diff mbox

Patch[2/2] Adding Interrupt Handling Support to Coretemp

Message ID D6D887BA8C9DFF48B5233887EF04654105C0E05F61@bgsmsx502.gar.corp.intel.com (mailing list archive)
State New, archived
Headers show

Commit Message

durgadoss.r@intel.com Dec. 8, 2010, 9:17 a.m. UTC
None
diff mbox

Patch

diff --git a/arch/x86/include/asm/mce.h b/arch/x86/include/asm/mce.h
index c62c13c..bbb17b9 100644
--- a/arch/x86/include/asm/mce.h
+++ b/arch/x86/include/asm/mce.h
@@ -223,6 +223,9 @@  void intel_init_thermal(struct cpuinfo_x86 *c);
 
 void mce_log_therm_throt_event(__u64 status);
 
+/* Interrupt Handler for core/package thermal thresholds */
+extern int (*platform_thermal_notify)(__u64 msr_val, int type);
+
 #ifdef CONFIG_X86_THERMAL_VECTOR
 extern void mcheck_intel_therm_init(void);
 #else
diff --git a/arch/x86/kernel/cpu/mcheck/therm_throt.c b/arch/x86/kernel/cpu/mcheck/therm_throt.c
index 4b68326..f4afd79 100644
--- a/arch/x86/kernel/cpu/mcheck/therm_throt.c
+++ b/arch/x86/kernel/cpu/mcheck/therm_throt.c
@@ -34,6 +34,11 @@ 
 /* How long to wait between reporting thermal events */
 #define CHECK_INTERVAL		(300 * HZ)
 
+/* Similar to CHECK_INTERVAL. But interval is reduced because
+ * these lower and upper threhsolds will be crossed frequently.
+ * Hence the reporting should be quick enough to handle the event */
+#define THRES_INTERVAL		(25 * HZ)
+
 #define THERMAL_THROTTLING_EVENT	0
 #define POWER_LIMIT_EVENT		1
 
@@ -53,8 +58,15 @@  struct thermal_state {
 	struct _thermal_state core_power_limit;
 	struct _thermal_state package_throttle;
 	struct _thermal_state package_power_limit;
+	struct _thermal_state core_thresh0;
+	struct _thermal_state core_thresh1;
+	struct _thermal_state package_thresh0;
+	struct _thermal_state package_thresh1;
 };
 
+/* Callback to handle core/package threshold interrupts */
+int (*platform_thermal_notify)(__u64 msr_val, int type);
+
 static DEFINE_PER_CPU(struct thermal_state, thermal_state);
 
 static atomic_t therm_throt_en	= ATOMIC_INIT(0);
@@ -200,6 +212,38 @@  static int therm_throt_process(bool new_event, int event, int level)
 	return 0;
 }
 
+static int thresh_event_valid(int event)
+{
+	struct _thermal_state *state = NULL;
+
+	unsigned int this_cpu = smp_processor_id();
+	struct thermal_state *pstate = &per_cpu(thermal_state, this_cpu);
+	u64 now = get_jiffies_64();
+
+	switch (event) {
+	case 0:
+		state = &pstate->core_thresh0;
+		break;
+	case 1:
+		state = &pstate->core_thresh1;
+		break;
+	case 2:
+		state = &pstate->package_thresh0;
+		break;
+	case 3:
+		state = &pstate->package_thresh1;
+		break;
+	default:
+		WARN_ON(1);
+	}
+
+	if (time_before64(now, state->next_check))
+		return 0;
+
+	state->next_check = now + THRES_INTERVAL;
+	return 1;
+}
+
 #ifdef CONFIG_SYSFS
 /* Add/Remove thermal_throttle interface for CPU device: */
 static __cpuinit int thermal_throttle_add_dev(struct sys_device *sys_dev,
@@ -313,6 +357,28 @@  device_initcall(thermal_throttle_init_device);
 #define PACKAGE_THROTTLED	((__u64)2 << 62)
 #define PACKAGE_POWER_LIMIT	((__u64)3 << 62)
 
+static void notify_thresholds(__u64 msr_val, int type)
+{
+	/* type 0 means core threshold
+	 * type 1 means package threshold
+	 */
+	int low_event = type * 2;
+	int high_event = (type * 2) + 1;
+
+	/* check whether the interrupt handler is defined;
+	 * otherwise simply return
+	 */
+	if (!platform_thermal_notify)
+		return;
+
+	/* lower threshold reached */
+	if ((msr_val & THERM_LOG_THRESHOLD0) &&	thresh_event_valid(low_event))
+		platform_thermal_notify(msr_val, type);
+	/* higher threshold reached */
+	if ((msr_val & THERM_LOG_THRESHOLD1) && thresh_event_valid(high_event))
+		platform_thermal_notify(msr_val, type);
+}
+
 /* Thermal transition interrupt handler */
 static void intel_thermal_interrupt(void)
 {
@@ -321,6 +387,10 @@  static void intel_thermal_interrupt(void)
 
 	rdmsrl(MSR_IA32_THERM_STATUS, msr_val);
 
+	/* Check for violation of core thermal thresholds
+	 * If so, send notification */
+	notify_thresholds(msr_val, 0);
+
 	if (therm_throt_process(msr_val & THERM_STATUS_PROCHOT,
 				THERMAL_THROTTLING_EVENT,
 				CORE_LEVEL) != 0)
@@ -334,6 +404,7 @@  static void intel_thermal_interrupt(void)
 
 	if (cpu_has(c, X86_FEATURE_PTS)) {
 		rdmsrl(MSR_IA32_PACKAGE_THERM_STATUS, msr_val);
+		notify_thresholds(msr_val, 1);
 		if (therm_throt_process(msr_val & PACKAGE_THERM_STATUS_PROCHOT,
 					THERMAL_THROTTLING_EVENT,
 					PACKAGE_LEVEL) != 0)
diff --git a/drivers/hwmon/coretemp.c b/drivers/hwmon/coretemp.c
index 7aea83e..6d8b21f 100644
--- a/drivers/hwmon/coretemp.c
+++ b/drivers/hwmon/coretemp.c
@@ -36,9 +36,17 @@ 
 #include <asm/msr.h>
 #include <asm/processor.h>
 #include <asm/smp.h>
+#include <asm/mce.h>
+#include <linux/thermal.h>
 
 #define DRVNAME	"coretemp"
 
+/* An identification number to the DTS sensor.
+ * This will help the user space to figure out which
+ * sensor caused the event
+ */
+#define DTS_ID	0
+
 typedef enum { SHOW_TEMP, SHOW_TJMAX, SHOW_TTARGET, SHOW_LABEL,
 		SHOW_NAME } SHOW;
 /* C indicates core thermal thresholds
@@ -75,6 +83,52 @@  static int set_core_threshold(struct coretemp_data *data, int val,
 						enum thresholds thresh);
 static int set_pkg_threshold(struct coretemp_data *data, int val,
 						enum thresholds thresh);
+
+/* Interrupt Handlers for core/package thresholds */
+struct work_struct *t0_netlink_handlr;
+struct work_struct *t1_netlink_handlr;
+
+/* Send netlink event for DTS sensor reaching threshold0 */
+static void gen_netlink_t0(struct work_struct *work)
+{
+	generate_netlink_event(DTS_ID, THERMAL_AUX0);
+}
+
+/* Send netlink event for DTS sensor reaching threshold1 */
+static void gen_netlink_t1(struct work_struct *work)
+{
+	generate_netlink_event(DTS_ID, THERMAL_AUX1);
+}
+
+/* Platform thermal Interrupt Handler.
+ * As of now, we handle both core/pkg threshold interrupts in the same
+ * way. If the user space is good enough to distinguish between these
+ * two and handle the events in a finer way, then we can improvise.
+ */
+static int coretemp_interrupt(__u64 msr_val, int type)
+{
+	u64 status_reg = (type == 0) ? MSR_IA32_THERM_STATUS :
+					MSR_IA32_PACKAGE_THERM_STATUS;
+
+	if (msr_val & THERM_LOG_THRESHOLD0) {
+		if (!(msr_val & THERM_STATUS_THRESHOLD0))
+			schedule_work(t0_netlink_handlr);
+
+		/* Reset the Threshold0 interrupt */
+		wrmsrl(status_reg, msr_val & ~THERM_LOG_THRESHOLD0);
+	}
+
+	if (msr_val & THERM_LOG_THRESHOLD1) {
+		if (msr_val & THERM_STATUS_THRESHOLD1)
+			schedule_work(t1_netlink_handlr);
+
+		/* Reset the Threshold1 interrupt */
+		wrmsrl(status_reg, msr_val & ~THERM_LOG_THRESHOLD1);
+	}
+
+	return 0;
+}
+
 /*
  * Sysfs stuff
  */
@@ -415,7 +469,7 @@  static int set_pkg_threshold(struct coretemp_data *data, int temp,
 	mutex_lock(&data->update_lock);
 
 	diff = (data->tjmax - temp)/1000;
-
+
 	/* Mask the thermal vector in the lapic */
 	l = apic_read(APIC_LVTTHMR);
 	apic_write(APIC_LVTTHMR, l | APIC_LVT_MASKED);
@@ -469,6 +523,8 @@  exit:
 	/* Unmask the thermal vector */
 	l = apic_read(APIC_LVTTHMR);
 	apic_write(APIC_LVTTHMR, l & ~APIC_LVT_MASKED);
+	/* Enable the Interrupt Handling Support */
+	platform_thermal_notify = coretemp_interrupt;
 	return 0;
 }
 
@@ -786,6 +842,20 @@  static int __init coretemp_init(void)
 #endif
 
 	register_hotcpu_notifier(&coretemp_cpu_notifier);
+
+	/* Initialize the Interrupt Handlers */
+	t0_netlink_handlr = kzalloc(sizeof(struct work_struct), GFP_KERNEL);
+	if (!t0_netlink_handlr)
+		return -ENOMEM;
+
+	t1_netlink_handlr = kzalloc(sizeof(struct work_struct), GFP_KERNEL);
+	if (!t1_netlink_handlr) {
+		kfree(t0_netlink_handlr);
+		return -ENOMEM;
+	}
+	INIT_WORK(t0_netlink_handlr, (void *)gen_netlink_t0);
+	INIT_WORK(t1_netlink_handlr, (void *)gen_netlink_t1);
+
 	return 0;
 
 #ifndef CONFIG_HOTPLUG_CPU