freebsd-dev/sys/dev/coretemp/coretemp.c
Mark Johnston e42412a966 Mark the coretemp(4) sysctls as MPSAFE, ensuring that Giant won't be held
unnecessarily by a user thread waiting to run on a specific CPU after
calling sched_bind().

Reviewed by:	rstone
Approved by:	emaste (co-mentor)
Sponsored by:	Sandvine Incorporated
MFC after:	1 week
2013-02-18 22:47:16 +00:00

447 lines
12 KiB
C

/*-
* Copyright (c) 2007, 2008 Rui Paulo <rpaulo@FreeBSD.org>
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
* IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
* WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
* DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT,
* INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
* (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
* SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
* STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
* ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
* POSSIBILITY OF SUCH DAMAGE.
*/
/*
* Device driver for Intel's On Die thermal sensor via MSR.
* First introduced in Intel's Core line of processors.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <sys/param.h>
#include <sys/bus.h>
#include <sys/systm.h>
#include <sys/types.h>
#include <sys/module.h>
#include <sys/conf.h>
#include <sys/kernel.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>
#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 {
device_t sc_dev;
int sc_tjmax;
unsigned int sc_throttle_log;
};
/*
* Device methods.
*/
static void coretemp_identify(driver_t *driver, device_t parent);
static int coretemp_probe(device_t dev);
static int coretemp_attach(device_t dev);
static int coretemp_detach(device_t dev);
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 */
DEVMETHOD(device_identify, coretemp_identify),
DEVMETHOD(device_probe, coretemp_probe),
DEVMETHOD(device_attach, coretemp_attach),
DEVMETHOD(device_detach, coretemp_detach),
DEVMETHOD_END
};
static driver_t coretemp_driver = {
"coretemp",
coretemp_methods,
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);
static void
coretemp_identify(driver_t *driver, device_t parent)
{
device_t child;
u_int regs[4];
/* Make sure we're not being doubly invoked. */
if (device_find_child(parent, "coretemp", -1) != NULL)
return;
/* Check that CPUID 0x06 is supported and the vendor is Intel.*/
if (cpu_high < 6 || cpu_vendor_id != CPU_VENDOR_INTEL)
return;
/*
* CPUID 0x06 returns 1 if the processor has on-die thermal
* sensors. EBX[0:3] contains the number of sensors.
*/
do_cpuid(0x06, regs);
if ((regs[0] & 0x1) != 1)
return;
/*
* We add a child for each CPU since settings must be performed
* on each CPU in the SMP case.
*/
child = device_add_child(parent, "coretemp", -1);
if (child == NULL)
device_printf(parent, "add coretemp child failed\n");
}
static int
coretemp_probe(device_t dev)
{
if (resource_disabled("coretemp", 0))
return (ENXIO);
device_set_desc(dev, "CPU On-Die Thermal Sensors");
return (BUS_PROBE_GENERIC);
}
static int
coretemp_attach(device_t dev)
{
struct coretemp_softc *sc = device_get_softc(dev);
device_t pdev;
uint64_t msr;
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 = CPUID_TO_MODEL(cpu_id);
cpu_stepping = cpu_id & CPUID_STEPPING;
/*
* Some CPUs, namely the PIII, don't have thermal sensors, but
* report them when the CPUID check is performed in
* coretemp_identify(). This leads to a later GPF when the sensor
* is queried via a MSR, so we stop here.
*/
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
* updating upon returning from C3/C4 state."
*
* Adapted from the Linux coretemp driver.
*/
if (cpu_model == 0xe && cpu_stepping < 0xc) {
msr = rdmsr(MSR_BIOS_SIGN);
msr = msr >> 32;
if (msr < 0x39) {
device_printf(dev, "not supported (Intel errata "
"AE18), try updating your BIOS\n");
return (ENXIO);
}
}
#endif
/*
* Use 100C as the initial value.
*/
sc->sc_tjmax = 100;
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 the MIBs to dev.cpu.N and dev.cpu.N.coretemp.
*/
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);
}
static int
coretemp_detach(device_t dev)
{
return (0);
}
static uint64_t
coretemp_get_thermal_msr(int cpu)
{
uint64_t msr;
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_val_sysctl(SYSCTL_HANDLER_ARGS)
{
device_t dev;
uint64_t msr;
int val, tmp;
struct coretemp_softc *sc;
enum therm_info type;
char stemp[16];
dev = (device_t) arg1;
msr = coretemp_get_thermal_msr(device_get_unit(dev));
sc = device_get_softc(dev);
type = arg2;
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;
}
}
if (msr & THERM_STATUS_LOG) {
coretemp_clear_thermal_msr(device_get_unit(dev));
sc->sc_throttle_log = 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 & 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 (sysctl_handle_int(oidp, &val, 0, req));
}
static int
coretemp_throttle_log_sysctl(SYSCTL_HANDLER_ARGS)
{
device_t dev;
uint64_t msr;
int error, val;
struct coretemp_softc *sc;
dev = (device_t) arg1;
msr = coretemp_get_thermal_msr(device_get_unit(dev));
sc = device_get_softc(dev);
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);
}