> VID_SHIFT) & VID_MASK;
+ u32 fid;
+ u32 vid;
+
+ if (data->exttype) {
+ fid = data->acpi_data.states[i].status & FID_MASK;
+ vid = (data->acpi_data.states[i].status >> VID_SHIFT) & VID_MASK;
+ } else {
+ fid = data->acpi_data.states[i].control & FID_MASK;
+ vid = (data->acpi_data.states[i].control >> VID_SHIFT) & VID_MASK;
+ }
dprintk(" %d : fid 0x%x, vid 0x%x\n", i, fid, vid);
}
/* verify voltage is OK - BIOSs are using "off" to indicate invalid */
- if (vid == 0x1f) {
+ if (vid == VID_OFF) {
dprintk("invalid vid %u, ignoring\n", vid);
powernow_table[i].frequency = CPUFREQ_ENTRY_INVALID;
continue;
}
- if (fid < HI_FID_TABLE_BOTTOM) {
- if (cntlofreq) {
- /* if both entries are the same, ignore this
- * one...
- */
- if ((powernow_table[i].frequency != powernow_table[cntlofreq].frequency) ||
- (powernow_table[i].index != powernow_table[cntlofreq].index)) {
- printk(KERN_ERR PFX "Too many lo freq table entries\n");
- goto err_out_mem;
- }
-
- dprintk("double low frequency table entry, ignoring it.\n");
- powernow_table[i].frequency = CPUFREQ_ENTRY_INVALID;
- continue;
- } else
- cntlofreq = i;
+ /* verify only 1 entry from the lo frequency table */
+ if (fid < HI_FID_TABLE_BOTTOM) {
+ if (cntlofreq) {
+ /* if both entries are the same, ignore this one ... */
+ if ((powernow_table[i].frequency != powernow_table[cntlofreq].frequency) ||
+ (powernow_table[i].index != powernow_table[cntlofreq].index)) {
+ printk(KERN_ERR PFX "Too many lo freq table entries\n");
+ goto err_out_mem;
+ }
+
+ dprintk("double low frequency table entry, ignoring it.\n");
+ powernow_table[i].frequency = CPUFREQ_ENTRY_INVALID;
+ continue;
+ } else
+ cntlofreq = i;
}
if (powernow_table[i].frequency != (data->acpi_data.states[i].core_frequency * 1000)) {
{
u32 fid;
u32 vid;
- int res;
+ int res, i;
struct cpufreq_freqs freqs;
dprintk("cpu %d transition to index %u\n", smp_processor_id(), index);
/* fid are the lower 8 bits of the index we stored into
- * the cpufreq frequency table in find_psb_table, vid are
+ * the cpufreq frequency table in find_psb_table, vid are
* the upper 8 bits.
*/
}
if ((fid < HI_FID_TABLE_BOTTOM) && (data->currfid < HI_FID_TABLE_BOTTOM)) {
- printk("ignoring illegal change in lo freq table-%x to 0x%x\n",
+ printk(KERN_ERR PFX
+ "ignoring illegal change in lo freq table-%x to 0x%x\n",
data->currfid, fid);
return 1;
}
smp_processor_id(), fid, vid);
freqs.cpu = data->cpu;
-
freqs.old = find_khz_freq_from_fid(data->currfid);
freqs.new = find_khz_freq_from_fid(fid);
- cpufreq_notify_transition(&freqs, CPUFREQ_PRECHANGE);
+ for_each_cpu_mask(i, cpu_core_map[data->cpu]) {
+ freqs.cpu = i;
+ cpufreq_notify_transition(&freqs, CPUFREQ_PRECHANGE);
+ }
- down(&fidvid_sem);
res = transition_fid_vid(data, fid, vid);
- up(&fidvid_sem);
freqs.new = find_khz_freq_from_fid(data->currfid);
- cpufreq_notify_transition(&freqs, CPUFREQ_POSTCHANGE);
-
+ for_each_cpu_mask(i, cpu_core_map[data->cpu]) {
+ freqs.cpu = i;
+ cpufreq_notify_transition(&freqs, CPUFREQ_POSTCHANGE);
+ }
return res;
}
{
cpumask_t oldmask = CPU_MASK_ALL;
struct powernow_k8_data *data = powernow_data[pol->cpu];
- u32 checkfid = data->currfid;
- u32 checkvid = data->currvid;
+ u32 checkfid;
+ u32 checkvid;
unsigned int newstate;
int ret = -EIO;
+ if (!data)
+ return -EINVAL;
+
+ checkfid = data->currfid;
+ checkvid = data->currvid;
+
/* only run on specific CPU from here on */
oldmask = current->cpus_allowed;
set_cpus_allowed(current, cpumask_of_cpu(pol->cpu));
- schedule();
if (smp_processor_id() != pol->cpu) {
- printk(KERN_ERR "limiting to cpu %u failed\n", pol->cpu);
+ printk(KERN_ERR PFX "limiting to cpu %u failed\n", pol->cpu);
goto err_out;
}
data->currfid, data->currvid);
if ((checkvid != data->currvid) || (checkfid != data->currfid)) {
- printk(KERN_ERR PFX
- "error - out of sync, fid 0x%x 0x%x, vid 0x%x 0x%x\n",
- checkfid, data->currfid, checkvid, data->currvid);
+ printk(KERN_INFO PFX
+ "error - out of sync, fix 0x%x 0x%x, vid 0x%x 0x%x\n",
+ checkfid, data->currfid, checkvid, data->currvid);
}
if (cpufreq_frequency_table_target(pol, data->powernow_table, targfreq, relation, &newstate))
goto err_out;
+ mutex_lock(&fidvid_mutex);
+
powernow_k8_acpi_pst_values(data, newstate);
if (transition_frequency(data, newstate)) {
printk(KERN_ERR PFX "transition frequency failed\n");
ret = 1;
+ mutex_unlock(&fidvid_mutex);
goto err_out;
}
+ mutex_unlock(&fidvid_mutex);
pol->cur = find_khz_freq_from_fid(data->currfid);
ret = 0;
err_out:
set_cpus_allowed(current, oldmask);
- schedule();
-
return ret;
}
{
struct powernow_k8_data *data = powernow_data[pol->cpu];
+ if (!data)
+ return -EINVAL;
+
return cpufreq_frequency_table_verify(pol, data->powernow_table);
}
/* per CPU init entry point to the driver */
-static int __init powernowk8_cpu_init(struct cpufreq_policy *pol)
+static int __cpuinit powernowk8_cpu_init(struct cpufreq_policy *pol)
{
struct powernow_k8_data *data;
cpumask_t oldmask = CPU_MASK_ALL;
int rc;
+ if (!cpu_online(pol->cpu))
+ return -ENODEV;
+
if (!check_supported_cpu(pol->cpu))
return -ENODEV;
- data = kmalloc(sizeof(struct powernow_k8_data), GFP_KERNEL);
+ data = kzalloc(sizeof(struct powernow_k8_data), GFP_KERNEL);
if (!data) {
printk(KERN_ERR PFX "unable to alloc powernow_k8_data");
return -ENOMEM;
}
- memset(data,0,sizeof(struct powernow_k8_data));
data->cpu = pol->cpu;
* an UP version, and is deprecated by AMD.
*/
- if ((num_online_cpus() != 1) || (num_possible_cpus() != 1)) {
- printk(KERN_INFO PFX "MP systems not supported by PSB BIOS structure\n");
+ if (num_online_cpus() != 1) {
+ printk(KERN_ERR PFX "MP systems not supported by PSB BIOS structure\n");
kfree(data);
return -ENODEV;
}
/* only run on specific CPU from here on */
oldmask = current->cpus_allowed;
set_cpus_allowed(current, cpumask_of_cpu(pol->cpu));
- schedule();
if (smp_processor_id() != pol->cpu) {
- printk(KERN_ERR "limiting to cpu %u failed\n", pol->cpu);
+ printk(KERN_ERR PFX "limiting to cpu %u failed\n", pol->cpu);
goto err_out;
}
/* run on any CPU again */
set_cpus_allowed(current, oldmask);
- schedule();
pol->governor = CPUFREQ_DEFAULT_GOVERNOR;
+ pol->cpus = cpu_core_map[pol->cpu];
- /* Take a crude guess here.
+ /* Take a crude guess here.
* That guess was in microseconds, so multiply with 1000 */
pol->cpuinfo.transition_latency = (((data->rvo + 8) * data->vstable * VST_UNITS_20US)
+ (3 * (1 << data->irt) * 10)) * 1000;
/* min/max the cpu is capable of */
if (cpufreq_frequency_table_cpuinfo(pol, data->powernow_table)) {
printk(KERN_ERR PFX "invalid powernow_table\n");
+ powernow_k8_cpu_exit_acpi(data);
kfree(data->powernow_table);
kfree(data);
return -EINVAL;
err_out:
set_cpus_allowed(current, oldmask);
- schedule();
+ powernow_k8_cpu_exit_acpi(data);
kfree(data);
return -ENODEV;
static unsigned int powernowk8_get (unsigned int cpu)
{
- struct powernow_k8_data *data = powernow_data[cpu];
+ struct powernow_k8_data *data;
cpumask_t oldmask = current->cpus_allowed;
unsigned int khz = 0;
+ data = powernow_data[first_cpu(cpu_core_map[cpu])];
+
+ if (!data)
+ return -EINVAL;
+
set_cpus_allowed(current, cpumask_of_cpu(cpu));
if (smp_processor_id() != cpu) {
printk(KERN_ERR PFX "limiting to CPU %d failed in powernowk8_get\n", cpu);
set_cpus_allowed(current, oldmask);
return 0;
}
- preempt_disable();
if (query_current_values_with_pending_wait(data))
goto out;
khz = find_khz_freq_from_fid(data->currfid);
- out:
- preempt_enable_no_resched();
+out:
set_cpus_allowed(current, oldmask);
-
return khz;
}
};
/* driver entry point for init */
-static int __init powernowk8_init(void)
+static int __cpuinit powernowk8_init(void)
{
unsigned int i, supported_cpus = 0;
- for (i=0; i<NR_CPUS; i++) {
- if (!cpu_online(i))
- continue;
+ for_each_online_cpu(i) {
if (check_supported_cpu(i))
supported_cpus++;
}
if (supported_cpus == num_online_cpus()) {
- printk(KERN_INFO PFX "Found %d AMD Athlon 64 / Opteron processors (" VERSION ")\n",
- supported_cpus);
+ printk(KERN_INFO PFX "Found %d AMD Athlon 64 / Opteron "
+ "processors (" VERSION ")\n", supported_cpus);
return cpufreq_register_driver(&cpufreq_amd64_driver);
}
cpufreq_unregister_driver(&cpufreq_amd64_driver);
}
-MODULE_AUTHOR("Paul Devriendt <paul.devriendt@amd.com>");
+MODULE_AUTHOR("Paul Devriendt <paul.devriendt@amd.com> and Mark Langsdorf <mark.langsdorf@amd.com>");
MODULE_DESCRIPTION("AMD Athlon 64 and Opteron processor frequency driver.");
MODULE_LICENSE("GPL");