Package Level Thermal Control and Power Limit Notification
Subsystems (3)
v3 cover Package Level Thermal Control and Power Limit Notification
From: Fenghua Yu <[email protected]>
The new MSRs PACKAGE_THERM_STATUS and PACKAGE_THERM_INTERRUPT and power limit notification are published in Intel 64 and IA32 SDM and first implemented on Intel Sandy Bridge platform. The following patch set supports the new features in Linux kernel.
Fenghua Yu (5):
Package Level Thermal Control and Power Limit Notification: enable
features
Package Level Thermal Control and Power Limit Notification: pkgtemp
hwmon driver
Package Level Thermal Control and Power Limit Notification: thermal
throttling handler
Package Level Thermal Control and Power Limit Notification: power
limit
Package Level Thermal Control and Power Limit Notification: pkgtemp
doc
Documentation/hwmon/pkgtemp | 36 +++ arch/x86/configs/i386_defconfig | 1 + arch/x86/configs/x86_64_defconfig | 1 + arch/x86/include/asm/cpufeature.h | 2 + arch/x86/include/asm/msr-index.h | 17 +- arch/x86/kernel/cpu/addon_cpuid_features.c | 2 + arch/x86/kernel/cpu/mcheck/therm_throt.c | 206 ++++++++++--- drivers/hwmon/Kconfig | 7 + drivers/hwmon/Makefile | 1 + drivers/hwmon/pkgtemp.c | 456 ++++++++++++++++++++++++++++ 10 files changed, 688 insertions(+), 41 deletions(-) create mode 100644 Documentation/hwmon/pkgtemp create mode 100644 drivers/hwmon/pkgtemp.c
_______________________________________________
lm-sensors mailing list
[email protected]
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
PATCH v3 1/5 Package Level Thermal Control and Power Limit Notification: enable features
From: Fenghua Yu <[email protected]>
Add package level thermal and power limit feature support in the kernel.
The two MSRs and features are new starting with Intel's Sandy Bridge processor.
Please check Intel 64 and IA-32 Architectures SDMV Vol 3A 14.5.6 Power Limit Notification and 14.6 Package Level Thermal Management.
This patch also fixes a bug which defines reverse THERM_INT_LOW_ENABLE bit and THERM_INT_HIGH_ENABLE bit.
--- arch/x86/include/asm/cpufeature.h | 2 ++ arch/x86/include/asm/msr-index.h | 17 +++++++++++++++-- arch/x86/kernel/cpu/addon_cpuid_features.c | 2 ++ 3 files changed, 19 insertions(+), 2 deletions(-)
arch/x86/include/asm/cpufeature.h +2 -0
@@ -162,6 +162,8 @@#define X86_FEATURE_IDA (7*32+ 0)/* Intel Dynamic Acceleration */ #define X86_FEATURE_ARAT (7*32+ 1)/* Always Running APIC Timer */ #define X86_FEATURE_CPB (7*32+ 2)/* AMD Core Performance Boost */ +#define X86_FEATURE_PLN (7*32+ 3)/* Intel Power Limit Notification */ +#define X86_FEATURE_PTS (7*32+ 4)/* Intel Package Thermal Status */ /* Virtualization flags: Linux defined */ #define X86_FEATURE_TPR_SHADOW (8*32+ 0)/* Intel TPR Shadow */
arch/x86/include/asm/msr-index.h +15 -2
@@ -224,12 +224,14 @@#define MSR_IA32_THERM_CONTROL 0x0000019a#define MSR_IA32_THERM_INTERRUPT 0x0000019b -#define THERM_INT_LOW_ENABLE (1 << 0) -#define THERM_INT_HIGH_ENABLE (1 << 1) +#define THERM_INT_HIGH_ENABLE (1 << 0) +#define THERM_INT_LOW_ENABLE (1 << 1) +#define THERM_INT_PLN_ENABLE (1 << 24)#define MSR_IA32_THERM_STATUS 0x0000019c#define THERM_STATUS_PROCHOT (1 << 0) +#define THERM_STATUS_POWER_LIMIT (1 << 10)#define MSR_THERM2_CTL 0x0000019d
@@ -239,6 +241,17 @@#define MSR_IA32_TEMPERATURE_TARGET 0x000001a2 +#define MSR_IA32_PACKAGE_THERM_STATUS 0x000001b1 + +#define PACKAGE_THERM_STATUS_PROCHOT (1 << 0) +#define PACKAGE_THERM_STATUS_POWER_LIMIT (1 << 10) + +#define MSR_IA32_PACKAGE_THERM_INTERRUPT 0x000001b2 + +#define PACKAGE_THERM_INT_HIGH_ENABLE (1 << 0) +#define PACKAGE_THERM_INT_LOW_ENABLE (1 << 1) +#define PACKAGE_THERM_INT_PLN_ENABLE (1 << 24) +/* MISC_ENABLE bits: architectural */ #define MSR_IA32_MISC_ENABLE_FAST_STRING (1ULL << 0)#define MSR_IA32_MISC_ENABLE_TCC (1ULL << 1)
arch/x86/kernel/cpu/addon_cpuid_features.c +2 -0
@@ -32,6 +32,8 @@ void __cpuinit init_scattered_cpuid_features(struct cpuinfo_x86 *c)static const struct cpuid_bit __cpuinitconst cpuid_bits[] = { {X86_FEATURE_IDA ,CR_EAX ,1 ,0x00000006 }, {X86_FEATURE_ARAT ,CR_EAX ,2 ,0x00000006 }, + {X86_FEATURE_PLN ,CR_EAX ,4 ,0x00000006 }, + {X86_FEATURE_PTS ,CR_EAX ,6 ,0x00000006 }, {X86_FEATURE_APERFMPERF ,CR_ECX ,0 ,0x00000006 }, {X86_FEATURE_CPB ,CR_EDX ,9 ,0x80000007 }, {X86_FEATURE_NPT ,CR_EDX ,0 ,0x8000000a },
Signature · 7 lines
1.6.0.3 _______________________________________________ lm-sensors mailing list [email protected] http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
[tip:x86/cpu] x86, cpu: Package Level Thermal Control, Power Limit Notification definitions
Commit-ID: 9792db6174d9927700ed288e6d74b9391bf785d1 Gitweb: http://git.kernel.org/tip/9792db6174d9927700ed288e6d74b9391bf785d1 Author: Fenghua Yu <[email protected]> AuthorDate: Thu, 29 Jul 2010 17:13:42 -0700 Committer: H. Peter Anvin <[email protected]> CommitDate: Fri, 30 Jul 2010 16:15:32 -0700
x86, cpu: Package Level Thermal Control, Power Limit Notification definitions
Add package level thermal and power limit feature support.
The two MSRs and features are new starting with Intel's Sandy Bridge processor.
Please check Intel 64 and IA-32 Architectures SDMV Vol 3A 14.5.6 Power Limit Notification and 14.6 Package Level Thermal Management.
This patch also fixes a bug which defines reverse THERM_INT_LOW_ENABLE bit and THERM_INT_HIGH_ENABLE bit.
[ hpa: fixed up against current tip:x86/cpu ]
LKML-Reference: <[email protected]>
--- arch/x86/include/asm/cpufeature.h | 2 ++ arch/x86/include/asm/msr-index.h | 17 +++++++++++++++-- arch/x86/kernel/cpu/scattered.c | 2 ++ 3 files changed, 19 insertions(+), 2 deletions(-)
arch/x86/include/asm/cpufeature.h +2 -0
@@ -166,6 +166,8 @@#define X86_FEATURE_CPB (7*32+ 2)/* AMD Core Performance Boost */ #define X86_FEATURE_EPB (7*32+ 3)/* IA32_ENERGY_PERF_BIAS support */ #define X86_FEATURE_XSAVEOPT (7*32+ 4)/* Optimized Xsave */ +#define X86_FEATURE_PLN (7*32+ 5)/* Intel Power Limit Notification */ +#define X86_FEATURE_PTS (7*32+ 6)/* Intel Package Thermal Status */ /* Virtualization flags: Linux defined, word 8 */ #define X86_FEATURE_TPR_SHADOW (8*32+ 0)/* Intel TPR Shadow */
arch/x86/include/asm/msr-index.h +15 -2
@@ -224,12 +224,14 @@#define MSR_IA32_THERM_CONTROL 0x0000019a#define MSR_IA32_THERM_INTERRUPT 0x0000019b -#define THERM_INT_LOW_ENABLE (1 << 0) -#define THERM_INT_HIGH_ENABLE (1 << 1) +#define THERM_INT_HIGH_ENABLE (1 << 0) +#define THERM_INT_LOW_ENABLE (1 << 1) +#define THERM_INT_PLN_ENABLE (1 << 24)#define MSR_IA32_THERM_STATUS 0x0000019c#define THERM_STATUS_PROCHOT (1 << 0) +#define THERM_STATUS_POWER_LIMIT (1 << 10)#define MSR_THERM2_CTL 0x0000019d
@@ -241,6 +243,17 @@#define MSR_IA32_ENERGY_PERF_BIAS 0x000001b0 +#define MSR_IA32_PACKAGE_THERM_STATUS 0x000001b1 + +#define PACKAGE_THERM_STATUS_PROCHOT (1 << 0) +#define PACKAGE_THERM_STATUS_POWER_LIMIT (1 << 10) + +#define MSR_IA32_PACKAGE_THERM_INTERRUPT 0x000001b2 + +#define PACKAGE_THERM_INT_HIGH_ENABLE (1 << 0) +#define PACKAGE_THERM_INT_LOW_ENABLE (1 << 1) +#define PACKAGE_THERM_INT_PLN_ENABLE (1 << 24) +/* MISC_ENABLE bits: architectural */ #define MSR_IA32_MISC_ENABLE_FAST_STRING (1ULL << 0)#define MSR_IA32_MISC_ENABLE_TCC (1ULL << 1)
arch/x86/kernel/cpu/scattered.c +2 -0
@@ -33,6 +33,8 @@ void __cpuinit init_scattered_cpuid_features(struct cpuinfo_x86 *c)static const struct cpuid_bit __cpuinitconst cpuid_bits[] = { {X86_FEATURE_IDA ,CR_EAX ,1 ,0x00000006 ,0 }, {X86_FEATURE_ARAT ,CR_EAX ,2 ,0x00000006 ,0 }, + {X86_FEATURE_PLN ,CR_EAX ,4 ,0x00000006 ,0 }, + {X86_FEATURE_PTS ,CR_EAX ,6 ,0x00000006 ,0 }, {X86_FEATURE_APERFMPERF ,CR_ECX ,0 ,0x00000006 ,0 }, {X86_FEATURE_EPB ,CR_ECX ,3 ,0x00000006 ,0 }, {X86_FEATURE_XSAVEOPT ,CR_EAX ,0 ,0x0000000d ,1 },
PATCH v3 2/5 Package Level Thermal Control and Power Limit Notification: pkgtemp hwmon driver
In reply to Fenghua YuFrom: Fenghua Yu <[email protected]>
This patch adds a hwmon driver for package level thermal control. The driver dumps package level thermal information through sysfs interface so that upper level application (e.g. lm_sensor) can retrive the information.
Instead of having the package level hwmon code in coretemp, I write a seperate driver pkgtemp because:
First, package level thermal sensors include not only sensors for each core, but also sensors for uncore, memory controller or other components in the package. Logically it will be clear to have a seperate hwmon driver for package level hwmon to monitor wider range of sensors in a package. Merging package thermal driver into core thermal driver doesn't make sense and may mislead.
Secondly, merging the two drivers together may cause coding mess. It's easier to include various package level sensors info if more sensor information is implemented. Coretemp code needs to consider a lot of legacy machine cases. Pkgtemp code only considers platform starting from Sandy Bridge.
On a 1Sx4Cx2T Sandy Bridge platform, lm-sensors dumps the pkgtemp and coretemp:
pkgtemp-isa-0000
Adapter: ISA adapter
physical id 0: +33.0°C (high = +79.0°C, crit = +99.0°C)
coretemp-isa-0000
Adapter: ISA adapter
Core 0: +32.0°C (high = +79.0°C, crit = +99.0°C)
coretemp-isa-0001
Adapter: ISA adapter
Core 1: +32.0°C (high = +79.0°C, crit = +99.0°C)
coretemp-isa-0002
Adapter: ISA adapter
Core 2: +32.0°C (high = +79.0°C, crit = +99.0°C)
coretemp-isa-0003
Adapter: ISA adapter
Core 3: +32.0°C (high = +79.0°C, crit = +99.0°C)
--- arch/x86/configs/i386_defconfig | 1 + arch/x86/configs/x86_64_defconfig | 1 + drivers/hwmon/Kconfig | 7 + drivers/hwmon/Makefile | 1 + drivers/hwmon/pkgtemp.c | 456 +++++++++++++++++++++++++++++++++++++ 5 files changed, 466 insertions(+), 0 deletions(-) create mode 100644 drivers/hwmon/pkgtemp.c
arch/x86/configs/i386_defconfig +1 -0
@@ -1471,6 +1471,7 @@ CONFIG_HWMON=y # CONFIG_SENSORS_GL518SM is not set # CONFIG_SENSORS_GL520SM is not set # CONFIG_SENSORS_CORETEMP is not set +# CONFIG_SENSORS_PKGTEMP is not set # CONFIG_SENSORS_IT87 is not set # CONFIG_SENSORS_LM63 is not set # CONFIG_SENSORS_LM75 is not set
arch/x86/configs/x86_64_defconfig +1 -0
@@ -1456,6 +1456,7 @@ CONFIG_HWMON=y # CONFIG_SENSORS_GL518SM is not set # CONFIG_SENSORS_GL520SM is not set # CONFIG_SENSORS_CORETEMP is not set +# CONFIG_SENSORS_PKGTEMP is not set # CONFIG_SENSORS_IT87 is not set # CONFIG_SENSORS_LM63 is not set # CONFIG_SENSORS_LM75 is not set
drivers/hwmon/Kconfig +7 -0
@@ -407,6 +407,13 @@ config SENSORS_CORETEMP sensor inside your CPU. Most of the family 6 CPUs are supported. Check documentation/driver for details. +config SENSORS_PKGTEMP + tristate "Intel processor package temperature sensor" + depends on X86 && PCI && EXPERIMENTAL + help + If you say yes here you get support for the package level temperature + sensor inside your CPU. Check documentation/driver for details. + config SENSORS_IBMAEM tristate "IBM Active Energy Manager temperature/power sensors and control" select IPMI_SI
drivers/hwmon/Makefile +1 -0
@@ -39,6 +39,7 @@ obj-$(CONFIG_SENSORS_AMS) += ams/ obj-$(CONFIG_SENSORS_ASC7621) += asc7621.o obj-$(CONFIG_SENSORS_ATXP1) += atxp1.o obj-$(CONFIG_SENSORS_CORETEMP) += coretemp.o +obj-$(CONFIG_SENSORS_PKGTEMP) += pkgtemp.o obj-$(CONFIG_SENSORS_DME1737) += dme1737.o obj-$(CONFIG_SENSORS_DS1621) += ds1621.o obj-$(CONFIG_SENSORS_EMC1403) += emc1403.o
drivers/hwmon/pkgtemp.c new file +456 -0
456 lines not on this page · raw message
Signature · 7 lines
1.6.0.3 _______________________________________________ lm-sensors mailing list [email protected] http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
H. Peter Anvin
reviewerRe: [PATCH V3 2/5] Package Level Thermal Control and Power Limit Notification: pkgtemp hwmon driver
drivers/hwmon/pkgtemp.c line 43
Quoted 3 lines · Fenghua Yu
drivers/hwmon/pkgtemp.c new file +3
-0+ +enum {SHOW_TEMP ,SHOW_TJMAX ,SHOW_TTARGET ,SHOW_LABEL ,SHOW_NAME }SHOW ; +
This conflicts with an equally poorly named global variable in drivers/hwmon/via-cputemp.c, and the conflict is causing a build failure.
I think both these drivers have the same bug: a missing "typedef" before the enum keyword, as present in coretemp.c. Of course, one can question if it should be given a typename at all since in none of these drivers they are actually referenced by type, and instead the enumeration is just used as a source of constants, which can perfectly well be handled with an unnamed enum:
enum { SHOW_TEMP, SHOW_TJMAX, SHOW_TTARGET, SHOW_LABEL, SHOW_NAME };
-hpa
_______________________________________________
lm-sensors mailing list
[email protected]
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
Re: [PATCH V3 2/5] Package Level Thermal Control and Power Limit Notification: pkgtemp hwmon driver
Quoted 11 lines · H. Peter Anvin
drivers/hwmon/pkgtemp.c line 43
Quoted 3 lines · Fenghua Yu
drivers/hwmon/pkgtemp.c new file +3
-0+ +enum {SHOW_TEMP ,SHOW_TJMAX ,SHOW_TTARGET ,SHOW_LABEL ,SHOW_NAME }SHOW ; +This conflicts with an equally poorly named global variable in
drivers/hwmon/via-cputemp.c, and the conflict is causing a build failure.I think both these drivers have the same bug: a missing "typedef" before the enum keyword, as present in
coretemp.c. Of course, one can question if it should be given a typename at all since in none of these drivers
Especially since it isn't really a type name, but a global variable named SHOW. Type name (also called tag) would be enum SHOW { ... }; .
Guenter
Quoted 7 lines
they are actually referenced by type, and instead the enumeration is just used as a source of constants, which can perfectly well be handled with an unnamed enum:
enum {
SHOW_TEMP,SHOW_TJMAX,SHOW_TTARGET,SHOW_LABEL,SHOW_NAME};-hpa
_______________________________________________
lm-sensors mailing list
[email protected]
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
H. Peter Anvin
reviewerRe: [PATCH V3 2/5] Package Level Thermal Control and Power Limit Notification: pkgtemp hwmon driver
With the typedef keyword (see coretemp.c) it would be a typename... which is different from a tag in C (but not C++).
Quoted 25 lines · Guenter Roeck
Quoted 11 lines · H. Peter Anvin
drivers/hwmon/pkgtemp.c line 43
Quoted 3 lines · Fenghua Yu
drivers/hwmon/pkgtemp.c new file +3
-0+ +enum {SHOW_TEMP ,SHOW_TJMAX ,SHOW_TTARGET ,SHOW_LABEL ,SHOW_NAME }SHOW ; +This conflicts with an equally poorly named global variable in
drivers/hwmon/via-cputemp.c, and the conflict is causing a build failure.I think both these drivers have the same bug: a missing "typedef" before the enum keyword, as present in
coretemp.c. Of course, one can question if it should be given a typename at all since in none of these driversEspecially since it isn't really a type name, but a global variable named SHOW. Type name (also called tag) would be enum SHOW { ... }; .
Guenter
Quoted 7 lines
they are actually referenced by type, and instead the enumeration is just used as a source of constants, which can perfectly well be handled with an unnamed enum:
enum {
SHOW_TEMP,SHOW_TJMAX,SHOW_TTARGET,SHOW_LABEL,SHOW_NAME};-hpa
Signature · 6 lines
Sent from my mobile phone. Please pardon any lack of formatting. _______________________________________________ lm-sensors mailing list [email protected] http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
Fenghua Yu
Reviewed-by Len BrownRe: [PATCH V3 2/5] Package Level Thermal Control and Power Limit Notification: pkgtemp hwmon driver
In reply to H. Peter AnvinQuoted 15 lines · H. Peter Anvin
Review
Quoted 3 lines · Fenghua Yu
(quoted diff) +3
-0+ +enum { SHOW_TEMP, SHOW_TJMAX, SHOW_TTARGET, SHOW_LABEL, SHOW_NAME } SHOW; +This conflicts with an equally poorly named global variable in
drivers/hwmon/via-cputemp.c, and the conflict is causing a build failure.I think both these drivers have the same bug: a missing "typedef" before the enum keyword, as present in
coretemp.c. Of course, one can question if it should be given a typename at all since in none of these drivers they are actually referenced by type, and instead the enumeration is just used as a source of constants, which can perfectly well be handled with an unnamed enum:
Quoted 3 lines
enum {
SHOW_TEMP,SHOW_TJMAX,SHOW_TTARGET,SHOW_LABEL,SHOW_NAME};-hpa
Yes, that global variable is definitely wrong. Belowing is updated patch which uses:
enum { SHOW_TEMP, SHOW_TJMAX, SHOW_TTARGET, SHOW_LABEL, SHOW_NAME };
This patch adds a hwmon driver for package level thermal control. The driver dumps package level thermal information through sysfs interface so that upper level application (e.g. lm_sensor) can retrive the information.
Instead of having the package level hwmon code in coretemp, I write a seperate driver pkgtemp because:
First, package level thermal sensors include not only sensors for each core, but also sensors for uncore, memory controller or other components in the package. Logically it will be clear to have a seperate hwmon driver for package level hwmon to monitor wider range of sensors in a package. Merging package thermal driver into core thermal driver doesn't make sense and may mislead.
Secondly, merging the two drivers together may cause coding mess. It's easier to include various package level sensors info if more sensor information is implemented. Coretemp code needs to consider a lot of legacy machine cases. Pkgtemp code only considers platform starting from Sandy Bridge.
On a 1Sx4Cx2T Sandy Bridge platform, lm-sensors dumps the pkgtemp and coretemp:
pkgtemp-isa-0000
Adapter: ISA adapter
physical id 0: +33.0°C (high = +79.0°C, crit = +99.0°C)
coretemp-isa-0000
Adapter: ISA adapter
Core 0: +32.0°C (high = +79.0°C, crit = +99.0°C)
coretemp-isa-0001
Adapter: ISA adapter
Core 1: +32.0°C (high = +79.0°C, crit = +99.0°C)
coretemp-isa-0002
Adapter: ISA adapter
Core 2: +32.0°C (high = +79.0°C, crit = +99.0°C)
coretemp-isa-0003
Adapter: ISA adapter
Core 3: +32.0°C (high = +79.0°C, crit = +99.0°C)
--- arch/x86/configs/i386_defconfig | 1 + arch/x86/configs/x86_64_defconfig | 1 + drivers/hwmon/Kconfig | 7 + drivers/hwmon/Makefile | 1 + drivers/hwmon/pkgtemp.c | 456 +++++++++++++++++++++++++++++++++++++ 5 files changed, 466 insertions(+), 0 deletions(-) create mode 100644 drivers/hwmon/pkgtemp.c
arch/x86/configs/i386_defconfig +1 -0
@@ -1471,6 +1471,7 @@ CONFIG_HWMON=y # CONFIG_SENSORS_GL518SM is not set # CONFIG_SENSORS_GL520SM is not set # CONFIG_SENSORS_CORETEMP is not set +# CONFIG_SENSORS_PKGTEMP is not set # CONFIG_SENSORS_IT87 is not set # CONFIG_SENSORS_LM63 is not set # CONFIG_SENSORS_LM75 is not set
arch/x86/configs/x86_64_defconfig +1 -0
@@ -1456,6 +1456,7 @@ CONFIG_HWMON=y # CONFIG_SENSORS_GL518SM is not set # CONFIG_SENSORS_GL520SM is not set # CONFIG_SENSORS_CORETEMP is not set +# CONFIG_SENSORS_PKGTEMP is not set # CONFIG_SENSORS_IT87 is not set # CONFIG_SENSORS_LM63 is not set # CONFIG_SENSORS_LM75 is not set
drivers/hwmon/Kconfig +7 -0
@@ -407,6 +407,13 @@ config SENSORS_CORETEMP sensor inside your CPU. Most of the family 6 CPUs are supported. Check documentation/driver for details. +config SENSORS_PKGTEMP + tristate "Intel processor package temperature sensor" + depends on X86 && PCI && EXPERIMENTAL + help + If you say yes here you get support for the package level temperature + sensor inside your CPU. Check documentation/driver for details. + config SENSORS_IBMAEM tristate "IBM Active Energy Manager temperature/power sensors and control" select IPMI_SI
drivers/hwmon/Makefile +1 -0
@@ -39,6 +39,7 @@ obj-$(CONFIG_SENSORS_AMS) += ams/ obj-$(CONFIG_SENSORS_ASC7621) += asc7621.o obj-$(CONFIG_SENSORS_ATXP1) += atxp1.o obj-$(CONFIG_SENSORS_CORETEMP) += coretemp.o +obj-$(CONFIG_SENSORS_PKGTEMP) += pkgtemp.o obj-$(CONFIG_SENSORS_DME1737) += dme1737.o obj-$(CONFIG_SENSORS_DS1621) += ds1621.o obj-$(CONFIG_SENSORS_EMC1403) += emc1403.o
drivers/hwmon/pkgtemp.c new file +456 -0
456 lines not on this page · raw message
Signature · 7 lines
1.6.0.3 _______________________________________________ lm-sensors mailing list [email protected] http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
PATCH v3 3/5 Package Level Thermal Control and Power Limit Notification: thermal throttling handler
In reply to Fenghua YuFrom: Fenghua Yu <[email protected]>
Add package level thermal throttle interrupt support. The interrupt handler increases package level thermal throttle count. It also logs the event in MCE log.
The package level thermal throttle interrupt happens across threads in a package. Each thread handles the interrupt individually. User level application is supposed to retrieve correct event count and log based on package/thread topology. This is the same situation for core level interrupt handler. In the future, interrupt may be reported only per package or per core.
core_throttle_count and package_throttle_count are used for user interface. Previously only throttle_count is used for core throttle count. If you think new core_throttle_count name breaks user interface, I can change this part.
--- arch/x86/kernel/cpu/mcheck/therm_throt.c | 89 ++++++++++++++++++++++++------ 1 files changed, 71 insertions(+), 18 deletions(-)
arch/x86/kernel/cpu/mcheck/therm_throt.c +71 -18
@@ -37,7 +37,7 @@/* * Current thermal throttling state: */ -struct thermal_state { +struct _thermal_state {bool is_throttled;u64 next_check;
@@ -45,6 +45,11 @@ struct thermal_state {unsigned long last_throttle_count; }; +struct thermal_state { +struct _thermal_state core; +struct _thermal_state package; +}; + static DEFINE_PER_CPU(struct thermal_state, thermal_state);static atomic_t therm_throt_en =ATOMIC_INIT (0 );
@@ -53,11 +58,13 @@ static u32 lvtthmr_init __read_mostly;#ifdef CONFIG_SYSFS #define define_therm_throt_sysdev_one_ro (_name) \ - static SYSDEV_ATTR(_name, 0444, therm_throt_sysdev_show_##_name, NULL) + static SYSDEV_ATTR(_name, 0444, \ + therm_throt_sysdev_show_##_name, \ + NULL) \ -#define define_therm_throt_sysdev_show_func (name) \ +#define define_therm_throt_sysdev_show_func (level, name) \ \ -static ssize_t therm_throt_sysdev_show_##name( \ +static ssize_t therm_throt_sysdev_show_##level##_##name( \ struct sys_device *dev, \ struct sysdev_attribute *attr, \ char *buf) \
@@ -66,21 +73,24 @@ static ssize_t therm_throt_sysdev_show_##name( \ssize_t ret; \ \preempt_disable ();/* CPU hotplug */ \ -if (cpu_online (cpu)) \ +if (cpu_online (cpu)) { \ ret =sprintf (buf,"%lu\n" , \ -per_cpu (thermal_state, cpu).name); \ -else \ +per_cpu (thermal_state, cpu).level.name); \ + }else \ ret =0 ; \preempt_enable (); \ \return ret; \ } -define_therm_throt_sysdev_show_func (throttle_count); -define_therm_throt_sysdev_one_ro (throttle_count); +define_therm_throt_sysdev_show_func (core, throttle_count); +define_therm_throt_sysdev_one_ro (core_throttle_count); + +define_therm_throt_sysdev_show_func (package, throttle_count); +define_therm_throt_sysdev_one_ro (package_throttle_count);static struct attribute *thermal_throttle_attrs[] = { - &attr_throttle_count.attr, + &attr_core_throttle_count.attr,NULL };
@@ -106,16 +116,21 @@ static struct attribute_group thermal_throttle_attr_group = {* 1 : Event should be logged further, and a message has been * printed to the syslog. */ -static int therm_throt_process (bool is_throttled) +#define CORE_LEVEL 0 +#define PACKAGE_LEVEL 1 +static int therm_throt_process (bool is_throttled,int level) { -struct thermal_state *state; +struct _thermal_state *state;unsigned int this_cpu;bool was_throttled;u64 now; this_cpu =smp_processor_id (); now =get_jiffies_64 (); - state = &per_cpu (thermal_state, this_cpu); +if (level =CORE_LEVEL ) + state = &per_cpu (thermal_state, this_cpu).core; +else + state = &per_cpu (thermal_state, this_cpu).package; was_throttled = state->is_throttled; state->is_throttled = is_throttled;
@@ -132,13 +147,18 @@ static int therm_throt_process(bool is_throttled)/* if we just entered the thermal event */ if (is_throttled) { -printk (KERN_CRIT "CPU%d: Temperature above threshold, cpu clock throttled (total events = %lu)\n" , this_cpu, state->throttle_count); +printk (KERN_CRIT "CPU%d: %s temperature above threshold, cpu clock throttled (total events = %lu)\n" , + this_cpu, + level =CORE_LEVEL ?"Core" :"Package" , + state->throttle_count);add_taint (TAINT_MACHINE_CHECK );return 1 ; }if (was_throttled) { -printk (KERN_INFO "CPU%d: Temperature/speed normal\n" , this_cpu); +printk (KERN_INFO "CPU%d: %s temperature/speed normal\n" , + this_cpu, + level =CORE_LEVEL ?"Core" :"Package" );return 1 ; }
@@ -149,8 +169,19 @@ static int therm_throt_process(bool is_throttled)/* Add/Remove thermal_throttle interface for CPU device: */ static __cpuinit int thermal_throttle_add_dev (struct sys_device *sys_dev) { -return sysfs_create_group (&sys_dev->kobj, - &thermal_throttle_attr_group); +int err; +struct cpuinfo_x86 *c = &cpu_data (smp_processor_id ()); + + err =sysfs_create_group (&sys_dev->kobj, &thermal_throttle_attr_group); +if (err) +return err; + +if (cpu_has (c,X86_FEATURE_PTS )) + err =sysfs_add_file_to_group (&sys_dev->kobj, + &attr_package_throttle_count.attr, + thermal_throttle_attr_group.name); + +return err; }static __cpuinit void thermal_throttle_remove_dev (struct sys_device *sys_dev)
@@ -230,10 +261,25 @@ device_initcall(thermal_throttle_init_device);static void intel_thermal_interrupt (void ) {__u64 msr_val; +struct cpuinfo_x86 *c = &cpu_data (smp_processor_id ());rdmsrl (MSR_IA32_THERM_STATUS , msr_val); -if (therm_throt_process ((msr_val &THERM_STATUS_PROCHOT ) !=0 )) +if (therm_throt_process (msr_val &THERM_STATUS_PROCHOT , +CORE_LEVEL ) !=0 )mce_log_therm_throt_event (msr_val); + +if (cpu_has (c,X86_FEATURE_PTS )) { +rdmsrl (MSR_IA32_PACKAGE_THERM_STATUS , msr_val); +if (therm_throt_process (msr_val &PACKAGE_THERM_STATUS_PROCHOT , +PACKAGE_LEVEL ) !=0 ) +/* +* Set up the most significant bit to notify mce log +* that this thermal event is a package level event. +* This is a temp solution. May be changed in the future +* with mce log infrasture. +*/ +mce_log_therm_throt_event (((__u64 )1 <<63 ) | msr_val); + } }static void unexpected_thermal_interrupt (void )
@@ -338,6 +384,13 @@ void intel_init_thermal(struct cpuinfo_x86 *c)wrmsr (MSR_IA32_THERM_INTERRUPT , l | (THERM_INT_LOW_ENABLE |THERM_INT_HIGH_ENABLE ), h); +if (cpu_has (c,X86_FEATURE_PTS )) { +rdmsr (MSR_IA32_PACKAGE_THERM_INTERRUPT , l, h); +wrmsr (MSR_IA32_PACKAGE_THERM_INTERRUPT , + l | (PACKAGE_THERM_INT_LOW_ENABLE + |PACKAGE_THERM_INT_HIGH_ENABLE ), h); + } + smp_thermal_vector = intel_thermal_interrupt;rdmsr (MSR_IA32_MISC_ENABLE , l, h);
Signature · 7 lines
1.6.0.3 _______________________________________________ lm-sensors mailing list [email protected] http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
[tip:x86/hwmon] x86, hwmon: Package Level Thermal/Power: thermal throttling handler
Commit-ID: 55d435a227bd28c77afab326de44dfacc0b15059 Gitweb: http://git.kernel.org/tip/55d435a227bd28c77afab326de44dfacc0b15059 Author: Fenghua Yu <[email protected]> AuthorDate: Thu, 29 Jul 2010 17:13:44 -0700 Committer: H. Peter Anvin <[email protected]> CommitDate: Tue, 3 Aug 2010 15:58:56 -0700
x86, hwmon: Package Level Thermal/Power: thermal throttling handler
Add package level thermal throttle interrupt support. The interrupt handler increases package level thermal throttle count. It also logs the event in MCE log.
The package level thermal throttle interrupt happens across threads in a package. Each thread handles the interrupt individually. User level application is supposed to retrieve correct event count and log based on package/thread topology. This is the same situation for core level interrupt handler. In the future, interrupt may be reported only per package or per core.
core_throttle_count and package_throttle_count are used for user interface. Previously only throttle_count is used for core throttle count. If you think new core_throttle_count name breaks user interface, I can change this part.
LKML-Reference: <[email protected]>
--- arch/x86/kernel/cpu/mcheck/therm_throt.c | 89 ++++++++++++++++++++++++------ 1 files changed, 71 insertions(+), 18 deletions(-)
arch/x86/kernel/cpu/mcheck/therm_throt.c +71 -18
@@ -37,7 +37,7 @@/* * Current thermal throttling state: */ -struct thermal_state { +struct _thermal_state {bool is_throttled;u64 next_check;
@@ -45,6 +45,11 @@ struct thermal_state {unsigned long last_throttle_count; }; +struct thermal_state { +struct _thermal_state core; +struct _thermal_state package; +}; + static DEFINE_PER_CPU(struct thermal_state, thermal_state);static atomic_t therm_throt_en =ATOMIC_INIT (0 );
@@ -53,11 +58,13 @@ static u32 lvtthmr_init __read_mostly;#ifdef CONFIG_SYSFS #define define_therm_throt_sysdev_one_ro (_name) \ - static SYSDEV_ATTR(_name, 0444, therm_throt_sysdev_show_##_name, NULL) + static SYSDEV_ATTR(_name, 0444, \ + therm_throt_sysdev_show_##_name, \ + NULL) \ -#define define_therm_throt_sysdev_show_func (name) \ +#define define_therm_throt_sysdev_show_func (level, name) \ \ -static ssize_t therm_throt_sysdev_show_##name( \ +static ssize_t therm_throt_sysdev_show_##level##_##name( \ struct sys_device *dev, \ struct sysdev_attribute *attr, \ char *buf) \
@@ -66,21 +73,24 @@ static ssize_t therm_throt_sysdev_show_##name( \ssize_t ret; \ \preempt_disable ();/* CPU hotplug */ \ -if (cpu_online (cpu)) \ +if (cpu_online (cpu)) { \ ret =sprintf (buf,"%lu\n" , \ -per_cpu (thermal_state, cpu).name); \ -else \ +per_cpu (thermal_state, cpu).level.name); \ + }else \ ret =0 ; \preempt_enable (); \ \return ret; \ } -define_therm_throt_sysdev_show_func (throttle_count); -define_therm_throt_sysdev_one_ro (throttle_count); +define_therm_throt_sysdev_show_func (core, throttle_count); +define_therm_throt_sysdev_one_ro (core_throttle_count); + +define_therm_throt_sysdev_show_func (package, throttle_count); +define_therm_throt_sysdev_one_ro (package_throttle_count);static struct attribute *thermal_throttle_attrs[] = { - &attr_throttle_count.attr, + &attr_core_throttle_count.attr,NULL };
@@ -106,16 +116,21 @@ static struct attribute_group thermal_throttle_attr_group = {* 1 : Event should be logged further, and a message has been * printed to the syslog. */ -static int therm_throt_process (bool is_throttled) +#define CORE_LEVEL 0 +#define PACKAGE_LEVEL 1 +static int therm_throt_process (bool is_throttled,int level) { -struct thermal_state *state; +struct _thermal_state *state;unsigned int this_cpu;bool was_throttled;u64 now; this_cpu =smp_processor_id (); now =get_jiffies_64 (); - state = &per_cpu (thermal_state, this_cpu); +if (level ==CORE_LEVEL ) + state = &per_cpu (thermal_state, this_cpu).core; +else + state = &per_cpu (thermal_state, this_cpu).package; was_throttled = state->is_throttled; state->is_throttled = is_throttled;
@@ -132,13 +147,18 @@ static int therm_throt_process(bool is_throttled)/* if we just entered the thermal event */ if (is_throttled) { -printk (KERN_CRIT "CPU%d: Temperature above threshold, cpu clock throttled (total events = %lu)\n" , this_cpu, state->throttle_count); +printk (KERN_CRIT "CPU%d: %s temperature above threshold, cpu clock throttled (total events = %lu)\n" , + this_cpu, + level ==CORE_LEVEL ?"Core" :"Package" , + state->throttle_count);add_taint (TAINT_MACHINE_CHECK );return 1 ; }if (was_throttled) { -printk (KERN_INFO "CPU%d: Temperature/speed normal\n" , this_cpu); +printk (KERN_INFO "CPU%d: %s temperature/speed normal\n" , + this_cpu, + level ==CORE_LEVEL ?"Core" :"Package" );return 1 ; }
@@ -149,8 +169,19 @@ static int therm_throt_process(bool is_throttled)/* Add/Remove thermal_throttle interface for CPU device: */ static __cpuinit int thermal_throttle_add_dev (struct sys_device *sys_dev) { -return sysfs_create_group (&sys_dev->kobj, - &thermal_throttle_attr_group); +int err; +struct cpuinfo_x86 *c = &cpu_data (smp_processor_id ()); + + err =sysfs_create_group (&sys_dev->kobj, &thermal_throttle_attr_group); +if (err) +return err; + +if (cpu_has (c,X86_FEATURE_PTS )) + err =sysfs_add_file_to_group (&sys_dev->kobj, + &attr_package_throttle_count.attr, + thermal_throttle_attr_group.name); + +return err; }static __cpuinit void thermal_throttle_remove_dev (struct sys_device *sys_dev)
@@ -230,10 +261,25 @@ device_initcall(thermal_throttle_init_device);static void intel_thermal_interrupt (void ) {__u64 msr_val; +struct cpuinfo_x86 *c = &cpu_data (smp_processor_id ());rdmsrl (MSR_IA32_THERM_STATUS , msr_val); -if (therm_throt_process ((msr_val &THERM_STATUS_PROCHOT ) !=0 )) +if (therm_throt_process (msr_val &THERM_STATUS_PROCHOT , +CORE_LEVEL ) !=0 )mce_log_therm_throt_event (msr_val); + +if (cpu_has (c,X86_FEATURE_PTS )) { +rdmsrl (MSR_IA32_PACKAGE_THERM_STATUS , msr_val); +if (therm_throt_process (msr_val &PACKAGE_THERM_STATUS_PROCHOT , +PACKAGE_LEVEL ) !=0 ) +/* +* Set up the most significant bit to notify mce log +* that this thermal event is a package level event. +* This is a temp solution. May be changed in the future +* with mce log infrasture. +*/ +mce_log_therm_throt_event (((__u64 )1 <<63 ) | msr_val); + } }static void unexpected_thermal_interrupt (void )
@@ -338,6 +384,13 @@ void intel_init_thermal(struct cpuinfo_x86 *c)wrmsr (MSR_IA32_THERM_INTERRUPT , l | (THERM_INT_LOW_ENABLE |THERM_INT_HIGH_ENABLE ), h); +if (cpu_has (c,X86_FEATURE_PTS )) { +rdmsr (MSR_IA32_PACKAGE_THERM_INTERRUPT , l, h); +wrmsr (MSR_IA32_PACKAGE_THERM_INTERRUPT , + l | (PACKAGE_THERM_INT_LOW_ENABLE + |PACKAGE_THERM_INT_HIGH_ENABLE ), h); + } + smp_thermal_vector = intel_thermal_interrupt;rdmsr (MSR_IA32_MISC_ENABLE , l, h);
PATCH v3 4/5 Package Level Thermal Control and Power Limit Notification: power limit
In reply to Fenghua YuFrom: Fenghua Yu <[email protected]>
Power limit notification feature is published in Intel 64 and IA-32 Architectures SDMV Vol 3A 14.5.6 Power Limit Notification.
It is implemented first on Intel Sandy Bridge platform.
The patch handles notification interrupt. Interrupt handler dumps power limit information in log_buf, logs the event in mce log, and increases the event counters (core_power_limit and package_power_limit). Upper level applications could use the data to detect system health or diagnose functionality/performance issues.
In the future, the event could be handled in a more fancy way.
--- arch/x86/kernel/cpu/mcheck/therm_throt.c | 183 +++++++++++++++++++++--------- 1 files changed, 129 insertions(+), 54 deletions(-)
arch/x86/kernel/cpu/mcheck/therm_throt.c +129 -54
@@ -34,20 +34,25 @@/* How long to wait between reporting thermal events */ #define CHECK_INTERVAL (300 * HZ) +#define THERMAL_THROTTLING_EVENT 0 +#define POWER_LIMIT_EVENT 1 +/* -* Current thermal throttling state: +* Current thermal event state: */ struct _thermal_state { -bool is_throttled; - +bool new_event; +int event;u64 next_check; -unsigned long throttle_count; -unsigned long last_throttle_count; +unsigned long count; +unsigned long last_count; };struct thermal_state { -struct _thermal_state core; -struct _thermal_state package; +struct _thermal_state core_throttle; +struct _thermal_state core_power_limit; +struct _thermal_state package_throttle; +struct _thermal_state package_power_limit; }; static DEFINE_PER_CPU(struct thermal_state, thermal_state);
@@ -62,9 +67,9 @@ static u32 lvtthmr_init __read_mostly; therm_throt_sysdev_show_##_name, \ NULL) \ -#define define_therm_throt_sysdev_show_func(level, name) \ +#define define_therm_throt_sysdev_show_func(event, name) \ \ -static ssize_t therm_throt_sysdev_show_##level##_##name( \ +static ssize_t therm_throt_sysdev_show_##event##_##name( \ struct sys_device *dev, \ struct sysdev_attribute *attr, \ char *buf) \
@@ -75,7 +80,7 @@ static ssize_t therm_throt_sysdev_show_##level##_##name( \preempt_disable ();/* CPU hotplug */ \if (cpu_online (cpu)) { \ ret =sprintf (buf,"%lu\n" , \ -per_cpu (thermal_state, cpu).level.name); \ +per_cpu (thermal_state, cpu).event.name); \ }else \ ret =0 ; \preempt_enable (); \
@@ -83,23 +88,32 @@ static ssize_t therm_throt_sysdev_show_##level##_##name( \ return ret; \ } -define_therm_throt_sysdev_show_func (core, throttle_count); +define_therm_throt_sysdev_show_func (core_throttle, count);define_therm_throt_sysdev_one_ro (core_throttle_count); -define_therm_throt_sysdev_show_func (package, throttle_count); +define_therm_throt_sysdev_show_func (core_power_limit, count); +define_therm_throt_sysdev_one_ro (core_power_limit_count); + +define_therm_throt_sysdev_show_func (package_throttle, count);define_therm_throt_sysdev_one_ro (package_throttle_count); +define_therm_throt_sysdev_show_func (package_power_limit, count); +define_therm_throt_sysdev_one_ro (package_power_limit_count); +static struct attribute *thermal_throttle_attrs[] = { &attr_core_throttle_count.attr,NULL }; -static struct attribute_group thermal_throttle_attr_group = { +static struct attribute_group thermal_attr_group = { .attrs = thermal_throttle_attrs, .name ="thermal_throttle" };#endif /* CONFIG_SYSFS */ +#define CORE_LEVEL 0 +#define PACKAGE_LEVEL 1 +/*** * therm_throt_process - Process thermal throttling event from interrupt * @curr: Whether the condition is current or not (boolean), since the
@@ -116,49 +130,70 @@ static struct attribute_group thermal_throttle_attr_group = {* 1 : Event should be logged further, and a message has been * printed to the syslog. */ -#define CORE_LEVEL 0 -#define PACKAGE_LEVEL 1 -static int therm_throt_process (bool is_throttled,int level) +static int therm_throt_process (bool new_event,int event,int level) {struct _thermal_state *state; -unsigned int this_cpu; -bool was_throttled; +unsigned int this_cpu =smp_processor_id (); +bool old_event;u64 now; +struct thermal_state *pstate = &per_cpu (thermal_state, this_cpu); - this_cpu =smp_processor_id (); now =get_jiffies_64 (); -if (level =CORE_LEVEL ) - state = &per_cpu (thermal_state, this_cpu).core; -else - state = &per_cpu (thermal_state, this_cpu).package; +if (level =CORE_LEVEL ) { +if (event =THERMAL_THROTTLING_EVENT ) + state = &pstate->core_throttle; +else if (event =POWER_LIMIT_EVENT ) + state = &pstate->core_power_limit; +else +return 0 ; + }else if (level =PACKAGE_LEVEL ) { +if (event =THERMAL_THROTTLING_EVENT ) + state = &pstate->package_throttle; +else if (event =POWER_LIMIT_EVENT ) + state = &pstate->package_power_limit; +else +return 0 ; + }else +return 0 ; - was_throttled = state->is_throttled; - state->is_throttled = is_throttled; + old_event = state->new_event; + state->new_event = new_event; -if (is_throttled) - state->throttle_count++; +if (new_event) + state->count++;if (time_before64 (now, state->next_check) && - state->throttle_count != state->last_throttle_count) + state->count != state->last_count)return 0 ; state->next_check = now +CHECK_INTERVAL ; - state->last_throttle_count = state->throttle_count; + state->last_count = state->count;/* if we just entered the thermal event */ -if (is_throttled) { -printk (KERN_CRIT "CPU%d: %s temperature above threshold, cpu clock throttled (total events = %lu)\n" , - this_cpu, - level =CORE_LEVEL ?"Core" :"Package" , - state->throttle_count); +if (new_event) { +if (event =THERMAL_THROTTLING_EVENT ) +printk (KERN_CRIT "CPU%d: %s temperature above threshold, cpu clock throttled (total events = %lu)\n" , + this_cpu, + level =CORE_LEVEL ?"Core" :"Package" , + state->count); +else +printk (KERN_CRIT "CPU%d: %s power limit notification (total events = %lu)\n" , + this_cpu, + level =CORE_LEVEL ?"Core" :"Package" , + state->count);add_taint (TAINT_MACHINE_CHECK );return 1 ; } -if (was_throttled) { -printk (KERN_INFO "CPU%d: %s temperature/speed normal\n" , - this_cpu, - level =CORE_LEVEL ?"Core" :"Package" ); +if (old_event) { +if (event =THERMAL_THROTTLING_EVENT ) +printk (KERN_INFO "CPU%d: %s temperature/speed normal\n" , + this_cpu, + level =CORE_LEVEL ?"Core" :"Package" ); +else +printk (KERN_INFO "CPU%d: %s power limit normal\n" , + this_cpu, + level =CORE_LEVEL ?"Core" :"Package" );return 1 ; }
@@ -172,21 +207,29 @@ static __cpuinit int thermal_throttle_add_dev(struct sys_device *sys_dev)int err;struct cpuinfo_x86 *c = &cpu_data (smp_processor_id ()); - err =sysfs_create_group (&sys_dev->kobj, &thermal_throttle_attr_group); + err =sysfs_create_group (&sys_dev->kobj, &thermal_attr_group);if (err)return err; +if (cpu_has (c,X86_FEATURE_PLN )) + err =sysfs_add_file_to_group (&sys_dev->kobj, + &attr_core_power_limit_count.attr, + thermal_attr_group.name);if (cpu_has (c,X86_FEATURE_PTS )) err =sysfs_add_file_to_group (&sys_dev->kobj, &attr_package_throttle_count.attr, - thermal_throttle_attr_group.name); + thermal_attr_group.name); +if (cpu_has (c,X86_FEATURE_PLN )) + err =sysfs_add_file_to_group (&sys_dev->kobj, + &attr_package_power_limit_count.attr, + thermal_attr_group.name);return err; }static __cpuinit void thermal_throttle_remove_dev (struct sys_device *sys_dev) { -sysfs_remove_group (&sys_dev->kobj, &thermal_throttle_attr_group); +sysfs_remove_group (&sys_dev->kobj, &thermal_attr_group); }/* Mutex protecting device creation against CPU hotplug: */
@@ -257,6 +300,17 @@ device_initcall(thermal_throttle_init_device);#endif /* CONFIG_SYSFS */ +/* +* Set up the most two significant bit to notify mce log that this thermal +* event type. +* This is a temp solution. May be changed in the future with mce log +* infrasture. +*/ +#define CORE_THROTTLED (0) +#define CORE_POWER_LIMIT ((__u64)1 << 62) +#define PACKAGE_THROTTLED ((__u64)2 << 62) +#define PACKAGE_POWER_LIMIT ((__u64)3 << 62) +/* Thermal transition interrupt handler */ static void intel_thermal_interrupt (void ) {
@@ -264,21 +318,31 @@ static void intel_thermal_interrupt(void)struct cpuinfo_x86 *c = &cpu_data (smp_processor_id ());rdmsrl (MSR_IA32_THERM_STATUS , msr_val); +if (therm_throt_process (msr_val &THERM_STATUS_PROCHOT , +THERMAL_THROTTLING_EVENT ,CORE_LEVEL ) !=0 ) -mce_log_therm_throt_event (msr_val); +mce_log_therm_throt_event (CORE_THROTTLED | msr_val); + +if (cpu_has (c,X86_FEATURE_PLN )) +if (therm_throt_process (msr_val &THERM_STATUS_POWER_LIMIT , +POWER_LIMIT_EVENT , +CORE_LEVEL ) !=0 ) +mce_log_therm_throt_event (CORE_POWER_LIMIT | msr_val);if (cpu_has (c,X86_FEATURE_PTS )) {rdmsrl (MSR_IA32_PACKAGE_THERM_STATUS , msr_val);if (therm_throt_process (msr_val &PACKAGE_THERM_STATUS_PROCHOT , +THERMAL_THROTTLING_EVENT ,PACKAGE_LEVEL ) !=0 ) -/* -* Set up the most significant bit to notify mce log -* that this thermal event is a package level event. -* This is a temp solution. May be changed in the future -* with mce log infrasture. -*/ -mce_log_therm_throt_event (((__u64 )1 <<63 ) | msr_val); +mce_log_therm_throt_event (PACKAGE_THROTTLED | msr_val); +if (cpu_has (c,X86_FEATURE_PLN )) +if (therm_throt_process (msr_val & +PACKAGE_THERM_STATUS_POWER_LIMIT , +POWER_LIMIT_EVENT , +PACKAGE_LEVEL ) !=0 ) +mce_log_therm_throt_event (PACKAGE_POWER_LIMIT + | msr_val); } }
@@ -381,14 +445,25 @@ void intel_init_thermal(struct cpuinfo_x86 *c)apic_write (APIC_LVTTHMR , h);rdmsr (MSR_IA32_THERM_INTERRUPT , l, h); -wrmsr (MSR_IA32_THERM_INTERRUPT , - l | (THERM_INT_LOW_ENABLE |THERM_INT_HIGH_ENABLE ), h); +if (cpu_has (c,X86_FEATURE_PLN )) +wrmsr (MSR_IA32_THERM_INTERRUPT , + l | (THERM_INT_LOW_ENABLE + |THERM_INT_HIGH_ENABLE |THERM_INT_PLN_ENABLE ), h); +else +wrmsr (MSR_IA32_THERM_INTERRUPT , + l | (THERM_INT_LOW_ENABLE |THERM_INT_HIGH_ENABLE ), h);if (cpu_has (c,X86_FEATURE_PTS )) {rdmsr (MSR_IA32_PACKAGE_THERM_INTERRUPT , l, h); -wrmsr (MSR_IA32_PACKAGE_THERM_INTERRUPT , - l | (PACKAGE_THERM_INT_LOW_ENABLE - |PACKAGE_THERM_INT_HIGH_ENABLE ), h); +if (cpu_has (c,X86_FEATURE_PLN )) +wrmsr (MSR_IA32_PACKAGE_THERM_INTERRUPT , + l | (PACKAGE_THERM_INT_LOW_ENABLE + |PACKAGE_THERM_INT_HIGH_ENABLE + |PACKAGE_THERM_INT_PLN_ENABLE ), h); +else +wrmsr (MSR_IA32_PACKAGE_THERM_INTERRUPT , + l | (PACKAGE_THERM_INT_LOW_ENABLE + |PACKAGE_THERM_INT_HIGH_ENABLE ), h); } smp_thermal_vector = intel_thermal_interrupt;
Signature · 7 lines
1.6.0.3 _______________________________________________ lm-sensors mailing list [email protected] http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
[tip:x86/hwmon] x86, hwmon: Package Level Thermal/Power: power limit
Commit-ID: 0199114c31798af5b83841b21759b64171060d9b Gitweb: http://git.kernel.org/tip/0199114c31798af5b83841b21759b64171060d9b Author: Fenghua Yu <[email protected]> AuthorDate: Thu, 29 Jul 2010 17:13:45 -0700 Committer: H. Peter Anvin <[email protected]> CommitDate: Tue, 3 Aug 2010 15:58:56 -0700
x86, hwmon: Package Level Thermal/Power: power limit
Power limit notification feature is published in Intel 64 and IA-32 Architectures SDMV Vol 3A 14.5.6 Power Limit Notification.
It is implemented first on Intel Sandy Bridge platform.
The patch handles notification interrupt. Interrupt handler dumps power limit information in log_buf, logs the event in mce log, and increases the event counters (core_power_limit and package_power_limit). Upper level applications could use the data to detect system health or diagnose functionality/performance issues.
In the future, the event could be handled in a more fancy way.
LKML-Reference: <[email protected]>
--- arch/x86/kernel/cpu/mcheck/therm_throt.c | 183 +++++++++++++++++++++--------- 1 files changed, 129 insertions(+), 54 deletions(-)
arch/x86/kernel/cpu/mcheck/therm_throt.c +129 -54
@@ -34,20 +34,25 @@/* How long to wait between reporting thermal events */ #define CHECK_INTERVAL (300 * HZ) +#define THERMAL_THROTTLING_EVENT 0 +#define POWER_LIMIT_EVENT 1 +/* -* Current thermal throttling state: +* Current thermal event state: */ struct _thermal_state { -bool is_throttled; - +bool new_event; +int event;u64 next_check; -unsigned long throttle_count; -unsigned long last_throttle_count; +unsigned long count; +unsigned long last_count; };struct thermal_state { -struct _thermal_state core; -struct _thermal_state package; +struct _thermal_state core_throttle; +struct _thermal_state core_power_limit; +struct _thermal_state package_throttle; +struct _thermal_state package_power_limit; }; static DEFINE_PER_CPU(struct thermal_state, thermal_state);
@@ -62,9 +67,9 @@ static u32 lvtthmr_init __read_mostly; therm_throt_sysdev_show_##_name, \ NULL) \ -#define define_therm_throt_sysdev_show_func(level, name) \ +#define define_therm_throt_sysdev_show_func(event, name) \ \ -static ssize_t therm_throt_sysdev_show_##level##_##name( \ +static ssize_t therm_throt_sysdev_show_##event##_##name( \ struct sys_device *dev, \ struct sysdev_attribute *attr, \ char *buf) \
@@ -75,7 +80,7 @@ static ssize_t therm_throt_sysdev_show_##level##_##name( \preempt_disable ();/* CPU hotplug */ \if (cpu_online (cpu)) { \ ret =sprintf (buf,"%lu\n" , \ -per_cpu (thermal_state, cpu).level.name); \ +per_cpu (thermal_state, cpu).event.name); \ }else \ ret =0 ; \preempt_enable (); \
@@ -83,23 +88,32 @@ static ssize_t therm_throt_sysdev_show_##level##_##name( \ return ret; \ } -define_therm_throt_sysdev_show_func (core, throttle_count); +define_therm_throt_sysdev_show_func (core_throttle, count);define_therm_throt_sysdev_one_ro (core_throttle_count); -define_therm_throt_sysdev_show_func (package, throttle_count); +define_therm_throt_sysdev_show_func (core_power_limit, count); +define_therm_throt_sysdev_one_ro (core_power_limit_count); + +define_therm_throt_sysdev_show_func (package_throttle, count);define_therm_throt_sysdev_one_ro (package_throttle_count); +define_therm_throt_sysdev_show_func (package_power_limit, count); +define_therm_throt_sysdev_one_ro (package_power_limit_count); +static struct attribute *thermal_throttle_attrs[] = { &attr_core_throttle_count.attr,NULL }; -static struct attribute_group thermal_throttle_attr_group = { +static struct attribute_group thermal_attr_group = { .attrs = thermal_throttle_attrs, .name ="thermal_throttle" };#endif /* CONFIG_SYSFS */ +#define CORE_LEVEL 0 +#define PACKAGE_LEVEL 1 +/*** * therm_throt_process - Process thermal throttling event from interrupt * @curr: Whether the condition is current or not (boolean), since the
@@ -116,49 +130,70 @@ static struct attribute_group thermal_throttle_attr_group = {* 1 : Event should be logged further, and a message has been * printed to the syslog. */ -#define CORE_LEVEL 0 -#define PACKAGE_LEVEL 1 -static int therm_throt_process (bool is_throttled,int level) +static int therm_throt_process (bool new_event,int event,int level) {struct _thermal_state *state; -unsigned int this_cpu; -bool was_throttled; +unsigned int this_cpu =smp_processor_id (); +bool old_event;u64 now; +struct thermal_state *pstate = &per_cpu (thermal_state, this_cpu); - this_cpu =smp_processor_id (); now =get_jiffies_64 (); -if (level ==CORE_LEVEL ) - state = &per_cpu (thermal_state, this_cpu).core; -else - state = &per_cpu (thermal_state, this_cpu).package; +if (level ==CORE_LEVEL ) { +if (event ==THERMAL_THROTTLING_EVENT ) + state = &pstate->core_throttle; +else if (event ==POWER_LIMIT_EVENT ) + state = &pstate->core_power_limit; +else +return 0 ; + }else if (level ==PACKAGE_LEVEL ) { +if (event ==THERMAL_THROTTLING_EVENT ) + state = &pstate->package_throttle; +else if (event ==POWER_LIMIT_EVENT ) + state = &pstate->package_power_limit; +else +return 0 ; + }else +return 0 ; - was_throttled = state->is_throttled; - state->is_throttled = is_throttled; + old_event = state->new_event; + state->new_event = new_event; -if (is_throttled) - state->throttle_count++; +if (new_event) + state->count++;if (time_before64 (now, state->next_check) && - state->throttle_count != state->last_throttle_count) + state->count != state->last_count)return 0 ; state->next_check = now +CHECK_INTERVAL ; - state->last_throttle_count = state->throttle_count; + state->last_count = state->count;/* if we just entered the thermal event */ -if (is_throttled) { -printk (KERN_CRIT "CPU%d: %s temperature above threshold, cpu clock throttled (total events = %lu)\n" , - this_cpu, - level ==CORE_LEVEL ?"Core" :"Package" , - state->throttle_count); +if (new_event) { +if (event ==THERMAL_THROTTLING_EVENT ) +printk (KERN_CRIT "CPU%d: %s temperature above threshold, cpu clock throttled (total events = %lu)\n" , + this_cpu, + level ==CORE_LEVEL ?"Core" :"Package" , + state->count); +else +printk (KERN_CRIT "CPU%d: %s power limit notification (total events = %lu)\n" , + this_cpu, + level ==CORE_LEVEL ?"Core" :"Package" , + state->count);add_taint (TAINT_MACHINE_CHECK );return 1 ; } -if (was_throttled) { -printk (KERN_INFO "CPU%d: %s temperature/speed normal\n" , - this_cpu, - level ==CORE_LEVEL ?"Core" :"Package" ); +if (old_event) { +if (event ==THERMAL_THROTTLING_EVENT ) +printk (KERN_INFO "CPU%d: %s temperature/speed normal\n" , + this_cpu, + level ==CORE_LEVEL ?"Core" :"Package" ); +else +printk (KERN_INFO "CPU%d: %s power limit normal\n" , + this_cpu, + level ==CORE_LEVEL ?"Core" :"Package" );return 1 ; }
@@ -172,21 +207,29 @@ static __cpuinit int thermal_throttle_add_dev(struct sys_device *sys_dev)int err;struct cpuinfo_x86 *c = &cpu_data (smp_processor_id ()); - err =sysfs_create_group (&sys_dev->kobj, &thermal_throttle_attr_group); + err =sysfs_create_group (&sys_dev->kobj, &thermal_attr_group);if (err)return err; +if (cpu_has (c,X86_FEATURE_PLN )) + err =sysfs_add_file_to_group (&sys_dev->kobj, + &attr_core_power_limit_count.attr, + thermal_attr_group.name);if (cpu_has (c,X86_FEATURE_PTS )) err =sysfs_add_file_to_group (&sys_dev->kobj, &attr_package_throttle_count.attr, - thermal_throttle_attr_group.name); + thermal_attr_group.name); +if (cpu_has (c,X86_FEATURE_PLN )) + err =sysfs_add_file_to_group (&sys_dev->kobj, + &attr_package_power_limit_count.attr, + thermal_attr_group.name);return err; }static __cpuinit void thermal_throttle_remove_dev (struct sys_device *sys_dev) { -sysfs_remove_group (&sys_dev->kobj, &thermal_throttle_attr_group); +sysfs_remove_group (&sys_dev->kobj, &thermal_attr_group); }/* Mutex protecting device creation against CPU hotplug: */
@@ -257,6 +300,17 @@ device_initcall(thermal_throttle_init_device);#endif /* CONFIG_SYSFS */ +/* +* Set up the most two significant bit to notify mce log that this thermal +* event type. +* This is a temp solution. May be changed in the future with mce log +* infrasture. +*/ +#define CORE_THROTTLED (0) +#define CORE_POWER_LIMIT ((__u64)1 << 62) +#define PACKAGE_THROTTLED ((__u64)2 << 62) +#define PACKAGE_POWER_LIMIT ((__u64)3 << 62) +/* Thermal transition interrupt handler */ static void intel_thermal_interrupt (void ) {
@@ -264,21 +318,31 @@ static void intel_thermal_interrupt(void)struct cpuinfo_x86 *c = &cpu_data (smp_processor_id ());rdmsrl (MSR_IA32_THERM_STATUS , msr_val); +if (therm_throt_process (msr_val &THERM_STATUS_PROCHOT , +THERMAL_THROTTLING_EVENT ,CORE_LEVEL ) !=0 ) -mce_log_therm_throt_event (msr_val); +mce_log_therm_throt_event (CORE_THROTTLED | msr_val); + +if (cpu_has (c,X86_FEATURE_PLN )) +if (therm_throt_process (msr_val &THERM_STATUS_POWER_LIMIT , +POWER_LIMIT_EVENT , +CORE_LEVEL ) !=0 ) +mce_log_therm_throt_event (CORE_POWER_LIMIT | msr_val);if (cpu_has (c,X86_FEATURE_PTS )) {rdmsrl (MSR_IA32_PACKAGE_THERM_STATUS , msr_val);if (therm_throt_process (msr_val &PACKAGE_THERM_STATUS_PROCHOT , +THERMAL_THROTTLING_EVENT ,PACKAGE_LEVEL ) !=0 ) -/* -* Set up the most significant bit to notify mce log -* that this thermal event is a package level event. -* This is a temp solution. May be changed in the future -* with mce log infrasture. -*/ -mce_log_therm_throt_event (((__u64 )1 <<63 ) | msr_val); +mce_log_therm_throt_event (PACKAGE_THROTTLED | msr_val); +if (cpu_has (c,X86_FEATURE_PLN )) +if (therm_throt_process (msr_val & +PACKAGE_THERM_STATUS_POWER_LIMIT , +POWER_LIMIT_EVENT , +PACKAGE_LEVEL ) !=0 ) +mce_log_therm_throt_event (PACKAGE_POWER_LIMIT + | msr_val); } }
@@ -381,14 +445,25 @@ void intel_init_thermal(struct cpuinfo_x86 *c)apic_write (APIC_LVTTHMR , h);rdmsr (MSR_IA32_THERM_INTERRUPT , l, h); -wrmsr (MSR_IA32_THERM_INTERRUPT , - l | (THERM_INT_LOW_ENABLE |THERM_INT_HIGH_ENABLE ), h); +if (cpu_has (c,X86_FEATURE_PLN )) +wrmsr (MSR_IA32_THERM_INTERRUPT , + l | (THERM_INT_LOW_ENABLE + |THERM_INT_HIGH_ENABLE |THERM_INT_PLN_ENABLE ), h); +else +wrmsr (MSR_IA32_THERM_INTERRUPT , + l | (THERM_INT_LOW_ENABLE |THERM_INT_HIGH_ENABLE ), h);if (cpu_has (c,X86_FEATURE_PTS )) {rdmsr (MSR_IA32_PACKAGE_THERM_INTERRUPT , l, h); -wrmsr (MSR_IA32_PACKAGE_THERM_INTERRUPT , - l | (PACKAGE_THERM_INT_LOW_ENABLE - |PACKAGE_THERM_INT_HIGH_ENABLE ), h); +if (cpu_has (c,X86_FEATURE_PLN )) +wrmsr (MSR_IA32_PACKAGE_THERM_INTERRUPT , + l | (PACKAGE_THERM_INT_LOW_ENABLE + |PACKAGE_THERM_INT_HIGH_ENABLE + |PACKAGE_THERM_INT_PLN_ENABLE ), h); +else +wrmsr (MSR_IA32_PACKAGE_THERM_INTERRUPT , + l | (PACKAGE_THERM_INT_LOW_ENABLE + |PACKAGE_THERM_INT_HIGH_ENABLE ), h); } smp_thermal_vector = intel_thermal_interrupt;
PATCH v3 5/5 Package Level Thermal Control and Power Limit Notification: pkgtemp doc
In reply to Fenghua YuFrom: Fenghua Yu <[email protected]>
Document for package level thermal hwmon driver.
--- Documentation/hwmon/pkgtemp | 36 ++++++++++++++++++++++++++++++++++++ 1 files changed, 36 insertions(+), 0 deletions(-) create mode 100644 Documentation/hwmon/pkgtemp
Documentation/hwmon/pkgtemp new file +36 -0
@@ -0,0 +1,36 @@ +Kernel driver pkgtemp +=========== + +Supported chips: + * Intel family + Prefix: 'pkgtemp' + CPUID: + Datasheet: Intel 64 and IA-32 Architectures Software Developer's Manual + Volume 3A: System Programming Guide + +Author: Fenghua Yu + +Description +----------- + +This driver permits reading package level temperature sensor embedded inside +Intel CPU package. The sensors can be in core, uncore, memory controller, or +other components in a package. The feature is first implemented in Intel Sandy +Bridge platform. + +Temperature is measured in degrees Celsius and measurement resolution is +1 degree C. Valid temperatures are from 0 to TjMax degrees C, because the actual +value of temperature register is in fact a delta from TjMax. + +Temperature known as TjMax is the maximum junction temperature of package. +We get this from MSR_IA32_TEMPERATURE_TARGET. If the MSR is not accessible, +we define TjMax as 100 degrees Celsius. At this temperature, protection +mechanism will perform actions to forcibly cool down the package. Alarm +may be raised, if the temperature grows enough (more than TjMax) to trigger +the Out-Of-Spec bit. Following table summarizes the exported sysfs files: + +temp1_input - Package temperature (in millidegrees Celsius). +temp1_max - All cooling devices should be turned on. +temp1_crit - Maximum junction temperature (in millidegrees Celsius). +temp1_crit_alarm - Set when Out-of-spec bit is set, never clears. + Correct CPU operation is no longer guaranteed.
Signature · 7 lines
1.6.0.3 _______________________________________________ lm-sensors mailing list [email protected] http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
[tip:x86/hwmon] x86, hwmon: Package Level Thermal/Power: pkgtemp documentation
Commit-ID: a3159864f2978bf2ace9cc787d73d315c98bbf9a Gitweb: http://git.kernel.org/tip/a3159864f2978bf2ace9cc787d73d315c98bbf9a Author: Fenghua Yu <[email protected]> AuthorDate: Thu, 29 Jul 2010 17:13:46 -0700 Committer: H. Peter Anvin <[email protected]> CommitDate: Tue, 3 Aug 2010 15:58:56 -0700
x86, hwmon: Package Level Thermal/Power: pkgtemp documentation
Document for package level thermal hwmon driver.
LKML-Reference: <[email protected]>
--- Documentation/hwmon/pkgtemp | 36 ++++++++++++++++++++++++++++++++++++ 1 files changed, 36 insertions(+), 0 deletions(-)
Documentation/hwmon/pkgtemp new file +36 -0
@@ -0,0 +1,36 @@ +Kernel driver pkgtemp +====================== + +Supported chips: + * Intel family + Prefix: 'pkgtemp' + CPUID: + Datasheet: Intel 64 and IA-32 Architectures Software Developer's Manual + Volume 3A: System Programming Guide + +Author: Fenghua Yu + +Description +----------- + +This driver permits reading package level temperature sensor embedded inside +Intel CPU package. The sensors can be in core, uncore, memory controller, or +other components in a package. The feature is first implemented in Intel Sandy +Bridge platform. + +Temperature is measured in degrees Celsius and measurement resolution is +1 degree C. Valid temperatures are from 0 to TjMax degrees C, because the actual +value of temperature register is in fact a delta from TjMax. + +Temperature known as TjMax is the maximum junction temperature of package. +We get this from MSR_IA32_TEMPERATURE_TARGET. If the MSR is not accessible, +we define TjMax as 100 degrees Celsius. At this temperature, protection +mechanism will perform actions to forcibly cool down the package. Alarm +may be raised, if the temperature grows enough (more than TjMax) to trigger +the Out-Of-Spec bit. Following table summarizes the exported sysfs files: + +temp1_input - Package temperature (in millidegrees Celsius). +temp1_max - All cooling devices should be turned on. +temp1_crit - Maximum junction temperature (in millidegrees Celsius). +temp1_crit_alarm - Set when Out-of-spec bit is set, never clears. + Correct CPU operation is no longer guaranteed.