cpufreq: intel_pstate: Correct the busy calculation for KNL

commit 6e34e1f2 upstream.

The busy percent calculated for the Knights Landing (KNL) platform
is 1024 times smaller than the correct busy value.  This causes
performance to get stuck at the lowest ratio.

The scaling algorithm used for KNL is performance-based, but it still
looks at the CPU load to set the scaled busy factor to 0 when the
load is less than 1 percent.  In this case, since the computed load
is 1024x smaller than it should be, the scaled busy factor will
always be 0, irrespective of CPU business.

This needs a fix similar to the turbostat one in commit b2b34dfe
(tools/power turbostat: KNL workaround for %Busy and Avg_MHz).

For this reason, add one more callback to processor-specific
callbacks to specify an MPERF multiplier represented by a number of
bit positions to shift the value of that register to the left to
copmensate for its rate difference with respect to the TSC.  This
shift value is used during CPU busy calculations.

Fixes: ffb81056 (intel_pstate: Avoid getting stuck in high P-states when idle)
Reported-and-tested-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
[ rjw: Changelog ]
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drivers/cpufreq/intel_pstate.c

@@ -225,6 +225,9 @@ struct global_params {
  * @vid:		Stores VID limits for this CPU
  * @pid:		Stores PID parameters for this CPU
  * @last_sample_time:	Last Sample time
+ * @aperf_mperf_shift:	Number of clock cycles after aperf, merf is incremented
+ *			This shift is a multiplier to mperf delta to
+ *			calculate CPU busy.
  * @prev_aperf:		Last APERF value read from APERF MSR
  * @prev_mperf:		Last MPERF value read from MPERF MSR
  * @prev_tsc:		Last timestamp counter (TSC) value
@@ -261,6 +264,7 @@ struct cpudata {
 
 	u64	last_update;
 	u64	last_sample_time;
+	u64	aperf_mperf_shift;
 	u64	prev_aperf;
 	u64	prev_mperf;
 	u64	prev_tsc;
@@ -323,6 +327,7 @@ struct pstate_funcs {
 	int (*get_min)(void);
 	int (*get_turbo)(void);
 	int (*get_scaling)(void);
+	int (*get_aperf_mperf_shift)(void);
 	u64 (*get_val)(struct cpudata*, int pstate);
 	void (*get_vid)(struct cpudata *);
 	void (*update_util)(struct update_util_data *data, u64 time,
@@ -1485,6 +1490,11 @@ static u64 core_get_val(struct cpudata *cpudata, int pstate)
 	return val;
 }
 
+static int knl_get_aperf_mperf_shift(void)
+{
+	return 10;
+}
+
 static int knl_get_turbo_pstate(void)
 {
 	u64 value;
@@ -1543,6 +1553,9 @@ static void intel_pstate_get_cpu_pstates(struct cpudata *cpu)
 	cpu->pstate.max_freq = cpu->pstate.max_pstate * cpu->pstate.scaling;
 	cpu->pstate.turbo_freq = cpu->pstate.turbo_pstate * cpu->pstate.scaling;
 
+	if (pstate_funcs.get_aperf_mperf_shift)
+		cpu->aperf_mperf_shift = pstate_funcs.get_aperf_mperf_shift();
+
 	if (pstate_funcs.get_vid)
 		pstate_funcs.get_vid(cpu);
 
@@ -1619,7 +1632,8 @@ static inline int32_t get_target_pstate_use_cpu_load(struct cpudata *cpu)
 	if (cpu->policy == CPUFREQ_POLICY_PERFORMANCE)
 		return cpu->pstate.turbo_pstate;
 
-	busy_frac = div_fp(sample->mperf, sample->tsc);
+	busy_frac = div_fp(sample->mperf << cpu->aperf_mperf_shift,
+			   sample->tsc);
 
 	boost = cpu->iowait_boost;
 	cpu->iowait_boost >>= 1;
@@ -1681,7 +1695,8 @@ static inline int32_t get_target_pstate_use_performance(struct cpudata *cpu)
 		sample_ratio = div_fp(pid_params.sample_rate_ns, duration_ns);
 		perf_scaled = mul_fp(perf_scaled, sample_ratio);
 	} else {
-		sample_ratio = div_fp(100 * cpu->sample.mperf, cpu->sample.tsc);
+		sample_ratio = div_fp(100 * (cpu->sample.mperf << cpu->aperf_mperf_shift),
+				      cpu->sample.tsc);
 		if (sample_ratio < int_tofp(1))
 			perf_scaled = 0;
 	}
@@ -1824,6 +1839,7 @@ static const struct pstate_funcs knl_funcs = {
 	.get_max_physical = core_get_max_pstate_physical,
 	.get_min = core_get_min_pstate,
 	.get_turbo = knl_get_turbo_pstate,
+	.get_aperf_mperf_shift = knl_get_aperf_mperf_shift,
 	.get_scaling = core_get_scaling,
 	.get_val = core_get_val,
 	.update_util = intel_pstate_update_util_pid,
@@ -2408,6 +2424,7 @@ static void __init copy_cpu_funcs(struct pstate_funcs *funcs)
 	pstate_funcs.get_val   = funcs->get_val;
 	pstate_funcs.get_vid   = funcs->get_vid;
 	pstate_funcs.update_util = funcs->update_util;
+	pstate_funcs.get_aperf_mperf_shift = funcs->get_aperf_mperf_shift;
 
 	intel_pstate_use_acpi_profile();
 }