intel_idle: disable HW auto-demotion

Commit Message

Len Brown Jan. 19, 2011, 2:52 a.m. UTC

None

Patch

diff --git a/arch/x86/include/asm/msr-index.h b/arch/x86/include/asm/msr-index.h
index 4d0dfa0..b75eeab 100644
--- a/arch/x86/include/asm/msr-index.h
+++ b/arch/x86/include/asm/msr-index.h
@@ -36,6 +36,10 @@ 
 #define MSR_IA32_PERFCTR1		0x000000c2
 #define MSR_FSB_FREQ			0x000000cd
 
+#define MSR_NHM_SNB_PKG_CST_CFG_CTL	0x000000e2
+#define NHM_C3_AUTO_DEMOTE		(1UL << 25)
+#define NHM_C1_AUTO_DEMOTE		(1UL << 26)
+
 #define MSR_MTRRcap			0x000000fe
 #define MSR_IA32_BBL_CR_CTL		0x00000119
 
diff --git a/drivers/idle/intel_idle.c b/drivers/idle/intel_idle.c
index 7acb32e..3ee8c38 100644
--- a/drivers/idle/intel_idle.c
+++ b/drivers/idle/intel_idle.c
@@ -62,6 +62,7 @@ 
 #include <linux/notifier.h>
 #include <linux/cpu.h>
 #include <asm/mwait.h>
+#include <asm/msr.h>
 
 #define INTEL_IDLE_VERSION "0.4"
 #define PREFIX "intel_idle: "
@@ -85,6 +86,16 @@  static int intel_idle(struct cpuidle_device *dev, struct cpuidle_state *state);
 static struct cpuidle_state *cpuidle_state_table;
 
 /*
+ * Disable HW auto demotion on tick-less idle kernels
+ */
+static unsigned int has_nhm_snb_hw_auto_demotion;
+#ifdef CONFIG_NO_HZ
+static unsigned int auto_demotion;
+#else
+static unsigned int auto_demotion = 1;
+#endif
+
+/*
  * Set this flag for states where the HW flushes the TLB for us
  * and so we don't need cross-calls to keep it consistent.
  * If this flag is set, SW flushes the TLB, so even if the
@@ -285,6 +296,20 @@  static struct notifier_block __cpuinitdata setup_broadcast_notifier = {
 	.notifier_call = setup_broadcast_cpuhp_notify,
 };
 
+static long nhm_snb_auto_demotion_off(void *unused)
+{
+	unsigned long long msr_bits;
+
+	rdmsrl(MSR_NHM_SNB_PKG_CST_CFG_CTL, msr_bits);
+
+	msr_bits &= ~(NHM_C1_AUTO_DEMOTE | NHM_C3_AUTO_DEMOTE);
+
+	wrmsrl(MSR_NHM_SNB_PKG_CST_CFG_CTL, msr_bits);
+
+	return 0;
+}
+
+
 /*
  * intel_idle_probe()
  */
@@ -328,6 +353,7 @@  static int intel_idle_probe(void)
 	case 0x25:	/* Westmere */
 	case 0x2C:	/* Westmere */
 		cpuidle_state_table = nehalem_cstates;
+		has_nhm_snb_hw_auto_demotion = 1;
 		break;
 
 	case 0x1C:	/* 28 - Atom Processor */
@@ -338,6 +364,7 @@  static int intel_idle_probe(void)
 	case 0x2A:	/* SNB */
 	case 0x2D:	/* SNB Xeon */
 		cpuidle_state_table = snb_cstates;
+		has_nhm_snb_hw_auto_demotion = 1;
 		break;
 
 	default:
@@ -439,6 +466,8 @@  static int intel_idle_cpuidle_devices_init(void)
 			intel_idle_cpuidle_devices_uninit();
 			return -EIO;
 		}
+		if (has_nhm_snb_hw_auto_demotion && (auto_demotion == 0))
+			work_on_cpu(i, nhm_snb_auto_demotion_off, 0);
 	}
 
 	return 0;
@@ -490,6 +519,7 @@  module_init(intel_idle_init);
 module_exit(intel_idle_exit);
 
 module_param(max_cstate, int, 0444);
+module_param(auto_demotion, int, 0444);
 
 MODULE_AUTHOR("Len Brown <[e-mail]>");
 MODULE_DESCRIPTION("Cpuidle driver for Intel Hardware v" INTEL_IDLE_VERSION);