Bug 57411

Created attachment 100431 [details]
.config

The patch below fixes this issue it is in Rafael's next branch and has been sent to stable.

Author: Dirk Brandewie <dirk.j.brandewie@intel.com>  2013-05-07 08:20:25
Committer: Rafael J. Wysocki <rafael.j.wysocki@intel.com>  2013-05-12 05:04:16
Parent: 885f925eef411f549f17bc64dd054a3269cf66cd (Merge branch 'pm-cpufreq')
Child:  d8f469e9cff3bc4a6317d923e9506be046aa7bdc (cpufreq / intel_pstate: use lowest requested max performance)
Branches: remotes/linux-pm/bleeding-edge, remotes/linux-pm/linux-next
Follows: v3.10-rc1
Precedes: 

    cpufreq / intel_pstate: remove idle time and duration from sample and calculations
    
    Idle time is taken into account in the APERF/MPERF ratio calculation
    there is no reason for the driver to track it seperately.  This
    reduces the work in the driver and makes the code more readable.
    
    Removal of the tracking of sample duration removes the possibility of
    the divide by zero exception when the duration is sub 1us
    
    References: https://bugzilla.kernel.org/show_bug.cgi?id=56691
    Reported-by: Mike Lothian <mike@fireburn.co.uk>
    Cc: 3.9+ <stable@vger.kernel.org>
    Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
    Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>

------------------------ drivers/cpufreq/intel_pstate.c ------------------------
index cc3a8e6..c6e10d0 100644
@@ -48,12 +48,7 @@ static inline int32_t div_fp(int32_t x, int32_t y)
 }
 
 struct sample {
-	ktime_t start_time;
-	ktime_t end_time;
 	int core_pct_busy;
-	int pstate_pct_busy;
-	u64 duration_us;
-	u64 idletime_us;
 	u64 aperf;
 	u64 mperf;
 	int freq;
@@ -91,8 +86,6 @@ struct cpudata {
 	int min_pstate_count;
 	int idle_mode;
 
-	ktime_t prev_sample;
-	u64	prev_idle_time_us;
 	u64	prev_aperf;
 	u64	prev_mperf;
 	int	sample_ptr;
@@ -450,48 +443,26 @@ static inline void intel_pstate_calc_busy(struct cpudata *cpu,
 					struct sample *sample)
 {
 	u64 core_pct;
-	sample->pstate_pct_busy = 100 - div64_u64(
-					sample->idletime_us * 100,
-					sample->duration_us);
 	core_pct = div64_u64(sample->aperf * 100, sample->mperf);
 	sample->freq = cpu->pstate.max_pstate * core_pct * 1000;
 
-	sample->core_pct_busy = div_s64((sample->pstate_pct_busy * core_pct),
-					100);
+	sample->core_pct_busy = core_pct;
 }
 
 static inline void intel_pstate_sample(struct cpudata *cpu)
 {
-	ktime_t now;
-	u64 idle_time_us;
 	u64 aperf, mperf;
 
-	now = ktime_get();
-	idle_time_us = get_cpu_idle_time_us(cpu->cpu, NULL);
-
 	rdmsrl(MSR_IA32_APERF, aperf);
 	rdmsrl(MSR_IA32_MPERF, mperf);
-	/* for the first sample, don't actually record a sample, just
-	 * set the baseline */
-	if (cpu->prev_idle_time_us > 0) {
-		cpu->sample_ptr = (cpu->sample_ptr + 1) % SAMPLE_COUNT;
-		cpu->samples[cpu->sample_ptr].start_time = cpu->prev_sample;
-		cpu->samples[cpu->sample_ptr].end_time = now;
-		cpu->samples[cpu->sample_ptr].duration_us =
-			ktime_us_delta(now, cpu->prev_sample);
-		cpu->samples[cpu->sample_ptr].idletime_us =
-			idle_time_us - cpu->prev_idle_time_us;
-
-		cpu->samples[cpu->sample_ptr].aperf = aperf;
-		cpu->samples[cpu->sample_ptr].mperf = mperf;
-		cpu->samples[cpu->sample_ptr].aperf -= cpu->prev_aperf;
-		cpu->samples[cpu->sample_ptr].mperf -= cpu->prev_mperf;
-
-		intel_pstate_calc_busy(cpu, &cpu->samples[cpu->sample_ptr]);
-	}
+	cpu->sample_ptr = (cpu->sample_ptr + 1) % SAMPLE_COUNT;
+	cpu->samples[cpu->sample_ptr].aperf = aperf;
+	cpu->samples[cpu->sample_ptr].mperf = mperf;
+	cpu->samples[cpu->sample_ptr].aperf -= cpu->prev_aperf;
+	cpu->samples[cpu->sample_ptr].mperf -= cpu->prev_mperf;
+
+	intel_pstate_calc_busy(cpu, &cpu->samples[cpu->sample_ptr]);
 
-	cpu->prev_sample = now;
-	cpu->prev_idle_time_us = idle_time_us;
 	cpu->prev_aperf = aperf;
 	cpu->prev_mperf = mperf;
 }

Hi Dirk, I somehow did not get the email about the patch. Now with 3.10-rc4 I don't see the issue. Thank you.

Fixed by commit 1abc4b2 (cpufreq / intel_pstate: remove idle time and duration from sample and calculations).