Skip to contentSkip to conversation

arm64: hyperv: Support Virtual Trust Level Boot

PATCH Superseded Roman Kisel 10 participants 55 messages Mar 7, 2025 – Mar 13, 2025

Subsystems (11)
Device tree bindings x86 ACPI PCI +7 more subsystems

v5 of 9 Superseded Latest is v9 →

lore
Share & tools
lore
#0 lore raw

v5 cover arm64: hyperv: Support Virtual Trust Level Boot

This patch set allows the Hyper-V code to boot on ARM64 inside a Virtual Trust Level. These levels are a part of the Virtual Secure Mode documented in the Top-Level Functional Specification available at
https://learn.microsoft.com/en-us/virtualization/hyper-v-on-windows/tlfs/vsm.

The OpenHCL paravisor https://github.com/microsoft/openvmm/tree/main/openhcl
can serve as a practical application of these patches on ARM64.

For validation, I built kernels for the {x86_64, ARM64} x {VTL0, VTL2} set with a small initrd embedded into the kernel and booted VMs managed by Hyper-V and OpenVMM off of that.

Starting from V5, the patch series includes a non-functional change to KVM on arm64 which I tested as well.

[V5]
- Provide and use a common SMCCC-based infra for the arm64 hypervisor guests to detect hypervisor presence.
** Thank you, Arnd! **

- Fix line wraps to follow the rest of the code.
- Open-code getting IRQ domain parent in the ACPI case to make the code better.
** Thank you, Bjorn! **

- Test the binding with the latest dtschema.
- Clean up the commit title and description.
- Use proper defines for known constants.
** Thank you, Krzysztof! **

- Extend comment on why ACPI v6 is checked for.
- Reorder patches to make sure that even with partial series application the compilation succeeds.
- Report VTL the kernel runs in.
- Use "X86_64" in Kconfig rather than "X86".
- Extract a non-functional change for hv_get_vmbus_root_device() into a separate patch.
** Thank you, Michael! **

[V4]
https://lore.kernel.org/linux-hyperv/[email protected]/
- Fixed wording to match acronyms defined in the "Terms and Abbreviations" section of the SMCCC specification throughout the patch series. **Thank you, Michael!**

- Replaced the hypervisor ID containing ASCII with an UUID as required by the specification.
**Thank you, Michael!**

- Added an explicit check for SMCCC_RET_NOT_SUPPORTED when discovering the hypervisor presence to make the backward compatibility obvious. **Thank you, Saurabh!**

- Split the fix for get_vtl(void) out to make it easier to backport.
- Refactored the configuration options as requested to eliminate the risk of building non-functional kernels with randomly selected options. **Thank you, Michael!**

- Refactored the changes not to introduce an additional file with a one-line function.
**Thank you, Wei!**

- Fixed change description for the VMBus DeviceTree changes, used scripts/get_maintainers.pl on the latest kernel to get the up-to-date list of maintainers as requested.
**Thank you, Krzysztof!**

- Removed the added (paranoidal+superfluous) checks for DMA coherence in the VMBus driver and instead relied on the DMA and the OF subsystem code. **Thank you, Arnd, Krzysztof, Michael!**

- Used another set of APIs for discovering the hardware interrupt number in the VMBus driver to be able to build the driver as a module. **Thank you, Michael, Saurabh!**

- Renamed the newly introduced get_vmbus_root_device(void) function to hv_get_vmbus_root_device(void) as requested.
**Thank you, Wei!**

- Applied the suggested small-scale refactoring to simplify changes to the Hyper-V PCI driver. Taking the offered liberty of doing the large scale refactoring in another patch series.
**Thank you, Michael!**

- Added a fix for the issue discovered internally where the CPU would not get the interrupt from a PCI device attached to VTL2 as the shared peripheral interrupt number (SPI) was not offset by 32 (the first valid SPI number). **Thank you, Brian!**

[V3]
https://lore.kernel.org/lkml/[email protected]/
- Employed the SMCCC function recently implemented in the Microsoft Hyper-V hypervisor to detect running on Hyper-V/arm64. No dependence on ACPI/DT is needed anymore although the source code still falls back to ACPI as the new hypervisor might be available only in the Windows Insiders channel just
yet.
- As a part of the above, refactored detecting the hypervisor via ACPI FADT.
- There was a suggestion to explore whether it is feasible or not to express
that ACPI must be absent for the VTL mode and present for the regular guests
in the Hyper-V Kconfig file.
My current conclusion is that this will require refactoring in many places. That becomes especially convoluted on x86_64 due to the MSI and APIC dependencies. I'd ask to let us tackle that in another patch series (or chalk up to nice-have's rather than fires to put out) to separate concerns and decrease chances of breakage.
- While refactoring get_vtl(void) and the related code, fixed the hypercall output address not to overlap with the input as the Hyper-V TLFS mandates:
"The input and output parameter lists cannot overlap or cross page boundaries." See https://learn.microsoft.com/en-us/virtualization/hyper-v-on-windows/tlfs/hypercall-interface
for more.
Some might argue that should've been a topic for a separate patch series;
I'd counter that the change is well-contained (one line), has no dependencies, and makes the code legal.
- Made the VTL boot code (c)leaner as was suggested.
- Set DMA cache coherency for the VMBus.
- Updated DT bindings in the VMBus documentation (separated out into a new patch).
- Fixed vmbus_set_irq to use the API that works both for the ACPI and OF.
- Reworked setting up the vPCI MSI IRQ domain in the non-ACPI case. The logic looks a bit fiddly/ad-hoc as I couldn't find the API that would fit the bill. Added comments to explain myself.

[V2]
https://lore.kernel.org/all/[email protected]/
- Decreased number of #ifdef's
- Updated the wording in the commit messages to adhere to the guidlines
- Sending to the correct set of maintainers and mail lists

[V1]

    https://lore.kernel.org/all/[email protected]/

Roman Kisel (11):
arm64: kvm, smccc: Introduce and use API for detectting hypervisor
presence
arm64: hyperv: Use SMCCC to detect hypervisor presence
Drivers: hv: Enable VTL mode for arm64
Drivers: hv: Provide arch-neutral implementation of get_vtl()
arm64: hyperv: Initialize the Virtual Trust Level field arm64, x86: hyperv: Report the VTL the system boots in
dt-bindings: microsoft,vmbus: Add interrupts and DMA coherence
Drivers: hv: vmbus: Get the IRQ number from DeviceTree
Drivers: hv: vmbus: Introduce hv_get_vmbus_root_device()
ACPI: irq: Introduce acpi_get_gsi_dispatcher()
PCI: hv: Get vPCI MSI IRQ domain from DeviceTree

 .../bindings/bus/microsoft,vmbus.yaml         |  8 +-
 arch/arm64/hyperv/mshyperv.c                  | 46 +++++++++--
 arch/arm64/kvm/hypercalls.c                   |  5 +-
 arch/x86/hyperv/hv_init.c                     | 34 --------
 arch/x86/hyperv/hv_vtl.c                      |  2 +-
 drivers/acpi/irq.c                            | 14 +++-
 drivers/firmware/smccc/kvm_guest.c            | 10 +--
 drivers/firmware/smccc/smccc.c                | 19 +++++
 drivers/hv/Kconfig                            | 10 ++-
 drivers/hv/hv_common.c                        | 31 ++++++++
 drivers/hv/vmbus_drv.c                        | 59 ++++++++++++--
 drivers/pci/controller/pci-hyperv.c           | 79 +++++++++++++++++--
 include/asm-generic/mshyperv.h                |  6 ++
 include/hyperv/hvgdk_mini.h                   |  2 +-
 include/linux/acpi.h                          |  5 +-
 include/linux/arm-smccc.h                     | 55 ++++++++++++-
 include/linux/hyperv.h                        |  2 +
 17 files changed, 308 insertions(+), 79 deletions(-)

base-commit: 3a7f7785eae7cf012af128ca9e383c91e4955354

Signature · 1 line
2.43.0
#1 lore raw

PATCH v5 1/11 arm64: kvm, smccc: Introduce and use API for detectting hypervisor presence

The KVM/arm64 uses SMCCC to detect hypervisor presence. That code is private, and it follows the SMCCC specification. Other existing and emerging hypervisor guest implementations can and should use that standard approach as well.

Factor out a common infrastructure that the guests can use, update KVM to employ the new API. The cenrtal notion of the SMCCC method is the UUID of the hypervisor, and the API follows that.

No functional changes. Validated with a KVM/arm64 guest.

---
 arch/arm64/kvm/hypercalls.c        |  5 +--
 drivers/firmware/smccc/kvm_guest.c | 10 ++----
 drivers/firmware/smccc/smccc.c     | 19 +++++++++++
 include/linux/arm-smccc.h          | 55 +++++++++++++++++++++++++++---
 4 files changed, 73 insertions(+), 16 deletions(-)
arch/arm64/kvm/hypercalls.c +1 -4
index 27ce4cb44904..92b9bc1ea8e8 100644
@@ -353,10 +353,7 @@ int kvm_smccc_call_handler(struct kvm_vcpu *vcpu)
 			val[0] = gpa;
 		break;
 	case ARM_SMCCC_VENDOR_HYP_CALL_UID_FUNC_ID:
-		val[0] = ARM_SMCCC_VENDOR_HYP_UID_KVM_REG_0;
-		val[1] = ARM_SMCCC_VENDOR_HYP_UID_KVM_REG_1;
-		val[2] = ARM_SMCCC_VENDOR_HYP_UID_KVM_REG_2;
-		val[3] = ARM_SMCCC_VENDOR_HYP_UID_KVM_REG_3;
+		UUID_TO_SMCCC_RES(ARM_SMCCC_VENDOR_HYP_UID_KVM, val);
 		break;
 	case ARM_SMCCC_VENDOR_HYP_KVM_FEATURES_FUNC_ID:
 		val[0] = smccc_feat->vendor_hyp_bmap;
drivers/firmware/smccc/kvm_guest.c +2 -8
index f3319be20b36..b5084b309ea0 100644
@@ -14,17 +14,11 @@ static DECLARE_BITMAP(__kvm_arm_hyp_services, ARM_SMCCC_KVM_NUM_FUNCS) __ro_afte
 
 void __init kvm_init_hyp_services(void)
 {
+	uuid_t kvm_uuid = ARM_SMCCC_VENDOR_HYP_UID_KVM;
 	struct arm_smccc_res res;
 	u32 val[4];
 
-	if (arm_smccc_1_1_get_conduit() != SMCCC_CONDUIT_HVC)
-		return;
-
-	arm_smccc_1_1_invoke(ARM_SMCCC_VENDOR_HYP_CALL_UID_FUNC_ID, &res);
-	if (res.a0 != ARM_SMCCC_VENDOR_HYP_UID_KVM_REG_0 ||
-	    res.a1 != ARM_SMCCC_VENDOR_HYP_UID_KVM_REG_1 ||
-	    res.a2 != ARM_SMCCC_VENDOR_HYP_UID_KVM_REG_2 ||
-	    res.a3 != ARM_SMCCC_VENDOR_HYP_UID_KVM_REG_3)
+	if (!arm_smccc_hyp_present(&kvm_uuid))
 		return;
 
 	memset(&res, 0, sizeof(res));
drivers/firmware/smccc/smccc.c +19 -0
index a74600d9f2d7..7399f27d58e5 100644
@@ -67,6 +67,25 @@ s32 arm_smccc_get_soc_id_revision(void)
 }
 EXPORT_SYMBOL_GPL(arm_smccc_get_soc_id_revision);
 
+bool arm_smccc_hyp_present(const uuid_t *hyp_uuid)
+{
+	struct arm_smccc_res res = {};
+
+	if (arm_smccc_1_1_get_conduit() != SMCCC_CONDUIT_HVC)
+		return false;
+	arm_smccc_1_1_hvc(ARM_SMCCC_VENDOR_HYP_CALL_UID_FUNC_ID, &res);
+	if (res.a0 == SMCCC_RET_NOT_SUPPORTED)
+		return false;
+
+	return ({
+		const uuid_t uuid = SMCCC_RES_TO_UUID(res.a0, res.a1, res.a2, res.a3);
+		const bool present = uuid_equal(&uuid, hyp_uuid);
+
+		present;
+	});
+}
+EXPORT_SYMBOL_GPL(arm_smccc_hyp_present);
+
 static int __init smccc_devices_init(void)
 {
 	struct platform_device *pdev;
include/linux/arm-smccc.h +51 -4

73 lines not on this page · raw message

Signature · 1 line
2.43.0
#7 lore raw

PATCH v5 2/11 arm64: hyperv: Use SMCCC to detect hypervisor presence

In reply to Roman Kisel

The arm64 Hyper-V startup path relies on ACPI to detect running under a Hyper-V compatible hypervisor. That doesn't work on non-ACPI systems.

Hoist the ACPI detection logic into a separate function. Then use the vendor-specific hypervisor service call (implemented recently in Hyper-V) via SMCCC in the non-ACPI case.

---
 arch/arm64/hyperv/mshyperv.c | 43 +++++++++++++++++++++++++++++++-----
 1 file changed, 38 insertions(+), 5 deletions(-)
arch/arm64/hyperv/mshyperv.c +38 -5
index 29fcfd595f48..c647db56fd6b 100644
@@ -27,6 +27,41 @@ int hv_get_hypervisor_version(union hv_hypervisor_version_info *info)
 	return 0;
 }
 
+static bool __init hyperv_detect_via_acpi(void)
+{
+	if (acpi_disabled)
+		return false;
+#if IS_ENABLED(CONFIG_ACPI)
+	/*
+	 * Hypervisor ID is only available in ACPI v6+, and the
+	 * structure layout was extended in v6 to accommodate that
+	 * new field.
+	 *
+	 * At the very minimum, this check makes sure not to read
+	 * past the FADT structure.
+	 *
+	 * It is also needed to catch running in some unknown
+	 * non-Hyper-V environment that has ACPI 5.x or less.
+	 * In such a case, it can't be Hyper-V.
+	 */
+	if (acpi_gbl_FADT.header.revision < 6)
+		return false;
+	return strncmp((char *)&acpi_gbl_FADT.hypervisor_id, "MsHyperV", 8) == 0;
+#else
+	return false;
+#endif
+}
+
+static bool __init hyperv_detect_via_smccc(void)
+{
+	uuid_t hyperv_uuid = UUID_INIT(
+		0x4d32ba58, 0x4764, 0xcd24,
+		0x75, 0x6c, 0xef, 0x8e,
+		0x24, 0x70, 0x59, 0x16);
+
+	return arm_smccc_hyp_present(&hyperv_uuid);
+}
+
 static int __init hyperv_init(void)
 {
 	struct hv_get_vp_registers_output	result;
@@ -35,13 +70,11 @@ static int __init hyperv_init(void)
 
 	/*
 	 * Allow for a kernel built with CONFIG_HYPERV to be running in
-	 * a non-Hyper-V environment, including on DT instead of ACPI.
+	 * a non-Hyper-V environment.
+	 *
 	 * In such cases, do nothing and return success.
 	 */
-	if (acpi_disabled)
-		return 0;
-
-	if (strncmp((char *)&acpi_gbl_FADT.hypervisor_id, "MsHyperV", 8))
+	if (!hyperv_detect_via_acpi() && !hyperv_detect_via_smccc())
 		return 0;
 
 	/* Setup the guest ID */
Signature · 1 line
2.43.0
#9 lore raw

PATCH v5 3/11 Drivers: hv: Enable VTL mode for arm64

In reply to Roman Kisel

Kconfig dependencies for arm64 guests on Hyper-V require that be ACPI enabled, and limit VTL mode to x86/x64. To enable VTL mode on arm64 as well, update the dependencies. Since VTL mode requires DeviceTree instead of ACPI, don’t require arm64 guests on Hyper-V to have ACPI unconditionally.

---
 drivers/hv/Kconfig | 10 ++++++----
 1 file changed, 6 insertions(+), 4 deletions(-)
drivers/hv/Kconfig +6 -4
index 862c47b191af..c37b1a44e580 100644
@@ -5,18 +5,20 @@ menu "Microsoft Hyper-V guest support"
 config HYPERV
 	tristate "Microsoft Hyper-V client drivers"
 	depends on (X86 && X86_LOCAL_APIC && HYPERVISOR_GUEST) \
-		|| (ACPI && ARM64 && !CPU_BIG_ENDIAN)
+		|| (ARM64 && !CPU_BIG_ENDIAN)
+	depends on (ACPI || HYPERV_VTL_MODE)
 	select PARAVIRT
 	select X86_HV_CALLBACK_VECTOR if X86
-	select OF_EARLY_FLATTREE if OF
 	help
 	  Select this option to run Linux as a Hyper-V client operating
 	  system.
 
 config HYPERV_VTL_MODE
 	bool "Enable Linux to boot in VTL context"
-	depends on X86_64 && HYPERV
+	depends on (X86_64 || ARM64)
 	depends on SMP
+	select OF_EARLY_FLATTREE
+	select OF
 	default n
 	help
 	  Virtual Secure Mode (VSM) is a set of hypervisor capabilities and
@@ -31,7 +33,7 @@ config HYPERV_VTL_MODE
 
 	  Select this option to build a Linux kernel to run at a VTL other than
 	  the normal VTL0, which currently is only VTL2.  This option
-	  initializes the x86 platform for VTL2, and adds the ability to boot
+	  initializes the kernel to run in VTL2, and adds the ability to boot
 	  secondary CPUs directly into 64-bit context as required for VTLs other
 	  than 0.  A kernel built with this option must run at VTL2, and will
 	  not run as a normal guest.
Signature · 1 line
2.43.0
#21 lore raw

PATCH v5 4/11 Drivers: hv: Provide arch-neutral implementation of get_vtl()

In reply to Roman Kisel

To run in the VTL mode, Hyper-V drivers have to know what VTL the system boots in, and the arm64/hyperv code does not have the means to compute that.

Refactor the code to hoist the function that detects VTL, make it arch-neutral to be able to employ it to get the VTL on arm64.

---
 arch/x86/hyperv/hv_init.c      | 34 ----------------------------------
 drivers/hv/hv_common.c         | 31 +++++++++++++++++++++++++++++++
 include/asm-generic/mshyperv.h |  6 ++++++
 include/hyperv/hvgdk_mini.h    |  2 +-
 4 files changed, 38 insertions(+), 35 deletions(-)
arch/x86/hyperv/hv_init.c +0 -34
index 9be1446f5bd3..e204cb9d41af 100644
@@ -390,40 +390,6 @@ static void __init hv_stimer_setup_percpu_clockev(void)
 		old_setup_percpu_clockev();
 }
 
-#if IS_ENABLED(CONFIG_HYPERV_VTL_MODE)
-static u8 __init get_vtl(void)
-{
-	u64 control = HV_HYPERCALL_REP_COMP_1 | HVCALL_GET_VP_REGISTERS;
-	struct hv_input_get_vp_registers *input;
-	struct hv_output_get_vp_registers *output;
-	unsigned long flags;
-	u64 ret;
-
-	local_irq_save(flags);
-	input = *this_cpu_ptr(hyperv_pcpu_input_arg);
-	output = *this_cpu_ptr(hyperv_pcpu_output_arg);
-
-	memset(input, 0, struct_size(input, names, 1));
-	input->partition_id = HV_PARTITION_ID_SELF;
-	input->vp_index = HV_VP_INDEX_SELF;
-	input->input_vtl.as_uint8 = 0;
-	input->names[0] = HV_REGISTER_VSM_VP_STATUS;
-
-	ret = hv_do_hypercall(control, input, output);
-	if (hv_result_success(ret)) {
-		ret = output->values[0].reg8 & HV_X64_VTL_MASK;
-	} else {
-		pr_err("Failed to get VTL(error: %lld) exiting...\n", ret);
-		BUG();
-	}
-
-	local_irq_restore(flags);
-	return ret;
-}
-#else
-static inline u8 get_vtl(void) { return 0; }
-#endif
-
 /*
  * This function is to be invoked early in the boot sequence after the
  * hypervisor has been detected.
drivers/hv/hv_common.c +31 -0
index ee3083937b4f..10983e854a4d 100644
@@ -304,6 +304,37 @@ void __init hv_get_partition_id(void)
 		pr_err("Hyper-V: failed to get partition ID: %#x\n",
 		       hv_result(status));
 }
+#if IS_ENABLED(CONFIG_HYPERV_VTL_MODE)
+u8 __init get_vtl(void)
+{
+	u64 control = HV_HYPERCALL_REP_COMP_1 | HVCALL_GET_VP_REGISTERS;
+	struct hv_input_get_vp_registers *input;
+	struct hv_output_get_vp_registers *output;
+	unsigned long flags;
+	u64 ret;
+
+	local_irq_save(flags);
+	input = *this_cpu_ptr(hyperv_pcpu_input_arg);
+	output = *this_cpu_ptr(hyperv_pcpu_output_arg);
+
+	memset(input, 0, struct_size(input, names, 1));
+	input->partition_id = HV_PARTITION_ID_SELF;
+	input->vp_index = HV_VP_INDEX_SELF;
+	input->input_vtl.as_uint8 = 0;
+	input->names[0] = HV_REGISTER_VSM_VP_STATUS;
+
+	ret = hv_do_hypercall(control, input, output);
+	if (hv_result_success(ret)) {
+		ret = output->values[0].reg8 & HV_VTL_MASK;
+	} else {
+		pr_err("Failed to get VTL(error: %lld) exiting...\n", ret);
+		BUG();
+	}
+
+	local_irq_restore(flags);
+	return ret;
+}
+#endif
 
 int __init hv_common_init(void)
 {
include/asm-generic/mshyperv.h +6 -0
index 7adc10a4fa3e..bf897c528976 100644
@@ -320,4 +320,10 @@ static inline enum hv_isolation_type hv_get_isolation_type(void)
 }
 #endif /* CONFIG_HYPERV */
 
+#if IS_ENABLED(CONFIG_HYPERV_VTL_MODE)
+u8 __init get_vtl(void);
+#else
+static inline u8 get_vtl(void) { return 0; }
+#endif
+
 #endif
include/hyperv/hvgdk_mini.h +1 -1
index 58895883f636..98aa64e8844c 100644
@@ -1202,7 +1202,7 @@ struct hv_send_ipi {	 /* HV_INPUT_SEND_SYNTHETIC_CLUSTER_IPI */
 	u64 cpu_mask;
 } __packed;
 
-#define	HV_X64_VTL_MASK			GENMASK(3, 0)
+#define	HV_VTL_MASK			GENMASK(3, 0)
 
 /* Hyper-V memory host visibility */
 enum hv_mem_host_visibility {
Signature · 1 line
2.43.0
#23 lore raw

PATCH v5 5/11 arm64: hyperv: Initialize the Virtual Trust Level field

In reply to Roman Kisel

Various parts of the hyperv code need to know what VTL the kernel runs at, most notably VMBus needs that to establish communication with the host.

Initialize the Virtual Trust Level field to enable booting in the Virtual Trust Level.

---
 arch/arm64/hyperv/mshyperv.c | 1 +
 1 file changed, 1 insertion(+)
arch/arm64/hyperv/mshyperv.c +1 -0
index c647db56fd6b..a7db03f5413d 100644
@@ -107,6 +107,7 @@ static int __init hyperv_init(void)
 
 	if (ms_hyperv.priv_high & HV_ACCESS_PARTITION_ID)
 		hv_get_partition_id();
+	ms_hyperv.vtl = get_vtl();
 
 	ms_hyperv_late_init();
 
Signature · 1 line
2.43.0
#25 lore raw

PATCH v5 6/11 arm64, x86: hyperv: Report the VTL the system boots in

In reply to Roman Kisel

The hyperv guest code might run in various Virtual Trust Levels.

Report the level when the kernel boots in the non-default (0) one.

---
 arch/arm64/hyperv/mshyperv.c | 2 ++
 arch/x86/hyperv/hv_vtl.c     | 2 +-
 2 files changed, 3 insertions(+), 1 deletion(-)
arch/arm64/hyperv/mshyperv.c +2 -0
index a7db03f5413d..3bc16dbee758 100644
@@ -108,6 +108,8 @@ static int __init hyperv_init(void)
 	if (ms_hyperv.priv_high & HV_ACCESS_PARTITION_ID)
 		hv_get_partition_id();
 	ms_hyperv.vtl = get_vtl();
+	if (ms_hyperv.vtl > 0) /* non default VTL */
+		pr_info("Linux runs in Hyper-V Virtual Trust Level %d\n", ms_hyperv.vtl);
 
 	ms_hyperv_late_init();
 
arch/x86/hyperv/hv_vtl.c +1 -1
index 4e1b1e3b5658..c21bee7e8ff3 100644
@@ -24,7 +24,7 @@ static bool __init hv_vtl_msi_ext_dest_id(void)
 
 void __init hv_vtl_init_platform(void)
 {
-	pr_info("Linux runs in Hyper-V Virtual Trust Level\n");
+	pr_info("Linux runs in Hyper-V Virtual Trust Level %d\n", ms_hyperv.vtl);
 
 	x86_platform.realmode_reserve = x86_init_noop;
 	x86_platform.realmode_init = x86_init_noop;
Signature · 1 line
2.43.0
#30 lore raw

PATCH v5 7/11 dt-bindings: microsoft,vmbus: Add interrupts and DMA coherence

In reply to Roman Kisel

To boot on ARM64, VMBus requires configuring interrupts. Missing DMA coherence property is sub-optimal as the VMBus transations are cache-coherent.

Add interrupts to be able to boot on ARM64. Add DMA coherence to avoid doing extra work on maintaining caches on ARM64.

---
 .../devicetree/bindings/bus/microsoft,vmbus.yaml          | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)
Documentation/devicetree/bindings/bus/microsoft,vmbus.yaml +7 -1
index a8d40c766dcd..3ab7d0116626 100644
@@ -28,13 +28,16 @@ properties:
 required:
   - compatible
   - ranges
+  - interrupts
   - '#address-cells'
   - '#size-cells'
 
-additionalProperties: false
+additionalProperties: true
 
 examples:
   - |
+    #include <dt-bindings/interrupt-controller/irq.h>
+    #include <dt-bindings/interrupt-controller/arm-gic.h>
     soc {
         #address-cells = <2>;
         #size-cells = <1>;
@@ -49,6 +52,9 @@ examples:
                 #address-cells = <2>;
                 #size-cells = <1>;
                 ranges = <0x0f 0xf0000000 0x0f 0xf0000000 0x10000000>;
+                dma-coherent;
+                interrupt-parent = <&gic>;
+                interrupts = <GIC_PPI 2 IRQ_TYPE_EDGE_RISING>;
             };
         };
     };
Signature · 1 line
2.43.0
#37 lore raw

PATCH v5 8/11 Drivers: hv: vmbus: Get the IRQ number from DeviceTree

In reply to Roman Kisel

The VMBus driver uses ACPI for interrupt assignment on arm64 hence it won't function in the VTL mode where only DeviceTree can be used.

Update the VMBus driver to discover interrupt configuration from DT.

---
 drivers/hv/vmbus_drv.c | 36 ++++++++++++++++++++++++++++++++++++
 1 file changed, 36 insertions(+)
drivers/hv/vmbus_drv.c +36 -0
index 75eb1390b45c..c8474b48dcd2 100644
@@ -2345,6 +2345,36 @@ static int vmbus_acpi_add(struct platform_device *pdev)
 }
 #endif
 
+static int __maybe_unused vmbus_set_irq(struct platform_device *pdev)
+{
+	struct irq_data *data;
+	int irq;
+	irq_hw_number_t hwirq;
+
+	irq = platform_get_irq(pdev, 0);
+	if (irq == 0) {
+		pr_err("VMBus interrupt mapping failure\n");
+		return -EINVAL;
+	}
+	if (irq < 0) {
+		pr_err("VMBus interrupt data can't be read from DeviceTree, error %d\n", irq);
+		return irq;
+	}
+
+	data = irq_get_irq_data(irq);
+	if (!data) {
+		pr_err("No interrupt data for VMBus virq %d\n", irq);
+		return -ENODEV;
+	}
+	hwirq = irqd_to_hwirq(data);
+
+	vmbus_irq = irq;
+	vmbus_interrupt = hwirq;
+	pr_debug("VMBus virq %d, hwirq %d\n", vmbus_irq, vmbus_interrupt);
+
+	return 0;
+}
+
 static int vmbus_device_add(struct platform_device *pdev)
 {
 	struct resource **cur_res = &hyperv_mmio;
@@ -2359,6 +2389,12 @@ static int vmbus_device_add(struct platform_device *pdev)
 	if (ret)
 		return ret;
 
+#ifndef HYPERVISOR_CALLBACK_VECTOR
+	ret = vmbus_set_irq(pdev);
+	if (ret)
+		return ret;
+#endif
+
 	for_each_of_range(&parser, &range) {
 		struct resource *res;
 
Signature · 1 line
2.43.0
#45 lore raw

PATCH v5 9/11 Drivers: hv: vmbus: Introduce hv_get_vmbus_root_device()

In reply to Roman Kisel

The ARM64 PCI code for hyperv needs to know the VMBus root device, and it is private.

Provide a function that returns it. Rename it from "hv_dev" as "hv_dev" as a symbol is very overloaded. No functional changes.

---
 drivers/hv/vmbus_drv.c | 23 +++++++++++++++--------
 include/linux/hyperv.h |  2 ++
 2 files changed, 17 insertions(+), 8 deletions(-)
drivers/hv/vmbus_drv.c +15 -8
index c8474b48dcd2..7bfafe702963 100644
@@ -45,7 +45,8 @@ struct vmbus_dynid {
 	struct hv_vmbus_device_id id;
 };
 
-static struct device  *hv_dev;
+/* VMBus Root Device */
+static struct device  *vmbus_root_device;
 
 static int hyperv_cpuhp_online;
 
@@ -80,9 +81,15 @@ static struct resource *fb_mmio;
 static struct resource *hyperv_mmio;
 static DEFINE_MUTEX(hyperv_mmio_lock);
 
+struct device *hv_get_vmbus_root_device(void)
+{
+	return vmbus_root_device;
+}
+EXPORT_SYMBOL_GPL(hv_get_vmbus_root_device);
+
 static int vmbus_exists(void)
 {
-	if (hv_dev == NULL)
+	if (vmbus_root_device == NULL)
 		return -ENODEV;
 
 	return 0;
@@ -861,7 +868,7 @@ static int vmbus_dma_configure(struct device *child_device)
 	 * On x86/x64 coherence is assumed and these calls have no effect.
 	 */
 	hv_setup_dma_ops(child_device,
-		device_get_dma_attr(hv_dev) == DEV_DMA_COHERENT);
+		device_get_dma_attr(vmbus_root_device) == DEV_DMA_COHERENT);
 	return 0;
 }
 
@@ -1930,7 +1937,7 @@ int vmbus_device_register(struct hv_device *child_device_obj)
 		     &child_device_obj->channel->offermsg.offer.if_instance);
 
 	child_device_obj->device.bus = &hv_bus;
-	child_device_obj->device.parent = hv_dev;
+	child_device_obj->device.parent = vmbus_root_device;
 	child_device_obj->device.release = vmbus_device_release;
 
 	child_device_obj->device.dma_parms = &child_device_obj->dma_parms;
@@ -2292,7 +2299,7 @@ static int vmbus_acpi_add(struct platform_device *pdev)
 	struct acpi_device *ancestor;
 	struct acpi_device *device = ACPI_COMPANION(&pdev->dev);
 
-	hv_dev = &device->dev;
+	vmbus_root_device = &device->dev;
 
 	/*
 	 * Older versions of Hyper-V for ARM64 fail to include the _CCA
@@ -2383,7 +2390,7 @@ static int vmbus_device_add(struct platform_device *pdev)
 	struct device_node *np = pdev->dev.of_node;
 	int ret;
 
-	hv_dev = &pdev->dev;
+	vmbus_root_device = &pdev->dev;
 
 	ret = of_range_parser_init(&parser, np);
 	if (ret)
@@ -2702,7 +2709,7 @@ static int __init hv_acpi_init(void)
 	if (ret)
 		return ret;
 
-	if (!hv_dev) {
+	if (!vmbus_root_device) {
 		ret = -ENODEV;
 		goto cleanup;
 	}
@@ -2733,7 +2740,7 @@ static int __init hv_acpi_init(void)
 
 cleanup:
 	platform_driver_unregister(&vmbus_platform_driver);
-	hv_dev = NULL;
+	vmbus_root_device = NULL;
 	return ret;
 }
 
include/linux/hyperv.h +2 -0
index 7f4f8d8bdf43..1f0851fde041 100644
@@ -1333,6 +1333,8 @@ static inline void *hv_get_drvdata(struct hv_device *dev)
 	return dev_get_drvdata(&dev->device);
 }
 
+struct device *hv_get_vmbus_root_device(void);
+
 struct hv_ring_buffer_debug_info {
 	u32 current_interrupt_mask;
 	u32 current_read_index;
Signature · 1 line
2.43.0
#49 lore raw

PATCH v5 10/11 ACPI: irq: Introduce acpi_get_gsi_dispatcher()

In reply to Roman Kisel

Using acpi_irq_create_hierarchy() in the cases where the code also handles OF leads to code duplication as the ACPI subsystem doesn't provide means to compute the IRQ domain parent whereas the OF does.

Introduce acpi_get_gsi_dispatcher() so that the drivers relying on both ACPI and OF may use irq_domain_create_hierarchy() in the common code paths.

No functional changes.

---
 drivers/acpi/irq.c   | 14 ++++++++++++--
 include/linux/acpi.h |  5 ++++-
 2 files changed, 16 insertions(+), 3 deletions(-)
drivers/acpi/irq.c +12 -2
index 1687483ff319..6243db610137 100644
@@ -12,7 +12,7 @@
 
 enum acpi_irq_model_id acpi_irq_model;
 
-static struct fwnode_handle *(*acpi_get_gsi_domain_id)(u32 gsi);
+static acpi_gsi_domain_disp_fn acpi_get_gsi_domain_id;
 static u32 (*acpi_gsi_to_irq_fallback)(u32 gsi);
 
 /**
@@ -307,12 +307,22 @@ EXPORT_SYMBOL_GPL(acpi_irq_get);
  *	for a given GSI
  */
 void __init acpi_set_irq_model(enum acpi_irq_model_id model,
-			       struct fwnode_handle *(*fn)(u32))
+	acpi_gsi_domain_disp_fn fn)
 {
 	acpi_irq_model = model;
 	acpi_get_gsi_domain_id = fn;
 }
 
+/**
+ * acpi_get_gsi_dispatcher - Returns dispatcher function that
+ *                           computes the domain fwnode for a
+ *                           given GSI.
+ */
+acpi_gsi_domain_disp_fn acpi_get_gsi_dispatcher(void)
+{
+	return acpi_get_gsi_domain_id;
+}
+
 /**
  * acpi_set_gsi_to_irq_fallback - Register a GSI transfer
  * callback to fallback to arch specified implementation.
include/linux/acpi.h +4 -1
index 4e495b29c640..abc51288e867 100644
@@ -336,8 +336,11 @@ int acpi_register_gsi (struct device *dev, u32 gsi, int triggering, int polarity
 int acpi_gsi_to_irq (u32 gsi, unsigned int *irq);
 int acpi_isa_irq_to_gsi (unsigned isa_irq, u32 *gsi);
 
+typedef struct fwnode_handle *(*acpi_gsi_domain_disp_fn)(u32);
+
 void acpi_set_irq_model(enum acpi_irq_model_id model,
-			struct fwnode_handle *(*)(u32));
+	acpi_gsi_domain_disp_fn fn);
+acpi_gsi_domain_disp_fn acpi_get_gsi_dispatcher(void);
 void acpi_set_gsi_to_irq_fallback(u32 (*)(u32));
 
 struct irq_domain *acpi_irq_create_hierarchy(unsigned int flags,
Signature · 1 line
2.43.0
#51 lore raw

PATCH v5 11/11 PCI: hv: Get vPCI MSI IRQ domain from DeviceTree

In reply to Roman Kisel

The hyperv-pci driver uses ACPI for MSI IRQ domain configuration on arm64. It won't be able to do that in the VTL mode where only DeviceTree can be used.

Update the hyperv-pci driver to get vPCI MSI IRQ domain in the DeviceTree case, too.

---
 drivers/pci/controller/pci-hyperv.c | 79 ++++++++++++++++++++++++++---
 1 file changed, 73 insertions(+), 6 deletions(-)
drivers/pci/controller/pci-hyperv.c +73 -6
index 6084b38bdda1..9740006a8c73 100644
@@ -50,6 +50,7 @@
 #include <linux/irqdomain.h>
 #include <linux/acpi.h>
 #include <linux/sizes.h>
+#include <linux/of_irq.h>
 #include <asm/mshyperv.h>
 
 /*
@@ -817,9 +818,17 @@ static int hv_pci_vec_irq_gic_domain_alloc(struct irq_domain *domain,
 	int ret;
 
 	fwspec.fwnode = domain->parent->fwnode;
-	fwspec.param_count = 2;
-	fwspec.param[0] = hwirq;
-	fwspec.param[1] = IRQ_TYPE_EDGE_RISING;
+	if (is_of_node(fwspec.fwnode)) {
+		/* SPI lines for OF translations start at offset 32 */
+		fwspec.param_count = 3;
+		fwspec.param[0] = 0;
+		fwspec.param[1] = hwirq - 32;
+		fwspec.param[2] = IRQ_TYPE_EDGE_RISING;
+	} else {
+		fwspec.param_count = 2;
+		fwspec.param[0] = hwirq;
+		fwspec.param[1] = IRQ_TYPE_EDGE_RISING;
+	}
 
 	ret = irq_domain_alloc_irqs_parent(domain, virq, 1, &fwspec);
 	if (ret)
@@ -887,10 +896,53 @@ static const struct irq_domain_ops hv_pci_domain_ops = {
 	.activate = hv_pci_vec_irq_domain_activate,
 };
 
+#ifdef CONFIG_OF
+
+static struct irq_domain *hv_pci_of_irq_domain_parent(void)
+{
+	struct device_node *parent;
+	struct irq_domain *domain;
+
+	parent = of_irq_find_parent(hv_get_vmbus_root_device()->of_node);
+	domain = NULL;
+	if (parent) {
+		domain = irq_find_host(parent);
+		of_node_put(parent);
+	}
+
+	return domain;
+}
+
+#endif
+
+#ifdef CONFIG_ACPI
+
+static struct irq_domain *hv_pci_acpi_irq_domain_parent(void)
+{
+	struct irq_domain *domain;
+	acpi_gsi_domain_disp_fn gsi_domain_disp_fn;
+
+	if (acpi_irq_model != ACPI_IRQ_MODEL_GIC)
+		return NULL;
+	gsi_domain_disp_fn = acpi_get_gsi_dispatcher();
+	if (!gsi_domain_disp_fn)
+		return NULL;
+	domain = irq_find_matching_fwnode(gsi_domain_disp_fn(0),
+				     DOMAIN_BUS_ANY);
+
+	if (!domain)
+		return NULL;
+
+	return domain;
+}
+
+#endif
+
 static int hv_pci_irqchip_init(void)
 {
 	static struct hv_pci_chip_data *chip_data;
 	struct fwnode_handle *fn = NULL;
+	struct irq_domain *irq_domain_parent = NULL;
 	int ret = -ENOMEM;
 
 	chip_data = kzalloc(sizeof(*chip_data), GFP_KERNEL);
@@ -907,9 +959,24 @@ static int hv_pci_irqchip_init(void)
 	 * way to ensure that all the corresponding devices are also gone and
 	 * no interrupts will be generated.
 	 */
-	hv_msi_gic_irq_domain = acpi_irq_create_hierarchy(0, HV_PCI_MSI_SPI_NR,
-							  fn, &hv_pci_domain_ops,
-							  chip_data);
+#ifdef CONFIG_ACPI
+	if (!acpi_disabled)
+		irq_domain_parent = hv_pci_acpi_irq_domain_parent();
+#endif
+#if defined(CONFIG_OF)
+	if (!irq_domain_parent)
+		irq_domain_parent = hv_pci_of_irq_domain_parent();
+#endif
+	if (!irq_domain_parent) {
+		WARN_ONCE(1, "Invalid firmware configuration for VMBus interrupts\n");
+		ret = -EINVAL;
+		goto free_chip;
+	}
+
+	hv_msi_gic_irq_domain = irq_domain_create_hierarchy(
+		irq_domain_parent, 0, HV_PCI_MSI_SPI_NR,
+		fn, &hv_pci_domain_ops,
+		chip_data);
 
 	if (!hv_msi_gic_irq_domain) {
 		pr_err("Failed to create Hyper-V arm64 vPCI MSI IRQ domain\n");
Signature · 1 line
2.43.0
maintainer
#10 lore raw

Re: [PATCH hyperv-next v5 03/11] Drivers: hv: Enable VTL mode for arm64

In reply to Roman Kisel
Quoted 13 lines · Roman Kisel
drivers/hv/Kconfig +2 -2
@@ -5,18 +5,20 @@ menu "Microsoft Hyper-V guest support"
 config HYPERV
 	tristate "Microsoft Hyper-V client drivers"
 	depends on (X86 && X86_LOCAL_APIC && HYPERVISOR_GUEST) \
-		|| (ACPI && ARM64 && !CPU_BIG_ENDIAN)
+		|| (ARM64 && !CPU_BIG_ENDIAN)
+	depends on (ACPI || HYPERV_VTL_MODE)
 	select PARAVIRT
 	select X86_HV_CALLBACK_VECTOR if X86
-	select OF_EARLY_FLATTREE if OF
 	help
 	  Select this option to run Linux as a Hyper-V client operating
 	  system.

drivers/hv/Kconfig menu "Microsoft Hyper-V guest support" · line 23

Quoted 10 lines
drivers/hv/Kconfig +3 -1
 
 config HYPERV_VTL_MODE
 	bool "Enable Linux to boot in VTL context"
-	depends on X86_64 && HYPERV
+	depends on (X86_64 || ARM64)
 	depends on SMP
+	select OF_EARLY_FLATTREE
+	select OF
 	default n
 	help

Having the dependency below the top-level Kconfig entry feels a little counterintuitive. You could flip that back as it was before by doing

select HYPERV_VTL_MODE if !ACPI
depends on ACPI || SMP

in the HYPERV option, leaving the dependency on HYPERV in HYPERV_VTL_MODE.

Is OF_EARLY_FLATTREE actually needed on x86?

Arnd

maintainer Acked-by on v5 1/11
#2 lore raw

Re: [PATCH hyperv-next v5 01/11] arm64: kvm, smccc: Introduce and use API for detectting hypervisor presence

In reply to Roman Kisel
Quoted 5 lines · Roman Kisel

The KVM/arm64 uses SMCCC to detect hypervisor presence. That code is private, and it follows the SMCCC specification. Other existing and emerging hypervisor guest implementations can and should use that standard approach as well.

Quoted 7 lines

Factor out a common infrastructure that the guests can use, update KVM to employ the new API. The cenrtal notion of the SMCCC method is the UUID of the hypervisor, and the API follows that.

No functional changes. Validated with a KVM/arm64 guest.

maintainer
#38 lore raw

Re: [PATCH hyperv-next v5 08/11] Drivers: hv: vmbus: Get the IRQ number from DeviceTree

In reply to Roman Kisel
Quoted 2 lines · Roman Kisel

+static int __maybe_unused vmbus_set_irq(struct platform_device *pdev)

Instead of the __maybe_unused annotation here

drivers/hv/vmbus_drv.c static int vmbus_device_add(struct platform_device *pdev) · line 2397

Quoted 7 lines
drivers/hv/vmbus_drv.c +6 -0
+#ifndef HYPERVISOR_CALLBACK_VECTOR
+	ret = vmbus_set_irq(pdev);
+	if (ret)
+		return ret;
+#endif
+

you can use

       if (!__is_defined(HYPERVISOR_CALLBACK_VECTOR))
                  ret = vmbus_set_irq(pdev);

and make it a little more readable.

Arnd

maintainer
#26 lore raw

Re: [PATCH hyperv-next v5 06/11] arm64, x86: hyperv: Report the VTL the system boots in

In reply to Roman Kisel
Quoted 16 lines · Roman Kisel

The hyperv guest code might run in various Virtual Trust Levels.

Report the level when the kernel boots in the non-default (0) one.

---
 arch/arm64/hyperv/mshyperv.c | 2 ++
 arch/x86/hyperv/hv_vtl.c     | 2 +-
 2 files changed, 3 insertions(+), 1 deletion(-)
arch/arm64/hyperv/mshyperv.c +0 -0
index a7db03f5413d..3bc16dbee758 100644

arch/arm64/hyperv/mshyperv.c static int __init hyperv_init(void) · line 111

Quoted 4 lines
arch/arm64/hyperv/mshyperv.c +1 -0
@@ -108,6 +108,8 @@ static int __init hyperv_init(void)
 	if (ms_hyperv.priv_high & HV_ACCESS_PARTITION_ID)
 		hv_get_partition_id();
 	ms_hyperv.vtl = get_vtl();
+	if (ms_hyperv.vtl > 0) /* non default VTL */

"non-default".

Quoted 9 lines
arch/arm64/hyperv/mshyperv.c +1 -0
+		pr_info("Linux runs in Hyper-V Virtual Trust Level %d\n", ms_hyperv.vtl);
 
 	ms_hyperv_late_init();
 
arch/x86/hyperv/hv_vtl.c +0 -0
index 4e1b1e3b5658..c21bee7e8ff3 100644

arch/x86/hyperv/hv_vtl.c static bool __init hv_vtl_msi_ext_dest_id(void) · line 27

Quoted 5 lines
arch/x86/hyperv/hv_vtl.c +1 -1
@@ -24,7 +24,7 @@ static bool __init hv_vtl_msi_ext_dest_id(void)
 
 void __init hv_vtl_init_platform(void)
 {
-	pr_info("Linux runs in Hyper-V Virtual Trust Level\n");
+	pr_info("Linux runs in Hyper-V Virtual Trust Level %d\n", ms_hyperv.vtl);

Where isn't there a check for ms_hyperv.vtl > 0 here?

Please be consistent across different architectures.

Quoted 7 lines
arch/x86/hyperv/hv_vtl.c +0 -0
 	x86_platform.realmode_reserve = x86_init_noop;
 	x86_platform.realmode_init = x86_init_noop;
Signature · 1 line
2.43.0
#31 lore raw

Re: [PATCH hyperv-next v5 07/11] dt-bindings: microsoft,vmbus: Add interrupts and DMA coherence

In reply to Roman Kisel
Quoted 6 lines · Roman Kisel

To boot on ARM64, VMBus requires configuring interrupts. Missing DMA coherence property is sub-optimal as the VMBus transations are cache-coherent.

Add interrupts to be able to boot on ARM64. Add DMA coherence to avoid doing extra work on maintaining caches on ARM64.

How do you add it?

Quoted 11 lines
---
 .../devicetree/bindings/bus/microsoft,vmbus.yaml          | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)
Documentation/devicetree/bindings/bus/microsoft,vmbus.yaml +0 -0
index a8d40c766dcd..3ab7d0116626 100644

Documentation/devicetree/bindings/bus/microsoft,vmbus.yaml properties: · line 35

Quoted 9 lines
Documentation/devicetree/bindings/bus/microsoft,vmbus.yaml +2 -1
@@ -28,13 +28,16 @@ properties:
 required:
   - compatible
   - ranges
+  - interrupts
   - '#address-cells'
   - '#size-cells'
 
-additionalProperties: false
+additionalProperties: true

This is neither explained in commit msg nor correct.

Drop the change. You cannot have device bindings ending with 'true' here - see talks, example-bindings, writing-schema and whatever resource is there.

Best regards,
Krzysztof

#46 lore raw

Re: [PATCH hyperv-next v5 09/11] Drivers: hv: vmbus: Introduce hv_get_vmbus_root_device()

In reply to Roman Kisel
Quoted 9 lines · Roman Kisel

The ARM64 PCI code for hyperv needs to know the VMBus root device, and it is private.

Provide a function that returns it. Rename it from "hv_dev" as "hv_dev" as a symbol is very overloaded. No functional changes.

Why change all device's parent to vmbus_root_device?

The ARM64 platform uses the device tree to enumerate vmbus devices.. Can we find the root device via device tree? vmbus code on the x86 use ACPI and it seems to work via ACPI.

Quoted 107 lines
---
 drivers/hv/vmbus_drv.c | 23 +++++++++++++++--------
 include/linux/hyperv.h |  2 ++
 2 files changed, 17 insertions(+), 8 deletions(-)
drivers/hv/vmbus_drv.c +15 -8
index c8474b48dcd2..7bfafe702963 100644
@@ -45,7 +45,8 @@ struct vmbus_dynid {
        struct hv_vmbus_device_id id;
 };
 
-static struct device  *hv_dev;
+/* VMBus Root Device */
+static struct device  *vmbus_root_device;
 
 static int hyperv_cpuhp_online;
 
@@ -80,9 +81,15 @@ static struct resource *fb_mmio;
 static struct resource *hyperv_mmio;
 static DEFINE_MUTEX(hyperv_mmio_lock);
 
+struct device *hv_get_vmbus_root_device(void)
+{
+       return vmbus_root_device;
+}
+EXPORT_SYMBOL_GPL(hv_get_vmbus_root_device);
+
 static int vmbus_exists(void)
 {
-       if (hv_dev == NULL)
+       if (vmbus_root_device == NULL)
                return -ENODEV;
 
        return 0;
@@ -861,7 +868,7 @@ static int vmbus_dma_configure(struct device *child_device)
         * On x86/x64 coherence is assumed and these calls have no effect.
         */
        hv_setup_dma_ops(child_device,
-               device_get_dma_attr(hv_dev) == DEV_DMA_COHERENT);
+               device_get_dma_attr(vmbus_root_device) == DEV_DMA_COHERENT);
        return 0;
 }
 
@@ -1930,7 +1937,7 @@ int vmbus_device_register(struct hv_device *child_device_obj)
                     &child_device_obj->channel->offermsg.offer.if_instance);
 
        child_device_obj->device.bus = &hv_bus;
-       child_device_obj->device.parent = hv_dev;
+       child_device_obj->device.parent = vmbus_root_device;
        child_device_obj->device.release = vmbus_device_release;
 
        child_device_obj->device.dma_parms = &child_device_obj->dma_parms;
@@ -2292,7 +2299,7 @@ static int vmbus_acpi_add(struct platform_device *pdev)
        struct acpi_device *ancestor;
        struct acpi_device *device = ACPI_COMPANION(&pdev->dev);
 
-       hv_dev = &device->dev;
+       vmbus_root_device = &device->dev;
 
        /*
         * Older versions of Hyper-V for ARM64 fail to include the _CCA
@@ -2383,7 +2390,7 @@ static int vmbus_device_add(struct platform_device *pdev)
        struct device_node *np = pdev->dev.of_node;
        int ret;
 
-       hv_dev = &pdev->dev;
+       vmbus_root_device = &pdev->dev;
 
        ret = of_range_parser_init(&parser, np);
        if (ret)
@@ -2702,7 +2709,7 @@ static int __init hv_acpi_init(void)
        if (ret)
                return ret;
 
-       if (!hv_dev) {
+       if (!vmbus_root_device) {
                ret = -ENODEV;
                goto cleanup;
        }
@@ -2733,7 +2740,7 @@ static int __init hv_acpi_init(void)
 
 cleanup:
        platform_driver_unregister(&vmbus_platform_driver);
-       hv_dev = NULL;
+       vmbus_root_device = NULL;
        return ret;
 }
 
include/linux/hyperv.h +2 -1
index 7f4f8d8bdf43..1f0851fde041 100644
@@ -1333,6 +1333,8 @@ static inline void *hv_get_drvdata(struct hv_device *dev)
        return dev_get_drvdata(&dev->device);
 }
 
+struct device *hv_get_vmbus_root_device(void);
+
 struct hv_ring_buffer_debug_info {
        u32 current_interrupt_mask;
        u32 current_read_index;
--

2.43.0

Signature · 2 lines
Thanks
Tianyu Lan
Reviewed-by on v5 4/11
#22 lore raw

Re: [PATCH hyperv-next v5 04/11] Drivers: hv: Provide arch-neutral implementation of get_vtl()

In reply to Roman Kisel
Quoted 5 lines · Roman Kisel

To run in the VTL mode, Hyper-V drivers have to know what VTL the system boots in, and the arm64/hyperv code does not have the means to compute that.

Quoted 7 lines

Refactor the code to hoist the function that detects VTL, make it arch-neutral to be able to employ it to get the VTL on arm64.

---

Signature · 2 lines
Thanks
Tianyu Lan
#52 lore raw

Re: [PATCH hyperv-next v5 11/11] PCI: hv: Get vPCI MSI IRQ domain from DeviceTree

In reply to Roman Kisel
Quoted 8 lines · Roman Kisel

The hyperv-pci driver uses ACPI for MSI IRQ domain configuration on arm64. It won't be able to do that in the VTL mode where only DeviceTree can be used.

Update the hyperv-pci driver to get vPCI MSI IRQ domain in the DeviceTree case, too.

A couple minor comments below, but I don't have any objection to this, so if it's OK with the pci-hyperv.c folks, it's OK with me.

Quoted 5 lines
drivers/pci/controller/pci-hyperv.c +5 -0
+#ifdef CONFIG_OF
+
+static struct irq_domain *hv_pci_of_irq_domain_parent(void)
+{
+	struct device_node *parent;

drivers/pci/controller/pci-hyperv.c static const struct irq_domain_ops hv_pci_domain_ops = { · line 913

Quoted 10 lines
drivers/pci/controller/pci-hyperv.c +10 -0
+	struct irq_domain *domain;
+
+	parent = of_irq_find_parent(hv_get_vmbus_root_device()->of_node);
+	domain = NULL;
+	if (parent) {
+		domain = irq_find_host(parent);
+		of_node_put(parent);
+	}
+
+	return domain;

I think this would be a little simpler as:

  parent = of_irq_find_parent(hv_get_vmbus_root_device()->of_node);
  if (!parent)
    return NULL;

  domain = irq_find_host(parent);
  of_node_put(parent);
  return domain;
Quoted 13 lines
drivers/pci/controller/pci-hyperv.c +13 -0
+}
+
+#endif
+
+#ifdef CONFIG_ACPI
+
+static struct irq_domain *hv_pci_acpi_irq_domain_parent(void)
+{
+	struct irq_domain *domain;
+	acpi_gsi_domain_disp_fn gsi_domain_disp_fn;
+
+	if (acpi_irq_model != ACPI_IRQ_MODEL_GIC)
+		return NULL;

drivers/pci/controller/pci-hyperv.c static const struct irq_domain_ops hv_pci_domain_ops = { · line 936

Quoted 10 lines
drivers/pci/controller/pci-hyperv.c +10 -0
+	gsi_domain_disp_fn = acpi_get_gsi_dispatcher();
+	if (!gsi_domain_disp_fn)
+		return NULL;
+	domain = irq_find_matching_fwnode(gsi_domain_disp_fn(0),
+				     DOMAIN_BUS_ANY);
+
+	if (!domain)
+		return NULL;
+
+	return domain;

if (!domain)

    return NULL;

return domain;

is the same as:

return domain;

or even just:

  return irq_find_matching_fwnode(gsi_domain_disp_fn(0), DOMAIN_BUS_ANY);
Quoted 1 line

+}

#27 lore raw

Re: [PATCH hyperv-next v5 06/11] arm64, x86: hyperv: Report the VTL the system boots in

In reply to Wei Liu
Quoted 12 lines · Wei Liu
Quoted 11 lines · Roman Kisel

The hyperv guest code might run in various Virtual Trust Levels.

Report the level when the kernel boots in the non-default (0) one.

---
  arch/arm64/hyperv/mshyperv.c | 2 ++
  arch/x86/hyperv/hv_vtl.c     | 2 +-
  2 files changed, 3 insertions(+), 1 deletion(-)
Quoted 12 lines

arch/arm64/hyperv/mshyperv.c static int __init hyperv_init(void) · line 111

Quoted 9 lines
arch/arm64/hyperv/mshyperv.c +1 -0
index a7db03f5413d..3bc16dbee758 100644
@@ -108,6 +108,8 @@ static int __init hyperv_init(void)
  	if (ms_hyperv.priv_high & HV_ACCESS_PARTITION_ID)
  		hv_get_partition_id();
  	ms_hyperv.vtl = get_vtl();
+	if (ms_hyperv.vtl > 0) /* non default VTL */

"non-default".

Thanks, will fix that!

Quoted 10 lines
Quoted 10 lines
arch/arm64/hyperv/mshyperv.c +1 -0
+		pr_info("Linux runs in Hyper-V Virtual Trust Level %d\n", ms_hyperv.vtl);
 
  	ms_hyperv_late_init();
 
arch/x86/hyperv/hv_vtl.c +0 -0
index 4e1b1e3b5658..c21bee7e8ff3 100644
@@ -24,7 +24,7 @@ static bool __init hv_vtl_msi_ext_dest_id(void)
 
Quoted 7 lines

arch/x86/hyperv/hv_vtl.c static bool __init hv_vtl_msi_ext_dest_id(void) · line 27

Quoted 4 lines
arch/x86/hyperv/hv_vtl.c +1 -1
  void __init hv_vtl_init_platform(void)
  {
-	pr_info("Linux runs in Hyper-V Virtual Trust Level\n");
+	pr_info("Linux runs in Hyper-V Virtual Trust Level %d\n", ms_hyperv.vtl);

Where isn't there a check for ms_hyperv.vtl > 0 here?

On x86, there is

#ifdef CONFIG_HYPERV_VTL_MODE
void __init hv_vtl_init_platform(void);
int __init hv_vtl_early_init(void);
#else
static inline void __init hv_vtl_init_platform(void) {} static inline int __init hv_vtl_early_init(void) { return 0; } #endif

Quoted 2 lines

Please be consistent across different architectures.

In the earlier versions of the patch series, I had that piece from the above mirrored in the arm64, and there was advice on removing the function as it contained just one statement. I'll revisit the approach, thanks for your help!

Quoted 7 lines
Quoted 7 lines
arch/x86/hyperv/hv_vtl.c +0 -0
  	x86_platform.realmode_reserve = x86_init_noop;
  	x86_platform.realmode_init = x86_init_noop;
Signature · 1 line
2.43.0
Signature · 2 lines
Thank you,
Roman
maintainer
#28 lore raw

Re: [PATCH hyperv-next v5 06/11] arm64, x86: hyperv: Report the VTL the system boots in

Quoted 55 lines · Roman Kisel
Quoted 12 lines · Wei Liu
Quoted 11 lines · Roman Kisel

The hyperv guest code might run in various Virtual Trust Levels.

Report the level when the kernel boots in the non-default (0) one.

---
  arch/arm64/hyperv/mshyperv.c | 2 ++
  arch/x86/hyperv/hv_vtl.c     | 2 +-
  2 files changed, 3 insertions(+), 1 deletion(-)
Quoted 12 lines

arch/arm64/hyperv/mshyperv.c static int __init hyperv_init(void) · line 111

Quoted 9 lines
arch/arm64/hyperv/mshyperv.c +1 -0
index a7db03f5413d..3bc16dbee758 100644
@@ -108,6 +108,8 @@ static int __init hyperv_init(void)
  	if (ms_hyperv.priv_high & HV_ACCESS_PARTITION_ID)
  		hv_get_partition_id();
  	ms_hyperv.vtl = get_vtl();
+	if (ms_hyperv.vtl > 0) /* non default VTL */

"non-default".

Thanks, will fix that!

Quoted 5 lines
Quoted 5 lines
arch/arm64/hyperv/mshyperv.c +1 -0
+		pr_info("Linux runs in Hyper-V Virtual Trust Level %d\n", ms_hyperv.vtl);
  	ms_hyperv_late_init();
arch/x86/hyperv/hv_vtl.c +0 -1
index 4e1b1e3b5658..c21bee7e8ff3 100644
--- a/arch/x86/hyperv/hv_vtl.c
Quoted 9 lines

arch/x86/hyperv/hv_vtl.c static bool __init hv_vtl_msi_ext_dest_id(void) · line 26

Quoted 6 lines

+++ b/arch/x86/hyperv/hv_vtl.c

arch/x86/hyperv/hv_vtl.c +1 -1
@@ -24,7 +24,7 @@ static bool __init hv_vtl_msi_ext_dest_id(void)
  void __init hv_vtl_init_platform(void)
  {
-	pr_info("Linux runs in Hyper-V Virtual Trust Level\n");
+	pr_info("Linux runs in Hyper-V Virtual Trust Level %d\n", ms_hyperv.vtl);

Where isn't there a check for ms_hyperv.vtl > 0 here?

On x86, there is

#ifdef CONFIG_HYPERV_VTL_MODE
void __init hv_vtl_init_platform(void);
int __init hv_vtl_early_init(void);
#else
static inline void __init hv_vtl_init_platform(void) {} static inline int __init hv_vtl_early_init(void) { return 0; } #endif

Quoted 7 lines
Quoted 2 lines

Please be consistent across different architectures.

In the earlier versions of the patch series, I had that piece from the above mirrored in the arm64, and there was advice on removing the function as it contained just one statement. I'll revisit the approach, thanks for your help!

As long as the output is consistent across different architectures, I'm good.

Wei.

Quoted 12 lines
Quoted 7 lines
Quoted 7 lines
arch/x86/hyperv/hv_vtl.c +0 -0
  	x86_platform.realmode_reserve = x86_init_noop;
  	x86_platform.realmode_init = x86_init_noop;
Signature · 1 line
2.43.0
Signature · 2 lines
Thank you,
Roman
#32 lore raw

Re: [PATCH hyperv-next v5 07/11] dt-bindings: microsoft,vmbus: Add interrupts and DMA coherence

In reply to Krzysztof Kozlowski
Quoted 10 lines · Krzysztof Kozlowski
Quoted 6 lines · Roman Kisel

To boot on ARM64, VMBus requires configuring interrupts. Missing DMA coherence property is sub-optimal as the VMBus transations are cache-coherent.

Add interrupts to be able to boot on ARM64. Add DMA coherence to avoid doing extra work on maintaining caches on ARM64.

How do you add it?

I added properties to the node. Should I fix the description, or I am misunderstanding the question?

Quoted 18 lines
Quoted 18 lines
---
  .../devicetree/bindings/bus/microsoft,vmbus.yaml          | 8 +++++++-
  1 file changed, 7 insertions(+), 1 deletion(-)
Documentation/devicetree/bindings/bus/microsoft,vmbus.yaml +1 -0
index a8d40c766dcd..3ab7d0116626 100644
@@ -28,13 +28,16 @@ properties:
  required:
    - compatible
    - ranges
+  - interrupts
    - '#address-cells'
    - '#size-cells'
 
Quoted 5 lines

Documentation/devicetree/bindings/bus/microsoft,vmbus.yaml properties: · line 35

Quoted 2 lines
Documentation/devicetree/bindings/bus/microsoft,vmbus.yaml +1 -1
-additionalProperties: false
+additionalProperties: true

This is neither explained in commit msg nor correct.

Not explained, as there is no good explanation as described below.

Quoted 4 lines

Drop the change. You cannot have device bindings ending with 'true' here - see talks, example-bindings, writing-schema and whatever resource is there.

Thanks, I'll put more effort into bringing this into a better form! If you have time, could you comment on the below?

The Documentation says

* additionalProperties: true
Rare case, used for schemas implementing common set of properties.
Such schemas are supposed to be referenced by other schemas, which then use 'unevaluatedProperties: false'. Typically bus or common-part schemas.

This is a bus so I added that line to the YAML, and I saw it in many other YAML files. Without that line, there was a warning from the local DT validation described in the Documentation about not having pin controls which was weird, and adding

"additionalProperties: true"

fixed the warnings (didn't debug much though). As a side note, there was a similar warning coming from another YAML during running DT schema validation as described in the Documentation so maybe warnings are fine.

Quoted 3 lines

Best regards,
Krzysztof

Signature · 2 lines
Thank you,
Roman
#47 lore raw

Re: [PATCH hyperv-next v5 09/11] Drivers: hv: vmbus: Introduce hv_get_vmbus_root_device()

In reply to Tianyu Lan
Quoted 5 lines · Tianyu Lan
Quoted 4 lines · Roman Kisel

The ARM64 PCI code for hyperv needs to know the VMBus root device, and it is private.

Quoted 8 lines
Quoted 5 lines

Provide a function that returns it. Rename it from "hv_dev" as "hv_dev" as a symbol is very overloaded. No functional changes.

Why change all device's parent to vmbus_root_device?

No changes from my account of the code. Please let me know if I am misunderstanding the question.

Quoted 5 lines

The ARM64 platform uses the device tree to enumerate vmbus devices.. Can we find the root device via device tree? vmbus code on the x86 use ACPI and it seems to work via ACPI.

Right, we find it from the DT as shown in the next patch:

+static struct irq_domain *hv_pci_of_irq_domain_parent(void) +{
+ struct device_node *parent;
+ struct irq_domain *domain;
+
+ parent = of_irq_find_parent(hv_get_vmbus_root_device()->of_node);
+ domain = NULL;
+ if (parent) {
+ domain = irq_find_host(parent);
+ of_node_put(parent);
+ }
+
+ return domain;
+}
+

and later use it for irq_create_hierarchy(). Please let me know if I missed anything in your question.

Quoted 109 lines
Quoted 107 lines
---
  drivers/hv/vmbus_drv.c | 23 +++++++++++++++--------
  include/linux/hyperv.h |  2 ++
  2 files changed, 17 insertions(+), 8 deletions(-)
drivers/hv/vmbus_drv.c +15 -8
index c8474b48dcd2..7bfafe702963 100644
@@ -45,7 +45,8 @@ struct vmbus_dynid {
         struct hv_vmbus_device_id id;
  };
 
-static struct device  *hv_dev;
+/* VMBus Root Device */
+static struct device  *vmbus_root_device;
 
  static int hyperv_cpuhp_online;
 
@@ -80,9 +81,15 @@ static struct resource *fb_mmio;
  static struct resource *hyperv_mmio;
  static DEFINE_MUTEX(hyperv_mmio_lock);
 
+struct device *hv_get_vmbus_root_device(void)
+{
+       return vmbus_root_device;
+}
+EXPORT_SYMBOL_GPL(hv_get_vmbus_root_device);
+
  static int vmbus_exists(void)
  {
-       if (hv_dev == NULL)
+       if (vmbus_root_device == NULL)
                 return -ENODEV;
 
         return 0;
@@ -861,7 +868,7 @@ static int vmbus_dma_configure(struct device *child_device)
          * On x86/x64 coherence is assumed and these calls have no effect.
          */
         hv_setup_dma_ops(child_device,
-               device_get_dma_attr(hv_dev) == DEV_DMA_COHERENT);
+               device_get_dma_attr(vmbus_root_device) == DEV_DMA_COHERENT);
         return 0;
  }
 
@@ -1930,7 +1937,7 @@ int vmbus_device_register(struct hv_device *child_device_obj)
                      &child_device_obj->channel->offermsg.offer.if_instance);
 
         child_device_obj->device.bus = &hv_bus;
-       child_device_obj->device.parent = hv_dev;
+       child_device_obj->device.parent = vmbus_root_device;
         child_device_obj->device.release = vmbus_device_release;
 
         child_device_obj->device.dma_parms = &child_device_obj->dma_parms;
@@ -2292,7 +2299,7 @@ static int vmbus_acpi_add(struct platform_device *pdev)
         struct acpi_device *ancestor;
         struct acpi_device *device = ACPI_COMPANION(&pdev->dev);
 
-       hv_dev = &device->dev;
+       vmbus_root_device = &device->dev;
 
         /*
          * Older versions of Hyper-V for ARM64 fail to include the _CCA
@@ -2383,7 +2390,7 @@ static int vmbus_device_add(struct platform_device *pdev)
         struct device_node *np = pdev->dev.of_node;
         int ret;
 
-       hv_dev = &pdev->dev;
+       vmbus_root_device = &pdev->dev;
 
         ret = of_range_parser_init(&parser, np);
         if (ret)
@@ -2702,7 +2709,7 @@ static int __init hv_acpi_init(void)
         if (ret)
                 return ret;
 
-       if (!hv_dev) {
+       if (!vmbus_root_device) {
                 ret = -ENODEV;
                 goto cleanup;
         }
@@ -2733,7 +2740,7 @@ static int __init hv_acpi_init(void)
 
  cleanup:
         platform_driver_unregister(&vmbus_platform_driver);
-       hv_dev = NULL;
+       vmbus_root_device = NULL;
         return ret;
  }
 
include/linux/hyperv.h +2 -1
index 7f4f8d8bdf43..1f0851fde041 100644
@@ -1333,6 +1333,8 @@ static inline void *hv_get_drvdata(struct hv_device *dev)
         return dev_get_drvdata(&dev->device);
  }
 
+struct device *hv_get_vmbus_root_device(void);
+
  struct hv_ring_buffer_debug_info {
         u32 current_interrupt_mask;
         u32 current_read_index;
--

2.43.0

Signature · 2 lines
Thank you,
Roman
#53 lore raw

Re: [PATCH hyperv-next v5 11/11] PCI: hv: Get vPCI MSI IRQ domain from DeviceTree

In reply to Bjorn Helgaas
Quoted 5 lines · Bjorn Helgaas
Quoted 4 lines · Roman Kisel

The hyperv-pci driver uses ACPI for MSI IRQ domain configuration on arm64. It won't be able to do that in the VTL mode where only DeviceTree can be used.

Quoted 8 lines
Quoted 4 lines

Update the hyperv-pci driver to get vPCI MSI IRQ domain in the DeviceTree case, too.

A couple minor comments below, but I don't have any objection to this, so if it's OK with the pci-hyperv.c folks, it's OK with me.

Bjorn, thanks a lot for your help and guidance! I'll be most happy to incorporate your suggestions into the next version of the series :)

Quoted 64 lines
Quoted 5 lines
drivers/pci/controller/pci-hyperv.c +5 -0
+#ifdef CONFIG_OF
+
+static struct irq_domain *hv_pci_of_irq_domain_parent(void)
+{
+	struct device_node *parent;

drivers/pci/controller/pci-hyperv.c static const struct irq_domain_ops hv_pci_domain_ops = { · line 913

Quoted 10 lines
drivers/pci/controller/pci-hyperv.c +10 -0
+	struct irq_domain *domain;
+
+	parent = of_irq_find_parent(hv_get_vmbus_root_device()->of_node);
+	domain = NULL;
+	if (parent) {
+		domain = irq_find_host(parent);
+		of_node_put(parent);
+	}
+
+	return domain;

I think this would be a little simpler as:

   parent = of_irq_find_parent(hv_get_vmbus_root_device()->of_node);
   if (!parent)
     return NULL;

   domain = irq_find_host(parent);
   of_node_put(parent);
   return domain;
Quoted 13 lines
drivers/pci/controller/pci-hyperv.c +13 -0
+}
+
+#endif
+
+#ifdef CONFIG_ACPI
+
+static struct irq_domain *hv_pci_acpi_irq_domain_parent(void)
+{
+	struct irq_domain *domain;
+	acpi_gsi_domain_disp_fn gsi_domain_disp_fn;
+
+	if (acpi_irq_model != ACPI_IRQ_MODEL_GIC)
+		return NULL;

drivers/pci/controller/pci-hyperv.c static const struct irq_domain_ops hv_pci_domain_ops = { · line 936

Quoted 10 lines
drivers/pci/controller/pci-hyperv.c +10 -0
+	gsi_domain_disp_fn = acpi_get_gsi_dispatcher();
+	if (!gsi_domain_disp_fn)
+		return NULL;
+	domain = irq_find_matching_fwnode(gsi_domain_disp_fn(0),
+				     DOMAIN_BUS_ANY);
+
+	if (!domain)
+		return NULL;
+
+	return domain;
   if (!domain)
     return NULL;

   return domain;

is the same as:

   return domain;

or even just:

   return irq_find_matching_fwnode(gsi_domain_disp_fn(0), DOMAIN_BUS_ANY);
Quoted 1 line

+}

Signature · 2 lines
Thank you,
Roman
#29 lore raw

Re: [PATCH hyperv-next v5 06/11] arm64, x86: hyperv: Report the VTL the system boots in

In reply to Wei Liu
Quoted 1 line · Wei Liu

On Mon, Mar 10, 2025 at 09:42:15AM -0700, Roman Kisel wrote:

[...]

Quoted 11 lines
Quoted 8 lines
Quoted 3 lines

Please be consistent across different architectures.

In the earlier versions of the patch series, I had that piece from the above mirrored in the arm64, and there was advice on removing the function as it contained just one statement. I'll revisit the approach, thanks for your help!

As long as the output is consistent across different architectures, I'm good.

I should add a comment most likely to save people some time grepping the code as the line does look like should always print that. IOW not printing for VTL0 is obscured by the preprocessor/#define cruft.

Quoted 15 lines

Wei.

Quoted 12 lines
Quoted 7 lines
Quoted 7 lines
   	x86_platform.realmode_reserve = x86_init_noop;
   	x86_platform.realmode_init = x86_init_noop;
Signature · 1 line
2.43.0
Signature · 2 lines
Thank you,
Roman
Signature · 2 lines
Thank you,
Roman
#11 lore raw

Re: [PATCH hyperv-next v5 03/11] Drivers: hv: Enable VTL mode for arm64

In reply to Arnd Bergmann
Quoted 28 lines · Arnd Bergmann
Quoted 13 lines · Roman Kisel
drivers/hv/Kconfig +2 -2
@@ -5,18 +5,20 @@ menu "Microsoft Hyper-V guest support"
  config HYPERV
  	tristate "Microsoft Hyper-V client drivers"
  	depends on (X86 && X86_LOCAL_APIC && HYPERVISOR_GUEST) \
-		|| (ACPI && ARM64 && !CPU_BIG_ENDIAN)
+		|| (ARM64 && !CPU_BIG_ENDIAN)
+	depends on (ACPI || HYPERV_VTL_MODE)
  	select PARAVIRT
  	select X86_HV_CALLBACK_VECTOR if X86
-	select OF_EARLY_FLATTREE if OF
  	help
  	  Select this option to run Linux as a Hyper-V client operating
  	  system.

drivers/hv/Kconfig menu "Microsoft Hyper-V guest support" · line 23

Quoted 10 lines
drivers/hv/Kconfig +3 -1
 
  config HYPERV_VTL_MODE
  	bool "Enable Linux to boot in VTL context"
-	depends on X86_64 && HYPERV
+	depends on (X86_64 || ARM64)
  	depends on SMP
+	select OF_EARLY_FLATTREE
+	select OF
  	default n
  	help

Having the dependency below the top-level Kconfig entry feels a little counterintuitive. You could flip that back as it was before by doing

Quoted 6 lines

select HYPERV_VTL_MODE if !ACPI
depends on ACPI || SMP

in the HYPERV option, leaving the dependency on HYPERV in HYPERV_VTL_MODE.

I was implementing Michael's suggestion, and might've gone a bit overboard, my bad. I'll fix this, thanks a lot for reviewing!

Quoted 2 lines

Is OF_EARLY_FLATTREE actually needed on x86?

No, it is not needed on x86. It is only needed when VTL mode is used.

Quoted 1 line

Arnd

Signature · 2 lines
Thank you,
Roman
#39 lore raw

Re: [PATCH hyperv-next v5 08/11] Drivers: hv: vmbus: Get the IRQ number from DeviceTree

In reply to Arnd Bergmann
Quoted 16 lines · Arnd Bergmann
Quoted 2 lines · Roman Kisel

+static int __maybe_unused vmbus_set_irq(struct platform_device *pdev)

Instead of the __maybe_unused annotation here

drivers/hv/vmbus_drv.c static int vmbus_device_add(struct platform_device *pdev) · line 2397

Quoted 7 lines
drivers/hv/vmbus_drv.c +6 -0
+#ifndef HYPERVISOR_CALLBACK_VECTOR
+	ret = vmbus_set_irq(pdev);
+	if (ret)
+		return ret;
+#endif
+

you can use

Quoted 5 lines
        if (!__is_defined(HYPERVISOR_CALLBACK_VECTOR))
                   ret = vmbus_set_irq(pdev);

and make it a little more readable.

Thanks you very much, will update! Very neat :)

Quoted 1 line

Arnd

Signature · 2 lines
Thank you,
Roman
#33 lore raw

Re: [PATCH hyperv-next v5 07/11] dt-bindings: microsoft,vmbus: Add interrupts and DMA coherence

In reply to Roman Kisel
Quoted 8 lines · Roman Kisel
Quoted 5 lines · Krzysztof Kozlowski
Quoted 4 lines · Roman Kisel

To boot on ARM64, VMBus requires configuring interrupts. Missing DMA coherence property is sub-optimal as the VMBus transations are cache-coherent.

Quoted 8 lines
Quoted 5 lines
Quoted 2 lines

Add interrupts to be able to boot on ARM64. Add DMA coherence to avoid doing extra work on maintaining caches on ARM64.

How do you add it?

I added properties to the node. Should I fix the description, or I am misunderstanding the question?

I saw interrupts in the schema, but I did not see dma-coherence. I also did not see any DTS patches here, so I don't understand what node you are referring to.

Quoted 35 lines
Quoted 19 lines
Quoted 19 lines
---
  .../devicetree/bindings/bus/microsoft,vmbus.yaml          | 8 +++++++-
  1 file changed, 7 insertions(+), 1 deletion(-)
Documentation/devicetree/bindings/bus/microsoft,vmbus.yaml +1 -0
index a8d40c766dcd..3ab7d0116626 100644
@@ -28,13 +28,16 @@ properties:
  required:
    - compatible
    - ranges
+  - interrupts
    - '#address-cells'
    - '#size-cells'
 
Quoted 5 lines

Documentation/devicetree/bindings/bus/microsoft,vmbus.yaml properties: · line 35

Quoted 2 lines
Documentation/devicetree/bindings/bus/microsoft,vmbus.yaml +1 -1
-additionalProperties: false
+additionalProperties: true

This is neither explained in commit msg nor correct.

Not explained, as there is no good explanation as described below.

Quoted 4 lines

Drop the change. You cannot have device bindings ending with 'true' here - see talks, example-bindings, writing-schema and whatever resource is there.

Thanks, I'll put more effort into bringing this into a better form! If you have time, could you comment on the below?

Quoted 8 lines

The Documentation says

* additionalProperties: true
Rare case, used for schemas implementing common set of properties.
Such schemas are supposed to be referenced by other schemas, which then use 'unevaluatedProperties: false'. Typically bus or common-part schemas.

This is a bus so I added that line to the YAML, and I saw it in many

If this is a bus, then where is schema using it for bus-attached-devices? You cannot have bus without devices.

You *must* fulfill that part:
"Such schemas are supposed to be referenced by other schemas, which then"

instead of calling it bus...

Please upstream bindings for the bus devices and extend the example here with these devices.

Or this is not bus (calling something vmpony does not make it a pony).

Best regards,
Krzysztof

#34 lore raw

Re: [PATCH hyperv-next v5 07/11] dt-bindings: microsoft,vmbus: Add interrupts and DMA coherence

Quoted 15 lines · Krzysztof Kozlowski
Quoted 14 lines · Roman Kisel
Quoted 10 lines · Krzysztof Kozlowski
Quoted 6 lines · Roman Kisel

To boot on ARM64, VMBus requires configuring interrupts. Missing DMA coherence property is sub-optimal as the VMBus transations are cache-coherent.

Add interrupts to be able to boot on ARM64. Add DMA coherence to avoid doing extra work on maintaining caches on ARM64.

How do you add it?

Quoted 7 lines
Quoted 2 lines

I added properties to the node. Should I fix the description, or I am misunderstanding the question?

I saw interrupts in the schema, but I did not see dma-coherence. I also did not see any DTS patches here, so I don't understand what node you are referring to.

I will refer to talks, example-bindings, writing-schema you've suggested to waste your time less. It appears there is some fundamental flaw in my understanding of how these YAML files work so much so that I cannot even write a commit description that can be understood, for the 5th time in the row, sorry about that.

Quoted 44 lines
Quoted 35 lines
Quoted 19 lines
Quoted 19 lines
---
   .../devicetree/bindings/bus/microsoft,vmbus.yaml          | 8 +++++++-
   1 file changed, 7 insertions(+), 1 deletion(-)
Documentation/devicetree/bindings/bus/microsoft,vmbus.yaml +1 -0
index a8d40c766dcd..3ab7d0116626 100644
@@ -28,13 +28,16 @@ properties:
   required:
     - compatible
     - ranges
+  - interrupts
     - '#address-cells'
     - '#size-cells'
 
Quoted 5 lines

Documentation/devicetree/bindings/bus/microsoft,vmbus.yaml properties: · line 35

Quoted 2 lines
Documentation/devicetree/bindings/bus/microsoft,vmbus.yaml +1 -1
-additionalProperties: false
+additionalProperties: true

This is neither explained in commit msg nor correct.

Not explained, as there is no good explanation as described below.

Quoted 4 lines

Drop the change. You cannot have device bindings ending with 'true' here - see talks, example-bindings, writing-schema and whatever resource is there.

Thanks, I'll put more effort into bringing this into a better form! If you have time, could you comment on the below?

Quoted 8 lines

The Documentation says

* additionalProperties: true
Rare case, used for schemas implementing common set of properties.
Such schemas are supposed to be referenced by other schemas, which then use 'unevaluatedProperties: false'. Typically bus or common-part schemas.

This is a bus so I added that line to the YAML, and I saw it in many

Quoted 8 lines

If this is a bus, then where is schema using it for bus-attached-devices? You cannot have bus without devices.

You *must* fulfill that part:
"Such schemas are supposed to be referenced by other schemas, which then"

instead of calling it bus...

It is modeled as a bus in the kernel:
https://www.kernel.org/doc/html/latest/virt/hyperv/vmbus.html

Quoted 2 lines

Please upstream bindings for the bus devices and extend the example here with these devices.

The set of synthetic devices that reside on the bus isn't fixed, and they don't require description neither in ACPI nor in DT as the devices negotiate their MMIO regions through the hyperv driver.

Perhaps, it is not as much bus as expected by the YAML files.

Quoted 5 lines

Or this is not bus (calling something vmpony does not make it a pony).

Quoted 1 line

Best regards,

Krzysztof

Signature · 2 lines
Thank you,
Roman
#12 lore raw

RE: [PATCH hyperv-next v5 03/11] Drivers: hv: Enable VTL mode for arm64

In reply to Arnd Bergmann

From: Arnd Bergmann <[email protected]> Sent: Saturday, March 8, 2025 1:05 PM

Quoted 29 lines · Roman Kisel
Quoted 13 lines
drivers/hv/Kconfig +2 -2
@@ -5,18 +5,20 @@ menu "Microsoft Hyper-V guest support"
 config HYPERV
 	tristate "Microsoft Hyper-V client drivers"
 	depends on (X86 && X86_LOCAL_APIC && HYPERVISOR_GUEST) \
-		|| (ACPI && ARM64 && !CPU_BIG_ENDIAN)
+		|| (ARM64 && !CPU_BIG_ENDIAN)
+	depends on (ACPI || HYPERV_VTL_MODE)
 	select PARAVIRT
 	select X86_HV_CALLBACK_VECTOR if X86
-	select OF_EARLY_FLATTREE if OF
 	help
 	  Select this option to run Linux as a Hyper-V client operating
 	  system.

drivers/hv/Kconfig menu "Microsoft Hyper-V guest support" · line 23

Quoted 10 lines
drivers/hv/Kconfig +3 -1
 
 config HYPERV_VTL_MODE
 	bool "Enable Linux to boot in VTL context"
-	depends on X86_64 && HYPERV
+	depends on (X86_64 || ARM64)
 	depends on SMP
+	select OF_EARLY_FLATTREE
+	select OF
 	default n
 	help

Having the dependency below the top-level Kconfig entry feels a little counterintuitive. You could flip that back as it was before by doing

Quoted 5 lines

select HYPERV_VTL_MODE if !ACPI
depends on ACPI || SMP

in the HYPERV option, leaving the dependency on HYPERV in HYPERV_VTL_MODE.

I would argue that we don't ever want to implicitly select HYPERV_VTL_MODE because of some other config setting or lack thereof. VTL mode is enough of a special case that it should only be explicitly selected. If someone omits ACPI, then HYPERV should not be selectable unless HYPERV_VTL_MODE is explicitly selected.

The last line of the comment for HYPERV_VTL_MODE says
"A kernel built with this option must run at VTL2, and will not run as a normal guest." In other words, don't choose this unless you 100% know that VTL2 is what you want.

Michael

Quoted 4 lines

Is OF_EARLY_FLATTREE actually needed on x86?

Arnd

#3 lore raw

RE: [PATCH hyperv-next v5 01/11] arm64: kvm, smccc: Introduce and use API for detectting hypervisor presence

In reply to Roman Kisel

From: Roman Kisel <[email protected]> Sent: Friday, March 7, 2025 2:03 PM

Quoted 1 line

In the Subject line,

s/detectting/detecting/

Quoted 7 lines

The KVM/arm64 uses SMCCC to detect hypervisor presence. That code is private, and it follows the SMCCC specification. Other existing and emerging hypervisor guest implementations can and should use that standard approach as well.

Factor out a common infrastructure that the guests can use, update KVM to employ the new API. The cenrtal notion of the SMCCC method is the

s/cenrtal/central/

Quoted 170 lines

UUID of the hypervisor, and the API follows that.

No functional changes. Validated with a KVM/arm64 guest.

---
 arch/arm64/kvm/hypercalls.c        |  5 +--
 drivers/firmware/smccc/kvm_guest.c | 10 ++----
 drivers/firmware/smccc/smccc.c     | 19 +++++++++++
 include/linux/arm-smccc.h          | 55 +++++++++++++++++++++++++++---
 4 files changed, 73 insertions(+), 16 deletions(-)
arch/arm64/kvm/hypercalls.c +1 -4
index 27ce4cb44904..92b9bc1ea8e8 100644
@@ -353,10 +353,7 @@ int kvm_smccc_call_handler(struct kvm_vcpu *vcpu)
 			val[0] = gpa;
 		break;
 	case ARM_SMCCC_VENDOR_HYP_CALL_UID_FUNC_ID:
-		val[0] = ARM_SMCCC_VENDOR_HYP_UID_KVM_REG_0;
-		val[1] = ARM_SMCCC_VENDOR_HYP_UID_KVM_REG_1;
-		val[2] = ARM_SMCCC_VENDOR_HYP_UID_KVM_REG_2;
-		val[3] = ARM_SMCCC_VENDOR_HYP_UID_KVM_REG_3;
+		UUID_TO_SMCCC_RES(ARM_SMCCC_VENDOR_HYP_UID_KVM, val);
 		break;
 	case ARM_SMCCC_VENDOR_HYP_KVM_FEATURES_FUNC_ID:
 		val[0] = smccc_feat->vendor_hyp_bmap;
drivers/firmware/smccc/kvm_guest.c +0 -0
index f3319be20b36..b5084b309ea0 100644
@@ -14,17 +14,11 @@ static DECLARE_BITMAP(__kvm_arm_hyp_services,

ARM_SMCCC_KVM_NUM_FUNCS) __ro_afte

void __init kvm_init_hyp_services(void)
{

drivers/firmware/smccc/kvm_guest.c +2 -8
+	uuid_t kvm_uuid = ARM_SMCCC_VENDOR_HYP_UID_KVM;
 	struct arm_smccc_res res;
 	u32 val[4];
 
-	if (arm_smccc_1_1_get_conduit() != SMCCC_CONDUIT_HVC)
-		return;
-
-	arm_smccc_1_1_invoke(ARM_SMCCC_VENDOR_HYP_CALL_UID_FUNC_ID, &res);
-	if (res.a0 != ARM_SMCCC_VENDOR_HYP_UID_KVM_REG_0 ||
-	    res.a1 != ARM_SMCCC_VENDOR_HYP_UID_KVM_REG_1 ||
-	    res.a2 != ARM_SMCCC_VENDOR_HYP_UID_KVM_REG_2 ||
-	    res.a3 != ARM_SMCCC_VENDOR_HYP_UID_KVM_REG_3)
+	if (!arm_smccc_hyp_present(&kvm_uuid))
 		return;
 
 	memset(&res, 0, sizeof(res));
drivers/firmware/smccc/smccc.c +19 -0
index a74600d9f2d7..7399f27d58e5 100644
@@ -67,6 +67,25 @@ s32 arm_smccc_get_soc_id_revision(void)
 }
 EXPORT_SYMBOL_GPL(arm_smccc_get_soc_id_revision);
 
+bool arm_smccc_hyp_present(const uuid_t *hyp_uuid)
+{
+	struct arm_smccc_res res = {};
+
+	if (arm_smccc_1_1_get_conduit() != SMCCC_CONDUIT_HVC)
+		return false;
+	arm_smccc_1_1_hvc(ARM_SMCCC_VENDOR_HYP_CALL_UID_FUNC_ID, &res);
+	if (res.a0 == SMCCC_RET_NOT_SUPPORTED)
+		return false;
+
+	return ({
+		const uuid_t uuid = SMCCC_RES_TO_UUID(res.a0, res.a1, res.a2, res.a3);
+		const bool present = uuid_equal(&uuid, hyp_uuid);
+
+		present;
+	});
+}
+EXPORT_SYMBOL_GPL(arm_smccc_hyp_present);
+
 static int __init smccc_devices_init(void)
 {
 	struct platform_device *pdev;
include/linux/arm-smccc.h +47 -4
index 67f6fdf2e7cd..726f18221f1c 100644
@@ -7,6 +7,11 @@
 
 #include <linux/args.h>
 #include <linux/init.h>
+
+#ifndef __ASSEMBLER__
+#include <linux/uuid.h>
+#endif
+
 #include <uapi/linux/const.h>
 
 /*
@@ -107,10 +112,10 @@
 			   ARM_SMCCC_FUNC_QUERY_CALL_UID)
 
 /* KVM UID value: 28b46fb6-2ec5-11e9-a9ca-4b564d003a74 */
-#define ARM_SMCCC_VENDOR_HYP_UID_KVM_REG_0	0xb66fb428U
-#define ARM_SMCCC_VENDOR_HYP_UID_KVM_REG_1	0xe911c52eU
-#define ARM_SMCCC_VENDOR_HYP_UID_KVM_REG_2	0x564bcaa9U
-#define ARM_SMCCC_VENDOR_HYP_UID_KVM_REG_3	0x743a004dU
+#define ARM_SMCCC_VENDOR_HYP_UID_KVM UUID_INIT(\
+	0xb66fb428, 0xc52e, 0xe911, \
+	0xa9, 0xca, 0x4b, 0x56, \
+	0x4d, 0x00, 0x3a, 0x74)
 
 /* KVM "vendor specific" services */
 #define ARM_SMCCC_KVM_FUNC_FEATURES		0
@@ -333,6 +338,48 @@ s32 arm_smccc_get_soc_id_version(void);
  */
 s32 arm_smccc_get_soc_id_revision(void);
 
+#ifndef __ASSEMBLER__
+
+/**
+ * arm_smccc_hyp_present(const uuid_t *hyp_uuid)
+ *
+ * Returns `true` if the hypervisor advertises its presence via SMCCC.
+ *
+ * When the function returns `false`, the caller shall not assume that
+ * there is no hypervisor running. Instead, the caller must fall back to
+ * other approaches if any are available.
+ */
+bool arm_smccc_hyp_present(const uuid_t *hyp_uuid);
+
+#define SMCCC_RES_TO_UUID(r0, r1, r2, r3) \
+	UUID_INIT( \
+		cpu_to_le32(lower_32_bits(r0)), \
+		cpu_to_le32(lower_32_bits(r1)) & 0xffff, \
+		cpu_to_le32(lower_32_bits(r1)) >> 16, \
+		cpu_to_le32(lower_32_bits(r2)) & 0xff, \
+		(cpu_to_le32(lower_32_bits(r2)) >> 8) & 0xff, \
+		(cpu_to_le32(lower_32_bits(r2)) >> 16) & 0xff, \
+		(cpu_to_le32(lower_32_bits(r2)) >> 24) & 0xff, \
+		cpu_to_le32(lower_32_bits(r3)) & 0xff, \
+		(cpu_to_le32(lower_32_bits(r3)) >> 8) & 0xff, \
+		(cpu_to_le32(lower_32_bits(r3)) >> 16) & 0xff, \
+		(cpu_to_le32(lower_32_bits(r3)) >> 24) & 0xff \
+	)
+
+#define UUID_TO_SMCCC_RES(uuid_init, regs) do { \
+		const uuid_t uuid = uuid_init; \
+		(regs)[0] = le32_to_cpu((u32)uuid.b[0] | (uuid.b[1] << 8) | \
+						((uuid.b[2]) << 16) | ((uuid.b[3]) << 24));
 \
+		(regs)[1] = le32_to_cpu((u32)uuid.b[4] | (uuid.b[5] << 8) | \
+						((uuid.b[6]) << 16) | ((uuid.b[7]) << 24));
 \
+		(regs)[2] = le32_to_cpu((u32)uuid.b[8] | (uuid.b[9] << 8) | \
+						((uuid.b[10]) << 16) | ((uuid.b[11]) <<
 24)); \
+		(regs)[3] = le32_to_cpu((u32)uuid.b[12] | (uuid.b[13] << 8) | \
+						((uuid.b[14]) << 16) | ((uuid.b[15]) <<

24)); \

include/linux/arm-smccc.h +4 -1
+	} while (0)
+
+#endif /* !__ASSEMBLER__ */
+
 /**
  * struct arm_smccc_res - Result from SMC/HVC call
  * @a0-a3 result values from registers 0 to 3
--

2.43.0

#35 lore raw

Re: [PATCH hyperv-next v5 07/11] dt-bindings: microsoft,vmbus: Add interrupts and DMA coherence

In reply to Roman Kisel
Quoted 4 lines · Roman Kisel

It is modeled as a bus in the kernel:
https://www.kernel.org/doc/html/latest/virt/hyperv/vmbus.html

Quoted 8 lines
Quoted 2 lines

Please upstream bindings for the bus devices and extend the example here with these devices.

The set of synthetic devices that reside on the bus isn't fixed, and they don't require description neither in ACPI nor in DT as the devices negotiate their MMIO regions through the hyperv driver.

Perhaps, it is not as much bus as expected by the YAML files.

OK, then this is not really a bus from the bindings point of view. It is a device schema which should end with additionalProperties: false.

If you have report about that pinctrl-0, it means you have undocumented properties in your DTS. Maybe that's the dma-coherence you mentioned in the commit msg.

Best regards,
Krzysztof

Reviewed-by on v5 2/11
#8 lore raw

RE: [PATCH hyperv-next v5 02/11] arm64: hyperv: Use SMCCC to detect hypervisor presence

In reply to Roman Kisel

From: Roman Kisel <[email protected]> Sent: Friday, March 7, 2025 2:03 PM

Quoted 79 lines

The arm64 Hyper-V startup path relies on ACPI to detect running under a Hyper-V compatible hypervisor. That doesn't work on non-ACPI systems.

Hoist the ACPI detection logic into a separate function. Then use the vendor-specific hypervisor service call (implemented recently in Hyper-V) via SMCCC in the non-ACPI case.

---
 arch/arm64/hyperv/mshyperv.c | 43 +++++++++++++++++++++++++++++++-----
 1 file changed, 38 insertions(+), 5 deletions(-)
arch/arm64/hyperv/mshyperv.c +35 -0
index 29fcfd595f48..c647db56fd6b 100644
@@ -27,6 +27,41 @@ int hv_get_hypervisor_version(union hv_hypervisor_version_info
 *info)
 	return 0;
 }
 
+static bool __init hyperv_detect_via_acpi(void)
+{
+	if (acpi_disabled)
+		return false;
+#if IS_ENABLED(CONFIG_ACPI)
+	/*
+	 * Hypervisor ID is only available in ACPI v6+, and the
+	 * structure layout was extended in v6 to accommodate that
+	 * new field.
+	 *
+	 * At the very minimum, this check makes sure not to read
+	 * past the FADT structure.
+	 *
+	 * It is also needed to catch running in some unknown
+	 * non-Hyper-V environment that has ACPI 5.x or less.
+	 * In such a case, it can't be Hyper-V.
+	 */
+	if (acpi_gbl_FADT.header.revision < 6)
+		return false;
+	return strncmp((char *)&acpi_gbl_FADT.hypervisor_id, "MsHyperV", 8) == 0;
+#else
+	return false;
+#endif
+}
+
+static bool __init hyperv_detect_via_smccc(void)
+{
+	uuid_t hyperv_uuid = UUID_INIT(
+		0x4d32ba58, 0x4764, 0xcd24,
+		0x75, 0x6c, 0xef, 0x8e,
+		0x24, 0x70, 0x59, 0x16);
+
+	return arm_smccc_hyp_present(&hyperv_uuid);
+}
+
 static int __init hyperv_init(void)
 {
arch/arm64/hyperv/mshyperv.c +3 -6
 	struct hv_get_vp_registers_output	result;
@@ -35,13 +70,11 @@ static int __init hyperv_init(void)
 
 	/*
 	 * Allow for a kernel built with CONFIG_HYPERV to be running in
-	 * a non-Hyper-V environment, including on DT instead of ACPI.
+	 * a non-Hyper-V environment.
+	 *
 	 * In such cases, do nothing and return success.
 	 */
-	if (acpi_disabled)
-		return 0;
-
-	if (strncmp((char *)&acpi_gbl_FADT.hypervisor_id, "MsHyperV", 8))
+	if (!hyperv_detect_via_acpi() && !hyperv_detect_via_smccc())
 		return 0;
 
 	/* Setup the guest ID */
--
Quoted 2 lines

2.43.0

maintainer
#13 lore raw

Re: [PATCH hyperv-next v5 03/11] Drivers: hv: Enable VTL mode for arm64

In reply to Michael Kelley
Quoted 27 lines · Michael Kelley

From: Arnd Bergmann <[email protected]> Sent: Saturday, March 8, 2025 1:05 PM

Quoted 13 lines

drivers/hv/Kconfig menu "Microsoft Hyper-V guest support" · line 23

Quoted 9 lines

config HYPERV_VTL_MODE

drivers/hv/Kconfig +3 -1
 	bool "Enable Linux to boot in VTL context"
-	depends on X86_64 && HYPERV
+	depends on (X86_64 || ARM64)
 	depends on SMP
+	select OF_EARLY_FLATTREE
+	select OF
 	default n
 	help

Having the dependency below the top-level Kconfig entry feels a little counterintuitive. You could flip that back as it was before by doing

Quoted 5 lines

select HYPERV_VTL_MODE if !ACPI
depends on ACPI || SMP

in the HYPERV option, leaving the dependency on HYPERV in HYPERV_VTL_MODE.

I would argue that we don't ever want to implicitly select HYPERV_VTL_MODE because of some other config setting or lack thereof. VTL mode is enough of a special case that it should only be explicitly selected. If someone omits ACPI, then HYPERV should not be selectable unless HYPERV_VTL_MODE is explicitly selected.

Quoted 4 lines

The last line of the comment for HYPERV_VTL_MODE says
"A kernel built with this option must run at VTL2, and will not run as a normal guest." In other words, don't choose this unless you 100% know that VTL2 is what you want.

It sounds like the latter is the real problem: enabling a feature should never prevent something else from working. Can you describe what VTL context is and why it requires an exception to a rather fundamental rule here? If you build a kernel that runs on every single piece of arm64 hardware and every hypervisor, why can't you add HYPERV_VTL_MODE to that as an option?

Arnd

#36 lore raw

Re: [PATCH hyperv-next v5 07/11] dt-bindings: microsoft,vmbus: Add interrupts and DMA coherence

In reply to Krzysztof Kozlowski
Quoted 12 lines · Krzysztof Kozlowski
Quoted 11 lines · Roman Kisel

It is modeled as a bus in the kernel:
https://www.kernel.org/doc/html/latest/virt/hyperv/vmbus.html

Quoted 2 lines

Please upstream bindings for the bus devices and extend the example here with these devices.

The set of synthetic devices that reside on the bus isn't fixed, and they don't require description neither in ACPI nor in DT as the devices negotiate their MMIO regions through the hyperv driver.

Quoted 9 lines
Quoted 1 line

Perhaps, it is not as much bus as expected by the YAML files.

OK, then this is not really a bus from the bindings point of view. It is a device schema which should end with additionalProperties: false.

If you have report about that pinctrl-0, it means you have undocumented properties in your DTS. Maybe that's the dma-coherence you mentioned in the commit msg.

Much appreciated! I started reviewing the learning materials you mentioned, and I think I already see where my understanding went
sideways: I perceived the example as the central part of the bindings whereas it seems to be just what the name suggests: an example. Yet, the example shall conform to the *schema* iiuc, and that is what the tooling validates.

Hopefully, I am starting to be getting what this is all about :) Thanks for your help again!

I've worked out what makes (more) sense (to me at least):

From 475fb74b49dc4987ca8b9117186941d848f0aacd Mon Sep 17 00:00:00 2001
From: Roman Kisel <[email protected]>
Date: Mon, 10 Mar 2025 14:39:41 -0700
Subject: [PATCH] dt-bindings: microsoft,vmbus: Add interrupt and DMA
coherence properties

To boot in the VTL mode, VMBus on arm64 needs interrupt description which the binding documentation lacks. The transactions on the bus are DMA coherent which is not mentioned as well.

Add the interrupt property and the DMA coherence property to the VMBus binding. Update the example to match that. Fix typos.

---
 .../bindings/bus/microsoft,vmbus.yaml           | 17 +++++++++++++++--
 1 file changed, 15 insertions(+), 2 deletions(-)
Documentation/devicetree/bindings/bus/microsoft,vmbus.yaml +15 -2
index a8d40c766dcd..b175ad01f219 100644
@@ -10,8 +10,8 @@ maintainers:
   - Saurabh Sengar <[email protected]>
 
 description:
-  VMBus is a software bus that implement the protocols for communication
-  between the root or host OS and guest OSs (virtual machines).
+  VMBus is a software bus that implements the protocols for communication
+  between the root or host OS and guest OS'es (virtual machines).
 
 properties:
   compatible:
@@ -25,9 +25,17 @@ properties:
   '#size-cells':
     const: 1
 
+  dma-coherent: true
+
+  interrupts:
+    maxItems: 1
+    description: |
+      This interrupt signals a message from the host.
+
 required:
   - compatible
   - ranges
+  - interrupts
   - '#address-cells'
   - '#size-cells'
 
@@ -35,6 +43,8 @@ additionalProperties: false
 
 examples:
   - |
+    #include <dt-bindings/interrupt-controller/irq.h>
+    #include <dt-bindings/interrupt-controller/arm-gic.h>
     soc {
         #address-cells = <2>;
         #size-cells = <1>;
@@ -49,6 +59,9 @@ examples:
                 #address-cells = <2>;
                 #size-cells = <1>;
                 ranges = <0x0f 0xf0000000 0x0f 0xf0000000 0x10000000>;
+                dma-coherent;
+                interrupt-parent = <&gic>;
+                interrupts = <GIC_PPI 2 IRQ_TYPE_EDGE_RISING>;
             };
         };
     };
Signature · 1 line
2.43.0
Quoted 3 lines

Best regards,
Krzysztof

Signature · 2 lines
Thank you,
Roman
#4 lore raw

Re: [PATCH hyperv-next v5 01/11] arm64: kvm, smccc: Introduce and use API for detectting hypervisor presence

In reply to Michael Kelley
Quoted 6 lines · Michael Kelley

From: Roman Kisel <[email protected]> Sent: Friday, March 7, 2025 2:03 PM

Quoted 1 line

In the Subject line,

s/detectting/detecting/

[...]

Quoted 2 lines

s/cenrtal/central/

My bad, will fix! Thanks for spotting that, I'll be sure to have the spellchecker on.

Signature · 2 lines
Thank you,
Roman
#14 lore raw

RE: [PATCH hyperv-next v5 03/11] Drivers: hv: Enable VTL mode for arm64

In reply to Arnd Bergmann

From: Arnd Bergmann <[email protected]> Sent: Monday, March 10, 2025 2:21 PM

Quoted 34 lines · Michael Kelley
Quoted 27 lines

From: Arnd Bergmann <[email protected]> Sent: Saturday, March 8, 2025 1:05 PM

Quoted 13 lines

drivers/hv/Kconfig menu "Microsoft Hyper-V guest support" · line 23

Quoted 9 lines

config HYPERV_VTL_MODE

drivers/hv/Kconfig +3 -1
 	bool "Enable Linux to boot in VTL context"
-	depends on X86_64 && HYPERV
+	depends on (X86_64 || ARM64)
 	depends on SMP
+	select OF_EARLY_FLATTREE
+	select OF
 	default n
 	help

Having the dependency below the top-level Kconfig entry feels a little counterintuitive. You could flip that back as it was before by doing

Quoted 5 lines

select HYPERV_VTL_MODE if !ACPI
depends on ACPI || SMP

in the HYPERV option, leaving the dependency on HYPERV in HYPERV_VTL_MODE.

I would argue that we don't ever want to implicitly select HYPERV_VTL_MODE because of some other config setting or lack thereof. VTL mode is enough of a special case that it should only be explicitly selected. If someone omits ACPI, then HYPERV should not be selectable unless HYPERV_VTL_MODE is explicitly selected.

Quoted 4 lines

The last line of the comment for HYPERV_VTL_MODE says
"A kernel built with this option must run at VTL2, and will not run as a normal guest." In other words, don't choose this unless you 100% know that VTL2 is what you want.

Quoted 7 lines

It sounds like the latter is the real problem: enabling a feature should never prevent something else from working. Can you describe what VTL context is and why it requires an exception to a rather fundamental rule here? If you build a kernel that runs on every single piece of arm64 hardware and every hypervisor, why can't you add HYPERV_VTL_MODE to that as an option?

VTL = Virtual Trust Level, and VSM = Virtual Secure Mode, are Hyper-V's terminology for offering multiple execution environments with hierarchical trust in the context of a single VM. A normal guest operating system runs at VTL 0, and there are no other VTLs in use. But in some environments, additional software may run as a paravisor layer between the normal guest OS and the hypervisor. This software runs at some other VTL > 0, and has a higher privilege level within the VM than software running at VTL 0 (which is the lowest privilege). VTL 2 is used today in the Azure cloud with CoCo VMs to run a paravisor, and there may be other uses in the future. See [1] if you want more details on VSM and VTLs. Also [2] for the CoCo VM use case.

Ideally, a Linux kernel image could detect at runtime what VTL it is running at, and "do the right thing". Unfortunately, on x86 Linux this has proved difficult (or perhaps impossible) because the amount of boot-time setup required to ask the question about the current VTL is significant. The idiosyncrasies and historical baggage of x86 requires that Linux do some x86-specific initialization steps for VTL > 0 before the question can be asked. Hence the introduction of CONFIG_HYPERV_VTL_MODE, and the behavior that when it is selected, the kernel image won't run normally in VTL 0.

I'll go out on a limb and say that I suspect on arm64 a runtime determination based on querying the VTL *could* be made (though I'm not the person writing the code). But taking advantage of that on arm64 produces an undesirable dichotomy with x86.

Roman may have further thoughts on the topic, but that's what I know about how we got here.

Michael

[1] https://learn.microsoft.com/en-us/virtualization/hyper-v-on-windows/tlfs/vsm
[2] https://techcommunity.microsoft.com/blog/windowsosplatform/openhcl-the-new-open-source-paravisor/4273172

Reviewed-by on v5 5/11
#24 lore raw

RE: [PATCH hyperv-next v5 05/11] arm64: hyperv: Initialize the Virtual Trust Level field

In reply to Roman Kisel

From: Roman Kisel <[email protected]> Sent: Friday, March 7, 2025 2:03 PM

Quoted 26 lines

Various parts of the hyperv code need to know what VTL the kernel runs at, most notably VMBus needs that to establish communication with the host.

Initialize the Virtual Trust Level field to enable booting in the Virtual Trust Level.

---
 arch/arm64/hyperv/mshyperv.c | 1 +
 1 file changed, 1 insertion(+)
arch/arm64/hyperv/mshyperv.c +1 -1
index c647db56fd6b..a7db03f5413d 100644
@@ -107,6 +107,7 @@ static int __init hyperv_init(void)
 
 	if (ms_hyperv.priv_high & HV_ACCESS_PARTITION_ID)
 		hv_get_partition_id();
+	ms_hyperv.vtl = get_vtl();
 
 	ms_hyperv_late_init();
 
--
Quoted 2 lines

2.43.0

Reviewed-by on v5 8/11
#40 lore raw

RE: [PATCH hyperv-next v5 08/11] Drivers: hv: vmbus: Get the IRQ number from DeviceTree

In reply to Roman Kisel

From: Roman Kisel <[email protected]> Sent: Friday, March 7, 2025 2:03 PM

Quoted 69 lines

The VMBus driver uses ACPI for interrupt assignment on arm64 hence it won't function in the VTL mode where only DeviceTree can be used.

Update the VMBus driver to discover interrupt configuration from DT.

---
 drivers/hv/vmbus_drv.c | 36 ++++++++++++++++++++++++++++++++++++
 1 file changed, 36 insertions(+)
drivers/hv/vmbus_drv.c +36 -1
index 75eb1390b45c..c8474b48dcd2 100644
@@ -2345,6 +2345,36 @@ static int vmbus_acpi_add(struct platform_device *pdev)
 }
 #endif
 
+static int __maybe_unused vmbus_set_irq(struct platform_device *pdev)
+{
+	struct irq_data *data;
+	int irq;
+	irq_hw_number_t hwirq;
+
+	irq = platform_get_irq(pdev, 0);
+	if (irq == 0) {
+		pr_err("VMBus interrupt mapping failure\n");
+		return -EINVAL;
+	}
+	if (irq < 0) {
+		pr_err("VMBus interrupt data can't be read from DeviceTree, error %d\n", irq);
+		return irq;
+	}
+
+	data = irq_get_irq_data(irq);
+	if (!data) {
+		pr_err("No interrupt data for VMBus virq %d\n", irq);
+		return -ENODEV;
+	}
+	hwirq = irqd_to_hwirq(data);
+
+	vmbus_irq = irq;
+	vmbus_interrupt = hwirq;
+	pr_debug("VMBus virq %d, hwirq %d\n", vmbus_irq, vmbus_interrupt);
+
+	return 0;
+}
+
 static int vmbus_device_add(struct platform_device *pdev)
 {
 	struct resource **cur_res = &hyperv_mmio;
@@ -2359,6 +2389,12 @@ static int vmbus_device_add(struct platform_device *pdev)
 	if (ret)
 		return ret;
 
+#ifndef HYPERVISOR_CALLBACK_VECTOR
+	ret = vmbus_set_irq(pdev);
+	if (ret)
+		return ret;
+#endif
+
 	for_each_of_range(&parser, &range) {
 		struct resource *res;
 
--
Quoted 2 lines

2.43.0

Modulo Arnd's suggestion for avoiding the #ifndef,

Reviewed-by on v5 9/11
#48 lore raw

RE: [PATCH hyperv-next v5 09/11] Drivers: hv: vmbus: Introduce hv_get_vmbus_root_device()

In reply to Roman Kisel

From: Roman Kisel <[email protected]> Sent: Friday, March 7, 2025 2:03 PM

Quoted 9 lines

The ARM64 PCI code for hyperv needs to know the VMBus root device, and it is private.

Provide a function that returns it. Rename it from "hv_dev" as "hv_dev" as a symbol is very overloaded. No functional changes.

Quoted 106 lines
---
 drivers/hv/vmbus_drv.c | 23 +++++++++++++++--------
 include/linux/hyperv.h |  2 ++
 2 files changed, 17 insertions(+), 8 deletions(-)
drivers/hv/vmbus_drv.c +15 -8
index c8474b48dcd2..7bfafe702963 100644
@@ -45,7 +45,8 @@ struct vmbus_dynid {
 	struct hv_vmbus_device_id id;
 };
 
-static struct device  *hv_dev;
+/* VMBus Root Device */
+static struct device  *vmbus_root_device;
 
 static int hyperv_cpuhp_online;
 
@@ -80,9 +81,15 @@ static struct resource *fb_mmio;
 static struct resource *hyperv_mmio;
 static DEFINE_MUTEX(hyperv_mmio_lock);
 
+struct device *hv_get_vmbus_root_device(void)
+{
+	return vmbus_root_device;
+}
+EXPORT_SYMBOL_GPL(hv_get_vmbus_root_device);
+
 static int vmbus_exists(void)
 {
-	if (hv_dev == NULL)
+	if (vmbus_root_device == NULL)
 		return -ENODEV;
 
 	return 0;
@@ -861,7 +868,7 @@ static int vmbus_dma_configure(struct device *child_device)
 	 * On x86/x64 coherence is assumed and these calls have no effect.
 	 */
 	hv_setup_dma_ops(child_device,
-		device_get_dma_attr(hv_dev) == DEV_DMA_COHERENT);
+		device_get_dma_attr(vmbus_root_device) == DEV_DMA_COHERENT);
 	return 0;
 }
 
@@ -1930,7 +1937,7 @@ int vmbus_device_register(struct hv_device *child_device_obj)
 		     &child_device_obj->channel->offermsg.offer.if_instance);
 
 	child_device_obj->device.bus = &hv_bus;
-	child_device_obj->device.parent = hv_dev;
+	child_device_obj->device.parent = vmbus_root_device;
 	child_device_obj->device.release = vmbus_device_release;
 
 	child_device_obj->device.dma_parms = &child_device_obj->dma_parms;
@@ -2292,7 +2299,7 @@ static int vmbus_acpi_add(struct platform_device *pdev)
 	struct acpi_device *ancestor;
 	struct acpi_device *device = ACPI_COMPANION(&pdev->dev);
 
-	hv_dev = &device->dev;
+	vmbus_root_device = &device->dev;
 
 	/*
 	 * Older versions of Hyper-V for ARM64 fail to include the _CCA
@@ -2383,7 +2390,7 @@ static int vmbus_device_add(struct platform_device *pdev)
 	struct device_node *np = pdev->dev.of_node;
 	int ret;
 
-	hv_dev = &pdev->dev;
+	vmbus_root_device = &pdev->dev;
 
 	ret = of_range_parser_init(&parser, np);
 	if (ret)
@@ -2702,7 +2709,7 @@ static int __init hv_acpi_init(void)
 	if (ret)
 		return ret;
 
-	if (!hv_dev) {
+	if (!vmbus_root_device) {
 		ret = -ENODEV;
 		goto cleanup;
 	}
@@ -2733,7 +2740,7 @@ static int __init hv_acpi_init(void)
 
 cleanup:
 	platform_driver_unregister(&vmbus_platform_driver);
-	hv_dev = NULL;
+	vmbus_root_device = NULL;
 	return ret;
 }
 
include/linux/hyperv.h +2 -1
index 7f4f8d8bdf43..1f0851fde041 100644
@@ -1333,6 +1333,8 @@ static inline void *hv_get_drvdata(struct hv_device *dev)
 	return dev_get_drvdata(&dev->device);
 }
 
+struct device *hv_get_vmbus_root_device(void);
+
 struct hv_ring_buffer_debug_info {
 	u32 current_interrupt_mask;
 	u32 current_read_index;
--

2.43.0

Reviewed-by on v5 10/11
#50 lore raw

RE: [PATCH hyperv-next v5 10/11] ACPI: irq: Introduce acpi_get_gsi_dispatcher()

In reply to Roman Kisel

From: Roman Kisel <[email protected]> Sent: Friday, March 7, 2025 2:03 PM

Quoted 41 lines

Using acpi_irq_create_hierarchy() in the cases where the code also handles OF leads to code duplication as the ACPI subsystem doesn't provide means to compute the IRQ domain parent whereas the OF does.

Introduce acpi_get_gsi_dispatcher() so that the drivers relying on both ACPI and OF may use irq_domain_create_hierarchy() in the common code paths.

No functional changes.

---
 drivers/acpi/irq.c   | 14 ++++++++++++--
 include/linux/acpi.h |  5 ++++-
 2 files changed, 16 insertions(+), 3 deletions(-)
drivers/acpi/irq.c +2 -2
index 1687483ff319..6243db610137 100644
@@ -12,7 +12,7 @@
 
 enum acpi_irq_model_id acpi_irq_model;
 
-static struct fwnode_handle *(*acpi_get_gsi_domain_id)(u32 gsi);
+static acpi_gsi_domain_disp_fn acpi_get_gsi_domain_id;
 static u32 (*acpi_gsi_to_irq_fallback)(u32 gsi);
 
 /**
@@ -307,12 +307,22 @@ EXPORT_SYMBOL_GPL(acpi_irq_get);
  *	for a given GSI
  */
 void __init acpi_set_irq_model(enum acpi_irq_model_id model,
-			       struct fwnode_handle *(*fn)(u32))
+	acpi_gsi_domain_disp_fn fn)
 {
 	acpi_irq_model = model;
 	acpi_get_gsi_domain_id = fn;
 }

drivers/acpi/irq.c EXPORT_SYMBOL_GPL(acpi_irq_get); · line 324

Quoted 10 lines
drivers/acpi/irq.c +9 -0
 
+/**
+ * acpi_get_gsi_dispatcher - Returns dispatcher function that
+ *                           computes the domain fwnode for a
+ *                           given GSI.
+ */
+acpi_gsi_domain_disp_fn acpi_get_gsi_dispatcher(void)
+{
+	return acpi_get_gsi_domain_id;
+}

This new function is needed by pci-hyperv.c. It will need to be marked as EXPORT_SYMBOL_GPL since pci-hyperv.c can be built as a module.

Quoted 23 lines
drivers/acpi/irq.c +1 -0
+
 /**
  * acpi_set_gsi_to_irq_fallback - Register a GSI transfer
  * callback to fallback to arch specified implementation.
include/linux/acpi.h +0 -0
index 4e495b29c640..abc51288e867 100644
@@ -336,8 +336,11 @@ int acpi_register_gsi (struct device *dev, u32 gsi, int triggering, int

polarity
int acpi_gsi_to_irq (u32 gsi, unsigned int *irq); int acpi_isa_irq_to_gsi (unsigned isa_irq, u32 *gsi);

include/linux/acpi.h +4 -2
+typedef struct fwnode_handle *(*acpi_gsi_domain_disp_fn)(u32);
+
 void acpi_set_irq_model(enum acpi_irq_model_id model,
-			struct fwnode_handle *(*)(u32));
+	acpi_gsi_domain_disp_fn fn);
+acpi_gsi_domain_disp_fn acpi_get_gsi_dispatcher(void);
 void acpi_set_gsi_to_irq_fallback(u32 (*)(u32));
 
 struct irq_domain *acpi_irq_create_hierarchy(unsigned int flags,
--
Quoted 2 lines

2.43.0

I'm not at all expert in ACPI code and IRQ domains, but the changes here look reasonable to me. Modulo adding the EXPORT_SYMBOL_GPL,

#54 lore raw

RE: [PATCH hyperv-next v5 11/11] PCI: hv: Get vPCI MSI IRQ domain from DeviceTree

In reply to Roman Kisel

From: Roman Kisel <[email protected]> Sent: Friday, March 7, 2025 2:03 PM

Quoted 56 lines

The hyperv-pci driver uses ACPI for MSI IRQ domain configuration on arm64. It won't be able to do that in the VTL mode where only DeviceTree can be used.

Update the hyperv-pci driver to get vPCI MSI IRQ domain in the DeviceTree case, too.

---
 drivers/pci/controller/pci-hyperv.c | 79 ++++++++++++++++++++++++++---
 1 file changed, 73 insertions(+), 6 deletions(-)
drivers/pci/controller/pci-hyperv.c +18 -3
index 6084b38bdda1..9740006a8c73 100644
@@ -50,6 +50,7 @@
 #include <linux/irqdomain.h>
 #include <linux/acpi.h>
 #include <linux/sizes.h>
+#include <linux/of_irq.h>
 #include <asm/mshyperv.h>
 
 /*
@@ -817,9 +818,17 @@ static int hv_pci_vec_irq_gic_domain_alloc(struct irq_domain *domain,
 	int ret;
 
 	fwspec.fwnode = domain->parent->fwnode;
-	fwspec.param_count = 2;
-	fwspec.param[0] = hwirq;
-	fwspec.param[1] = IRQ_TYPE_EDGE_RISING;
+	if (is_of_node(fwspec.fwnode)) {
+		/* SPI lines for OF translations start at offset 32 */
+		fwspec.param_count = 3;
+		fwspec.param[0] = 0;
+		fwspec.param[1] = hwirq - 32;
+		fwspec.param[2] = IRQ_TYPE_EDGE_RISING;
+	} else {
+		fwspec.param_count = 2;
+		fwspec.param[0] = hwirq;
+		fwspec.param[1] = IRQ_TYPE_EDGE_RISING;
+	}
 
 	ret = irq_domain_alloc_irqs_parent(domain, virq, 1, &fwspec);
 	if (ret)
@@ -887,10 +896,53 @@ static const struct irq_domain_ops hv_pci_domain_ops = {
 	.activate = hv_pci_vec_irq_domain_activate,
 };
 
+#ifdef CONFIG_OF
+
+static struct irq_domain *hv_pci_of_irq_domain_parent(void)
+{
+	struct device_node *parent;
+	struct irq_domain *domain;

drivers/pci/controller/pci-hyperv.c static const struct irq_domain_ops hv_pci_domain_ops = { · line 914

Quoted 10 lines
drivers/pci/controller/pci-hyperv.c +10 -0
+
+	parent = of_irq_find_parent(hv_get_vmbus_root_device()->of_node);
+	domain = NULL;
+	if (parent) {
+		domain = irq_find_host(parent);
+		of_node_put(parent);
+	}
+
+	return domain;
+}

You could insert the following and avoid the #ifdef's in hv_pci_irqchip_init():

#else
static struct irq_domain *hv_pci_of_irq_domain_parent(void) {return NULL;}

Quoted 13 lines
drivers/pci/controller/pci-hyperv.c +13 -0
+
+#endif
+
+#ifdef CONFIG_ACPI
+
+static struct irq_domain *hv_pci_acpi_irq_domain_parent(void)
+{
+	struct irq_domain *domain;
+	acpi_gsi_domain_disp_fn gsi_domain_disp_fn;
+
+	if (acpi_irq_model != ACPI_IRQ_MODEL_GIC)
+		return NULL;
+	gsi_domain_disp_fn = acpi_get_gsi_dispatcher();

drivers/pci/controller/pci-hyperv.c static const struct irq_domain_ops hv_pci_domain_ops = { · line 937

Quoted 10 lines
drivers/pci/controller/pci-hyperv.c +10 -0
+	if (!gsi_domain_disp_fn)
+		return NULL;
+	domain = irq_find_matching_fwnode(gsi_domain_disp_fn(0),
+				     DOMAIN_BUS_ANY);
+
+	if (!domain)
+		return NULL;
+
+	return domain;
+}

Same here:

#else
static struct irq_domain *hv_pci_acpi_irq_domain_parent(void) {return NULL;}

I don't know if it's better or not. Just a suggestion -- keep what you have if you prefer.

Quoted 40 lines
drivers/pci/controller/pci-hyperv.c +22 -4
+
+#endif
+
 static int hv_pci_irqchip_init(void)
 {
 	static struct hv_pci_chip_data *chip_data;
 	struct fwnode_handle *fn = NULL;
+	struct irq_domain *irq_domain_parent = NULL;
 	int ret = -ENOMEM;
 
 	chip_data = kzalloc(sizeof(*chip_data), GFP_KERNEL);
@@ -907,9 +959,24 @@ static int hv_pci_irqchip_init(void)
 	 * way to ensure that all the corresponding devices are also gone and
 	 * no interrupts will be generated.
 	 */
-	hv_msi_gic_irq_domain = acpi_irq_create_hierarchy(0, HV_PCI_MSI_SPI_NR,
-							  fn, &hv_pci_domain_ops,
-							  chip_data);
+#ifdef CONFIG_ACPI
+	if (!acpi_disabled)
+		irq_domain_parent = hv_pci_acpi_irq_domain_parent();
+#endif
+#if defined(CONFIG_OF)
+	if (!irq_domain_parent)
+		irq_domain_parent = hv_pci_of_irq_domain_parent();
+#endif
+	if (!irq_domain_parent) {
+		WARN_ONCE(1, "Invalid firmware configuration for VMBus interrupts\n");
+		ret = -EINVAL;
+		goto free_chip;
+	}
+
+	hv_msi_gic_irq_domain = irq_domain_create_hierarchy(
+		irq_domain_parent, 0, HV_PCI_MSI_SPI_NR,
+		fn, &hv_pci_domain_ops,
+		chip_data);
 
 	if (!hv_msi_gic_irq_domain) {
 		pr_err("Failed to create Hyper-V arm64 vPCI MSI IRQ domain\n");
--
Quoted 2 lines

2.43.0

Overall, I think the code looks right, though I'm not an expert in this area. I'll give my Reviewed-by: once Bjorn's coding suggestions are incorporated.

Michael

#5 lore raw

Re: [PATCH hyperv-next v5 01/11] arm64: kvm, smccc: Introduce and use API for detectting hypervisor presence

In reply to Roman Kisel
kernel test robot noticed the following build warnings:

Hi Roman,

kernel test robot noticed the following build warnings:

[auto build test WARNING on 3a7f7785eae7cf012af128ca9e383c91e4955354]

url:    https://github.com/intel-lab-lkp/linux/commits/Roman-Kisel/arm64-kvm-smccc-Introduce-and-use-API-for-detectting-hypervisor-presence/20250308-060639
base:   3a7f7785eae7cf012af128ca9e383c91e4955354
patch link:    https://lore.kernel.org/r/20250307220304.247725-2-romank%40linux.microsoft.com
patch subject: [PATCH hyperv-next v5 01/11] arm64: kvm, smccc: Introduce and use API for detectting hypervisor presence
config: arm64-randconfig-r131-20250312 (https://download.01.org/0day-ci/archive/20250312/[email protected]/config)
compiler: aarch64-linux-gcc (GCC) 14.2.0
reproduce: (https://download.01.org/0day-ci/archive/20250312/[email protected]/reproduce)

If you fix the issue in a separate patch/commit (i.e. not just a new version of the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <[email protected]>
| Closes: https://lore.kernel.org/oe-kbuild-all/[email protected]/

sparse warnings: (new ones prefixed by >>)
drivers/firmware/smccc/smccc.c:19:21: sparse: sparse: symbol 'smccc_soc_id_version' was not declared. Should it be static? drivers/firmware/smccc/smccc.c:20:21: sparse: sparse: symbol 'smccc_soc_id_revision' was not declared. Should it be static?

>> drivers/firmware/smccc/smccc.c:81:37: sparse: sparse: restricted __le32 degrades to integer
>> drivers/firmware/smccc/smccc.c:81:37: sparse: sparse: restricted __le32 degrades to integer
>> drivers/firmware/smccc/smccc.c:81:37: sparse: sparse: restricted __le32 degrades to integer
>> drivers/firmware/smccc/smccc.c:81:37: sparse: sparse: restricted __le32 degrades to integer
>> drivers/firmware/smccc/smccc.c:81:37: sparse: sparse: restricted __le32 degrades to integer
>> drivers/firmware/smccc/smccc.c:81:37: sparse: sparse: restricted __le32 degrades to integer
>> drivers/firmware/smccc/smccc.c:81:37: sparse: sparse: restricted __le32 degrades to integer
>> drivers/firmware/smccc/smccc.c:81:37: sparse: sparse: restricted __le32 degrades to integer
>> drivers/firmware/smccc/smccc.c:81:37: sparse: sparse: restricted __le32 degrades to integer
>> drivers/firmware/smccc/smccc.c:81:37: sparse: sparse: restricted __le32 degrades to integer
>> drivers/firmware/smccc/smccc.c:81:37: sparse: sparse: restricted __le32 degrades to integer
>> drivers/firmware/smccc/smccc.c:81:37: sparse: sparse: restricted __le32 degrades to integer
>> drivers/firmware/smccc/smccc.c:81:37: sparse: sparse: restricted __le32 degrades to integer
>> drivers/firmware/smccc/smccc.c:81:37: sparse: sparse: restricted __le32 degrades to integer
>> drivers/firmware/smccc/smccc.c:81:37: sparse: sparse: restricted __le32 degrades to integer
>> drivers/firmware/smccc/smccc.c:81:37: sparse: sparse: restricted __le32 degrades to integer

vim +81 drivers/firmware/smccc/smccc.c

    69	
    70	bool arm_smccc_hyp_present(const uuid_t *hyp_uuid)
    71	{
    72		struct arm_smccc_res res = {};
    73	
    74		if (arm_smccc_1_1_get_conduit() != SMCCC_CONDUIT_HVC)
    75			return false;
    76		arm_smccc_1_1_hvc(ARM_SMCCC_VENDOR_HYP_CALL_UID_FUNC_ID, &res);
    77		if (res.a0 == SMCCC_RET_NOT_SUPPORTED)
    78			return false;
    79	
    80		return ({
  > 81			const uuid_t uuid = SMCCC_RES_TO_UUID(res.a0, res.a1, res.a2, res.a3);
    82			const bool present = uuid_equal(&uuid, hyp_uuid);
    83	
    84			present;
    85		});
    86	}
    87	EXPORT_SYMBOL_GPL(arm_smccc_hyp_present);
    88	
Signature · 2 lines
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki
#15 lore raw

Re: [PATCH hyperv-next v5 03/11] Drivers: hv: Enable VTL mode for arm64

In reply to Michael Kelley
Quoted 42 lines · Michael Kelley

From: Arnd Bergmann <[email protected]> Sent: Monday, March 10, 2025 2:21 PM

Quoted 41 lines · Michael Kelley
Quoted 27 lines

From: Arnd Bergmann <[email protected]> Sent: Saturday, March 8, 2025 1:05 PM

Quoted 13 lines

drivers/hv/Kconfig menu "Microsoft Hyper-V guest support" · line 23

Quoted 9 lines

config HYPERV_VTL_MODE
bool "Enable Linux to boot in VTL context"

drivers/hv/Kconfig +3 -1
-	depends on X86_64 && HYPERV
+	depends on (X86_64 || ARM64)
  	depends on SMP
+	select OF_EARLY_FLATTREE
+	select OF
  	default n
  	help

Having the dependency below the top-level Kconfig entry feels a little counterintuitive. You could flip that back as it was before by doing

Quoted 5 lines

select HYPERV_VTL_MODE if !ACPI
depends on ACPI || SMP

in the HYPERV option, leaving the dependency on HYPERV in HYPERV_VTL_MODE.

I would argue that we don't ever want to implicitly select HYPERV_VTL_MODE because of some other config setting or lack thereof. VTL mode is enough of a special case that it should only be explicitly selected. If someone omits ACPI, then HYPERV should not be selectable unless HYPERV_VTL_MODE is explicitly selected.

Quoted 4 lines

The last line of the comment for HYPERV_VTL_MODE says
"A kernel built with this option must run at VTL2, and will not run as a normal guest." In other words, don't choose this unless you 100% know that VTL2 is what you want.

It sounds like the latter is the real problem: enabling a feature should never prevent something else from working. Can you describe what VTL context is and why it requires an exception to a rather fundamental rule here? If you build a kernel that runs on every single piece of arm64 hardware and every hypervisor, why can't you add HYPERV_VTL_MODE to that as an option?

In the VTL mode, we're running the kernel as secure firmware inside the guest (one might see VTL2 working as Intel SMM or Secure World on ARM).

[...]

Quoted 11 lines

Ideally, a Linux kernel image could detect at runtime what VTL it is running at, and "do the right thing". Unfortunately, on x86 Linux this has proved difficult (or perhaps impossible) because the amount of boot-time setup required to ask the question about the current VTL is significant. The idiosyncrasies and historical baggage of x86 requires that Linux do some x86-specific initialization steps for VTL > 0 before the question can be asked. Hence the introduction of CONFIG_HYPERV_VTL_MODE, and the behavior that when it is selected, the kernel image won't run normally in VTL 0.

Quoted 4 lines

I'll go out on a limb and say that I suspect on arm64 a runtime determination based on querying the VTL *could* be made (though I'm not the person writing the code). But taking advantage of that on arm64 produces an undesirable dichotomy with x86.

On arm64 that is much easier, I agree. On x86 we'd need a kludge of

static void __naked __init __aligned(4096) early_hvcall_pg(void)

{
	/*
	 * Fill the early hvcall page with `0xF1` aka `INT1` to catch
	 * programming errors. The hypervisor will overlay the page with
	 * the vendor-specific code sequences to make hypercalls on x86(_64).
	 */
	asm (".skip 4096, 0xf1");
}

static u8 __init early_hvcall_pg_input[4096] __attribute__((aligned(4096))); static u8 __init early_hvcall_pg_output[4096] __attribute__((aligned(4096)));

static void __init early_connect_to_hv(void)

{
	union hv_x64_msr_hypercall_contents hypercall_msr;
	u64 guest_id;

	guest_id = hv_generate_guest_id(LINUX_VERSION_CODE);
	wrmsrl(HV_X64_MSR_GUEST_OS_ID, guest_id);
	rdmsrl(HV_X64_MSR_HYPERCALL, hypercall_msr.as_uint64);
	hypercall_msr.enable = 1;
	hypercall_msr.guest_physical_address = 
__phys_to_pfn(virt_to_phys(early_hvcall_pg));
	wrmsrl(HV_X64_MSR_HYPERCALL, hypercall_msr.as_uint64);
}

or variations thereof.

What's very nice about arm64 in this case at least, it's got SMCCC, hvc, OF/DT and a history of options of being power-efficient and embedded. Conversely, on x86(_64) the code sequences for hyeprcalls vary from the first vendor to the second one so we have to have the hvcall page to make this regular in the code. Support for OF/DT on x86 was added for Intel set top boxes (MID, ~2015 iirc), and it took a bit of huffing and puffing to make that work for us on the large/NUMA systems (and there might be something about supporting x2apic that had to be figured out).

All told, we can have nicer things in our arm64 code yet diverging the code much from x86(_64) is not very desirable. I am not sure yet what the tradeoff should be, and my knowledge of Kconfig is rather basic. Certainly I cannot propose to arm64 maintainers that we'd like to do quirky things in Kconfig because of x86(-64), legacy specs, etc.

Perhaps, we could go back to the V2's option of

config HYPERV
tristate "Microsoft Hyper-V client drivers" depends on (X86 && X86_LOCAL_APIC && HYPERVISOR_GUEST) \

-		|| (ACPI && ARM64 && !CPU_BIG_ENDIAN)
+		|| (ARM64 && !CPU_BIG_ENDIAN)

select PARAVIRT
select X86_HV_CALLBACK_VECTOR if X86
select OF_EARLY_FLATTREE if OF
@@ -15,7 +15,7 @@ config HYPERV

config HYPERV_VTL_MODE
bool "Enable Linux to boot in VTL context"
- depends on X86_64 && HYPERV
+ depends on HYPERV
depends on SMP
default n
help
@@ -31,7 +31,7 @@ config HYPERV_VTL_MODE

Select this option to build a Linux kernel to run at a VTL other than the normal VTL0, which currently is only VTL2. This option
- initializes the x86 platform for VTL2, and adds the ability to boot
+ initializes the kernel to run in VTL2, and adds the ability to boot
secondary CPUs directly into 64-bit context as required for VTLs other than 0. A kernel built with this option must run at VTL2, and will not run as a normal guest.

That's a minimal extension, its surprise factor is very low. It has not been seen to cause issues. If no one has strong opinions against that, I'd send that in V6.

Quoted 8 lines

Roman may have further thoughts on the topic, but that's what I know about how we got here.

Michael

[1] https://learn.microsoft.com/en-us/virtualization/hyper-v-on-windows/tlfs/vsm
[2] https://techcommunity.microsoft.com/blog/windowsosplatform/openhcl-the-new-open-source-paravisor/4273172

Signature · 2 lines
Thank you,
Roman
maintainer
#16 lore raw

Re: [PATCH hyperv-next v5 03/11] Drivers: hv: Enable VTL mode for arm64

Quoted 6 lines · Roman Kisel

On 3/10/2025 3:18 PM, Michael Kelley wrote:

That's a minimal extension, its surprise factor is very low. It has not been seen to cause issues. If no one has strong opinions against that, I'd send that in V6.

Works for me. Thanks for your detailed explanations.

Arnd

maintainer
#19 lore raw

Re: [PATCH hyperv-next v5 03/11] Drivers: hv: Enable VTL mode for arm64

In reply to Roman Kisel
Quoted 97 lines · Roman Kisel
Quoted 42 lines · Michael Kelley

From: Arnd Bergmann <[email protected]> Sent: Monday, March 10, 2025 2:21 PM

Quoted 41 lines · Michael Kelley
Quoted 27 lines

From: Arnd Bergmann <[email protected]> Sent: Saturday, March 8, 2025 1:05 PM

Quoted 13 lines

drivers/hv/Kconfig menu "Microsoft Hyper-V guest support" · line 23

Quoted 9 lines

config HYPERV_VTL_MODE
bool "Enable Linux to boot in VTL context"

drivers/hv/Kconfig +3 -1
-	depends on X86_64 && HYPERV
+	depends on (X86_64 || ARM64)
  	depends on SMP
+	select OF_EARLY_FLATTREE
+	select OF
  	default n
  	help

Having the dependency below the top-level Kconfig entry feels a little counterintuitive. You could flip that back as it was before by doing

Quoted 5 lines

select HYPERV_VTL_MODE if !ACPI
depends on ACPI || SMP

in the HYPERV option, leaving the dependency on HYPERV in HYPERV_VTL_MODE.

I would argue that we don't ever want to implicitly select HYPERV_VTL_MODE because of some other config setting or lack thereof. VTL mode is enough of a special case that it should only be explicitly selected. If someone omits ACPI, then HYPERV should not be selectable unless HYPERV_VTL_MODE is explicitly selected.

Quoted 4 lines

The last line of the comment for HYPERV_VTL_MODE says
"A kernel built with this option must run at VTL2, and will not run as a normal guest." In other words, don't choose this unless you 100% know that VTL2 is what you want.

It sounds like the latter is the real problem: enabling a feature should never prevent something else from working. Can you describe what VTL context is and why it requires an exception to a rather fundamental rule here? If you build a kernel that runs on every single piece of arm64 hardware and every hypervisor, why can't you add HYPERV_VTL_MODE to that as an option?

In the VTL mode, we're running the kernel as secure firmware inside the guest (one might see VTL2 working as Intel SMM or Secure World on ARM).

[...]

Quoted 11 lines

Ideally, a Linux kernel image could detect at runtime what VTL it is running at, and "do the right thing". Unfortunately, on x86 Linux this has proved difficult (or perhaps impossible) because the amount of boot-time setup required to ask the question about the current VTL is significant. The idiosyncrasies and historical baggage of x86 requires that Linux do some x86-specific initialization steps for VTL > 0 before the question can be asked. Hence the introduction of CONFIG_HYPERV_VTL_MODE, and the behavior that when it is selected, the kernel image won't run normally in VTL 0.

Quoted 4 lines

I'll go out on a limb and say that I suspect on arm64 a runtime determination based on querying the VTL *could* be made (though I'm not the person writing the code). But taking advantage of that on arm64 produces an undesirable dichotomy with x86.

On arm64 that is much easier, I agree. On x86 we'd need a kludge of

static void __naked __init __aligned(4096) early_hvcall_pg(void)

{
	/*
	 * Fill the early hvcall page with `0xF1` aka `INT1` to catch
	 * programming errors. The hypervisor will overlay the page with
	 * the vendor-specific code sequences to make hypercalls on x86(_64).
	 */
	asm (".skip 4096, 0xf1");
}

static u8 __init early_hvcall_pg_input[4096] __attribute__((aligned(4096))); static u8 __init early_hvcall_pg_output[4096] __attribute__((aligned(4096)));

static void __init early_connect_to_hv(void)

{
	union hv_x64_msr_hypercall_contents hypercall_msr;
	u64 guest_id;

	guest_id = hv_generate_guest_id(LINUX_VERSION_CODE);
	wrmsrl(HV_X64_MSR_GUEST_OS_ID, guest_id);
	rdmsrl(HV_X64_MSR_HYPERCALL, hypercall_msr.as_uint64);
	hypercall_msr.enable = 1;
	hypercall_msr.guest_physical_address =
__phys_to_pfn(virt_to_phys(early_hvcall_pg));
	wrmsrl(HV_X64_MSR_HYPERCALL, hypercall_msr.as_uint64);
}
Quoted 1 line

or variations thereof.

OT here but what's stopping us from doing this on x86?

It seems to me there is some value in setting up the hypercall page as early as possible. The same page can be used through the lifetime of the partition. The early input and output pages should be reclaimed.

Also, since the hypervisor will insert an overlay page, it makes sense to not allocate a page from Linux at all. When I ported Xen to run as a guest on Hyper-V, I used that approach. The setup worked just fine.

All being said, things work today, so I'm in no hurry to change things.

Wei.

#6 lore raw

Re: [PATCH hyperv-next v5 01/11] arm64: kvm, smccc: Introduce and use API for detectting hypervisor presence

In reply to Roman Kisel
kernel test robot noticed the following build warnings:

Hi Roman,

kernel test robot noticed the following build warnings:

[auto build test WARNING on 3a7f7785eae7cf012af128ca9e383c91e4955354]

url:    https://github.com/intel-lab-lkp/linux/commits/Roman-Kisel/arm64-kvm-smccc-Introduce-and-use-API-for-detectting-hypervisor-presence/20250308-060639
base:   3a7f7785eae7cf012af128ca9e383c91e4955354
patch link:    https://lore.kernel.org/r/20250307220304.247725-2-romank%40linux.microsoft.com
patch subject: [PATCH hyperv-next v5 01/11] arm64: kvm, smccc: Introduce and use API for detectting hypervisor presence
config: arm64-randconfig-r131-20250312 (https://download.01.org/0day-ci/archive/20250313/[email protected]/config)
compiler: aarch64-linux-gcc (GCC) 14.2.0
reproduce: (https://download.01.org/0day-ci/archive/20250313/[email protected]/reproduce)

If you fix the issue in a separate patch/commit (i.e. not just a new version of the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <[email protected]>
| Closes: https://lore.kernel.org/oe-kbuild-all/[email protected]/

sparse warnings: (new ones prefixed by >>)
drivers/firmware/smccc/smccc.c:19:21: sparse: sparse: symbol 'smccc_soc_id_version' was not declared. Should it be static? drivers/firmware/smccc/smccc.c:20:21: sparse: sparse: symbol 'smccc_soc_id_revision' was not declared. Should it be static?

>> drivers/firmware/smccc/smccc.c:81:37: sparse: sparse: restricted __le32 degrades to integer
>> drivers/firmware/smccc/smccc.c:81:37: sparse: sparse: restricted __le32 degrades to integer
>> drivers/firmware/smccc/smccc.c:81:37: sparse: sparse: restricted __le32 degrades to integer
>> drivers/firmware/smccc/smccc.c:81:37: sparse: sparse: restricted __le32 degrades to integer
>> drivers/firmware/smccc/smccc.c:81:37: sparse: sparse: restricted __le32 degrades to integer
>> drivers/firmware/smccc/smccc.c:81:37: sparse: sparse: restricted __le32 degrades to integer
>> drivers/firmware/smccc/smccc.c:81:37: sparse: sparse: restricted __le32 degrades to integer
>> drivers/firmware/smccc/smccc.c:81:37: sparse: sparse: restricted __le32 degrades to integer
>> drivers/firmware/smccc/smccc.c:81:37: sparse: sparse: restricted __le32 degrades to integer
>> drivers/firmware/smccc/smccc.c:81:37: sparse: sparse: restricted __le32 degrades to integer
>> drivers/firmware/smccc/smccc.c:81:37: sparse: sparse: restricted __le32 degrades to integer
>> drivers/firmware/smccc/smccc.c:81:37: sparse: sparse: restricted __le32 degrades to integer
>> drivers/firmware/smccc/smccc.c:81:37: sparse: sparse: restricted __le32 degrades to integer
>> drivers/firmware/smccc/smccc.c:81:37: sparse: sparse: restricted __le32 degrades to integer
>> drivers/firmware/smccc/smccc.c:81:37: sparse: sparse: restricted __le32 degrades to integer
>> drivers/firmware/smccc/smccc.c:81:37: sparse: sparse: restricted __le32 degrades to integer

vim +81 drivers/firmware/smccc/smccc.c

    69	
    70	bool arm_smccc_hyp_present(const uuid_t *hyp_uuid)
    71	{
    72		struct arm_smccc_res res = {};
    73	
    74		if (arm_smccc_1_1_get_conduit() != SMCCC_CONDUIT_HVC)
    75			return false;
    76		arm_smccc_1_1_hvc(ARM_SMCCC_VENDOR_HYP_CALL_UID_FUNC_ID, &res);
    77		if (res.a0 == SMCCC_RET_NOT_SUPPORTED)
    78			return false;
    79	
    80		return ({
  > 81			const uuid_t uuid = SMCCC_RES_TO_UUID(res.a0, res.a1, res.a2, res.a3);
    82			const bool present = uuid_equal(&uuid, hyp_uuid);
    83	
    84			present;
    85		});
    86	}
    87	EXPORT_SYMBOL_GPL(arm_smccc_hyp_present);
    88	
Signature · 2 lines
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki
#17 lore raw

Re: [PATCH hyperv-next v5 03/11] Drivers: hv: Enable VTL mode for arm64

In reply to Arnd Bergmann
Quoted 10 lines · Arnd Bergmann
Quoted 6 lines · Roman Kisel

On 3/10/2025 3:18 PM, Michael Kelley wrote:

That's a minimal extension, its surprise factor is very low. It has not been seen to cause issues. If no one has strong opinions against that, I'd send that in V6.

Works for me. Thanks for your detailed explanations.

Thank you for your review very much!

Quoted 1 line

Arnd

Signature · 2 lines
Thank you,
Roman
#20 lore raw

Re: [PATCH hyperv-next v5 03/11] Drivers: hv: Enable VTL mode for arm64

In reply to Wei Liu
Quoted 98 lines · Wei Liu
Quoted 97 lines · Roman Kisel
Quoted 42 lines · Michael Kelley

From: Arnd Bergmann <[email protected]> Sent: Monday, March 10, 2025 2:21 PM

Quoted 41 lines · Michael Kelley
Quoted 27 lines

From: Arnd Bergmann <[email protected]> Sent: Saturday, March 8, 2025 1:05 PM

Quoted 13 lines

Review

Quoted 9 lines

config HYPERV_VTL_MODE
bool "Enable Linux to boot in VTL context"

(quoted diff) +3 -1
-	depends on X86_64 && HYPERV
+	depends on (X86_64 || ARM64)
   	depends on SMP
+	select OF_EARLY_FLATTREE
+	select OF
   	default n
   	help

Having the dependency below the top-level Kconfig entry feels a little counterintuitive. You could flip that back as it was before by doing

Quoted 5 lines

select HYPERV_VTL_MODE if !ACPI
depends on ACPI || SMP

in the HYPERV option, leaving the dependency on HYPERV in HYPERV_VTL_MODE.

I would argue that we don't ever want to implicitly select HYPERV_VTL_MODE because of some other config setting or lack thereof. VTL mode is enough of a special case that it should only be explicitly selected. If someone omits ACPI, then HYPERV should not be selectable unless HYPERV_VTL_MODE is explicitly selected.

Quoted 4 lines

The last line of the comment for HYPERV_VTL_MODE says
"A kernel built with this option must run at VTL2, and will not run as a normal guest." In other words, don't choose this unless you 100% know that VTL2 is what you want.

It sounds like the latter is the real problem: enabling a feature should never prevent something else from working. Can you describe what VTL context is and why it requires an exception to a rather fundamental rule here? If you build a kernel that runs on every single piece of arm64 hardware and every hypervisor, why can't you add HYPERV_VTL_MODE to that as an option?

In the VTL mode, we're running the kernel as secure firmware inside the guest (one might see VTL2 working as Intel SMM or Secure World on ARM).

[...]

Quoted 11 lines

Ideally, a Linux kernel image could detect at runtime what VTL it is running at, and "do the right thing". Unfortunately, on x86 Linux this has proved difficult (or perhaps impossible) because the amount of boot-time setup required to ask the question about the current VTL is significant. The idiosyncrasies and historical baggage of x86 requires that Linux do some x86-specific initialization steps for VTL > 0 before the question can be asked. Hence the introduction of CONFIG_HYPERV_VTL_MODE, and the behavior that when it is selected, the kernel image won't run normally in VTL 0.

Quoted 4 lines

I'll go out on a limb and say that I suspect on arm64 a runtime determination based on querying the VTL *could* be made (though I'm not the person writing the code). But taking advantage of that on arm64 produces an undesirable dichotomy with x86.

On arm64 that is much easier, I agree. On x86 we'd need a kludge of

static void __naked __init __aligned(4096) early_hvcall_pg(void)

{
	/*
	 * Fill the early hvcall page with `0xF1` aka `INT1` to catch
	 * programming errors. The hypervisor will overlay the page with
	 * the vendor-specific code sequences to make hypercalls on x86(_64).
	 */
	asm (".skip 4096, 0xf1");
}

static u8 __init early_hvcall_pg_input[4096] __attribute__((aligned(4096))); static u8 __init early_hvcall_pg_output[4096] __attribute__((aligned(4096)));

static void __init early_connect_to_hv(void)

{
	union hv_x64_msr_hypercall_contents hypercall_msr;
	u64 guest_id;

	guest_id = hv_generate_guest_id(LINUX_VERSION_CODE);
	wrmsrl(HV_X64_MSR_GUEST_OS_ID, guest_id);
	rdmsrl(HV_X64_MSR_HYPERCALL, hypercall_msr.as_uint64);
	hypercall_msr.enable = 1;
	hypercall_msr.guest_physical_address =
__phys_to_pfn(virt_to_phys(early_hvcall_pg));
	wrmsrl(HV_X64_MSR_HYPERCALL, hypercall_msr.as_uint64);
}
Quoted 4 lines
Quoted 1 line

or variations thereof.

OT here but what's stopping us from doing this on x86?

At the first glance, seems like nothing I think. For the conf scenarios like TDX and SEV-SNP, due to the early hvcall I/O pages above allocated in BSS, might need to mark the pages as decrypted and zero them out so they look like proper BSS section (the page contents are scrambled after flipping the page encryption bit iirc).

Quoted 4 lines

It seems to me there is some value in setting up the hypercall page as early as possible. The same page can be used through the lifetime of the partition. The early input and output pages should be reclaimed.

Wholeheartedly agree!

Quoted 6 lines

Also, since the hypervisor will insert an overlay page, it makes sense to not allocate a page from Linux at all. When I ported Xen to run as a guest on Hyper-V, I used that approach. The setup worked just fine.

All being said, things work today, so I'm in no hurry to change things.

I'll try fleshing this out soon-ish if no one beats me to that :)

Quoted 1 line

Wei.

Signature · 2 lines
Thank you,
Roman
#18 lore raw

RE: [PATCH hyperv-next v5 03/11] Drivers: hv: Enable VTL mode for arm64

In reply to Roman Kisel

From: Roman Kisel <[email protected]> Sent: Wednesday, March 12, 2025 2:21 PM

Quoted 9 lines · Arnd Bergmann
Quoted 7 lines · Roman Kisel
Quoted 5 lines

That's a minimal extension, its surprise factor is very low. It has not been seen to cause issues. If no one has strong opinions against that, I'd send that in V6.

Quoted 5 lines
Quoted 2 lines

Works for me. Thanks for your detailed explanations.

Thank you for your review very much!

My original concern [1] with this minimal change is that it allows building a normal Linux kernel (i.e., not for VTL 2) for Hyper-V with CONFIG_ACPI=n. Such a kernel will not run in a Hyper-V VM since ACPI is required unless building for and running in VTL 2. Current upstream code disallows CONFIG_HYPERV=y with CONFIG_ACPI=n.

However, I don't want to make too big of a deal about now allowing this misconfiguration. Arguably it's not likely to happen, and the solution is "don't do that". So if we want to go back to the minimal set of changes to drivers/hv/Kconfig as Roman proposes, I won't object further. I just want to sure everyone is clear on the tradeoffs.

Michael

[1] https://lore.kernel.org/linux-hyperv/SN6PR02MB4157E15EFE263BBA3D8DFC51D4EC2@SN6PR02MB4157.namprd02.prod.outlook.com/

#41 lore raw

Re: [PATCH hyperv-next v5 08/11] Drivers: hv: vmbus: Get the IRQ number from DeviceTree

In reply to Roman Kisel

Hi Roman,

kernel test robot noticed the following build warnings:

url:    https://github.com/intel-lab-lkp/linux/commits/Roman-Kisel/arm64-kvm-smccc-Introduce-and-use-API-for-detectting-hypervisor-presence/20250308-060639
base:   3a7f7785eae7cf012af128ca9e383c91e4955354
patch link:    https://lore.kernel.org/r/20250307220304.247725-9-romank%40linux.microsoft.com
patch subject: [PATCH hyperv-next v5 08/11] Drivers: hv: vmbus: Get the IRQ number from DeviceTree
config: arm64-randconfig-r071-20250312 (https://download.01.org/0day-ci/archive/20250313/[email protected]/config)
compiler: aarch64-linux-gcc (GCC) 14.2.0

If you fix the issue in a separate patch/commit (i.e. not just a new version of the same patch/commit), kindly add following tags

| Reported-by: kernel test robot <[email protected]>
| Reported-by: Dan Carpenter <[email protected]>
| Closes: https://lore.kernel.org/r/[email protected]/

smatch warnings:
drivers/hv/vmbus_drv.c:2355 vmbus_set_irq() warn: platform_get_irq() does not return zero

vim +2355 drivers/hv/vmbus_drv.c

2e494cbd89c51d1 Roman Kisel    2025-03-07  2348  static int __maybe_unused vmbus_set_irq(struct platform_device *pdev)
2e494cbd89c51d1 Roman Kisel    2025-03-07  2349  {
2e494cbd89c51d1 Roman Kisel    2025-03-07  2350  	struct irq_data *data;
2e494cbd89c51d1 Roman Kisel    2025-03-07  2351  	int irq;
2e494cbd89c51d1 Roman Kisel    2025-03-07  2352  	irq_hw_number_t hwirq;
2e494cbd89c51d1 Roman Kisel    2025-03-07  2353  
2e494cbd89c51d1 Roman Kisel    2025-03-07  2354  	irq = platform_get_irq(pdev, 0);
2e494cbd89c51d1 Roman Kisel    2025-03-07 @2355  	if (irq == 0) {
2e494cbd89c51d1 Roman Kisel    2025-03-07  2356  		pr_err("VMBus interrupt mapping failure\n");
2e494cbd89c51d1 Roman Kisel    2025-03-07  2357  		return -EINVAL;

You can delete this if statement. It's dead code. See the comment next to platform_get_irq(). I wrote a blog about the history of this:
https://staticthinking.wordpress.com/2023/08/07/writing-a-check-for-zero-irq-error-codes/

2e494cbd89c51d1 Roman Kisel    2025-03-07  2358  	}
2e494cbd89c51d1 Roman Kisel    2025-03-07  2359  	if (irq < 0) {
2e494cbd89c51d1 Roman Kisel    2025-03-07  2360  		pr_err("VMBus interrupt data can't be read from DeviceTree, error %d\n", irq);
2e494cbd89c51d1 Roman Kisel    2025-03-07  2361  		return irq;
2e494cbd89c51d1 Roman Kisel    2025-03-07  2362  	}
2e494cbd89c51d1 Roman Kisel    2025-03-07  2363  
2e494cbd89c51d1 Roman Kisel    2025-03-07  2364  	data = irq_get_irq_data(irq);
2e494cbd89c51d1 Roman Kisel    2025-03-07  2365  	if (!data) {
2e494cbd89c51d1 Roman Kisel    2025-03-07  2366  		pr_err("No interrupt data for VMBus virq %d\n", irq);
2e494cbd89c51d1 Roman Kisel    2025-03-07  2367  		return -ENODEV;
2e494cbd89c51d1 Roman Kisel    2025-03-07  2368  	}
2e494cbd89c51d1 Roman Kisel    2025-03-07  2369  	hwirq = irqd_to_hwirq(data);
2e494cbd89c51d1 Roman Kisel    2025-03-07  2370  
2e494cbd89c51d1 Roman Kisel    2025-03-07  2371  	vmbus_irq = irq;
2e494cbd89c51d1 Roman Kisel    2025-03-07  2372  	vmbus_interrupt = hwirq;
2e494cbd89c51d1 Roman Kisel    2025-03-07  2373  	pr_debug("VMBus virq %d, hwirq %d\n", vmbus_irq, vmbus_interrupt);
2e494cbd89c51d1 Roman Kisel    2025-03-07  2374  
2e494cbd89c51d1 Roman Kisel    2025-03-07  2375  	return 0;
2e494cbd89c51d1 Roman Kisel    2025-03-07  2376  }
Signature · 2 lines
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki
#42 lore raw

Re: [PATCH hyperv-next v5 08/11] Drivers: hv: vmbus: Get the IRQ number from DeviceTree

Quoted 1 line · Dan Carpenter

Hi Roman,

Hi Dan,

Appreciate helping with that very much!

Signature · 2 lines
Thank you,
Roman
Quoted 57 lines

kernel test robot noticed the following build warnings:

url:    https://github.com/intel-lab-lkp/linux/commits/Roman-Kisel/arm64-kvm-smccc-Introduce-and-use-API-for-detectting-hypervisor-presence/20250308-060639
base:   3a7f7785eae7cf012af128ca9e383c91e4955354
patch link:    https://lore.kernel.org/r/20250307220304.247725-9-romank%40linux.microsoft.com
patch subject: [PATCH hyperv-next v5 08/11] Drivers: hv: vmbus: Get the IRQ number from DeviceTree
config: arm64-randconfig-r071-20250312 (https://download.01.org/0day-ci/archive/20250313/[email protected]/config)
compiler: aarch64-linux-gcc (GCC) 14.2.0

If you fix the issue in a separate patch/commit (i.e. not just a new version of the same patch/commit), kindly add following tags

| Reported-by: kernel test robot <[email protected]>
| Reported-by: Dan Carpenter <[email protected]>
| Closes: https://lore.kernel.org/r/[email protected]/

smatch warnings:
drivers/hv/vmbus_drv.c:2355 vmbus_set_irq() warn: platform_get_irq() does not return zero

vim +2355 drivers/hv/vmbus_drv.c

2e494cbd89c51d1 Roman Kisel    2025-03-07  2348  static int __maybe_unused vmbus_set_irq(struct platform_device *pdev)
2e494cbd89c51d1 Roman Kisel    2025-03-07  2349  {
2e494cbd89c51d1 Roman Kisel    2025-03-07  2350  	struct irq_data *data;
2e494cbd89c51d1 Roman Kisel    2025-03-07  2351  	int irq;
2e494cbd89c51d1 Roman Kisel    2025-03-07  2352  	irq_hw_number_t hwirq;
2e494cbd89c51d1 Roman Kisel    2025-03-07  2353
2e494cbd89c51d1 Roman Kisel    2025-03-07  2354  	irq = platform_get_irq(pdev, 0);
2e494cbd89c51d1 Roman Kisel    2025-03-07 @2355  	if (irq == 0) {
2e494cbd89c51d1 Roman Kisel    2025-03-07  2356  		pr_err("VMBus interrupt mapping failure\n");
2e494cbd89c51d1 Roman Kisel    2025-03-07  2357  		return -EINVAL;

You can delete this if statement. It's dead code. See the comment next to platform_get_irq(). I wrote a blog about the history of this:
https://staticthinking.wordpress.com/2023/08/07/writing-a-check-for-zero-irq-error-codes/

2e494cbd89c51d1 Roman Kisel    2025-03-07  2358  	}
2e494cbd89c51d1 Roman Kisel    2025-03-07  2359  	if (irq < 0) {
2e494cbd89c51d1 Roman Kisel    2025-03-07  2360  		pr_err("VMBus interrupt data can't be read from DeviceTree, error %d\n", irq);
2e494cbd89c51d1 Roman Kisel    2025-03-07  2361  		return irq;
2e494cbd89c51d1 Roman Kisel    2025-03-07  2362  	}
2e494cbd89c51d1 Roman Kisel    2025-03-07  2363
2e494cbd89c51d1 Roman Kisel    2025-03-07  2364  	data = irq_get_irq_data(irq);
2e494cbd89c51d1 Roman Kisel    2025-03-07  2365  	if (!data) {
2e494cbd89c51d1 Roman Kisel    2025-03-07  2366  		pr_err("No interrupt data for VMBus virq %d\n", irq);
2e494cbd89c51d1 Roman Kisel    2025-03-07  2367  		return -ENODEV;
2e494cbd89c51d1 Roman Kisel    2025-03-07  2368  	}
2e494cbd89c51d1 Roman Kisel    2025-03-07  2369  	hwirq = irqd_to_hwirq(data);
2e494cbd89c51d1 Roman Kisel    2025-03-07  2370
2e494cbd89c51d1 Roman Kisel    2025-03-07  2371  	vmbus_irq = irq;
2e494cbd89c51d1 Roman Kisel    2025-03-07  2372  	vmbus_interrupt = hwirq;
2e494cbd89c51d1 Roman Kisel    2025-03-07  2373  	pr_debug("VMBus virq %d, hwirq %d\n", vmbus_irq, vmbus_interrupt);
2e494cbd89c51d1 Roman Kisel    2025-03-07  2374
2e494cbd89c51d1 Roman Kisel    2025-03-07  2375  	return 0;
2e494cbd89c51d1 Roman Kisel    2025-03-07  2376  }
maintainer
#43 lore raw

Re: [PATCH hyperv-next v5 08/11] Drivers: hv: vmbus: Get the IRQ number from DeviceTree

In reply to Roman Kisel
Quoted 27 lines · Roman Kisel

The VMBus driver uses ACPI for interrupt assignment on arm64 hence it won't function in the VTL mode where only DeviceTree can be used.

Update the VMBus driver to discover interrupt configuration from DT.

---
 drivers/hv/vmbus_drv.c | 36 ++++++++++++++++++++++++++++++++++++
 1 file changed, 36 insertions(+)
drivers/hv/vmbus_drv.c +5 -0
index 75eb1390b45c..c8474b48dcd2 100644
@@ -2345,6 +2345,36 @@ static int vmbus_acpi_add(struct platform_device *pdev)
 }
 #endif
 
+static int __maybe_unused vmbus_set_irq(struct platform_device *pdev)
+{
+       struct irq_data *data;
+       int irq;
+       irq_hw_number_t hwirq;

drivers/hv/vmbus_drv.c static int vmbus_acpi_add(struct platform_device *pdev) · line 2362

Quoted 10 lines
drivers/hv/vmbus_drv.c +10 -0
+
+       irq = platform_get_irq(pdev, 0);
+       if (irq == 0) {
+               pr_err("VMBus interrupt mapping failure\n");
+               return -EINVAL;
+       }
+       if (irq < 0) {
+               pr_err("VMBus interrupt data can't be read from DeviceTree, error %d\n", irq);
+               return irq;
+       }

I don't think why you couldn't get the interrupt is important. Just check for (irq <= 0) and be done with it. I'm not even sure if returning 0 is possible now. There's a long history to that and NO_IRQ.

Rob

#44 lore raw

Re: [PATCH hyperv-next v5 08/11] Drivers: hv: vmbus: Get the IRQ number from DeviceTree

Quoted 1 line · Rob Herring

On Fri, Mar 7, 2025 at 4:03 PM Roman Kisel <[email protected]> wrote:

[...]

Quoted 10 lines

drivers/hv/vmbus_drv.c static int vmbus_acpi_add(struct platform_device *pdev) · line 2362

Quoted 9 lines
drivers/hv/vmbus_drv.c +9 -0
+       irq = platform_get_irq(pdev, 0);
+       if (irq == 0) {
+               pr_err("VMBus interrupt mapping failure\n");
+               return -EINVAL;
+       }
+       if (irq < 0) {
+               pr_err("VMBus interrupt data can't be read from DeviceTree, error %d\n", irq);
+               return irq;
+       }
Quoted 5 lines

I don't think why you couldn't get the interrupt is important. Just check for (irq <= 0) and be done with it. I'm not even sure if returning 0 is possible now. There's a long history to that and NO_IRQ.

That will certainly make the code look much better! Thank you very much for the idea!

Quoted 1 line

Rob

Signature · 2 lines
Thank you,
Roman