Re: [2.5.53 - Oops] CPU Frequency scaling

Dominik Brodowski (linux@brodo.de)
Sun, 29 Dec 2002 22:03:41 +0100


--ikeVEW9yuYc//A+q
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline

Hi,

Thanks for testing cpufreq - and sorry for that oops. The attached patches
should resolve this problem. For further cpufreq updates, check
http://www.brodo.de/cpufreq/advanced.html

Dominik

--ikeVEW9yuYc//A+q
Content-Type: text/plain; charset=us-ascii
Content-Disposition: attachment; filename="cpufreq-2.5.53-1-p4-1"

[PATCH 2.5][RESEND] cpufreq: p4-clockmod bugfixes

The "get current state" algorithm wasn't aware of the disable/enable bit,
and the policy verification function wasn't aware of the N44 / O17 bug.
Also, some unused code is removed.

Dominik

diff -ruN linux-original/arch/i386/kernel/cpu/cpufreq/p4-clockmod.c linux/arch/i386/kernel/cpu/cpufreq/p4-clockmod.c
--- linux-original/arch/i386/kernel/cpu/cpufreq/p4-clockmod.c 2002-12-21 14:53:44.000000000 +0100
+++ linux/arch/i386/kernel/cpu/cpufreq/p4-clockmod.c 2002-12-21 18:24:29.000000000 +0100
@@ -82,12 +82,17 @@

/* get current state */
rdmsr(MSR_IA32_THERM_CONTROL, l, h);
- l = l >> 1;
- l &= 0x7;
-
+ 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 */
@@ -141,13 +146,18 @@
unsigned int i;
unsigned int newstate = 0;
unsigned int number_states = 0;
+ unsigned int minstate = 1;

- if (!cpufreq_p4_driver || !stock_freq || !policy)
+ if (!cpufreq_p4_driver || !stock_freq ||
+ !policy || !cpu_online(policy->cpu))
return -EINVAL;

+ if (has_N44_O17_errata)
+ minstate = 3;
+
if (policy->policy == CPUFREQ_POLICY_POWERSAVE)
{
- for (i=8; i>0; i--)
+ for (i=8; i>=minstate; i--)
if ((policy->min <= ((stock_freq / 8) * i)) &&
(policy->max >= ((stock_freq / 8) * i)))
{
@@ -155,7 +165,7 @@
number_states++;
}
} else {
- for (i=1; i<=8; i++)
+ for (i=minstate; i<=8; i++)
if ((policy->min <= ((stock_freq / 8) * i)) &&
(policy->max >= ((stock_freq / 8) * i)))
{
@@ -164,25 +174,8 @@
}
}

- /* if (number_states == 1) */
- {
- if (policy->cpu == CPUFREQ_ALL_CPUS) {
- for (i=0; i<NR_CPUS; i++)
- if (cpu_online(i))
- cpufreq_p4_setdc(i, newstate);
- } else {
- cpufreq_p4_setdc(policy->cpu, newstate);
- }
- }
- /* else {
- if (policy->policy == CPUFREQ_POLICY_POWERSAVE) {
- min_state = newstate;
- max_state = newstate + (number_states - 1);
- } else {
- max_state = newstate;
- min_state = newstate - (number_states - 1);
- }
- } */
+ cpufreq_p4_setdc(policy->cpu, newstate);
+
return 0;
}

@@ -190,17 +183,21 @@
static int cpufreq_p4_verify(struct cpufreq_policy *policy)
{
unsigned int number_states = 0;
- unsigned int i;
+ unsigned int i = 1;

- if (!cpufreq_p4_driver || !stock_freq || !policy)
+ if (!cpufreq_p4_driver || !stock_freq ||
+ !policy || !cpu_online(policy->cpu))
return -EINVAL;

- if (!cpu_online(policy->cpu))
- policy->cpu = CPUFREQ_ALL_CPUS;
- cpufreq_verify_within_limits(policy, (stock_freq / 8), stock_freq);
+ cpufreq_verify_within_limits(policy,
+ policy->cpuinfo.min_freq,
+ policy->cpuinfo.max_freq);
+
+ if (has_N44_O17_errata)
+ i = 3;

- /* is there at least one state within limit? */
- for (i=1; i<=8; i++)
+ /* is there at least one state within the limit? */
+ for (; i<=8; i++)
if ((policy->min <= ((stock_freq / 8) * i)) &&
(policy->max >= ((stock_freq / 8) * i)))
number_states++;
@@ -209,6 +206,9 @@
return 0;

policy->max = (stock_freq / 8) * (((unsigned int) ((policy->max * 8) / stock_freq)) + 1);
+ cpufreq_verify_within_limits(policy,
+ policy->cpuinfo.min_freq,
+ policy->cpuinfo.max_freq);
return 0;
}

@@ -292,13 +292,14 @@

void __exit cpufreq_p4_exit(void)
{
- u32 l, h;
+ unsigned int i;

if (cpufreq_p4_driver) {
+ for (i=0; i<NR_CPUS; i++) {
+ if (cpu_online(i))
+ cpufreq_p4_setdc(i, DC_DISABLE);
+ }
cpufreq_unregister();
- /* return back to a non modulated state */
- rdmsr(MSR_IA32_THERM_CONTROL, l, h);
- wrmsr(MSR_IA32_THERM_CONTROL, l & ~(1<<4), h);
kfree(cpufreq_p4_driver);
}
}

--ikeVEW9yuYc//A+q
Content-Type: text/plain; charset=us-ascii
Content-Disposition: attachment; filename="cpufreq-2.5.53-2-timer-2"

[PATCH] cpufreq: update timer notifiers

- The global loops_per_jiffy can only be safely adjusted on UP.
- x86 per-CPU loops_per_jiffy is not dependend of TSC, so move it to
non-TSC code.
- Save reference values so that rounding errors do not accumulate

arch/i386/kernel/cpu/common.c | 47 ++++++++++++++++++++++++++++++++++++
arch/i386/kernel/timers/timer_tsc.c | 44 ++++++++++++++++-----------------
kernel/cpufreq.c | 20 +++++++++++----
3 files changed, 84 insertions(+), 27 deletions(-)
diff -ruN linux-original/arch/i386/kernel/cpu/common.c linux/arch/i386/kernel/cpu/common.c
--- linux-original/arch/i386/kernel/cpu/common.c 2002-12-24 13:37:12.000000000 +0100
+++ linux/arch/i386/kernel/cpu/common.c 2002-12-24 14:44:14.000000000 +0100
@@ -1,5 +1,7 @@
#include <linux/init.h>
#include <linux/string.h>
+#include <linux/cpufreq.h>
+#include <linux/notifier.h>
#include <linux/delay.h>
#include <linux/smp.h>
#include <asm/semaphore.h>
@@ -64,6 +66,41 @@
#endif
__setup("notsc", tsc_setup);

+#ifdef CONFIG_CPU_FREQ
+static unsigned long loops_per_jiffy_ref = 0;
+static unsigned int ref_freq = 0;
+
+static int
+loops_per_jiffy_cpufreq_notifier(struct notifier_block *nb, unsigned long val,
+ void *data)
+{
+ struct cpufreq_freqs *freq = data;
+
+ if (!loops_per_jiffy_ref) {
+ loops_per_jiffy_ref = cpu_data[freq->cpu].loops_per_jiffy;
+ ref_freq = freq->old;
+ }
+
+ switch (val) {
+ case CPUFREQ_PRECHANGE:
+ if (freq->old < freq->new)
+ cpu_data[freq->cpu].loops_per_jiffy = cpufreq_scale(loops_per_jiffy_ref, ref_freq, freq->new);
+ break;
+ case CPUFREQ_POSTCHANGE:
+ if (freq->new < freq->old)
+ cpu_data[freq->cpu].loops_per_jiffy = cpufreq_scale(loops_per_jiffy_ref, ref_freq, freq->new);
+ break;
+ }
+
+ return 0;
+}
+
+static struct notifier_block loops_per_jiffy_cpufreq_notifier_block = {
+ .notifier_call = loops_per_jiffy_cpufreq_notifier
+};
+#endif
+
+
int __init get_model_name(struct cpuinfo_x86 *c)
{
unsigned int *v;
@@ -373,6 +410,16 @@
boot_cpu_data.x86_capability[i] &= c->x86_capability[i];
}

+ /*
+ * Update the per-CPU loops_per_jiffy count on CPU frequency
+ * transitions
+ */
+#ifdef CONFIG_CPU_FREQ
+ if (c == &boot_cpu_data) {
+ cpufreq_register_notifier(&loops_per_jiffy_cpufreq_notifier_block, CPUFREQ_TRANSITION_NOTIFIER);
+ }
+#endif
+
/* Init Machine Check Exception if available. */
#ifdef CONFIG_X86_MCE
mcheck_init(c);
diff -ruN linux-original/arch/i386/kernel/timers/timer_tsc.c linux/arch/i386/kernel/timers/timer_tsc.c
--- linux-original/arch/i386/kernel/timers/timer_tsc.c 2002-12-24 13:37:12.000000000 +0100
+++ linux/arch/i386/kernel/timers/timer_tsc.c 2002-12-24 14:41:28.000000000 +0100
@@ -186,45 +186,45 @@

#ifdef CONFIG_CPU_FREQ

+static unsigned long fast_gettimeoffset_ref = 0;
+static unsigned long cpu_khz_ref = 0;
+static unsigned int ref_freq = 0;
+
static int
-time_cpufreq_notifier(struct notifier_block *nb, unsigned long val,
+tsc_cpufreq_notifier(struct notifier_block *nb, unsigned long val,
void *data)
{
struct cpufreq_freqs *freq = data;
- unsigned int i;

- if (!cpu_has_tsc)
+ if (!use_tsc)
return 0;

+ if (!fast_gettimeoffset_ref) {
+ fast_gettimeoffset_ref = fast_gettimeoffset_quotient;
+ cpu_khz_ref = cpu_khz;
+ ref_freq = freq->old;
+ }
+
switch (val) {
case CPUFREQ_PRECHANGE:
- if ((freq->old < freq->new) &&
- ((freq->cpu == CPUFREQ_ALL_CPUS) || (freq->cpu == 0))) {
- cpu_khz = cpufreq_scale(cpu_khz, freq->old, freq->new);
- fast_gettimeoffset_quotient = cpufreq_scale(fast_gettimeoffset_quotient, freq->new, freq->old);
+ if (freq->old < freq->new) {
+ fast_gettimeoffset_quotient = cpufreq_scale(fast_gettimeoffset_ref, freq->new, ref_freq);
+ cpu_khz = cpufreq_scale(cpu_khz_ref, ref_freq, freq->new);
}
- for (i=0; i<NR_CPUS; i++)
- if ((freq->cpu == CPUFREQ_ALL_CPUS) || (freq->cpu == i))
- cpu_data[i].loops_per_jiffy = cpufreq_scale(cpu_data[i].loops_per_jiffy, freq->old, freq->new);
break;
-
case CPUFREQ_POSTCHANGE:
- if ((freq->new < freq->old) &&
- ((freq->cpu == CPUFREQ_ALL_CPUS) || (freq->cpu == 0))) {
- cpu_khz = cpufreq_scale(cpu_khz, freq->old, freq->new);
- fast_gettimeoffset_quotient = cpufreq_scale(fast_gettimeoffset_quotient, freq->new, freq->old);
+ if (freq->new < freq->old) {
+ fast_gettimeoffset_quotient = cpufreq_scale(fast_gettimeoffset_ref, freq->new, ref_freq);
+ cpu_khz = cpufreq_scale(cpu_khz_ref, ref_freq, freq->new);
}
- for (i=0; i<NR_CPUS; i++)
- if ((freq->cpu == CPUFREQ_ALL_CPUS) || (freq->cpu == i))
- cpu_data[i].loops_per_jiffy = cpufreq_scale(cpu_data[i].loops_per_jiffy, freq->old, freq->new);
break;
}

return 0;
}

-static struct notifier_block time_cpufreq_notifier_block = {
- .notifier_call = time_cpufreq_notifier
+static struct notifier_block tsc_cpufreq_notifier_block = {
+ .notifier_call = tsc_cpufreq_notifier
};
#endif

@@ -278,8 +278,8 @@
"0" (eax), "1" (edx));
printk("Detected %lu.%03lu MHz processor.\n", cpu_khz / 1000, cpu_khz % 1000);
}
-#ifdef CONFIG_CPU_FREQ
- cpufreq_register_notifier(&time_cpufreq_notifier_block, CPUFREQ_TRANSITION_NOTIFIER);
+#if defined(CONFIG_CPU_FREQ) && !defined(CONFIG_SMP)
+ cpufreq_register_notifier(&tsc_cpufreq_notifier_block, CPUFREQ_TRANSITION_NOTIFIER);
#endif
return 0;
}
diff -ruN linux-original/kernel/cpufreq.c linux/kernel/cpufreq.c
--- linux-original/kernel/cpufreq.c 2002-12-24 13:38:25.000000000 +0100
+++ linux/kernel/cpufreq.c 2002-12-24 14:46:25.000000000 +0100
@@ -936,17 +936,27 @@
* adjust_jiffies - adjust the system "loops_per_jiffy"
*
* This function alters the system "loops_per_jiffy" for the clock
- * speed change. Note that loops_per_jiffy is only updated if all
- * CPUs are affected - else there is a need for per-CPU loops_per_jiffy
- * values which are provided by various architectures.
+ * speed change. Note that loops_per_jiffy cannot be updated on SMP
+ * systems as each CPU might be scaled differently. So, use the arch
+ * per-CPU loops_per_jiffy value wherever possible.
*/
+#ifndef CONFIG_SMP
+static unsigned long l_p_j_ref = 0;
+static unsigned int l_p_j_ref_freq = 0;
+
static inline void adjust_jiffies(unsigned long val, struct cpufreq_freqs *ci)
{
+ if (!l_p_j_ref_freq) {
+ l_p_j_ref = loops_per_jiffy;
+ l_p_j_ref_freq = ci->old;
+ }
if ((val == CPUFREQ_PRECHANGE && ci->old < ci->new) ||
(val == CPUFREQ_POSTCHANGE && ci->old > ci->new))
- if (ci->cpu == CPUFREQ_ALL_CPUS)
- loops_per_jiffy = cpufreq_scale(loops_per_jiffy, ci->old, ci->new);
+ loops_per_jiffy = cpufreq_scale(l_p_j_ref, l_p_j_ref_freq, ci->new);
}
+#else
+#define adjust_jiffies(...)
+#endif


/**

--ikeVEW9yuYc//A+q--
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/