Skip to contentSkip to conversation

Support for Intel temperature sensors (DTS)

PATCH Superseded Teddy Astie 3 participants 11 messages Dec 19, 2025 – Jan 12

v4 of 5 Superseded Latest is v5 →

lore
Share & tools
lore
#1 lore raw

PATCH v4 1/4 x86/cpu-policy: define bits of leaf 6

In reply to Teddy Astie

From: Jan Beulich <[email protected]>

... as far as we presently use them in the codebase.

---
I don't know to which extend Reviewed-by applies as I only applied Andrew's proposal.

 xen/include/xen/lib/x86/cpu-policy.h | 27 ++++++++++++++++++++++++++-
 1 file changed, 26 insertions(+), 1 deletion(-)
xen/include/xen/lib/x86/cpu-policy.h +26 -1
index f94f23e159..8772ef80e3 100644
@@ -121,7 +121,32 @@ struct cpu_policy
             uint64_t :64, :64; /* Leaf 0x3 - PSN. */
             uint64_t :64, :64; /* Leaf 0x4 - Structured Cache. */
             uint64_t :64, :64; /* Leaf 0x5 - MONITOR. */
-            uint64_t :64, :64; /* Leaf 0x6 - Therm/Perf. */
+
+            /* Leaf 0x6 - Therm/Perf. */
+            bool :1,
+                turbo:1,
+                arat:1,
+                :1,
+                :1,
+                :1,
+                :1,
+                hwp:1,
+                hwp_notification:1,
+                hwp_activity_window:1,
+                hwp_epp:1,
+                hwp_plr:1,
+                :1,
+                hdc:1,
+                :1,
+                :1,
+                hwp_peci:1,
+                :1,
+                :1,
+                hw_feedback:1;
+            uint32_t :32; /* b */
+            bool aperfmperf:1;
+            uint32_t :32; /* d */
+
             uint64_t :64, :64; /* Leaf 0x7 - Structured Features. */
             uint64_t :64, :64; /* Leaf 0x8 - rsvd */
             uint64_t :64, :64; /* Leaf 0x9 - DCA */
Signature · 10 lines
2.51.2



--
Teddy Astie | Vates XCP-ng Developer

XCP-ng & Xen Orchestra - Vates solutions

web: https://vates.tech
#0 lore raw

v4 cover Support for Intel temperature sensors (DTS)

The idea here is to expose the DTS sensors through XENPF_resource_op and expose it for the user through xenpm.

v4:
- reuse original cpu-policy patch
- remove xenpm building on non-x86
- add changelog notes
- reviewed xc_resource_op return code handling
- only print missing MSR_TEMPERATURE_TARGET message once
- formatting changes

v3:
- use msr-index.h instead of adding defines to MSRs in xenpm

v2:
- moved userland part to xenpm
- use cpu policy infrastructure instead of inline cpuid_*

Jan Beulich (1):
x86/cpu-policy: define bits of leaf 6

Teddy Astie (3):
xenpm: Don't build outside of x86
x86/platform: Expose DTS sensors MSR
xenpm: Add get-intel-temp subcommand

 CHANGELOG.md                         |   6 ++
 tools/misc/Makefile                  |   2 +-
 tools/misc/xenpm.c                   | 113 ++++++++++++++++++++++++++-
 xen/arch/x86/include/asm/msr-index.h |   3 +
 xen/arch/x86/platform_hypercall.c    |   6 ++
 xen/include/xen/lib/x86/cpu-policy.h |  27 ++++++-
 6 files changed, 154 insertions(+), 3 deletions(-)
Signature · 10 lines
2.51.2



--
Teddy Astie | Vates XCP-ng Developer

XCP-ng & Xen Orchestra - Vates solutions

web: https://vates.tech
#7 lore raw

PATCH v4 3/4 x86/platform: Expose DTS sensors MSR

Intel provide CPU sensors through "DTS" MSRs. As these MSR are core-specific (or package-specific), we can't reliably fetch them from Dom0 directly. Expose these MSR (if supported) through XENPF_resource_op so that it is accessible through hypercall.

---
 xen/arch/x86/include/asm/msr-index.h | 3 +++
 xen/arch/x86/platform_hypercall.c    | 6 ++++++
 xen/include/xen/lib/x86/cpu-policy.h | 2 +-
 3 files changed, 10 insertions(+), 1 deletion(-)
xen/arch/x86/include/asm/msr-index.h +3 -0
index df52587c85..b92a278611 100644
@@ -115,6 +115,9 @@
 #define  MCU_OPT_CTRL_GDS_MIT_DIS           (_AC(1, ULL) <<  4)
 #define  MCU_OPT_CTRL_GDS_MIT_LOCK          (_AC(1, ULL) <<  5)
 
+#define MSR_TEMPERATURE_TARGET              0x000001a2
+#define MSR_PACKAGE_THERM_STATUS            0x000001b1
+
 #define MSR_FRED_RSP_SL0                    0x000001cc
 #define MSR_FRED_RSP_SL1                    0x000001cd
 #define MSR_FRED_RSP_SL2                    0x000001ce
xen/arch/x86/platform_hypercall.c +6 -0
index 79bb99e0b6..2f84494759 100644
@@ -27,6 +27,7 @@
 #include <asm/current.h>
 #include <public/platform.h>
 #include <acpi/cpufreq/processor_perf.h>
+#include <asm/cpu-policy.h>
 #include <asm/edd.h>
 #include <asm/microcode.h>
 #include <asm/mtrr.h>
@@ -86,6 +87,11 @@ static bool msr_read_allowed(unsigned int msr)
 
     case MSR_MCU_OPT_CTRL:
         return cpu_has_srbds_ctrl;
+    
+    case MSR_IA32_THERM_STATUS:
+    case MSR_TEMPERATURE_TARGET:
+    case MSR_PACKAGE_THERM_STATUS:
+        return host_cpu_policy.basic.dts;
     }
 
     if ( ppin_msr && msr == ppin_msr )
xen/include/xen/lib/x86/cpu-policy.h +1 -1
index 8772ef80e3..0362f1cb24 100644
@@ -123,7 +123,7 @@ struct cpu_policy
             uint64_t :64, :64; /* Leaf 0x5 - MONITOR. */
 
             /* Leaf 0x6 - Therm/Perf. */
-            bool :1,
+            bool dts:1,
                 turbo:1,
                 arat:1,
                 :1,
Signature · 10 lines
2.51.2



--
Teddy Astie | Vates XCP-ng Developer

XCP-ng & Xen Orchestra - Vates solutions

web: https://vates.tech
#3 lore raw

PATCH v4 2/4 xenpm: Don't build outside of x86

In reply to Teddy Astie

xenpm doesn't provide any interesting usable features outside of x86, skip building it if we are not x86.

---
 CHANGELOG.md        | 3 +++
 tools/misc/Makefile | 2 +-
 2 files changed, 4 insertions(+), 1 deletion(-)
CHANGELOG.md +3 -0
index 3aaf598623..1fa58ce848 100644
@@ -16,6 +16,9 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/)
      deprecated in Xen 4.7 and noted not to work correctly with AMD CPUs from
      2011 onwards.
 
+ - Removed xenpm on non-x86 platforms as it doesn't actually provide anything
+   useful outside of x86.
+
 ## [4.21.0](https://xenbits.xenproject.org/gitweb/?p=xen.git;a=shortlog;h=RELEASE-4.21.0) - 2025-11-19
 
 ### Changed
tools/misc/Makefile +1 -1
index c26e544e83..672df02c3b 100644
@@ -23,13 +23,13 @@ INSTALL_SBIN-$(CONFIG_X86)     += xen-lowmemd
 INSTALL_SBIN-$(CONFIG_X86)     += xen-mceinj
 INSTALL_SBIN-$(CONFIG_X86)     += xen-memshare
 INSTALL_SBIN-$(CONFIG_X86)     += xen-mfndump
+INSTALL_SBIN-$(CONFIG_X86)     += xenpm
 INSTALL_SBIN-$(CONFIG_X86)     += xen-ucode
 INSTALL_SBIN-$(CONFIG_X86)     += xen-vmtrace
 INSTALL_SBIN                   += xencov
 INSTALL_SBIN                   += xenhypfs
 INSTALL_SBIN                   += xenlockprof
 INSTALL_SBIN                   += xenperf
-INSTALL_SBIN                   += xenpm
 INSTALL_SBIN                   += xenwatchdogd
 INSTALL_SBIN                   += xen-access
 INSTALL_SBIN                   += xen-livepatch
Signature · 10 lines
2.51.2



--
Teddy Astie | Vates XCP-ng Developer

XCP-ng & Xen Orchestra - Vates solutions

web: https://vates.tech
#9 lore raw

PATCH v4 4/4 xenpm: Add get-intel-temp subcommand

In reply to Teddy Astie

get-intel-temp allows querying the per-core CPU temperature and per-package one on Intel processors (as usual Dom0 drivers cannot work due to misalignment between Dom0 vCPU and pCPUs).

---
 CHANGELOG.md       |   3 ++
 tools/misc/xenpm.c | 113 ++++++++++++++++++++++++++++++++++++++++++++-
 2 files changed, 115 insertions(+), 1 deletion(-)
CHANGELOG.md +3 -0
index 1fa58ce848..f55d77016c 100644
@@ -9,6 +9,9 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/)
 ### Changed
 
 ### Added
+ - On x86:
+   - Introduce get-intel-temp to xenpm to query CPU temperatures on Intel
+     platforms.
 
 ### Removed
  - On x86:
tools/misc/xenpm.c +112 -1

147 lines not on this page · raw message

--
2.51.2

--
Teddy Astie | Vates XCP-ng Developer

XCP-ng & Xen Orchestra - Vates solutions

web: https://vates.tech

#2 lore raw

Re: [PATCH v4 1/4] x86/cpu-policy: define bits of leaf 6

In reply to Teddy Astie
Quoted 9 lines · Teddy Astie

From: Jan Beulich <[email protected]>

... as far as we presently use them in the codebase.

---
I don't know to which extend Reviewed-by applies as I only applied Andrew's proposal.

Which I pointed out has a latent weakness. I'm intending to submit a new version of my patch, but instead with the names change to better match the SDM (as was requested by Andrew in reply to a change later in that other series of mine).

Jan

#4 lore raw

Re: [PATCH v4 2/4] xenpm: Don't build outside of x86

In reply to Teddy Astie
Quoted 9 lines · Teddy Astie

xenpm doesn't provide any interesting usable features outside of x86, skip building it if we are not x86.

---
 CHANGELOG.md        | 3 +++
 tools/misc/Makefile | 2 +-
 2 files changed, 4 insertions(+), 1 deletion(-)

Arm being what is mainly affected, I think you may have wanted to Cc Arm folks, for them to ack this (alongside Anthony).

CHANGELOG.md The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/) · line 20

Quoted 8 lines
CHANGELOG.md +2 -0
@@ -16,6 +16,9 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/)
      deprecated in Xen 4.7 and noted not to work correctly with AMD CPUs from
      2011 onwards.
 
+ - Removed xenpm on non-x86 platforms as it doesn't actually provide anything
+   useful outside of x86.

Maybe add the word "tool" for those who don't recognize "xenpm" at the first glance?

Jan

#10 lore raw

Re: [PATCH v4 4/4] xenpm: Add get-intel-temp subcommand

In reply to Teddy Astie
Quoted 41 lines · Teddy Astie
tools/misc/xenpm.c +37 -0
@@ -1354,6 +1358,113 @@ 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 -1:
+        /* xc_resource_op returns -1 in out of memory scenarios */
+        return -ENOMEM;
+    case 0:
+        /* This CPU isn't online or can't query this MSR */
+        return -ENODATA;
+    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;

tools/misc/xenpm.c void enable_turbo_mode(int argc, char *argv[]) · line 1407

Quoted 10 lines
tools/misc/xenpm.c +10 -0
+        }
+
+        tjmax = 100;
+        break;
+    }
+    case 2:
+    {
+        tjmax = (entries[1].val >> 16) & 0xff;
+        break;
+    }

No need for figure braces here; you don't have ones either in cases further up where there are no new local variables. Instead a blank line between non-fall- through case blocks may be nice.

Jan

Reviewed-by on v4 3/4
#8 lore raw

Re: [PATCH v4 3/4] x86/platform: Expose DTS sensors MSR

In reply to Teddy Astie
Quoted 7 lines · Teddy Astie

Intel provide CPU sensors through "DTS" MSRs. As these MSR are core-specific (or package-specific), we can't reliably fetch them from Dom0 directly. Expose these MSR (if supported) through XENPF_resource_op so that it is accessible through hypercall.

I wonder though whether ...

xen/include/xen/lib/x86/cpu-policy.h struct cpu_policy · line 126

Quoted 8 lines
xen/include/xen/lib/x86/cpu-policy.h +1 -1
@@ -123,7 +123,7 @@ struct cpu_policy
             uint64_t :64, :64; /* Leaf 0x5 - MONITOR. */
 
             /* Leaf 0x6 - Therm/Perf. */
-            bool :1,
+            bool dts:1,

... we really want to go with the acronym here, when for other bits the plan now is to stay close to the SDM names.

Jan

Acked-by on v4 2/4
#5 lore raw

Re: [PATCH v4 2/4] xenpm: Don't build outside of x86

In reply to Teddy Astie
Quoted 5 lines · Teddy Astie

xenpm doesn't provide any interesting usable features outside of x86, skip building it if we are not x86.

Happy with that on the Arm side? Julien suggested so on #XenDevel.

For me, this seems fine:

Quoted 24 lines
---
 CHANGELOG.md        | 3 +++
 tools/misc/Makefile | 2 +-
 2 files changed, 4 insertions(+), 1 deletion(-)
CHANGELOG.md +3 -0
index 3aaf598623..1fa58ce848 100644
@@ -16,6 +16,9 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/)
      deprecated in Xen 4.7 and noted not to work correctly with AMD CPUs from
      2011 onwards.
 
+ - Removed xenpm on non-x86 platforms as it doesn't actually provide anything
+   useful outside of x86.
+
 ## [4.21.0](https://xenbits.xenproject.org/gitweb/?p=xen.git;a=shortlog;h=RELEASE-4.21.0) - 2025-11-19
 
 ### Changed
tools/misc/Makefile +0 -0
index c26e544e83..672df02c3b 100644

tools/misc/Makefile INSTALL_SBIN-$(CONFIG_X86) += xen-lowmemd · line 26

Quoted 4 lines
tools/misc/Makefile +1 -0
@@ -23,13 +23,13 @@ INSTALL_SBIN-$(CONFIG_X86)     += xen-lowmemd
 INSTALL_SBIN-$(CONFIG_X86)     += xen-mceinj
 INSTALL_SBIN-$(CONFIG_X86)     += xen-memshare
 INSTALL_SBIN-$(CONFIG_X86)     += xen-mfndump
+INSTALL_SBIN-$(CONFIG_X86)     += xenpm

Nit: I would sort this by taking the dash - into account since we do so for the arch-common list, so xenpm would go after xen-vmtrace.

Quoted 10 lines
tools/misc/Makefile +0 -1
 INSTALL_SBIN-$(CONFIG_X86)     += xen-ucode
 INSTALL_SBIN-$(CONFIG_X86)     += xen-vmtrace
 INSTALL_SBIN                   += xencov
 INSTALL_SBIN                   += xenhypfs
 INSTALL_SBIN                   += xenlockprof
 INSTALL_SBIN                   += xenperf
-INSTALL_SBIN                   += xenpm
 INSTALL_SBIN                   += xenwatchdogd
 INSTALL_SBIN                   += xen-access
 INSTALL_SBIN                   += xen-livepatch
Signature · 1 line
Anthony PERARD
#6 lore raw

Re: [PATCH v4 2/4] xenpm: Don't build outside of x86

Quoted 11 lines · Anthony PERARD

tools/misc/Makefile INSTALL_SBIN-$(CONFIG_X86) += xen-lowmemd · line 26

Quoted 7 lines · Teddy Astie
tools/misc/Makefile +1 -0
@@ -23,13 +23,13 @@ INSTALL_SBIN-$(CONFIG_X86)     += xen-lowmemd
 INSTALL_SBIN-$(CONFIG_X86)     += xen-mceinj
 INSTALL_SBIN-$(CONFIG_X86)     += xen-memshare
 INSTALL_SBIN-$(CONFIG_X86)     += xen-mfndump
+INSTALL_SBIN-$(CONFIG_X86)     += xenpm

Nit: I would sort this by taking the dash - into account since we do so for the arch-common list, so xenpm would go after xen-vmtrace.

Not really; imo it's coincidental that all present x86-only ones have a dash. I think this simply wants to ...

Quoted 7 lines

tools/misc/Makefile INSTALL_SBIN-$(CONFIG_X86) += xen-lowmemd · line 32

Quoted 7 lines
tools/misc/Makefile +0 -1
 INSTALL_SBIN-$(CONFIG_X86)     += xen-ucode
 INSTALL_SBIN-$(CONFIG_X86)     += xen-vmtrace
 INSTALL_SBIN                   += xencov
 INSTALL_SBIN                   += xenhypfs
 INSTALL_SBIN                   += xenlockprof
 INSTALL_SBIN                   += xenperf
-INSTALL_SBIN                   += xenpm

... stay in its present place, merely becoming conditional.

Quoted 3 lines

tools/misc/Makefile INSTALL_SBIN-$(CONFIG_X86) += xen-lowmemd · line 35

Quoted 3 lines
tools/misc/Makefile +0 -0
 INSTALL_SBIN                   += xenwatchdogd
 INSTALL_SBIN                   += xen-access
 INSTALL_SBIN                   += xen-livepatch

The three entries with dashes quite likely were blindly added to the tail, without paying attention to sorting. Imo they want moving up into their respective positions.

Jan