CHANGELOG.md | 2 + tools/misc/xenpm.c | 121 ++++++++++++++++++++++++++++++++++++++++++++- 2 files changed, 122 insertions(+), 1 deletion(-)
get-core-temp allows querying the per-core CPU temperature and
per-package one on processors that supports Digital Temperature Sensors
(most Intel processors; as usual Dom0 drivers cannot work due to
misalignment between Dom0 vCPU and pCPUs).
Signed-off-by: Teddy Astie <teddy.astie@vates.tech>
---
v4: https://lore.kernel.org/xen-devel/cover.1766158766.git.teddy.astie@vates.tech/
v5: Removed trailing whitespace.
v6: Report errors through errno and use strerror() to display them
v7:
- Rename get-intel-temp with get-dts-temp
- handle properly errno
- make process return a error code if no data
v8:
- update Changelog
- improve error handling
- rename core-dts-temp with get-core-temp
v9:
- exit with EXIT_{SUCCESS,FAILURE} instead of errno
v10:
- make error handling more uniform
- removed "No data" message
- stop querying package temperature on failure
CHANGELOG.md | 2 +
tools/misc/xenpm.c | 121 ++++++++++++++++++++++++++++++++++++++++++++-
2 files changed, 122 insertions(+), 1 deletion(-)
diff --git a/CHANGELOG.md b/CHANGELOG.md
index 166df43c63..c8cb125fae 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -16,6 +16,8 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/)
mitigate (by rate-limiting) the system wide impact of an HVM guest
misusing atomic instructions.
- Support for CPIO microcode in discrete multiboot modules.
+ - Introduce get-core-temp to xenpm to query CPU temperatures on Intel
+ platforms.
- On Arm:
- Support for guest suspend and resume to/from RAM via vPSCI.
diff --git a/tools/misc/xenpm.c b/tools/misc/xenpm.c
index e4902d2e82..3371bb5be7 100644
--- a/tools/misc/xenpm.c
+++ b/tools/misc/xenpm.c
@@ -32,11 +32,14 @@
#include <xen-tools/common-macros.h>
+#include <xen/asm/msr-index.h>
+
#define MAX_PKG_RESIDENCIES 12
#define MAX_CORE_RESIDENCIES 8
static xc_interface *xc_handle;
static unsigned int max_cpu_nr;
+static xc_physinfo_t physinfo;
/* help message */
void show_help(void)
@@ -93,6 +96,7 @@ void show_help(void)
" units default to \"us\" if unspecified.\n"
" truncates un-representable values.\n"
" 0 lets the hardware decide.\n"
+ " get-core-temp [cpuid] get CPU temperature for <cpuid> or all (Intel only)\n"
" start [seconds] start collect Cx/Px statistics,\n"
" output after CTRL-C or SIGINT or several seconds.\n"
" enable-turbo-mode [cpuid] enable Turbo Mode for processors that support it.\n"
@@ -1354,6 +1358,121 @@ void enable_turbo_mode(int argc, char *argv[])
errno, strerror(errno));
}
+static int fetch_dts_temp(xc_interface *xch, uint32_t cpu, bool package, int *temp)
+{
+ xc_resource_entry_t entries[] = {
+ { .idx = package ? MSR_PACKAGE_THERM_STATUS : MSR_IA32_THERM_STATUS },
+ { .idx = MSR_TEMPERATURE_TARGET },
+ };
+ struct xc_resource_op ops = {
+ .cpu = cpu,
+ .entries = entries,
+ .nr_entries = ARRAY_SIZE(entries),
+ };
+ int tjmax;
+
+ int ret = xc_resource_op(xch, 1, &ops);
+
+ switch ( ret )
+ {
+ case 0:
+ /* This CPU isn't online or can't query this MSR */
+ errno = ENODATA;
+ return -1;
+
+ case 1:
+ {
+ /*
+ * The CPU doesn't support MSR_TEMPERATURE_TARGET, we assume it's 100
+ * which is correct aside a few selected Atom CPUs. Check Linux
+ * kernel's coretemp.c for more information.
+ */
+ static bool has_reported_once = false;
+
+ if ( !has_reported_once )
+ {
+ fprintf(stderr, "MSR_TEMPERATURE_TARGET is not supported, assume "
+ "tjmax = 100, readings may be incorrect.\n");
+ has_reported_once = true;
+ }
+
+ tjmax = 100;
+ break;
+ }
+
+ case 2:
+ tjmax = (entries[1].val >> 16) & 0xff;
+ break;
+
+ default:
+ if ( ret > 0 )
+ {
+ fprintf(stderr, "Got unexpected xc_resource_op return value: %d", ret);
+ errno = EINVAL;
+ }
+ return -1;
+ }
+
+ *temp = tjmax - ((entries[0].val >> 16) & 0xff);
+ return 0;
+}
+
+static void get_core_temp(int argc, char *argv[])
+{
+ int temp = -1, cpu = -1;
+ unsigned int socket;
+ bool has_data = false;
+
+ if ( argc > 0 )
+ parse_cpuid(argv[0], &cpu);
+
+ if ( cpu != -1 )
+ {
+ if ( fetch_dts_temp(xc_handle, cpu, false, &temp) )
+ {
+ fprintf(stderr, "Unable to fetch temperature (%d - %s)\n",
+ errno, strerror(errno));
+ exit(EXIT_FAILURE);
+ }
+ else
+ printf("CPU%d: %d°C\n", cpu, temp);
+ return;
+ }
+
+ /* Per socket measurement */
+ for ( socket = 0, cpu = 0; cpu < max_cpu_nr;
+ socket++, cpu += physinfo.cores_per_socket * physinfo.threads_per_core )
+ {
+ if ( fetch_dts_temp(xc_handle, cpu, true, &temp) )
+ {
+ fprintf(stderr,
+ "[Package%u] Unable to fetch temperature (%d - %s)\n",
+ cpu, errno, strerror(errno));
+ /* CPU may not support package temperatures, but still support DTS */
+ break;
+ }
+
+ has_data = true;
+ printf("Package%u: %d°C\n", socket, temp);
+ }
+
+ for ( cpu = 0; cpu < max_cpu_nr; cpu += physinfo.threads_per_core )
+ {
+ if ( fetch_dts_temp(xc_handle, cpu, false, &temp) )
+ {
+ fprintf(stderr, "[CPU%d] Unable to fetch temperature (%d - %s)\n",
+ cpu, errno, strerror(errno));
+ continue;
+ }
+
+ has_data = true;
+ printf("CPU%d: %d°C\n", cpu, temp);
+ }
+
+ if ( !has_data )
+ exit(EXIT_FAILURE);
+}
+
void disable_turbo_mode(int argc, char *argv[])
{
int cpuid = -1;
@@ -1618,12 +1737,12 @@ struct {
{ "set-max-cstate", set_max_cstate_func},
{ "enable-turbo-mode", enable_turbo_mode },
{ "disable-turbo-mode", disable_turbo_mode },
+ { "get-core-temp", get_core_temp },
};
int main(int argc, char *argv[])
{
int i, ret = 0;
- xc_physinfo_t physinfo;
int nr_matches = 0;
int matches_main_options[ARRAY_SIZE(main_options)];
--
2.52.0
--
Teddy Astie | Vates XCP-ng Developer
XCP-ng & Xen Orchestra - Vates solutions
web: https://vates.techOn 07.04.2026 16:10, Teddy Astie wrote:
> --- a/CHANGELOG.md
> +++ b/CHANGELOG.md
> @@ -16,6 +16,8 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/)
> mitigate (by rate-limiting) the system wide impact of an HVM guest
> misusing atomic instructions.
> - Support for CPIO microcode in discrete multiboot modules.
> + - Introduce get-core-temp to xenpm to query CPU temperatures on Intel
> + platforms.
Would you mind inserting "command" or "option" before "to xenpm"?
> @@ -1354,6 +1358,121 @@ void enable_turbo_mode(int argc, char *argv[])
> errno, strerror(errno));
> }
>
> +static int fetch_dts_temp(xc_interface *xch, uint32_t cpu, bool package, int *temp)
> +{
> + xc_resource_entry_t entries[] = {
> + { .idx = package ? MSR_PACKAGE_THERM_STATUS : MSR_IA32_THERM_STATUS },
> + { .idx = MSR_TEMPERATURE_TARGET },
> + };
> + struct xc_resource_op ops = {
> + .cpu = cpu,
> + .entries = entries,
> + .nr_entries = ARRAY_SIZE(entries),
> + };
> + int tjmax;
> +
> + int ret = xc_resource_op(xch, 1, &ops);
> +
> + switch ( ret )
> + {
> + case 0:
> + /* This CPU isn't online or can't query this MSR */
> + errno = ENODATA;
> + return -1;
> +
> + case 1:
> + {
> + /*
> + * The CPU doesn't support MSR_TEMPERATURE_TARGET, we assume it's 100
> + * which is correct aside a few selected Atom CPUs. Check Linux
> + * kernel's coretemp.c for more information.
> + */
> + static bool has_reported_once = false;
> +
> + if ( !has_reported_once )
> + {
> + fprintf(stderr, "MSR_TEMPERATURE_TARGET is not supported, assume "
> + "tjmax = 100, readings may be incorrect.\n");
> + has_reported_once = true;
> + }
> +
> + tjmax = 100;
> + break;
> + }
> +
> + case 2:
> + tjmax = (entries[1].val >> 16) & 0xff;
> + break;
> +
> + default:
> + if ( ret > 0 )
> + {
> + fprintf(stderr, "Got unexpected xc_resource_op return value: %d", ret);
> + errno = EINVAL;
> + }
> + return -1;
> + }
> +
> + *temp = tjmax - ((entries[0].val >> 16) & 0xff);
> + return 0;
> +}
> +
> +static void get_core_temp(int argc, char *argv[])
> +{
> + int temp = -1, cpu = -1;
cpu's initializer is needed, but why would temp need one? You rely on ...
> + unsigned int socket;
> + bool has_data = false;
> +
> + if ( argc > 0 )
> + parse_cpuid(argv[0], &cpu);
> +
> + if ( cpu != -1 )
> + {
> + if ( fetch_dts_temp(xc_handle, cpu, false, &temp) )
> + {
> + fprintf(stderr, "Unable to fetch temperature (%d - %s)\n",
> + errno, strerror(errno));
> + exit(EXIT_FAILURE);
> + }
> + else
> + printf("CPU%d: %d°C\n", cpu, temp);
> + return;
> + }
> +
> + /* Per socket measurement */
> + for ( socket = 0, cpu = 0; cpu < max_cpu_nr;
> + socket++, cpu += physinfo.cores_per_socket * physinfo.threads_per_core )
> + {
> + if ( fetch_dts_temp(xc_handle, cpu, true, &temp) )
... fetch_dts_temp() to always update it in the success case anyway, both here and
in the other loop further down.
Other than this (happy to adjust while committing, provided you agree):
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Jan
Le 08/04/2026 à 14:36, Jan Beulich a écrit :
> On 07.04.2026 16:10, Teddy Astie wrote:
>> --- a/CHANGELOG.md
>> +++ b/CHANGELOG.md
>> @@ -16,6 +16,8 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/)
>> mitigate (by rate-limiting) the system wide impact of an HVM guest
>> misusing atomic instructions.
>> - Support for CPIO microcode in discrete multiboot modules.
>> + - Introduce get-core-temp to xenpm to query CPU temperatures on Intel
>> + platforms.
>
> Would you mind inserting "command" or "option" before "to xenpm"?
>
Some like
> Introduce get-core-temp option to xenpm command ... ?
(or something like that)
I don't have any issue with rewording it.
>> @@ -1354,6 +1358,121 @@ void enable_turbo_mode(int argc, char *argv[])
>> errno, strerror(errno));
>> }
>>
>> +static int fetch_dts_temp(xc_interface *xch, uint32_t cpu, bool package, int *temp)
>> +{
>> + xc_resource_entry_t entries[] = {
>> + { .idx = package ? MSR_PACKAGE_THERM_STATUS : MSR_IA32_THERM_STATUS },
>> + { .idx = MSR_TEMPERATURE_TARGET },
>> + };
>> + struct xc_resource_op ops = {
>> + .cpu = cpu,
>> + .entries = entries,
>> + .nr_entries = ARRAY_SIZE(entries),
>> + };
>> + int tjmax;
>> +
>> + int ret = xc_resource_op(xch, 1, &ops);
>> +
>> + switch ( ret )
>> + {
>> + case 0:
>> + /* This CPU isn't online or can't query this MSR */
>> + errno = ENODATA;
>> + return -1;
>> +
>> + case 1:
>> + {
>> + /*
>> + * The CPU doesn't support MSR_TEMPERATURE_TARGET, we assume it's 100
>> + * which is correct aside a few selected Atom CPUs. Check Linux
>> + * kernel's coretemp.c for more information.
>> + */
>> + static bool has_reported_once = false;
>> +
>> + if ( !has_reported_once )
>> + {
>> + fprintf(stderr, "MSR_TEMPERATURE_TARGET is not supported, assume "
>> + "tjmax = 100, readings may be incorrect.\n");
>> + has_reported_once = true;
>> + }
>> +
>> + tjmax = 100;
>> + break;
>> + }
>> +
>> + case 2:
>> + tjmax = (entries[1].val >> 16) & 0xff;
>> + break;
>> +
>> + default:
>> + if ( ret > 0 )
>> + {
>> + fprintf(stderr, "Got unexpected xc_resource_op return value: %d", ret);
>> + errno = EINVAL;
>> + }
>> + return -1;
>> + }
>> +
>> + *temp = tjmax - ((entries[0].val >> 16) & 0xff);
>> + return 0;
>> +}
>> +
>> +static void get_core_temp(int argc, char *argv[])
>> +{
>> + int temp = -1, cpu = -1;
>
> cpu's initializer is needed, but why would temp need one? You rely on ...
>
>> + unsigned int socket;
>> + bool has_data = false;
>> +
>> + if ( argc > 0 )
>> + parse_cpuid(argv[0], &cpu);
>> +
>> + if ( cpu != -1 )
>> + {
>> + if ( fetch_dts_temp(xc_handle, cpu, false, &temp) )
>> + {
>> + fprintf(stderr, "Unable to fetch temperature (%d - %s)\n",
>> + errno, strerror(errno));
>> + exit(EXIT_FAILURE);
>> + }
>> + else
>> + printf("CPU%d: %d°C\n", cpu, temp);
>> + return;
>> + }
>> +
>> + /* Per socket measurement */
>> + for ( socket = 0, cpu = 0; cpu < max_cpu_nr;
>> + socket++, cpu += physinfo.cores_per_socket * physinfo.threads_per_core )
>> + {
>> + if ( fetch_dts_temp(xc_handle, cpu, true, &temp) )
>
> ... fetch_dts_temp() to always update it in the success case anyway, both here and
> in the other loop further down.
>
Indeed, that's not really required anymore.
> Other than this (happy to adjust while committing, provided you agree):
> Reviewed-by: Jan Beulich <jbeulich@suse.com>
>
Looks good to me with the changes.
> Jan
>
Teddy
--
Teddy Astie | Vates XCP-ng Developer
XCP-ng & Xen Orchestra - Vates solutions
web: https://vates.tech
© 2016 - 2026 Red Hat, Inc.