[Midnightbsd-cvs] src [7760] stable/0.8/sys/dev/coretemp/coretemp.c: switch over to freebsd implementation until sensors framework issue is sorted.

laffer1 at midnightbsd.org laffer1 at midnightbsd.org
Sun Aug 21 11:08:55 EDT 2016


Revision: 7760
          http://svnweb.midnightbsd.org/src/?rev=7760
Author:   laffer1
Date:     2016-08-21 11:08:55 -0400 (Sun, 21 Aug 2016)
Log Message:
-----------
switch over to freebsd implementation until sensors framework issue is sorted.

Modified Paths:
--------------
    stable/0.8/sys/dev/coretemp/coretemp.c

Modified: stable/0.8/sys/dev/coretemp/coretemp.c
===================================================================
--- stable/0.8/sys/dev/coretemp/coretemp.c	2016-08-21 15:05:28 UTC (rev 7759)
+++ stable/0.8/sys/dev/coretemp/coretemp.c	2016-08-21 15:08:55 UTC (rev 7760)
@@ -1,4 +1,3 @@
-/* $MidnightBSD: src/sys/dev/coretemp/coretemp.c,v 1.5 2010/03/05 03:40:43 laffer1 Exp $ */
 /*-
  * Copyright (c) 2007, 2008 Rui Paulo <rpaulo at FreeBSD.org>
  * All rights reserved.
@@ -31,7 +30,7 @@
  */
 
 #include <sys/cdefs.h>
-__FBSDID("$FreeBSD: src/sys/dev/coretemp/coretemp.c,v 1.2.4.3.2.1 2008/11/25 02:59:29 kensmith Exp $");
+__FBSDID("$FreeBSD$");
 
 #include <sys/param.h>
 #include <sys/bus.h>
@@ -40,21 +39,30 @@
 #include <sys/module.h>
 #include <sys/conf.h>
 #include <sys/kernel.h>
-#include <sys/sensors.h>
+#include <sys/sysctl.h>
 #include <sys/proc.h>	/* for curthread */
 #include <sys/sched.h>
 
 #include <machine/specialreg.h>
 #include <machine/cpufunc.h>
+#include <machine/cputypes.h>
 #include <machine/md_var.h>
 
-extern int smp_cpus;
+#define	TZ_ZEROC			2732
 
+#define	THERM_STATUS_LOG		0x02
+#define	THERM_STATUS			0x01
+#define	THERM_STATUS_TEMP_SHIFT		16
+#define	THERM_STATUS_TEMP_MASK		0x7f
+#define	THERM_STATUS_RES_SHIFT		27
+#define	THERM_STATUS_RES_MASK		0x0f
+#define	THERM_STATUS_VALID_SHIFT	31
+#define	THERM_STATUS_VALID_MASK		0x01
+
 struct coretemp_softc {
-	struct ksensordev	sc_sensordev;
-	struct ksensor		sc_sensor;
-	device_t		sc_dev;
-	int			sc_tjmax;
+	device_t	sc_dev;
+	int		sc_tjmax;
+	unsigned int	sc_throttle_log;
 };
 
 /*
@@ -65,8 +73,10 @@
 static int	coretemp_attach(device_t dev);
 static int	coretemp_detach(device_t dev);
 
-static int	coretemp_get_temp(device_t dev);
-static void	coretemp_refresh(void *arg);
+static uint64_t	coretemp_get_thermal_msr(int cpu);
+static void	coretemp_clear_thermal_msr(int cpu);
+static int	coretemp_get_val_sysctl(SYSCTL_HANDLER_ARGS);
+static int	coretemp_throttle_log_sysctl(SYSCTL_HANDLER_ARGS);
 
 static device_method_t coretemp_methods[] = {
 	/* Device interface */
@@ -84,8 +94,16 @@
 	sizeof(struct coretemp_softc),
 };
 
+enum therm_info {
+	CORETEMP_TEMP,
+	CORETEMP_DELTA,
+	CORETEMP_RESOLUTION,
+	CORETEMP_TJMAX,
+};
+
 static devclass_t coretemp_devclass;
-DRIVER_MODULE(coretemp, cpu, coretemp_driver, coretemp_devclass, NULL, NULL);
+DRIVER_MODULE(coretemp, cpu, coretemp_driver, coretemp_devclass, NULL,
+    NULL);
 
 static void
 coretemp_identify(driver_t *driver, device_t parent)
@@ -98,7 +116,7 @@
 		return;
 
 	/* Check that CPUID 0x06 is supported and the vendor is Intel.*/
-	if (cpu_high < 6 || strcmp(cpu_vendor, "GenuineIntel"))
+	if (cpu_high < 6 || cpu_vendor_id != CPU_VENDOR_INTEL)
 		return;
 	/*
 	 * CPUID 0x06 returns 1 if the processor has on-die thermal
@@ -134,21 +152,16 @@
 	struct coretemp_softc *sc = device_get_softc(dev);
 	device_t pdev;
 	uint64_t msr;
-	int cpu_model;
-	int cpu_mask;
+	int cpu_model, cpu_stepping;
+	int ret, tjtarget;
+	struct sysctl_oid *oid;
+	struct sysctl_ctx_list *ctx;
 
 	sc->sc_dev = dev;
 	pdev = device_get_parent(dev);
-	cpu_model = (cpu_id >> 4) & 15;
-	/* extended model */
-	cpu_model += ((cpu_id >> 16) & 0xf) << 4;
-	cpu_mask = cpu_id & 15;
+	cpu_model = CPUID_TO_MODEL(cpu_id);
+	cpu_stepping = cpu_id & CPUID_STEPPING;
 
-#if 0 /*
-       * XXXrpaulo: I have this CPU model and when it returns from C3
-       * coretemp continues to function properly.
-       */
-	 
 	/*
 	 * Some CPUs, namely the PIII, don't have thermal sensors, but
 	 * report them when the CPUID check is performed in
@@ -158,6 +171,11 @@
 	if (cpu_model < 0xe)
 		return (ENXIO);
 
+#if 0 /*
+       * XXXrpaulo: I have this CPU model and when it returns from C3
+       * coretemp continues to function properly.
+       */
+	 
 	/*
 	 * Check for errata AE18.
 	 * "Processor Digital Thermal Sensor (DTS) Readout stops
@@ -165,7 +183,7 @@
 	 *
 	 * Adapted from the Linux coretemp driver.
 	 */
-	if (cpu_model == 0xe && cpu_mask < 0xc) {
+	if (cpu_model == 0xe && cpu_stepping < 0xc) {
 		msr = rdmsr(MSR_BIOS_SIGN);
 		msr = msr >> 32;
 		if (msr < 0x39) {
@@ -175,32 +193,106 @@
 		}
 	}
 #endif
+
 	/*
-	 * On some Core 2 CPUs, there's an undocumented MSR that
-	 * can tell us if Tj(max) is 100 or 85.
-	 *
-	 * The if-clause for CPUs having the MSR_IA32_EXT_CONFIG was adapted
-	 * from the Linux coretemp driver.
+	 * Use 100C as the initial value.
 	 */
 	sc->sc_tjmax = 100;
-	if ((cpu_model == 0xf && cpu_mask >= 2) || cpu_model == 0xe) {
+
+	if ((cpu_model == 0xf && cpu_stepping >= 2) || cpu_model == 0xe) {
+		/*
+		 * On some Core 2 CPUs, there's an undocumented MSR that
+		 * can tell us if Tj(max) is 100 or 85.
+		 *
+		 * The if-clause for CPUs having the MSR_IA32_EXT_CONFIG was adapted
+		 * from the Linux coretemp driver.
+		 */
 		msr = rdmsr(MSR_IA32_EXT_CONFIG);
 		if (msr & (1 << 30))
 			sc->sc_tjmax = 85;
+	} else if (cpu_model == 0x17) {
+		switch (cpu_stepping) {
+		case 0x6:	/* Mobile Core 2 Duo */
+			sc->sc_tjmax = 105;
+			break;
+		default:	/* Unknown stepping */
+			break;
+		}
+	} else if (cpu_model == 0x1c) {
+		switch (cpu_stepping) {
+		case 0xa:	/* 45nm Atom D400, N400 and D500 series */
+			sc->sc_tjmax = 100;
+			break;
+		default:
+			sc->sc_tjmax = 90;
+			break;
+		}
+	} else {
+		/*
+		 * Attempt to get Tj(max) from MSR IA32_TEMPERATURE_TARGET.
+		 *
+		 * This method is described in Intel white paper "CPU
+		 * Monitoring With DTS/PECI". (#322683)
+		 */
+		ret = rdmsr_safe(MSR_IA32_TEMPERATURE_TARGET, &msr);
+		if (ret == 0) {
+			tjtarget = (msr >> 16) & 0xff;
+
+			/*
+			 * On earlier generation of processors, the value
+			 * obtained from IA32_TEMPERATURE_TARGET register is
+			 * an offset that needs to be summed with a model
+			 * specific base.  It is however not clear what
+			 * these numbers are, with the publicly available
+			 * documents from Intel.
+			 *
+			 * For now, we consider [70, 110]C range, as
+			 * described in #322683, as "reasonable" and accept
+			 * these values whenever the MSR is available for
+			 * read, regardless the CPU model.
+			 */
+			if (tjtarget >= 70 && tjtarget <= 110)
+				sc->sc_tjmax = tjtarget;
+			else
+				device_printf(dev, "Tj(target) value %d "
+				    "does not seem right.\n", tjtarget);
+		} else
+			device_printf(dev, "Can not get Tj(target) "
+			    "from your CPU, using 100C.\n");
 	}
 
+	if (bootverbose)
+		device_printf(dev, "Setting TjMax=%d\n", sc->sc_tjmax);
+
+	ctx = device_get_sysctl_ctx(dev);
+
+	oid = SYSCTL_ADD_NODE(ctx,
+	    SYSCTL_CHILDREN(device_get_sysctl_tree(pdev)), OID_AUTO,
+	    "coretemp", CTLFLAG_RD, NULL, "Per-CPU thermal information");
+
 	/*
-	 * Add hw.sensors.cpuN.temp0 MIB.
+	 * Add the MIBs to dev.cpu.N and dev.cpu.N.coretemp.
 	 */
-	strlcpy(sc->sc_sensordev.xname, device_get_nameunit(pdev),
-	    sizeof(sc->sc_sensordev.xname));
-	sc->sc_sensor.type = SENSOR_TEMP;
-	sensor_attach(&sc->sc_sensordev, &sc->sc_sensor);
-	if (sensor_task_register(sc, coretemp_refresh, 2)) {
-		device_printf(dev, "unable to register update task\n");
-		return (ENXIO);
-	}
-	sensordev_install(&sc->sc_sensordev);
+	SYSCTL_ADD_PROC(ctx, SYSCTL_CHILDREN(device_get_sysctl_tree(pdev)),
+	    OID_AUTO, "temperature", CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
+	    dev, CORETEMP_TEMP, coretemp_get_val_sysctl, "IK",
+	    "Current temperature");
+	SYSCTL_ADD_PROC(ctx, SYSCTL_CHILDREN(oid), OID_AUTO, "delta",
+	    CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, dev, CORETEMP_DELTA,
+	    coretemp_get_val_sysctl, "I",
+	    "Delta between TCC activation and current temperature");
+	SYSCTL_ADD_PROC(ctx, SYSCTL_CHILDREN(oid), OID_AUTO, "resolution",
+	    CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, dev, CORETEMP_RESOLUTION,
+	    coretemp_get_val_sysctl, "I",
+	    "Resolution of CPU thermal sensor");
+	SYSCTL_ADD_PROC(ctx, SYSCTL_CHILDREN(oid), OID_AUTO, "tjmax",
+	    CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, dev, CORETEMP_TJMAX,
+	    coretemp_get_val_sysctl, "IK",
+	    "TCC activation temperature");
+	SYSCTL_ADD_PROC(ctx, SYSCTL_CHILDREN(oid), OID_AUTO,
+	    "throttle_log", CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, dev, 0,
+	    coretemp_throttle_log_sysctl, "I",
+	    "Set to 1 if the thermal sensor has tripped");
 
 	return (0);
 }
@@ -208,93 +300,147 @@
 static int
 coretemp_detach(device_t dev)
 {
-	struct coretemp_softc *sc = device_get_softc(dev);
+	return (0);
+}
 
-	sensordev_deinstall(&sc->sc_sensordev);
-	sensor_task_unregister(sc);
+static uint64_t
+coretemp_get_thermal_msr(int cpu)
+{
+	uint64_t msr;
 
-	return (0);
+	thread_lock(curthread);
+	sched_bind(curthread, cpu);
+	thread_unlock(curthread);
+
+	/*
+	 * The digital temperature reading is located at bit 16
+	 * of MSR_THERM_STATUS.
+	 *
+	 * There is a bit on that MSR that indicates whether the
+	 * temperature is valid or not.
+	 *
+	 * The temperature is computed by subtracting the temperature
+	 * reading by Tj(max).
+	 */
+	msr = rdmsr(MSR_THERM_STATUS);
+
+	thread_lock(curthread);
+	sched_unbind(curthread);
+	thread_unlock(curthread);
+
+	return (msr);
 }
 
+static void
+coretemp_clear_thermal_msr(int cpu)
+{
+	thread_lock(curthread);
+	sched_bind(curthread, cpu);
+	thread_unlock(curthread);
 
+	wrmsr(MSR_THERM_STATUS, 0);
+
+	thread_lock(curthread);
+	sched_unbind(curthread);
+	thread_unlock(curthread);
+}
+
 static int
-coretemp_get_temp(device_t dev)
+coretemp_get_val_sysctl(SYSCTL_HANDLER_ARGS)
 {
+	device_t dev;
 	uint64_t msr;
-	int temp;
-	int cpu = device_get_unit(dev);
-	struct coretemp_softc *sc = device_get_softc(dev);
+	int val, tmp;
+	struct coretemp_softc *sc;
+	enum therm_info type;
 	char stemp[16];
 
-	/*
-	 * Bind to specific CPU to read the correct temperature.
-	 * If not all CPUs are initialised, then only read from
-	 * cpu0, returning -1 on all other CPUs.
-	 */
-	if (smp_cpus > 1) {
-		thread_lock(curthread);
-		sched_bind(curthread, cpu);
-		msr = rdmsr(MSR_THERM_STATUS);
-		sched_unbind(curthread);
-		thread_unlock(curthread);
-	} else if (cpu != 0)
-		return (-1);
-	else
-		msr = rdmsr(MSR_THERM_STATUS);
+	dev = (device_t) arg1;
+	msr = coretemp_get_thermal_msr(device_get_unit(dev));
+	sc = device_get_softc(dev);
+	type = arg2;
 
-	/*
-	 * Check for Thermal Status and Thermal Status Log.
-	 */
-	if ((msr & 0x3) == 0x3)
-		device_printf(dev, "PROCHOT asserted\n");
+	if (((msr >> THERM_STATUS_VALID_SHIFT) & THERM_STATUS_VALID_MASK) != 1) {
+		val = -1;
+	} else {
+		switch (type) {
+		case CORETEMP_TEMP:
+			tmp = (msr >> THERM_STATUS_TEMP_SHIFT) &
+			    THERM_STATUS_TEMP_MASK;
+			val = (sc->sc_tjmax - tmp) * 10 + TZ_ZEROC;
+			break;
+		case CORETEMP_DELTA:
+			val = (msr >> THERM_STATUS_TEMP_SHIFT) &
+			    THERM_STATUS_TEMP_MASK;
+			break;
+		case CORETEMP_RESOLUTION:
+			val = (msr >> THERM_STATUS_RES_SHIFT) &
+			    THERM_STATUS_RES_MASK;
+			break;
+		case CORETEMP_TJMAX:
+			val = sc->sc_tjmax * 10 + TZ_ZEROC;
+			break;
+		}
+	}
 
-	/*
-	 * Bit 31 contains "Reading valid"
-	 */
-	if (((msr >> 31) & 0x1) == 1) {
+	if (msr & THERM_STATUS_LOG) {
+		coretemp_clear_thermal_msr(device_get_unit(dev));
+		sc->sc_throttle_log = 1;
+
 		/*
-		 * Starting on bit 16 and ending on bit 22.
+		 * Check for Critical Temperature Status and Critical
+		 * Temperature Log.  It doesn't really matter if the
+		 * current temperature is invalid because the "Critical
+		 * Temperature Log" bit will tell us if the Critical
+		 * Temperature has * been reached in past. It's not
+		 * directly related to the current temperature.
+		 *
+		 * If we reach a critical level, allow devctl(4)
+		 * to catch this and shutdown the system.
 		 */
-		temp = sc->sc_tjmax - ((msr >> 16) & 0x7f);
-	} else
-		temp = -1;
-
-	/*
-	 * Check for Critical Temperature Status and Critical
-	 * Temperature Log.
-	 * It doesn't really matter if the current temperature is
-	 * invalid because the "Critical Temperature Log" bit will
-	 * tell us if the Critical Temperature has been reached in
-	 * past. It's not directly related to the current temperature.
-	 *
-	 * If we reach a critical level, allow devctl(4) to catch this
-	 * and shutdown the system.
-	 */
-	if (((msr >> 4) & 0x3) == 0x3) {
-		device_printf(dev, "critical temperature detected, "
-		    "suggest system shutdown\n");
-		snprintf(stemp, sizeof(stemp), "%d", temp);
-		devctl_notify("coretemp", "Thermal", stemp, "notify=0xcc");
+		if (msr & THERM_STATUS) {
+			tmp = (msr >> THERM_STATUS_TEMP_SHIFT) &
+			    THERM_STATUS_TEMP_MASK;
+			tmp = (sc->sc_tjmax - tmp) * 10 + TZ_ZEROC;
+			device_printf(dev, "critical temperature detected, "
+			    "suggest system shutdown\n");
+			snprintf(stemp, sizeof(stemp), "%d", tmp);
+			devctl_notify("coretemp", "Thermal", stemp,
+			    "notify=0xcc");
+		}
 	}
 
-	return (temp);
+	return (sysctl_handle_int(oidp, &val, 0, req));
 }
 
-static void
-coretemp_refresh(void *arg)
+static int
+coretemp_throttle_log_sysctl(SYSCTL_HANDLER_ARGS)
 {
-	struct coretemp_softc *sc = arg;
-	device_t dev = sc->sc_dev;
-	struct ksensor *s = &sc->sc_sensor;
-	int temp;
+	device_t dev;
+	uint64_t msr;
+	int error, val;
+	struct coretemp_softc *sc;
 
-	temp = coretemp_get_temp(dev);
+	dev = (device_t) arg1;
+	msr = coretemp_get_thermal_msr(device_get_unit(dev));
+	sc = device_get_softc(dev);
 
-	if (temp == -1) {
-		s->flags |= SENSOR_FINVALID;
-		s->value = 0;
-	} else {
-		s->flags &= ~SENSOR_FINVALID;
-		s->value = temp * 1000000 + 273150000;
+	if (msr & THERM_STATUS_LOG) {
+		coretemp_clear_thermal_msr(device_get_unit(dev));
+		sc->sc_throttle_log = 1;
 	}
+
+	val = sc->sc_throttle_log;
+
+	error = sysctl_handle_int(oidp, &val, 0, req);
+
+	if (error || !req->newptr)
+		return (error);
+	else if (val != 0)
+		return (EINVAL);
+
+	coretemp_clear_thermal_msr(device_get_unit(dev));
+	sc->sc_throttle_log = 0;
+
+	return (0);
 }



More information about the Midnightbsd-cvs mailing list