Skip to contentSkip to conversation

platform/x86: don't unconditionally attach Intel PMC when virtualized

PATCH In review Roger Pau Monne 3 participants 3 messages Nov 10, 2022 – Nov 15, 2022

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH v3 1/1 platform/x86: don't unconditionally attach Intel PMC when virtualized

The current logic in the Intel PMC driver will forcefully attach it when detecting any CPU on the intel_pmc_core_platform_ids array, even if the matching ACPI device is not present.

There's no checking in pmc_core_probe() to assert that the PMC device is present, and hence on virtualized environments the PMC device probes successfully, even if the underlying registers are not present. Previous to 21ae43570940 the driver would check for the presence of a specific PCI device, and that prevented the driver from attaching when running virtualized.

Fix by only forcefully attaching the PMC device when not running virtualized. Note that virtualized platforms can still get the device to load if the appropriate ACPI device is present on the tables provided to the VM.

Make an exception for the Xen initial domain, which does have full hardware access, and hence can attach to the PMC if present.

Fixes 21ae43570940 ('platform/x86: intel_pmc_core: Substitute PCI with CPUID enumeration')

---
Changes since v2:
- Don't split condition line.

Changes since v1:
- Use cpu_feature_enabled() instead of boot_cpu_has().

---
 drivers/platform/x86/intel/pmc/pltdrv.c | 9 +++++++++
 1 file changed, 9 insertions(+)
drivers/platform/x86/intel/pmc/pltdrv.c +9 -0
index 15ca8afdd973..ddfba38c2104 100644
@@ -18,6 +18,8 @@
 #include <asm/cpu_device_id.h>
 #include <asm/intel-family.h>
 
+#include <xen/xen.h>
+
 static void intel_pmc_core_release(struct device *dev)
 {
 	kfree(dev);
@@ -53,6 +55,13 @@ static int __init pmc_core_platform_init(void)
 	if (acpi_dev_present("INT33A1", NULL, -1))
 		return -ENODEV;
 
+	/*
+	 * Skip forcefully attaching the device for VMs. Make an exception for
+	 * Xen dom0, which does have full hardware access.
+	 */
+	if (cpu_feature_enabled(X86_FEATURE_HYPERVISOR) && !xen_initial_domain())
+		return -ENODEV;
+
 	if (!x86_match_cpu(intel_pmc_core_platform_ids))
 		return -ENODEV;
 
Signature · 1 line
2.37.3
Reviewed-by on v3 1/1
#1 lore raw
Quoted 11 lines · Roger Pau Monne

The current logic in the Intel PMC driver will forcefully attach it when detecting any CPU on the intel_pmc_core_platform_ids array, even if the matching ACPI device is not present.

There's no checking in pmc_core_probe() to assert that the PMC device is present, and hence on virtualized environments the PMC device probes successfully, even if the underlying registers are not present. Previous to 21ae43570940 the driver would check for the presence of a specific PCI device, and that prevented the driver from attaching when running virtualized.

Quoted 7 lines

Fix by only forcefully attaching the PMC device when not running virtualized. Note that virtualized platforms can still get the device to load if the appropriate ACPI device is present on the tables provided to the VM.

Make an exception for the Xen initial domain, which does have full hardware access, and hence can attach to the PMC if present.

Quoted 43 lines
Fixes 21ae43570940 ('platform/x86: intel_pmc_core: Substitute PCI with CPUID enumeration')

---
Changes since v2:
- Don't split condition line.

Changes since v1:
- Use cpu_feature_enabled() instead of boot_cpu_has().

---
 drivers/platform/x86/intel/pmc/pltdrv.c | 9 +++++++++
 1 file changed, 9 insertions(+)
drivers/platform/x86/intel/pmc/pltdrv.c +9 -0
index 15ca8afdd973..ddfba38c2104 100644
@@ -18,6 +18,8 @@
 #include <asm/cpu_device_id.h>
 #include <asm/intel-family.h>
 
+#include <xen/xen.h>
+
 static void intel_pmc_core_release(struct device *dev)
 {
 	kfree(dev);
@@ -53,6 +55,13 @@ static int __init pmc_core_platform_init(void)
 	if (acpi_dev_present("INT33A1", NULL, -1))
 		return -ENODEV;
 
+	/*
+	 * Skip forcefully attaching the device for VMs. Make an exception for
+	 * Xen dom0, which does have full hardware access.
+	 */
+	if (cpu_feature_enabled(X86_FEATURE_HYPERVISOR) && !xen_initial_domain())
+		return -ENODEV;
+
 	if (!x86_match_cpu(intel_pmc_core_platform_ids))
 		return -ENODEV;
 
Signature · 1 line
2.37.3
Signature · 2 lines
With Best Regards,
Andy Shevchenko
#2 lore raw
In reply to Roger Pau Monne

Hi,

Quoted 16 lines · Roger Pau Monne

The current logic in the Intel PMC driver will forcefully attach it when detecting any CPU on the intel_pmc_core_platform_ids array, even if the matching ACPI device is not present.

There's no checking in pmc_core_probe() to assert that the PMC device is present, and hence on virtualized environments the PMC device probes successfully, even if the underlying registers are not present. Previous to 21ae43570940 the driver would check for the presence of a specific PCI device, and that prevented the driver from attaching when running virtualized.

Fix by only forcefully attaching the PMC device when not running virtualized. Note that virtualized platforms can still get the device to load if the appropriate ACPI device is present on the tables provided to the VM.

Quoted 6 lines

Make an exception for the Xen initial domain, which does have full hardware access, and hence can attach to the PMC if present.

Fixes 21ae43570940 ('platform/x86: intel_pmc_core: Substitute PCI with CPUID enumeration')

Thank you for your patch, I've applied this patch to my fixes branch:
https://git.kernel.org/pub/scm/linux/kernel/git/pdx86/platform-drivers-x86.git/log/?h=fixes

Note it will show up in my fixes branch once I've pushed my local branch there, which might take a while.

I will include this patch in my next fixes pull-req to Linus for the current kernel development cycle.

Regards,

Hans

Quoted 37 lines

---
Changes since v2:
- Don't split condition line.

Changes since v1:
- Use cpu_feature_enabled() instead of boot_cpu_has().

---
 drivers/platform/x86/intel/pmc/pltdrv.c | 9 +++++++++
 1 file changed, 9 insertions(+)
drivers/platform/x86/intel/pmc/pltdrv.c +9 -0
index 15ca8afdd973..ddfba38c2104 100644
@@ -18,6 +18,8 @@
 #include <asm/cpu_device_id.h>
 #include <asm/intel-family.h>
 
+#include <xen/xen.h>
+
 static void intel_pmc_core_release(struct device *dev)
 {
 	kfree(dev);
@@ -53,6 +55,13 @@ static int __init pmc_core_platform_init(void)
 	if (acpi_dev_present("INT33A1", NULL, -1))
 		return -ENODEV;
 
+	/*
+	 * Skip forcefully attaching the device for VMs. Make an exception for
+	 * Xen dom0, which does have full hardware access.
+	 */
+	if (cpu_feature_enabled(X86_FEATURE_HYPERVISOR) && !xen_initial_domain())
+		return -ENODEV;
+
 	if (!x86_match_cpu(intel_pmc_core_platform_ids))
 		return -ENODEV;