cpufreq: Linux support for the architectural pstate driver

With the third generation Opteron parts, AMD switched to an
architecturally defined interface for PowerNow! that uses
different MSRs than previous versions.

This patch brings the PowerNow! driver up to match the mainline
Linux driver and provide support for all AMD parts that use
or will use the architectural pstate interface.

It also removes a WARN_ON statement in kernel/cpu.c that
highlights a cpu hotplug locking issue in the ondemand cpufreq
governor.  It is only a warning message and the scope of
the changes to properly surpress it is a bit large to add to
the Xen 2.6.18.8 kernel.  I will backport them if people
think that's a better idea.

Signed-off-by: Mark Langsdorf <mark.langsdorf@amd.com>

diff --git a/arch/i386/kernel/cpu/cpufreq/powernow-k8.c b/arch/i386/kernel/cpu/cpufreq/powernow-k8.c
index e474a3e277c3464bb70e39799e4cc6bbdae94f9e..9985b5ffef68ac4a66be8a66421fdf3be84857ee 100644 (file)
--- a/arch/i386/kernel/cpu/cpufreq/powernow-k8.c
+++ b/arch/i386/kernel/cpu/cpufreq/powernow-k8.c
@@ -46,7 +46,7 @@
 
 #define PFX "powernow-k8: "
 #define BFX PFX "BIOS error: "
-#define VERSION "version 2.00.00"
+#define VERSION "version 2.20.00"
 #include "powernow-k8.h"
 
 /* serialize freq changes  */
@@ -66,36 +66,15 @@ static u32 find_freq_from_fid(u32 fid)
        return 800 + (fid * 100);
 }
 
-
 /* Return a frequency in KHz, given an input fid */
 static u32 find_khz_freq_from_fid(u32 fid)
 {
        return 1000 * find_freq_from_fid(fid);
 }
 
-/* Return a frequency in MHz, given an input fid and did */
-static u32 find_freq_from_fiddid(u32 fid, u32 did)
-{
-       return 100 * (fid + 0x10) >> did;
-}
-
-static u32 find_khz_freq_from_fiddid(u32 fid, u32 did)
-{
-       return 1000 * find_freq_from_fiddid(fid, did);
-}
-
-static u32 find_fid_from_pstate(u32 pstate)
+static u32 find_khz_freq_from_pstate(struct cpufreq_frequency_table *data, u32 pstate)
 {
-       u32 hi, lo;
-       rdmsr(MSR_PSTATE_DEF_BASE + pstate, lo, hi);
-       return lo & HW_PSTATE_FID_MASK;
-}
-
-static u32 find_did_from_pstate(u32 pstate)
-{
-       u32 hi, lo;
-       rdmsr(MSR_PSTATE_DEF_BASE + pstate, lo, hi);
-       return (lo & HW_PSTATE_DID_MASK) >> HW_PSTATE_DID_SHIFT;
+       return data[pstate].frequency;
 }
 
 /* Return the vco fid for an input fid
@@ -139,9 +118,7 @@ static int query_current_values_with_pending_wait(struct powernow_k8_data *data)
        if (cpu_family == CPU_HW_PSTATE) {
                rdmsr(MSR_PSTATE_STATUS, lo, hi);
                i = lo & HW_PSTATE_MASK;
-               rdmsr(MSR_PSTATE_DEF_BASE + i, lo, hi);
-               data->currfid = lo & HW_PSTATE_FID_MASK;
-               data->currdid = (lo & HW_PSTATE_DID_MASK) >> HW_PSTATE_DID_SHIFT;
+               data->currpstate = i;
                return 0;
        }
        do {
@@ -292,7 +269,7 @@ static int decrease_vid_code_by_step(struct powernow_k8_data *data, u32 reqvid,
 static int transition_pstate(struct powernow_k8_data *data, u32 pstate)
 {
        wrmsr(MSR_PSTATE_CTRL, pstate, 0);
-       data->currfid = find_fid_from_pstate(pstate);
+       data->currpstate = pstate;
        return 0;
 }
 
@@ -882,41 +859,23 @@ err_out:
 static int fill_powernow_table_pstate(struct powernow_k8_data *data, struct cpufreq_frequency_table *powernow_table)
 {
        int i;
+       u32 hi = 0, lo = 0;
+       rdmsr(MSR_PSTATE_CUR_LIMIT, hi, lo);
+       data->max_hw_pstate = (hi & HW_PSTATE_MAX_MASK) >> HW_PSTATE_MAX_SHIFT;
 
        for (i = 0; i < data->acpi_data->state_count; i++) {
                u32 index;
-               u32 hi = 0, lo = 0;
-               u32 fid;
-               u32 did;
 
                index = data->acpi_data->states[i].control & HW_PSTATE_MASK;
-               if (index > MAX_HW_PSTATE) {
+               if (index > data->max_hw_pstate) {
                        printk(KERN_ERR PFX "invalid pstate %d - bad value %d.\n", i, index);
                        printk(KERN_ERR PFX "Please report to BIOS manufacturer\n");
-               }
-               rdmsr(MSR_PSTATE_DEF_BASE + index, lo, hi);
-               if (!(hi & HW_PSTATE_VALID_MASK)) {
-                       dprintk("invalid pstate %d, ignoring\n", index);
-                       powernow_table[i].frequency = CPUFREQ_ENTRY_INVALID;
                        continue;
                }
 
-               fid = lo & HW_PSTATE_FID_MASK;
-               did = (lo & HW_PSTATE_DID_MASK) >> HW_PSTATE_DID_SHIFT;
-
-               dprintk("   %d : fid 0x%x, did 0x%x\n", index, fid, did);
+               powernow_table[i].index = index;
+               powernow_table[i].frequency = data->acpi_data->states[i].core_frequency * 1000;
 
-               powernow_table[i].index = index | (fid << HW_FID_INDEX_SHIFT) | (did << HW_DID_INDEX_SHIFT);
-
-               powernow_table[i].frequency = find_khz_freq_from_fiddid(fid, did);
-
-               if (powernow_table[i].frequency != (data->acpi_data->states[i].core_frequency * 1000)) {
-                       printk(KERN_INFO PFX "invalid freq entries %u kHz vs. %u kHz\n",
-                               powernow_table[i].frequency,
-                               (unsigned int) (data->acpi_data->states[i].core_frequency * 1000));
-                       powernow_table[i].frequency = CPUFREQ_ENTRY_INVALID;
-                       continue;
-               }
        }
        return 0;
 }
@@ -1058,22 +1017,18 @@ static int transition_frequency_fidvid(struct powernow_k8_data *data, unsigned i
 /* Take a frequency, and issue the hardware pstate transition command */
 static int transition_frequency_pstate(struct powernow_k8_data *data, unsigned int index)
 {
-       u32 fid = 0;
-       u32 did = 0;
        u32 pstate = 0;
        int res, i;
        struct cpufreq_freqs freqs;
 
        dprintk("cpu %d transition to index %u\n", smp_processor_id(), index);
 
-       /* get fid did for hardware pstate transition */
+       /* get MSR index for hardware pstate transition */
        pstate = index & HW_PSTATE_MASK;
-       if (pstate > MAX_HW_PSTATE)
+       if (pstate > data->max_hw_pstate)
                return 0;
-       fid = (index & HW_FID_INDEX_MASK) >> HW_FID_INDEX_SHIFT;
-       did = (index & HW_DID_INDEX_MASK) >> HW_DID_INDEX_SHIFT;
-       freqs.old = find_khz_freq_from_fiddid(data->currfid, data->currdid);
-       freqs.new = find_khz_freq_from_fiddid(fid, did);
+       freqs.old = find_khz_freq_from_pstate(data->powernow_table, data->currpstate);
+       freqs.new = find_khz_freq_from_pstate(data->powernow_table, pstate);
 
        for_each_cpu_mask(i, *(data->available_cores)) {
                freqs.cpu = i;
@@ -1081,9 +1036,7 @@ static int transition_frequency_pstate(struct powernow_k8_data *data, unsigned i
        }
 
        res = transition_pstate(data, pstate);
-       data->currfid = find_fid_from_pstate(pstate);
-       data->currdid = find_did_from_pstate(pstate);
-       freqs.new = find_khz_freq_from_fiddid(data->currfid, data->currdid);
+       freqs.new = find_khz_freq_from_pstate(data->powernow_table, pstate);
 
        for_each_cpu_mask(i, *(data->available_cores)) {
                freqs.cpu = i;
@@ -1128,10 +1081,7 @@ static int powernowk8_target(struct cpufreq_policy *pol, unsigned targfreq, unsi
        if (query_current_values_with_pending_wait(data))
                goto err_out;
 
-       if (cpu_family == CPU_HW_PSTATE)
-               dprintk("targ: curr fid 0x%x, did 0x%x\n",
-                       data->currfid, data->currvid);
-       else {
+       if (cpu_family != CPU_HW_PSTATE) {
                dprintk("targ: curr fid 0x%x, vid 0x%x\n",
                data->currfid, data->currvid);
 
@@ -1162,7 +1112,7 @@ static int powernowk8_target(struct cpufreq_policy *pol, unsigned targfreq, unsi
        mutex_unlock(&fidvid_mutex);
 
        if (cpu_family == CPU_HW_PSTATE)
-               pol->cur = find_khz_freq_from_fiddid(data->currfid, data->currdid);
+               pol->cur = find_khz_freq_from_pstate(data->powernow_table, newstate);
        else
                pol->cur = find_khz_freq_from_fid(data->currfid);
        ret = 0;
@@ -1259,7 +1209,7 @@ static int __cpuinit powernowk8_cpu_init(struct cpufreq_policy *pol)
            + (3 * (1 << data->irt) * 10)) * 1000;
 
        if (cpu_family == CPU_HW_PSTATE)
-               pol->cur = find_khz_freq_from_fiddid(data->currfid, data->currdid);
+               pol->cur = find_khz_freq_from_pstate(data->powernow_table, data->currpstate);
        else
                pol->cur = find_khz_freq_from_fid(data->currfid);
        dprintk("policy current frequency %d kHz\n", pol->cur);
@@ -1276,8 +1226,7 @@ static int __cpuinit powernowk8_cpu_init(struct cpufreq_policy *pol)
        cpufreq_frequency_table_get_attr(data->powernow_table, pol->cpu);
 
        if (cpu_family == CPU_HW_PSTATE)
-               dprintk("cpu_init done, current fid 0x%x, did 0x%x\n",
-                       data->currfid, data->currdid);
+               dprintk("cpu_init done, current pstate 0x%x\n", data->currpstate);
        else
                dprintk("cpu_init done, current fid 0x%x, vid 0x%x\n",
                        data->currfid, data->currvid);
@@ -1332,7 +1281,10 @@ static unsigned int powernowk8_get (unsigned int cpu)
        if (query_current_values_with_pending_wait(data))
                goto out;
 
-       khz = find_khz_freq_from_fid(data->currfid);
+       if (cpu_family == CPU_HW_PSTATE)
+               khz = find_khz_freq_from_pstate(data->powernow_table, data->currpstate);
+       else
+               khz = find_khz_freq_from_fid(data->currfid);
 
 out:
        set_cpus_allowed(current, oldmask);

diff --git a/kernel/cpu.c b/kernel/cpu.c
index f230f9ae01c239f21ea53e54087310ad4b2a7a0f..ea06709d7b74279087214cf8d85b2ed70969b3e8 100644 (file)
--- a/kernel/cpu.c
+++ b/kernel/cpu.c
@@ -48,7 +48,10 @@ EXPORT_SYMBOL_GPL(lock_cpu_hotplug);
 
 void unlock_cpu_hotplug(void)
 {
+/* cpufreq lock-takers fixed in mainline; shut up until dom0 kernel catches up*/
+#ifdef CONFIG_XEN
        WARN_ON(recursive != current);
+#endif
        if (recursive_depth) {
                recursive_depth--;
                return;