Skip to contentSkip to conversation

WHPX support for Arm

PATCH Superseded Mohamed Mediouni 6 participants 38 messages Aug 8, 2025 – Sep 4, 2025 Arm/Amlogic Meson SoC support

v5 of 22 Superseded Latest is v22 →

lore
Share & tools
lore
#0 lore raw

v5 cover WHPX support for Arm

Link to branch: https://github.com/mediouni-m/qemu whpx (tag for this submission: whpx-v5)

Missing features:
- PSCI state sync with Hyper-V: notably breaks reboots when multiple cores are enabled
- Interrupt controller save-restore
- SVE register sync

To do:
- Switching ITS to off by default when using HVF w/ vGIC or WHPX (will probably do when both the series are merged)
- After that, perhaps exit if ITS is explicitly enabled.

Note:

"accel/system: Introduce hwaccel_enabled() helper" taken from the mailing list, added here as part of this series to make it compilable as a whole.

"hw/arm: virt: add GICv2m for the case when ITS is not available" present in both the HVF vGIC and this series.

Updates since v4:
- Taking into account review comments
- Add migration blocker in the vGICv3 code due to missing interrupt controller save/restore
- Debug register sync

Updates since v3:
- Disabling SVE on WHPX
- Taking into account review comments incl:

- fixing x86 support
- reduce the amount of __x86_64__ checks in common code to the minimum (winhvemulation) which can be reduced even further down the road.
- generalize get_physical_address_range into something common between hvf and whpx

Updates since v2:
- Fixed up a rebase screwup for whpx-internal.h
- Fixed ID_AA64ISAR1_EL1 and ID_AA64ISAR2_EL1 feature probe for -cpu host
- Switched to ID_AA64PFR1_EL1/ID_AA64DFR0_EL1 instead of their non-AA64 variant

Updates since v1:
- Shutdowns and reboots
- MPIDR_EL1 register sync
- Fixing GICD_TYPER_LPIS value
- IPA size clamping
- -cpu host now implemented

Mohamed Mediouni (17):
hw/arm: virt: add GICv2m for the case when ITS is not available
whpx: Move around files before introducing AArch64 support
whpx: reshuffle common code
whpx: ifdef out winhvemulation on non-x86_64
whpx: common: add WHPX_INTERCEPT_DEBUG_TRAPS define
hw, target, accel: whpx: change apic_in_platform to kernel_irqchip
whpx: interrupt controller support
whpx: add arm64 support
whpx: copy over memory management logic from hvf target/arm: cpu: mark WHPX as supporting PSCI 1.1
hw/arm: virt: cleanly fail on attempt to use the platform vGIC
together with ITS
whpx: arm64: clamp down IPA size
hw/arm, accel/hvf, whpx: unify get_physical_address_range between WHPX
and HVF
whpx: arm64: implement -cpu host
target/arm: whpx: instantiate GIC early
whpx: arm64: gicv3: add migration blocker
MAINTAINERS: Add myself as a maintainer for WHPX

Philippe Mathieu-Daudé (1):
accel/system: Introduce hwaccel_enabled() helper

 MAINTAINERS                                   |    9 +-
 accel/hvf/hvf-all.c                           |    7 +-
 accel/meson.build                             |    1 +
 accel/whpx/meson.build                        |    7 +
 {target/i386 => accel}/whpx/whpx-accel-ops.c  |    6 +-
 accel/whpx/whpx-common.c                      |  666 +++++++++++
 hw/arm/virt-acpi-build.c                      |    4 +-
 hw/arm/virt.c                                 |   34 +-
 hw/i386/x86-cpu.c                             |    4 +-
 hw/intc/arm_gicv3_common.c                    |    3 +
 hw/intc/arm_gicv3_whpx.c                      |  272 +++++
 hw/intc/meson.build                           |    1 +
 include/hw/arm/virt.h                         |    2 +
 include/hw/boards.h                           |    3 +-
 include/hw/intc/arm_gicv3_common.h            |    3 +
 include/system/hvf_int.h                      |    2 +
 include/system/hw_accel.h                     |   13 +
 .../whpx => include/system}/whpx-accel-ops.h  |    4 +-
 include/system/whpx-all.h                     |   20 +
 include/system/whpx-common.h                  |   26 +
 .../whpx => include/system}/whpx-internal.h   |   14 +-
 include/system/whpx.h                         |    4 +-
 meson.build                                   |   21 +-
 target/arm/cpu.c                              |    3 +-
 target/arm/cpu64.c                            |   19 +-
 target/arm/hvf-stub.c                         |   20 -
 target/arm/hvf/hvf.c                          |    6 +-
 target/arm/hvf_arm.h                          |    3 -
 target/arm/meson.build                        |    2 +-
 target/arm/whpx/meson.build                   |    5 +
 target/arm/whpx/whpx-all.c                    | 1025 +++++++++++++++++
 target/arm/whpx/whpx-stub.c                   |   15 +
 target/arm/whpx_arm.h                         |   17 +
 target/i386/cpu-apic.c                        |    2 +-
 target/i386/hvf/hvf.c                         |   11 +
 target/i386/whpx/meson.build                  |    1 -
 target/i386/whpx/whpx-all.c                   |  569 +--------
 target/i386/whpx/whpx-apic.c                  |    2 +-
 38 files changed, 2207 insertions(+), 619 deletions(-)
 create mode 100644 accel/whpx/meson.build
 rename {target/i386 => accel}/whpx/whpx-accel-ops.c (96%)
 create mode 100644 accel/whpx/whpx-common.c
 create mode 100644 hw/intc/arm_gicv3_whpx.c
 rename {target/i386/whpx => include/system}/whpx-accel-ops.h (92%)
 create mode 100644 include/system/whpx-all.h
 create mode 100644 include/system/whpx-common.h
 rename {target/i386/whpx => include/system}/whpx-internal.h (96%)
 delete mode 100644 target/arm/hvf-stub.c
 create mode 100644 target/arm/whpx/meson.build
 create mode 100644 target/arm/whpx/whpx-all.c
 create mode 100644 target/arm/whpx/whpx-stub.c
 create mode 100644 target/arm/whpx_arm.h
Signature · 1 line
2.39.5 (Apple Git-154)
#1 lore raw

PATCH v5 1/18 accel/system: Introduce hwaccel_enabled() helper

From: Philippe Mathieu-Daudé <[email protected]>

hwaccel_enabled() return whether any hardware accelerator is enabled.

---
 include/system/hw_accel.h | 13 +++++++++++++
 1 file changed, 13 insertions(+)
include/system/hw_accel.h +13 -0
index fa9228d5d2..49556b026e 100644
@@ -39,4 +39,17 @@ void cpu_synchronize_pre_loadvm(CPUState *cpu);
 void cpu_synchronize_post_reset(CPUState *cpu);
 void cpu_synchronize_post_init(CPUState *cpu);
 
+/**
+ * hwaccel_enabled:
+ *
+ * Returns: %true if a hardware accelerator is enabled, %false otherwise.
+ */
+static inline bool hwaccel_enabled(void)
+{
+    return hvf_enabled()
+        || kvm_enabled()
+        || nvmm_enabled()
+        || whpx_enabled();
+}
+
 #endif /* QEMU_HW_ACCEL_H */
Signature · 1 line
2.39.5 (Apple Git-154)
#2 lore raw

PATCH v5 2/18 hw/arm: virt: add GICv2m for the case when ITS is not available

In reply to Mohamed Mediouni

On Hypervisor.framework for macOS and WHPX for Windows, the provided environment is a GICv3 without ITS.

As such, support a GICv3 w/ GICv2m for that scenario.

---
 hw/arm/virt-acpi-build.c | 4 +++-
 hw/arm/virt.c            | 8 ++++++++
 include/hw/arm/virt.h    | 2 ++
 3 files changed, 13 insertions(+), 1 deletion(-)
hw/arm/virt-acpi-build.c +3 -1
index b01fc4f8ef..969fa3f686 100644
@@ -848,7 +848,9 @@ build_madt(GArray *table_data, BIOSLinker *linker, VirtMachineState *vms)
             build_append_int_noprefix(table_data, memmap[VIRT_GIC_ITS].base, 8);
             build_append_int_noprefix(table_data, 0, 4);    /* Reserved */
         }
-    } else {
+    }
+
+    if (!vms->its && !vms->no_gicv3_with_gicv2m) {
         const uint16_t spi_base = vms->irqmap[VIRT_GIC_V2M] + ARM_SPI_BASE;
 
         /* 5.2.12.16 GIC MSI Frame Structure */
hw/arm/virt.c +8 -0
index ef6be3660f..5951b331f3 100644
@@ -953,6 +953,8 @@ static void create_gic(VirtMachineState *vms, MemoryRegion *mem)
 
     if (vms->gic_version != VIRT_GIC_VERSION_2 && vms->its) {
         create_its(vms);
+    } else if (vms->gic_version != VIRT_GIC_VERSION_2 && !vms->no_gicv3_with_gicv2m) {
+        create_v2m(vms);
     } else if (vms->gic_version == VIRT_GIC_VERSION_2) {
         create_v2m(vms);
     }
@@ -2402,6 +2404,8 @@ static void machvirt_init(MachineState *machine)
     vms->ns_el2_virt_timer_irq = ns_el2_virt_timer_present() &&
         !vmc->no_ns_el2_virt_timer_irq;
 
+    vms->no_gicv3_with_gicv2m = vmc->no_gicv3_with_gicv2m;
+
     fdt_add_timer_nodes(vms);
     fdt_add_cpu_nodes(vms);
 
@@ -3410,6 +3414,7 @@ static void virt_instance_init(Object *obj)
     vms->its = true;
     /* Allow ITS emulation if the machine version supports it */
     vms->tcg_its = !vmc->no_tcg_its;
+    vms->no_gicv3_with_gicv2m = false;
 
     /* Default disallows iommu instantiation */
     vms->iommu = VIRT_IOMMU_NONE;
@@ -3462,8 +3467,11 @@ DEFINE_VIRT_MACHINE_AS_LATEST(10, 1)
 
 static void virt_machine_10_0_options(MachineClass *mc)
 {
+    VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
+
     virt_machine_10_1_options(mc);
     compat_props_add(mc->compat_props, hw_compat_10_0, hw_compat_10_0_len);
+    vmc->no_gicv3_with_gicv2m = true;
 }
 DEFINE_VIRT_MACHINE(10, 0)
 
include/hw/arm/virt.h +2 -0
index 365a28b082..725ec18fd2 100644
@@ -131,6 +131,7 @@ struct VirtMachineClass {
     bool no_cpu_topology;
     bool no_tcg_lpa2;
     bool no_ns_el2_virt_timer_irq;
+    bool no_gicv3_with_gicv2m;
     bool no_nested_smmu;
 };
 
@@ -178,6 +179,7 @@ struct VirtMachineState {
     char *oem_id;
     char *oem_table_id;
     bool ns_el2_virt_timer_irq;
+    bool no_gicv3_with_gicv2m;
     CXLState cxl_devices_state;
 };
 
Signature · 1 line
2.39.5 (Apple Git-154)
#3 lore raw

Re: [PATCH v5 02/18] hw/arm: virt: add GICv2m for the case when ITS is not available

Quoted 8 lines · Mohamed Mediouni

On Hypervisor.framework for macOS and WHPX for Windows, the provided environment is a GICv3 without ITS.

As such, support a GICv3 w/ GICv2m for that scenario.

This breaks the bios-tables-test which is run as part of "make check", because the ACPI tables generated no longer match the golden reference output. If the ACPI table change here is intentional then the comment at the top of tests/qtest/bios-tables-test.c describes the process and sequence of patches you need to update the reference versions.

$ make -C build/arm-clang/ -j6 && (cd build/arm-clang/ &&
QTEST_QEMU_BINARY=./qemu-system-aarch64
./tests/qtest/bios-tables-test)
make: Entering directory
'/mnt/nvmedisk/linaro/qemu-from-laptop/qemu/build/arm-clang'
[1/28] Generating qemu-version.h with a custom command (wrapped by
meson to capture output)
[2/8] Compiling C object libsystem.a.p/hw_pci-host_gpex-acpi.c.o
[3/8] Compiling C object libqemu-arm-softmmu.a.p/hw_arm_virt-acpi-build.c.o
[4/8] Compiling C object libqemu-aarch64-softmmu.a.p/hw_arm_virt-acpi-build.c.o
[5/8] Compiling C object libqemu-arm-softmmu.a.p/hw_arm_virt.c.o
[6/8] Compiling C object libqemu-aarch64-softmmu.a.p/hw_arm_virt.c.o
[7/8] Linking target qemu-system-arm
[8/8] Linking target qemu-system-aarch64
make: Leaving directory
'/mnt/nvmedisk/linaro/qemu-from-laptop/qemu/build/arm-clang'
TAP version 13
# random seed: R02Sbd44a41112c6e2df4635231a2a1f3510
# starting QEMU: exec ./qemu-system-aarch64 -qtest
unix:/tmp/qtest-2147871.sock -qtest-log /dev/null -chardev
socket,path=/tmp/qtest-2147871.qmp,id=char0 -mon
chardev=char0,mode=control -display none -audio none -machine none
-accel qtest
# starting QEMU: exec ./qemu-system-aarch64 -qtest
unix:/tmp/qtest-2147871.sock -qtest-log /dev/null -chardev
socket,path=/tmp/qtest-2147871.qmp,id=char0 -mon
chardev=char0,mode=control -display none -audio none -machine none
-accel qtest
1..12
# Start of aarch64 tests
# Start of acpi tests
# starting QEMU: exec ./qemu-system-aarch64 -qtest
unix:/tmp/qtest-2147871.sock -qtest-log /dev/null -chardev
socket,path=/tmp/qtest-2147871.qmp,id=char0 -mon
chardev=char0,mode=control -display none -audio none -machine virt
-accel tcg -nodefaults -nographic -drive
if=pflash,format=raw,file=pc-bios/edk2-aarch64-code.fd,readonly=on
-drive if=pflash,format=raw,file=pc-bios/edk2-arm-vars.fd,snapshot=on
-cdrom tests/data/uefi-boot-images/bios-tables-test.aarch64.iso.qcow2
-cpu cortex-a57 -machine ras=on -smbios
type=4,max-speed=2900,current-speed=2700 -accel qtest
acpi-test: Warning! APIC binary file mismatch. Actual
[aml:/tmp/aml-0EHGB3], Expected
[aml:tests/data/acpi/aarch64/virt/APIC].
See source file tests/qtest/bios-tables-test.c for instructions on how
to update expected files.
acpi-test: Warning! APIC mismatch. Actual [asl:/tmp/asl-48FGB3.dsl,
aml:/tmp/aml-0EHGB3], Expected [asl:/tmp/asl-D5NHB3.dsl,
aml:tests/data/acpi/aarch64/virt/APIC].
**
ERROR:../../tests/qtest/bios-tables-test.c:554:test_acpi_asl:
assertion failed: (all_tables_match)
not ok /aarch64/acpi/virt -
ERROR:../../tests/qtest/bios-tables-test.c:554:test_acpi_asl:
assertion failed: (all_tables_match)
Bail out!
Aborted (core dumped)

thanks
-- PMM

#4 lore raw

Re: [PATCH v5 02/18] hw/arm: virt: add GICv2m for the case when ITS is not available

In reply to Mohamed Mediouni
Quoted 6 lines · Mohamed Mediouni

On Hypervisor.framework for macOS and WHPX for Windows, the provided environment is a GICv3 without ITS.

As such, support a GICv3 w/ GICv2m for that scenario.

We should also update the docs in docs/system/arm/virt.rst. These are already incorrect in that we never updated them when we implemented the ITS:

- An MSI controller (GICv2M or ITS). GICv2M is selected by default along with GICv2. ITS is selected by default with GICv3 (>= virt-2.7). Note that ITS is not modeled in TCG mode.

so we should update that bullet point to match current reality. (If it is looking a bit big then we can have the bullet point be just "An MSI controller (GICv2M or ITS)" and move the explanation of when we go with GICv2M vs ITS to a paragraph below the bullet list, in the same way we have for the explanation about the second NS UART.

-- PMM

#5 lore raw

PATCH v5 3/18 whpx: Move around files before introducing AArch64 support

In reply to Mohamed Mediouni

Switch to a design where we can share whpx code between x86 and AArch64 when it makes sense to do so.

---
 MAINTAINERS                                           | 2 ++
 accel/meson.build                                     | 1 +
 accel/whpx/meson.build                                | 6 ++++++
 {target/i386 => accel}/whpx/whpx-accel-ops.c          | 4 ++--
 {target/i386/whpx => include/system}/whpx-accel-ops.h | 4 ++--
 {target/i386/whpx => include/system}/whpx-internal.h  | 5 +++--
 target/i386/whpx/meson.build                          | 1 -
 target/i386/whpx/whpx-all.c                           | 4 ++--
 target/i386/whpx/whpx-apic.c                          | 2 +-
 9 files changed, 19 insertions(+), 10 deletions(-)
 create mode 100644 accel/whpx/meson.build
 rename {target/i386 => accel}/whpx/whpx-accel-ops.c (97%)
 rename {target/i386/whpx => include/system}/whpx-accel-ops.h (92%)
 rename {target/i386/whpx => include/system}/whpx-internal.h (98%)
MAINTAINERS +2 -0
index a07086ed76..3d28509be4 100644
@@ -542,9 +542,11 @@ F: include/system/hvf_int.h
 WHPX CPUs
 M: Sunil Muthuswamy <[email protected]>
 S: Supported
+F: accel/whpx/
 F: target/i386/whpx/
 F: accel/stubs/whpx-stub.c
 F: include/system/whpx.h
+F: include/system/whpx-accel-ops.h
 
 X86 Instruction Emulator
 M: Cameron Esfahani <[email protected]>
accel/meson.build +1 -0
index 25b0f100b5..de927a3b37 100644
@@ -6,6 +6,7 @@ user_ss.add(files('accel-user.c'))
 subdir('tcg')
 if have_system
   subdir('hvf')
+  subdir('whpx')
   subdir('qtest')
   subdir('kvm')
   subdir('xen')
accel/whpx/meson.build new file +6 -0
new file mode 100644
index 0000000000..7b3d6f1c1c
@@ -0,0 +1,6 @@
+whpx_ss = ss.source_set()
+whpx_ss.add(files(
+  'whpx-accel-ops.c',
+))
+
+specific_ss.add_all(when: 'CONFIG_WHPX', if_true: whpx_ss)
accel/whpx/whpx-accel-ops.c renamed from target/i386/whpx/whpx-accel-ops.c +2 -2
similarity index 97%
rename from target/i386/whpx/whpx-accel-ops.c
rename to accel/whpx/whpx-accel-ops.c
index da58805b1a..18488421bc 100644
@@ -16,8 +16,8 @@
 #include "qemu/guest-random.h"
 
 #include "system/whpx.h"
-#include "whpx-internal.h"
-#include "whpx-accel-ops.h"
+#include "system/whpx-internal.h"
+#include "system/whpx-accel-ops.h"
 
 static void *whpx_cpu_thread_fn(void *arg)
 {
include/system/whpx-accel-ops.h renamed from target/i386/whpx/whpx-accel-ops.h +2 -2
similarity index 92%
rename from target/i386/whpx/whpx-accel-ops.h
rename to include/system/whpx-accel-ops.h
index 54cfc25a14..ed9d4c49f4 100644
@@ -7,8 +7,8 @@
  * See the COPYING file in the top-level directory.
  */
 
-#ifndef TARGET_I386_WHPX_ACCEL_OPS_H
-#define TARGET_I386_WHPX_ACCEL_OPS_H
+#ifndef SYSTEM_WHPX_ACCEL_OPS_H
+#define SYSTEM_WHPX_ACCEL_OPS_H
 
 #include "system/cpus.h"
 
include/system/whpx-internal.h renamed from target/i386/whpx/whpx-internal.h +3 -2
similarity index 98%
rename from target/i386/whpx/whpx-internal.h
rename to include/system/whpx-internal.h
index 6633e9c4ca..e61375d554 100644
@@ -1,5 +1,6 @@
-#ifndef TARGET_I386_WHPX_INTERNAL_H
-#define TARGET_I386_WHPX_INTERNAL_H
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+#ifndef SYSTEM_WHPX_INTERNAL_H
+#define SYSTEM_WHPX_INTERNAL_H
 
 #include <windows.h>
 #include <winhvplatform.h>
target/i386/whpx/meson.build +0 -1
index 9c54aaad39..c3aaaff9fd 100644
@@ -1,5 +1,4 @@
 i386_system_ss.add(when: 'CONFIG_WHPX', if_true: files(
   'whpx-all.c',
   'whpx-apic.c',
-  'whpx-accel-ops.c',
 ))
target/i386/whpx/whpx-all.c +2 -2
index b72dcff3c8..5a431fc3c7 100644
@@ -31,8 +31,8 @@
 #include "accel/accel-cpu-target.h"
 #include <winerror.h>
 
-#include "whpx-internal.h"
-#include "whpx-accel-ops.h"
+#include "system/whpx-internal.h"
+#include "system/whpx-accel-ops.h"
 
 #include <winhvplatform.h>
 #include <winhvemulation.h>
target/i386/whpx/whpx-apic.c +1 -1
index e1ef6d4e6d..badb404b63 100644
@@ -18,7 +18,7 @@
 #include "hw/pci/msi.h"
 #include "system/hw_accel.h"
 #include "system/whpx.h"
-#include "whpx-internal.h"
+#include "system/whpx-internal.h"
 
 struct whpx_lapic_state {
     struct {
Signature · 1 line
2.39.5 (Apple Git-154)
#6 lore raw

PATCH v5 4/18 whpx: reshuffle common code

In reply to Mohamed Mediouni

Some code can be shared between x86_64 and arm64 WHPX. Do so as much as reasonable.

---
 MAINTAINERS                  |   2 +
 accel/whpx/meson.build       |   1 +
 accel/whpx/whpx-common.c     | 562 +++++++++++++++++++++++++++++++++++
 include/system/whpx-all.h    |  20 ++
 include/system/whpx-common.h |  21 ++
 target/i386/whpx/whpx-all.c  | 551 +---------------------------------
 6 files changed, 616 insertions(+), 541 deletions(-)
 create mode 100644 accel/whpx/whpx-common.c
 create mode 100644 include/system/whpx-all.h
 create mode 100644 include/system/whpx-common.h
MAINTAINERS +2 -0
index 3d28509be4..070ba2e9cb 100644
@@ -547,6 +547,8 @@ F: target/i386/whpx/
 F: accel/stubs/whpx-stub.c
 F: include/system/whpx.h
 F: include/system/whpx-accel-ops.h
+F: include/system/whpx-common.h
+F: include/system/whpx-internal.h
 
 X86 Instruction Emulator
 M: Cameron Esfahani <[email protected]>
accel/whpx/meson.build +1 -0
index 7b3d6f1c1c..fad28dddcb 100644
@@ -1,6 +1,7 @@
 whpx_ss = ss.source_set()
 whpx_ss.add(files(
   'whpx-accel-ops.c',
+  'whpx-common.c'
 ))
 
 specific_ss.add_all(when: 'CONFIG_WHPX', if_true: whpx_ss)
accel/whpx/whpx-common.c new file +562 -0

562 lines not on this page · raw message

include/system/whpx-all.h new file +20 -0
new file mode 100644
index 0000000000..f13cdf7f66
@@ -0,0 +1,20 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+#ifndef SYSTEM_WHPX_ALL_H
+#define SYSTEM_WHPX_ALL_H
+
+/* Called by whpx-common */
+int whpx_vcpu_run(CPUState *cpu);
+void whpx_get_registers(CPUState *cpu);
+void whpx_set_registers(CPUState *cpu, int level);
+int whpx_accel_init(AccelState *as, MachineState *ms);
+void whpx_cpu_instance_init(CPUState *cs);
+HRESULT whpx_set_exception_exit_bitmap(UINT64 exceptions);
+void whpx_apply_breakpoints(
+struct whpx_breakpoint_collection *breakpoints,
+    CPUState *cpu,
+    bool resuming);
+void whpx_translate_cpu_breakpoints(
+    struct whpx_breakpoints *breakpoints,
+    CPUState *cpu,
+    int cpu_breakpoint_count);
+#endif
include/system/whpx-common.h new file +21 -0
new file mode 100644
index 0000000000..e549c7539c
@@ -0,0 +1,21 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+#ifndef SYSTEM_WHPX_COMMON_H
+#define SYSTEM_WHPX_COMMON_H
+
+struct AccelCPUState {
+    WHV_EMULATOR_HANDLE emulator;
+    bool window_registered;
+    bool interruptable;
+    bool ready_for_pic_interrupt;
+    uint64_t tpr;
+    uint64_t apic_base;
+    bool interruption_pending;
+    /* Must be the last field as it may have a tail */
+    WHV_RUN_VP_EXIT_CONTEXT exit_ctx;
+};
+
+int whpx_first_vcpu_starting(CPUState *cpu);
+int whpx_last_vcpu_stopping(CPUState *cpu);
+void whpx_memory_init(void);
+struct whpx_breakpoint *whpx_lookup_breakpoint_by_addr(uint64_t address);
+#endif
target/i386/whpx/whpx-all.c +10 -541

639 lines not on this page · raw message

Signature · 1 line
2.39.5 (Apple Git-154)
#7 lore raw

PATCH v5 5/18 whpx: ifdef out winhvemulation on non-x86_64

In reply to Mohamed Mediouni

winhvemulation is x86_64 only.

In the future, we might want to get rid of winhvemulation usage entirely.

---
 accel/whpx/whpx-common.c       | 14 ++++++++++++--
 include/system/whpx-common.h   |  2 ++
 include/system/whpx-internal.h |  7 ++++++-
 3 files changed, 20 insertions(+), 3 deletions(-)
accel/whpx/whpx-common.c +12 -2
index 66c9238586..e2d692126a 100644
@@ -41,7 +41,9 @@
 bool whpx_allowed;
 static bool whp_dispatch_initialized;
 static HMODULE hWinHvPlatform;
+#ifdef __x86_64__
 static HMODULE hWinHvEmulation;
+#endif
 
 struct whpx_state whpx_global;
 struct WHPDispatch whp_dispatch;
@@ -236,8 +238,10 @@ void whpx_destroy_vcpu(CPUState *cpu)
     struct whpx_state *whpx = &whpx_global;
 
     whp_dispatch.WHvDeleteVirtualProcessor(whpx->partition, cpu->cpu_index);
+#ifdef __x86_64__
     AccelCPUState *vcpu = cpu->accel;
     whp_dispatch.WHvEmulatorDestroyEmulator(vcpu->emulator);
+#endif
     g_free(cpu->accel);
 }
 
@@ -412,8 +416,12 @@ static bool load_whp_dispatch_fns(HMODULE *handle,
         LIST_WINHVPLATFORM_FUNCTIONS(WHP_LOAD_FIELD)
         break;
     case WINHV_EMULATION_FNS_DEFAULT:
+#ifdef __x86_64__
         WHP_LOAD_LIB(WINHV_EMULATION_DLL, hLib)
         LIST_WINHVEMULATION_FUNCTIONS(WHP_LOAD_FIELD)
+#else
+        abort();
+#endif
         break;
     case WINHV_PLATFORM_FNS_SUPPLEMENTAL:
         WHP_LOAD_LIB(WINHV_PLATFORM_DLL, hLib)
@@ -539,11 +547,11 @@ bool init_whp_dispatch(void)
     if (!load_whp_dispatch_fns(&hWinHvPlatform, WINHV_PLATFORM_FNS_DEFAULT)) {
         goto error;
     }
-
+#ifdef __x86_64__
     if (!load_whp_dispatch_fns(&hWinHvEmulation, WINHV_EMULATION_FNS_DEFAULT)) {
         goto error;
     }
-
+#endif
     assert(load_whp_dispatch_fns(&hWinHvPlatform,
         WINHV_PLATFORM_FNS_SUPPLEMENTAL));
     whp_dispatch_initialized = true;
@@ -553,9 +561,11 @@ error:
     if (hWinHvPlatform) {
         FreeLibrary(hWinHvPlatform);
     }
+#ifdef __x86_64__
     if (hWinHvEmulation) {
         FreeLibrary(hWinHvEmulation);
     }
+#endif
     return false;
 }
 
include/system/whpx-common.h +2 -0
index e549c7539c..7a7c607e0a 100644
@@ -3,7 +3,9 @@
 #define SYSTEM_WHPX_COMMON_H
 
 struct AccelCPUState {
+#ifdef __x86_64__
     WHV_EMULATOR_HANDLE emulator;
+#endif
     bool window_registered;
     bool interruptable;
     bool ready_for_pic_interrupt;
include/system/whpx-internal.h +6 -1
index e61375d554..e57d2c8526 100644
@@ -4,8 +4,9 @@
 
 #include <windows.h>
 #include <winhvplatform.h>
+#ifdef __x86_64__
 #include <winhvemulation.h>
-
+#endif
 typedef enum WhpxBreakpointState {
     WHPX_BP_CLEARED = 0,
     WHPX_BP_SET_PENDING,
@@ -98,12 +99,16 @@ void whpx_apic_get(DeviceState *s);
 
 /* Define function typedef */
 LIST_WINHVPLATFORM_FUNCTIONS(WHP_DEFINE_TYPE)
+#ifdef __x86_64__
 LIST_WINHVEMULATION_FUNCTIONS(WHP_DEFINE_TYPE)
+#endif
 LIST_WINHVPLATFORM_FUNCTIONS_SUPPLEMENTAL(WHP_DEFINE_TYPE)
 
 struct WHPDispatch {
     LIST_WINHVPLATFORM_FUNCTIONS(WHP_DECLARE_MEMBER)
+#ifdef __x86_64__
     LIST_WINHVEMULATION_FUNCTIONS(WHP_DECLARE_MEMBER)
+#endif
     LIST_WINHVPLATFORM_FUNCTIONS_SUPPLEMENTAL(WHP_DECLARE_MEMBER)
 };
 
Signature · 1 line
2.39.5 (Apple Git-154)
#8 lore raw

Re: [PATCH v5 05/18] whpx: ifdef out winhvemulation on non-x86_64

Am 8. August 2025 06:54:06 UTC schrieb Mohamed Mediouni <[email protected]>:

Quoted 40 lines

winhvemulation is x86_64 only.

In the future, we might want to get rid of winhvemulation usage entirely.

---

accel/whpx/whpx-common.c       | 14 ++++++++++++--
include/system/whpx-common.h   |  2 ++
include/system/whpx-internal.h |  7 ++++++-
3 files changed, 20 insertions(+), 3 deletions(-)
accel/whpx/whpx-common.c +4 -0
index 66c9238586..e2d692126a 100644
@@ -41,7 +41,9 @@
 bool whpx_allowed;
 static bool whp_dispatch_initialized;
 static HMODULE hWinHvPlatform;
+#ifdef __x86_64__
 static HMODULE hWinHvEmulation;
+#endif
 
 struct whpx_state whpx_global;
 struct WHPDispatch whp_dispatch;
@@ -236,8 +238,10 @@ void whpx_destroy_vcpu(CPUState *cpu)
    struct whpx_state *whpx = &whpx_global;
 
    whp_dispatch.WHvDeleteVirtualProcessor(whpx->partition, cpu->cpu_index);
+#ifdef __x86_64__
    AccelCPUState *vcpu = cpu->accel;
    whp_dispatch.WHvEmulatorDestroyEmulator(vcpu->emulator);
+#endif
    g_free(cpu->accel);
 }
 

accel/whpx/whpx-common.c static bool load_whp_dispatch_fns(HMODULE *handle, · line 423

Quoted 8 lines
accel/whpx/whpx-common.c +3 -0
@@ -412,8 +416,12 @@ static bool load_whp_dispatch_fns(HMODULE *handle,
        LIST_WINHVPLATFORM_FUNCTIONS(WHP_LOAD_FIELD)
        break;
    case WINHV_EMULATION_FNS_DEFAULT:
+#ifdef __x86_64__
        WHP_LOAD_LIB(WINHV_EMULATION_DLL, hLib)
        LIST_WINHVEMULATION_FUNCTIONS(WHP_LOAD_FIELD)
+#else
+        abort();

In the sense of g_assert_not_reached()? Then it is preferred.

Quoted 76 lines
accel/whpx/whpx-common.c +5 -2
+#endif
        break;
    case WINHV_PLATFORM_FNS_SUPPLEMENTAL:
        WHP_LOAD_LIB(WINHV_PLATFORM_DLL, hLib)
@@ -539,11 +547,11 @@ bool init_whp_dispatch(void)
    if (!load_whp_dispatch_fns(&hWinHvPlatform, WINHV_PLATFORM_FNS_DEFAULT)) {
        goto error;
    }
-
+#ifdef __x86_64__
    if (!load_whp_dispatch_fns(&hWinHvEmulation, WINHV_EMULATION_FNS_DEFAULT)) {
        goto error;
    }
-
+#endif
    assert(load_whp_dispatch_fns(&hWinHvPlatform,
        WINHV_PLATFORM_FNS_SUPPLEMENTAL));
    whp_dispatch_initialized = true;
@@ -553,9 +561,11 @@ error:
    if (hWinHvPlatform) {
        FreeLibrary(hWinHvPlatform);
    }
+#ifdef __x86_64__
    if (hWinHvEmulation) {
        FreeLibrary(hWinHvEmulation);
    }
+#endif
    return false;
 }
 
include/system/whpx-common.h +2 -0
index e549c7539c..7a7c607e0a 100644
@@ -3,7 +3,9 @@
 #define SYSTEM_WHPX_COMMON_H
 
 struct AccelCPUState {
+#ifdef __x86_64__
    WHV_EMULATOR_HANDLE emulator;
+#endif
    bool window_registered;
    bool interruptable;
    bool ready_for_pic_interrupt;
include/system/whpx-internal.h +6 -1
index e61375d554..e57d2c8526 100644
@@ -4,8 +4,9 @@
 
 #include <windows.h>
 #include <winhvplatform.h>
+#ifdef __x86_64__
 #include <winhvemulation.h>
-
+#endif
 typedef enum WhpxBreakpointState {
    WHPX_BP_CLEARED = 0,
    WHPX_BP_SET_PENDING,
@@ -98,12 +99,16 @@ void whpx_apic_get(DeviceState *s);
 
 /* Define function typedef */
 LIST_WINHVPLATFORM_FUNCTIONS(WHP_DEFINE_TYPE)
+#ifdef __x86_64__
 LIST_WINHVEMULATION_FUNCTIONS(WHP_DEFINE_TYPE)
+#endif
 LIST_WINHVPLATFORM_FUNCTIONS_SUPPLEMENTAL(WHP_DEFINE_TYPE)
 
 struct WHPDispatch {
    LIST_WINHVPLATFORM_FUNCTIONS(WHP_DECLARE_MEMBER)
+#ifdef __x86_64__
    LIST_WINHVEMULATION_FUNCTIONS(WHP_DECLARE_MEMBER)
+#endif
    LIST_WINHVPLATFORM_FUNCTIONS_SUPPLEMENTAL(WHP_DECLARE_MEMBER)
 };
 
#9 lore raw

PATCH v5 6/18 whpx: common: add WHPX_INTERCEPT_DEBUG_TRAPS define

In reply to Mohamed Mediouni

As of why: WHPX on arm64 doesn't have debug trap support as of today.

Keep the exception bitmap interface for now - despite that being entirely unavailable on arm64 too.

---
 accel/whpx/whpx-common.c     | 2 +-
 include/system/whpx-common.h | 3 +++
 2 files changed, 4 insertions(+), 1 deletion(-)
accel/whpx/whpx-common.c +1 -1
index e2d692126a..5c2da9de4d 100644
@@ -123,7 +123,7 @@ int whpx_first_vcpu_starting(CPUState *cpu)
          * have one or more breakpoints enabled. Both require intercepting
          * the WHvX64ExceptionTypeBreakpointTrap exception.
          */
-        exception_mask = 1UL << WHvX64ExceptionTypeDebugTrapOrFault;
+        exception_mask = 1UL << WHPX_INTERCEPT_DEBUG_TRAPS;
     } else {
         /* Let the guest handle all exceptions. */
         exception_mask = 0;
include/system/whpx-common.h +3 -0
index 7a7c607e0a..73b9f7c119 100644
@@ -20,4 +20,7 @@ int whpx_first_vcpu_starting(CPUState *cpu);
 int whpx_last_vcpu_stopping(CPUState *cpu);
 void whpx_memory_init(void);
 struct whpx_breakpoint *whpx_lookup_breakpoint_by_addr(uint64_t address);
+
+/* On x64: same as WHvX64ExceptionTypeDebugTrapOrFault */
+#define WHPX_INTERCEPT_DEBUG_TRAPS 1
 #endif
Signature · 1 line
2.39.5 (Apple Git-154)
#10 lore raw

PATCH v5 7/18 hw, target, accel: whpx: change apic_in_platform to kernel_irqchip

In reply to Mohamed Mediouni

Change terminology to match the KVM one, as APIC is x86-specific.

---
 accel/whpx/whpx-accel-ops.c    |  2 +-
 accel/whpx/whpx-common.c       |  4 ++--
 hw/i386/x86-cpu.c              |  4 ++--
 include/system/whpx-internal.h |  2 +-
 include/system/whpx.h          |  4 ++--
 target/i386/cpu-apic.c         |  2 +-
 target/i386/whpx/whpx-all.c    | 14 +++++++-------
 7 files changed, 16 insertions(+), 16 deletions(-)
accel/whpx/whpx-accel-ops.c +1 -1
index 18488421bc..ef1fe52860 100644
@@ -80,7 +80,7 @@ static void whpx_kick_vcpu_thread(CPUState *cpu)
 
 static bool whpx_vcpu_thread_is_idle(CPUState *cpu)
 {
-    return !whpx_apic_in_platform();
+    return !whpx_irqchip_in_kernel();
 }
 
 static void whpx_accel_ops_class_init(ObjectClass *oc, const void *data)
accel/whpx/whpx-common.c +2 -2
index 5c2da9de4d..0dbfb4b242 100644
@@ -496,9 +496,9 @@ static const TypeInfo whpx_cpu_accel_type = {
  * Partition support
  */
 
-bool whpx_apic_in_platform(void)
+bool whpx_irqchip_in_kernel(void)
 {
-    return whpx_global.apic_in_platform;
+    return whpx_global.kernel_irqchip;
 }
 
 static void whpx_accel_class_init(ObjectClass *oc, const void *data)
hw/i386/x86-cpu.c +2 -2
index c876e6709e..778607e7ca 100644
@@ -45,7 +45,7 @@ static void pic_irq_request(void *opaque, int irq, int level)
 
     trace_x86_pic_interrupt(irq, level);
     if (cpu_is_apic_enabled(cpu->apic_state) && !kvm_irqchip_in_kernel() &&
-        !whpx_apic_in_platform()) {
+        !whpx_irqchip_in_kernel()) {
         CPU_FOREACH(cs) {
             cpu = X86_CPU(cs);
             if (apic_accept_pic_intr(cpu->apic_state)) {
@@ -71,7 +71,7 @@ int cpu_get_pic_interrupt(CPUX86State *env)
     X86CPU *cpu = env_archcpu(env);
     int intno;
 
-    if (!kvm_irqchip_in_kernel() && !whpx_apic_in_platform()) {
+    if (!kvm_irqchip_in_kernel() && !whpx_irqchip_in_kernel()) {
         intno = apic_get_interrupt(cpu->apic_state);
         if (intno >= 0) {
             return intno;
include/system/whpx-internal.h +1 -1
index e57d2c8526..366bc525a3 100644
@@ -42,7 +42,7 @@ struct whpx_state {
 
     bool kernel_irqchip_allowed;
     bool kernel_irqchip_required;
-    bool apic_in_platform;
+    bool kernel_irqchip;
 };
 
 extern struct whpx_state whpx_global;
include/system/whpx.h +2 -2
index 00f6a3e523..98fe045ba1 100644
@@ -26,10 +26,10 @@
 #ifdef CONFIG_WHPX_IS_POSSIBLE
 extern bool whpx_allowed;
 #define whpx_enabled() (whpx_allowed)
-bool whpx_apic_in_platform(void);
+bool whpx_irqchip_in_kernel(void);
 #else /* !CONFIG_WHPX_IS_POSSIBLE */
 #define whpx_enabled() 0
-#define whpx_apic_in_platform() (0)
+#define whpx_irqchip_in_kernel() (0)
 #endif /* !CONFIG_WHPX_IS_POSSIBLE */
 
 #endif /* QEMU_WHPX_H */
target/i386/cpu-apic.c +1 -1
index 242a05fdbe..d4d371a616 100644
@@ -32,7 +32,7 @@ APICCommonClass *apic_get_class(Error **errp)
         apic_type = "kvm-apic";
     } else if (xen_enabled()) {
         apic_type = "xen-apic";
-    } else if (whpx_apic_in_platform()) {
+    } else if (whpx_irqchip_in_kernel()) {
         apic_type = "whpx-apic";
     }
 
target/i386/whpx/whpx-all.c +7 -7
index 9f671cc0a6..d7d55e0d19 100644
@@ -607,7 +607,7 @@ void whpx_get_registers(CPUState *cpu)
                      hr);
     }
 
-    if (whpx_apic_in_platform()) {
+    if (whpx_irqchip_in_kernel()) {
         /*
          * Fetch the TPR value from the emulated APIC. It may get overwritten
          * below with the value from CR8 returned by
@@ -749,7 +749,7 @@ void whpx_get_registers(CPUState *cpu)
 
     assert(idx == RTL_NUMBER_OF(whpx_register_names));
 
-    if (whpx_apic_in_platform()) {
+    if (whpx_irqchip_in_kernel()) {
         whpx_apic_get(x86_cpu->apic_state);
     }
 
@@ -1376,7 +1376,7 @@ static void whpx_vcpu_pre_run(CPUState *cpu)
     }
 
     /* Get pending hard interruption or replay one that was overwritten */
-    if (!whpx_apic_in_platform()) {
+    if (!whpx_irqchip_in_kernel()) {
         if (!vcpu->interruption_pending &&
             vcpu->interruptable && (env->eflags & IF_MASK)) {
             assert(!new_int.InterruptionPending);
@@ -1549,7 +1549,7 @@ int whpx_vcpu_run(CPUState *cpu)
 
     if (exclusive_step_mode == WHPX_STEP_NONE) {
         whpx_vcpu_process_async_events(cpu);
-        if (cpu->halted && !whpx_apic_in_platform()) {
+        if (cpu->halted && !whpx_irqchip_in_kernel()) {
             cpu->exception_index = EXCP_HLT;
             qatomic_set(&cpu->exit_request, false);
             return 0;
@@ -1637,7 +1637,7 @@ int whpx_vcpu_run(CPUState *cpu)
             break;
 
         case WHvRunVpExitReasonX64ApicEoi:
-            assert(whpx_apic_in_platform());
+            assert(whpx_irqchip_in_kernel());
             ioapic_eoi_broadcast(vcpu->exit_ctx.ApicEoi.InterruptVector);
             break;
 
@@ -2184,7 +2184,7 @@ int whpx_accel_init(AccelState *as, MachineState *ms)
                 goto error;
             }
         } else {
-            whpx->apic_in_platform = true;
+            whpx->kernel_irqchip = true;
         }
     }
 
@@ -2193,7 +2193,7 @@ int whpx_accel_init(AccelState *as, MachineState *ms)
     prop.ExtendedVmExits.X64MsrExit = 1;
     prop.ExtendedVmExits.X64CpuidExit = 1;
     prop.ExtendedVmExits.ExceptionExit = 1;
-    if (whpx_apic_in_platform()) {
+    if (whpx_irqchip_in_kernel()) {
         prop.ExtendedVmExits.X64ApicInitSipiExitTrap = 1;
     }
 
Signature · 1 line
2.39.5 (Apple Git-154)
Reviewed-by on v5 7/18
#11 lore raw

Re: [PATCH v5 07/18] hw, target, accel: whpx: change apic_in_platform to kernel_irqchip

Quoted 6 lines · Mohamed Mediouni

Change terminology to match the KVM one, as APIC is x86-specific.

Quoted 9 lines
---
  accel/whpx/whpx-accel-ops.c    |  2 +-
  accel/whpx/whpx-common.c       |  4 ++--
  hw/i386/x86-cpu.c              |  4 ++--
  include/system/whpx-internal.h |  2 +-
  include/system/whpx.h          |  4 ++--
  target/i386/cpu-apic.c         |  2 +-
  target/i386/whpx/whpx-all.c    | 14 +++++++-------
  7 files changed, 16 insertions(+), 16 deletions(-)
#12 lore raw

PATCH v5 8/18 whpx: interrupt controller support

In reply to Mohamed Mediouni
---
 hw/arm/virt.c                      |   3 +
 hw/intc/arm_gicv3_common.c         |   3 +
 hw/intc/arm_gicv3_whpx.c           | 262 +++++++++++++++++++++++++++++
 hw/intc/meson.build                |   1 +
 include/hw/intc/arm_gicv3_common.h |   3 +
 5 files changed, 272 insertions(+)
 create mode 100644 hw/intc/arm_gicv3_whpx.c
hw/arm/virt.c +3 -0
index 5951b331f3..98a1c74c42 100644
@@ -49,6 +49,7 @@
 #include "system/tcg.h"
 #include "system/kvm.h"
 #include "system/hvf.h"
+#include "system/whpx.h"
 #include "system/qtest.h"
 #include "hw/loader.h"
 #include "qapi/error.h"
@@ -2060,6 +2061,8 @@ static void finalize_gic_version(VirtMachineState *vms)
         /* KVM w/o kernel irqchip can only deal with GICv2 */
         gics_supported |= VIRT_GIC_VERSION_2_MASK;
         accel_name = "KVM with kernel-irqchip=off";
+    } else if (whpx_enabled()) {
+        gics_supported |= VIRT_GIC_VERSION_3_MASK;
     } else if (tcg_enabled() || hvf_enabled() || qtest_enabled())  {
         gics_supported |= VIRT_GIC_VERSION_2_MASK;
         if (module_object_class_by_name("arm-gicv3")) {
hw/intc/arm_gicv3_common.c +3 -0
index e438d8c042..8b85b60c9b 100644
@@ -32,6 +32,7 @@
 #include "gicv3_internal.h"
 #include "hw/arm/linux-boot-if.h"
 #include "system/kvm.h"
+#include "system/whpx.h"
 
 
 static void gicv3_gicd_no_migration_shift_bug_post_load(GICv3State *cs)
@@ -662,6 +663,8 @@ const char *gicv3_class_name(void)
 {
     if (kvm_irqchip_in_kernel()) {
         return "kvm-arm-gicv3";
+    } else if (whpx_enabled()) {
+        return TYPE_WHPX_GICV3;
     } else {
         if (kvm_enabled()) {
             error_report("Userspace GICv3 is not supported with KVM");
hw/intc/arm_gicv3_whpx.c new file +262 -0

262 lines not on this page · raw message

hw/intc/meson.build +1 -0
index 3137521a4a..4fc6b78a04 100644
@@ -41,6 +41,7 @@ specific_ss.add(when: 'CONFIG_APIC', if_true: files('apic.c', 'apic_common.c'))
 specific_ss.add(when: 'CONFIG_ARM_GIC', if_true: files('arm_gicv3_cpuif_common.c'))
 specific_ss.add(when: 'CONFIG_ARM_GICV3', if_true: files('arm_gicv3_cpuif.c'))
 specific_ss.add(when: 'CONFIG_ARM_GIC_KVM', if_true: files('arm_gic_kvm.c'))
+specific_ss.add(when: ['CONFIG_WHPX', 'TARGET_AARCH64'], if_true: files('arm_gicv3_whpx.c'))
 specific_ss.add(when: ['CONFIG_ARM_GIC_KVM', 'TARGET_AARCH64'], if_true: files('arm_gicv3_kvm.c', 'arm_gicv3_its_kvm.c'))
 specific_ss.add(when: 'CONFIG_ARM_V7M', if_true: files('armv7m_nvic.c'))
 specific_ss.add(when: 'CONFIG_GRLIB', if_true: files('grlib_irqmp.c'))
include/hw/intc/arm_gicv3_common.h +3 -0
index c18503869f..7776558a0e 100644
@@ -306,6 +306,9 @@ typedef struct ARMGICv3CommonClass ARMGICv3CommonClass;
 DECLARE_OBJ_CHECKERS(GICv3State, ARMGICv3CommonClass,
                      ARM_GICV3_COMMON, TYPE_ARM_GICV3_COMMON)
 
+/* Types for GICv3 kernel-irqchip */
+#define TYPE_WHPX_GICV3 "whpx-arm-gicv3"
+
 struct ARMGICv3CommonClass {
     /*< private >*/
     SysBusDeviceClass parent_class;
Signature · 1 line
2.39.5 (Apple Git-154)
#13 lore raw

Re: [PATCH v5 08/18] whpx: interrupt controller support

Quoted 11 lines · Mohamed Mediouni
hw/intc/arm_gicv3_whpx.c new file +5 -0
+static void whpx_gicv3_put_cpu(CPUState *cpu_state, run_on_cpu_data arg)
+{
+}
+
+static void whpx_gicv3_put(GICv3State *s)

hw/intc/arm_gicv3_whpx.c line 59

Quoted 10 lines
hw/intc/arm_gicv3_whpx.c new file +10 -0
+{
+    int ncpu;
+
+    whpx_gicv3_check(s);
+
+    for (ncpu = 0; ncpu < s->num_cpu; ncpu++) {
+        run_on_cpu_data data;
+        data.host_ptr = &s->cpu[ncpu];
+        run_on_cpu(s->cpu[ncpu].cpu, whpx_gicv3_put_cpu, data);
+    }

Why are we calling run_on_cpu() and telling it to do nothing?

Also, run_on_cpu() implicitly drops the big kernel lock -- I'm not sure that is valid to do from the pre-save/post-load routines or in a reset handler.

thanks
-- PMM

#14 lore raw

PATCH v5 9/18 whpx: add arm64 support

In reply to Mohamed Mediouni
---
 accel/whpx/whpx-common.c    |   1 +
 meson.build                 |  21 +-
 target/arm/meson.build      |   1 +
 target/arm/whpx/meson.build |   3 +
 target/arm/whpx/whpx-all.c  | 846 ++++++++++++++++++++++++++++++++++++
 5 files changed, 865 insertions(+), 7 deletions(-)
 create mode 100644 target/arm/whpx/meson.build
 create mode 100644 target/arm/whpx/whpx-all.c
accel/whpx/whpx-common.c +1 -0
index 0dbfb4b242..86750c304d 100644
@@ -16,6 +16,7 @@
 #include "gdbstub/helpers.h"
 #include "qemu/accel.h"
 #include "accel/accel-ops.h"
+#include "system/memory.h"
 #include "system/whpx.h"
 #include "system/cpus.h"
 #include "system/runstate.h"
meson.build +14 -7
index a7b3c683ce..d907c5f28e 100644
@@ -327,7 +327,8 @@ accelerator_targets += { 'CONFIG_XEN': xen_targets }
 
 if cpu == 'aarch64'
   accelerator_targets += {
-    'CONFIG_HVF': ['aarch64-softmmu']
+    'CONFIG_HVF': ['aarch64-softmmu'],
+    'CONFIG_WHPX': ['aarch64-softmmu']
   }
 elif cpu == 'x86_64'
   accelerator_targets += {
@@ -884,14 +885,20 @@ accelerators = []
 if get_option('kvm').allowed() and host_os == 'linux'
   accelerators += 'CONFIG_KVM'
 endif
+
 if get_option('whpx').allowed() and host_os == 'windows'
-  if get_option('whpx').enabled() and host_machine.cpu() != 'x86_64'
-    error('WHPX requires 64-bit host')
-  elif cc.has_header('winhvplatform.h', required: get_option('whpx')) and \
-       cc.has_header('winhvemulation.h', required: get_option('whpx'))
-    accelerators += 'CONFIG_WHPX'
+  if cpu == 'i386'
+    if get_option('whpx').enabled()
+     error('WHPX requires 64-bit host')
+    endif
+   # Leave CONFIG_WHPX disabled
+  else
+    if cc.has_header('winhvplatform.h', required: get_option('whpx')) and \
+      cc.has_header('winhvemulation.h', required: get_option('whpx'))
+      accelerators += 'CONFIG_WHPX'
+    endif
   endif
-endif
+ endif
 
 hvf = not_found
 if get_option('hvf').allowed()
target/arm/meson.build +1 -0
index 07d9271aa4..e28bd3f8e2 100644
@@ -46,6 +46,7 @@ arm_common_system_ss.add(files(
 ))
 
 subdir('hvf')
+subdir('whpx')
 
 if 'CONFIG_TCG' in config_all_accel
    subdir('tcg')
target/arm/whpx/meson.build new file +3 -0
new file mode 100644
index 0000000000..1de2ef0283
@@ -0,0 +1,3 @@
+arm_system_ss.add(when: 'CONFIG_WHPX', if_true: files(
+  'whpx-all.c',
+))
target/arm/whpx/whpx-all.c new file +846 -0

846 lines not on this page · raw message

Signature · 1 line
2.39.5 (Apple Git-154)
#15 lore raw

PATCH v5 10/18 whpx: copy over memory management logic from hvf

In reply to Mohamed Mediouni

This allows edk2 to work, although u-boot is still not functional.

---
 accel/whpx/whpx-common.c | 201 ++++++++++++++++++++++++++++-----------
 1 file changed, 147 insertions(+), 54 deletions(-)
accel/whpx/whpx-common.c +147 -54
index 86750c304d..752a57170e 100644
@@ -258,89 +258,174 @@ void whpx_vcpu_kick(CPUState *cpu)
  * Memory support.
  */
 
-static void whpx_update_mapping(hwaddr start_pa, ram_addr_t size,
-                                void *host_va, int add, int rom,
-                                const char *name)
+ /* whpx_slot flags */
+#define WHPX_SLOT_LOG (1 << 0)
+typedef struct whpx_slot {
+    uint64_t start;
+    uint64_t size;
+    uint8_t *mem;
+    int slot_id;
+    uint32_t flags;
+    MemoryRegion *region;
+} whpx_slot;
+
+typedef struct WHPXState {
+    whpx_slot slots[32];
+    int num_slots;
+} WHPXState;
+
+ WHPXState *whpx_state;
+
+ struct mac_slot {
+    int present;
+    uint64_t size;
+    uint64_t gpa_start;
+    uint64_t gva;
+};
+
+struct mac_slot mac_slots[32];
+
+static int do_whpx_set_memory(whpx_slot *slot, WHV_MAP_GPA_RANGE_FLAGS flags)
 {
     struct whpx_state *whpx = &whpx_global;
+    struct mac_slot *macslot;
     HRESULT hr;
 
-    /*
-    if (add) {
-        printf("WHPX: ADD PA:%p Size:%p, Host:%p, %s, '%s'\n",
-               (void*)start_pa, (void*)size, host_va,
-               (rom ? "ROM" : "RAM"), name);
-    } else {
-        printf("WHPX: DEL PA:%p Size:%p, Host:%p,      '%s'\n",
-               (void*)start_pa, (void*)size, host_va, name);
+    macslot = &mac_slots[slot->slot_id];
+
+    if (macslot->present) {
+        if (macslot->size != slot->size) {
+            macslot->present = 0;
+            hr = whp_dispatch.WHvUnmapGpaRange(whpx->partition,
+                 macslot->gpa_start, macslot->size);
+            if (FAILED(hr)) {
+                abort();
+            }
+        }
     }
-    */
-
-    if (add) {
-        hr = whp_dispatch.WHvMapGpaRange(whpx->partition,
-                                         host_va,
-                                         start_pa,
-                                         size,
-                                         (WHvMapGpaRangeFlagRead |
-                                          WHvMapGpaRangeFlagExecute |
-                                          (rom ? 0 : WHvMapGpaRangeFlagWrite)));
-    } else {
-        hr = whp_dispatch.WHvUnmapGpaRange(whpx->partition,
-                                           start_pa,
-                                           size);
+
+    if (!slot->size) {
+        return 0;
     }
 
-    if (FAILED(hr)) {
-        error_report("WHPX: Failed to %s GPA range '%s' PA:%p, Size:%p bytes,"
-                     " Host:%p, hr=%08lx",
-                     (add ? "MAP" : "UNMAP"), name,
-                     (void *)(uintptr_t)start_pa, (void *)size, host_va, hr);
+    macslot->present = 1;
+    macslot->gpa_start = slot->start;
+    macslot->size = slot->size;
+    hr = whp_dispatch.WHvMapGpaRange(whpx->partition,
+         slot->mem, slot->start, slot->size, flags);
+    return 0;
+}
+
+static whpx_slot *whpx_find_overlap_slot(uint64_t start, uint64_t size)
+{
+    whpx_slot *slot;
+    int x;
+    for (x = 0; x < whpx_state->num_slots; ++x) {
+        slot = &whpx_state->slots[x];
+        if (slot->size && start < (slot->start + slot->size) &&
+            (start + size) > slot->start) {
+            return slot;
+        }
     }
+    return NULL;
 }
 
-static void whpx_process_section(MemoryRegionSection *section, int add)
+static void whpx_set_phys_mem(MemoryRegionSection *section, bool add)
 {
-    MemoryRegion *mr = section->mr;
-    hwaddr start_pa = section->offset_within_address_space;
-    ram_addr_t size = int128_get64(section->size);
-    unsigned int delta;
-    uint64_t host_va;
+    whpx_slot *mem;
+    MemoryRegion *area = section->mr;
+    bool writable = !area->readonly && !area->rom_device;
+    WHV_MAP_GPA_RANGE_FLAGS flags;
+    uint64_t page_size = qemu_real_host_page_size();
+
+    if (!memory_region_is_ram(area)) {
+        if (writable) {
+            return;
+        } else if (!memory_region_is_romd(area)) {
+            /*
+             * If the memory device is not in romd_mode, then we actually want
+             * to remove the whpx memory slot so all accesses will trap.
+             */
+             add = false;
+        }
+    }
 
-    if (!memory_region_is_ram(mr)) {
-        return;
+    if (!QEMU_IS_ALIGNED(int128_get64(section->size), page_size) ||
+        !QEMU_IS_ALIGNED(section->offset_within_address_space, page_size)) {
+        /* Not page aligned, so we can not map as RAM */
+        add = false;
     }
 
-    delta = qemu_real_host_page_size() - (start_pa & ~qemu_real_host_page_mask());
-    delta &= ~qemu_real_host_page_mask();
-    if (delta > size) {
-        return;
+    mem = whpx_find_overlap_slot(
+            section->offset_within_address_space,
+            int128_get64(section->size));
+
+    if (mem && add) {
+        if (mem->size == int128_get64(section->size) &&
+            mem->start == section->offset_within_address_space &&
+            mem->mem == (memory_region_get_ram_ptr(area) +
+            section->offset_within_region)) {
+            return; /* Same region was attempted to register, go away. */
+        }
+    }
+
+    /* Region needs to be reset. set the size to 0 and remap it. */
+    if (mem) {
+        mem->size = 0;
+        if (do_whpx_set_memory(mem, 0)) {
+            error_report("Failed to reset overlapping slot");
+            abort();
+        }
     }
-    start_pa += delta;
-    size -= delta;
-    size &= qemu_real_host_page_mask();
-    if (!size || (start_pa & ~qemu_real_host_page_mask())) {
+
+    if (!add) {
         return;
     }
 
-    host_va = (uintptr_t)memory_region_get_ram_ptr(mr)
-            + section->offset_within_region + delta;
+    if (area->readonly ||
+        (!memory_region_is_ram(area) && memory_region_is_romd(area))) {
+        flags = WHvMapGpaRangeFlagRead | WHvMapGpaRangeFlagExecute;
+    } else {
+        flags = WHvMapGpaRangeFlagRead | WHvMapGpaRangeFlagWrite
+         | WHvMapGpaRangeFlagExecute;
+    }
+
+    /* Now make a new slot. */
+    int x;
+
+    for (x = 0; x < whpx_state->num_slots; ++x) {
+        mem = &whpx_state->slots[x];
+        if (!mem->size) {
+            break;
+        }
+    }
+
+    if (x == whpx_state->num_slots) {
+        error_report("No free slots");
+        abort();
+    }
 
-    whpx_update_mapping(start_pa, size, (void *)(uintptr_t)host_va, add,
-                        memory_region_is_rom(mr), mr->name);
+    mem->size = int128_get64(section->size);
+    mem->mem = memory_region_get_ram_ptr(area) + section->offset_within_region;
+    mem->start = section->offset_within_address_space;
+    mem->region = area;
+
+    if (do_whpx_set_memory(mem, flags)) {
+        error_report("Error registering new memory slot");
+        abort();
+    }
 }
 
 static void whpx_region_add(MemoryListener *listener,
                            MemoryRegionSection *section)
 {
-    memory_region_ref(section->mr);
-    whpx_process_section(section, 1);
+    whpx_set_phys_mem(section, true);
 }
 
 static void whpx_region_del(MemoryListener *listener,
                            MemoryRegionSection *section)
 {
-    whpx_process_section(section, 0);
-    memory_region_unref(section->mr);
+    whpx_set_phys_mem(section, false);
 }
 
 static void whpx_transaction_begin(MemoryListener *listener)
@@ -524,6 +609,14 @@ static void whpx_accel_instance_init(Object *obj)
     memset(whpx, 0, sizeof(struct whpx_state));
     /* Turn on kernel-irqchip, by default */
     whpx->kernel_irqchip_allowed = true;
+
+    int x;
+    whpx_state = malloc(sizeof(WHPXState));
+    whpx_state->num_slots = ARRAY_SIZE(whpx_state->slots);
+    for (x = 0; x < whpx_state->num_slots; ++x) {
+        whpx_state->slots[x].size = 0;
+        whpx_state->slots[x].slot_id = x;
+    }
 }
 
 static const TypeInfo whpx_accel_type = {
Signature · 1 line
2.39.5 (Apple Git-154)
#16 lore raw

PATCH v5 11/18 target/arm: cpu: mark WHPX as supporting PSCI 1.1

In reply to Mohamed Mediouni

Hyper-V supports PSCI 1.3, and that implementation is exposed through WHPX.

---
 target/arm/cpu.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
target/arm/cpu.c +2 -1
index e2b2337399..3b69c9786a 100644
@@ -23,6 +23,7 @@
 #include "qemu/timer.h"
 #include "qemu/log.h"
 #include "exec/page-vary.h"
+#include "system/whpx.h"
 #include "target/arm/idau.h"
 #include "qemu/module.h"
 #include "qapi/error.h"
@@ -1496,7 +1497,7 @@ static void arm_cpu_initfn(Object *obj)
     cpu->psci_version = QEMU_PSCI_VERSION_0_1; /* By default assume PSCI v0.1 */
     cpu->kvm_target = QEMU_KVM_ARM_TARGET_NONE;
 
-    if (tcg_enabled() || hvf_enabled()) {
+    if (tcg_enabled() || hvf_enabled() || whpx_enabled()) {
         /* TCG and HVF implement PSCI 1.1 */
         cpu->psci_version = QEMU_PSCI_VERSION_1_1;
     }
Signature · 1 line
2.39.5 (Apple Git-154)
#17 lore raw

Re: [PATCH v5 11/18] target/arm: cpu: mark WHPX as supporting PSCI 1.1

Quoted 24 lines · Mohamed Mediouni

Hyper-V supports PSCI 1.3, and that implementation is exposed through WHPX.

---
  target/arm/cpu.c | 3 ++-
  1 file changed, 2 insertions(+), 1 deletion(-)
target/arm/cpu.c +1 -0
index e2b2337399..3b69c9786a 100644
@@ -23,6 +23,7 @@
  #include "qemu/timer.h"
  #include "qemu/log.h"
  #include "exec/page-vary.h"
+#include "system/whpx.h"
  #include "target/arm/idau.h"
  #include "qemu/module.h"
  #include "qapi/error.h"

target/arm/cpu.c static void arm_cpu_initfn(Object *obj) · line 1501

Quoted 6 lines
target/arm/cpu.c +1 -1
@@ -1496,7 +1497,7 @@ static void arm_cpu_initfn(Object *obj)
      cpu->psci_version = QEMU_PSCI_VERSION_0_1; /* By default assume PSCI v0.1 */
      cpu->kvm_target = QEMU_KVM_ARM_TARGET_NONE;
 
-    if (tcg_enabled() || hvf_enabled()) {
+    if (tcg_enabled() || hvf_enabled() || whpx_enabled()) {
          /* TCG and HVF implement PSCI 1.1 */

(Update comment, or simply remove it?)

target/arm/cpu.c static void arm_cpu_initfn(Object *obj) · line 1502

Quoted 1 line
target/arm/cpu.c +0 -0
          cpu->psci_version = QEMU_PSCI_VERSION_1_1;

Maybe we could implement SYSTEM_RESET2, SYSTEM_OFF2 and MEM_PROTECT* and bump to QEMU_PSCI_VERSION_1_3. Or add UNIMP stubs and here directly start with:

       } else if (whpx_enabled()) {
           cpu->psci_version = QEMU_PSCI_VERSION_1_3;
Quoted 1 line
target/arm/cpu.c +0 -0
      }
#18 lore raw

PATCH v5 12/18 hw/arm: virt: cleanly fail on attempt to use the platform vGIC together with ITS

In reply to Mohamed Mediouni

Windows Hypervisor Platform's vGIC doesn't support ITS.
Deal with this by reporting to the user and not creating the ITS device.

Regular configuration: GICv3 + ITS
Resulting configuration here: GICv3 with no MSIs
And its=off explicitly for the newest machine version: GICv3 + GICv2m

---
 hw/arm/virt.c | 10 ++++++++++
 1 file changed, 10 insertions(+)
hw/arm/virt.c +10 -0
index 98a1c74c42..0039f6a12b 100644
@@ -741,6 +741,16 @@ static void create_its(VirtMachineState *vms)
         return;
     }
 
+    if (whpx_enabled() && vms->tcg_its) {
+        /*
+         * Signal to the user when ITS is neither supported by the host
+         * nor emulated by the machine.
+         */
+        info_report("ITS not supported on WHPX.");
+        info_report("To support MSIs, use its=off to enable GICv3 + GICv2m.");
+        return;
+    }
+
     dev = qdev_new(its_class_name());
 
     object_property_set_link(OBJECT(dev), "parent-gicv3", OBJECT(vms->gic),
Signature · 1 line
2.39.5 (Apple Git-154)
#19 lore raw

PATCH v5 13/18 whpx: arm64: clamp down IPA size

In reply to Mohamed Mediouni

Code taken from HVF and adapted for WHPX use. Note that WHPX doesn't have a default vs maximum IPA distinction.

---
 hw/arm/virt.c               | 32 ++++++++++++++++++++++++++
 include/hw/boards.h         |  1 +
 target/arm/whpx/meson.build |  2 ++
 target/arm/whpx/whpx-all.c  | 45 +++++++++++++++++++++++++++++++++++++
 target/arm/whpx/whpx-stub.c | 15 +++++++++++++
 target/arm/whpx_arm.h       | 16 +++++++++++++
 6 files changed, 111 insertions(+)
 create mode 100644 target/arm/whpx/whpx-stub.c
 create mode 100644 target/arm/whpx_arm.h
hw/arm/virt.c +32 -0
index 0039f6a12b..f90c3fc113 100644
@@ -70,6 +70,7 @@
 #include "hw/irq.h"
 #include "kvm_arm.h"
 #include "hvf_arm.h"
+#include "whpx_arm.h"
 #include "hw/firmware/smbios.h"
 #include "qapi/visitor.h"
 #include "qapi/qapi-visit-common.h"
@@ -3162,6 +3163,36 @@ static int virt_kvm_type(MachineState *ms, const char *type_str)
     return fixed_ipa ? 0 : requested_pa_size;
 }
 
+static int virt_whpx_get_physical_address_range(MachineState *ms)
+{
+    VirtMachineState *vms = VIRT_MACHINE(ms);
+
+    int max_ipa_size = whpx_arm_get_ipa_bit_size();
+
+    /* We freeze the memory map to compute the highest gpa */
+    virt_set_memmap(vms, max_ipa_size);
+
+    int requested_ipa_size = 64 - clz64(vms->highest_gpa);
+
+    /*
+     * If we're <= the default IPA size just use the default.
+     * If we're above the default but below the maximum, round up to
+     * the maximum. hvf_arm_get_max_ipa_bit_size() conveniently only
+     * returns values that are valid ARM PARange values.
+     */
+    if (requested_ipa_size <= max_ipa_size) {
+        requested_ipa_size = max_ipa_size;
+    } else {
+        error_report("-m and ,maxmem option values "
+                     "require an IPA range (%d bits) larger than "
+                     "the one supported by the host (%d bits)",
+                     requested_ipa_size, max_ipa_size);
+        return -1;
+    }
+
+    return requested_ipa_size;
+}
+
 static int virt_hvf_get_physical_address_range(MachineState *ms)
 {
     VirtMachineState *vms = VIRT_MACHINE(ms);
@@ -3256,6 +3287,7 @@ static void virt_machine_class_init(ObjectClass *oc, const void *data)
     mc->get_default_cpu_node_id = virt_get_default_cpu_node_id;
     mc->kvm_type = virt_kvm_type;
     mc->hvf_get_physical_address_range = virt_hvf_get_physical_address_range;
+    mc->whpx_get_physical_address_range = virt_whpx_get_physical_address_range;
     assert(!mc->get_hotplug_handler);
     mc->get_hotplug_handler = virt_machine_get_hotplug_handler;
     hc->pre_plug = virt_machine_device_pre_plug_cb;
include/hw/boards.h +1 -0
index f94713e6e2..dde8013bd7 100644
@@ -276,6 +276,7 @@ struct MachineClass {
     void (*wakeup)(MachineState *state);
     int (*kvm_type)(MachineState *machine, const char *arg);
     int (*hvf_get_physical_address_range)(MachineState *machine);
+    int (*whpx_get_physical_address_range)(MachineState *machine);
 
     BlockInterfaceType block_default_type;
     int units_per_default_bus;
target/arm/whpx/meson.build +2 -0
index 1de2ef0283..3df632c9d3 100644
@@ -1,3 +1,5 @@
 arm_system_ss.add(when: 'CONFIG_WHPX', if_true: files(
   'whpx-all.c',
 ))
+
+arm_common_system_ss.add(when: 'CONFIG_WHPX', if_false: files('whpx-stub.c'))
target/arm/whpx/whpx-all.c +45 -0

75 lines not on this page · raw message

target/arm/whpx/whpx-stub.c new file +15 -0
new file mode 100644
index 0000000000..32e434a5f6
@@ -0,0 +1,15 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * WHPX stubs for ARM
+ *
+ *  Copyright (c) 2025 Mohamed Mediouni
+ *
+ */
+
+#include "qemu/osdep.h"
+#include "whpx_arm.h"
+
+uint32_t whpx_arm_get_ipa_bit_size(void)
+{
+    g_assert_not_reached();
+}
target/arm/whpx_arm.h new file +16 -0
new file mode 100644
index 0000000000..de7406b66f
@@ -0,0 +1,16 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * WHPX support -- ARM specifics
+ *
+ * Copyright (c) 2025 Mohamed Mediouni
+ *
+ */
+
+#ifndef QEMU_WHPX_ARM_H
+#define QEMU_WHPX_ARM_H
+
+#include "target/arm/cpu-qom.h"
+
+uint32_t whpx_arm_get_ipa_bit_size(void);
+
+#endif
Signature · 1 line
2.39.5 (Apple Git-154)
#20 lore raw

PATCH v5 14/18 hw/arm, accel/hvf, whpx: unify get_physical_address_range between WHPX and HVF

In reply to Mohamed Mediouni
---
 accel/hvf/hvf-all.c        |  7 +++++--
 hw/arm/virt.c              | 41 ++++----------------------------------
 include/hw/boards.h        |  4 ++--
 include/system/hvf_int.h   |  2 ++
 target/arm/hvf-stub.c      | 20 -------------------
 target/arm/hvf/hvf.c       |  6 +++---
 target/arm/hvf_arm.h       |  3 ---
 target/arm/meson.build     |  1 -
 target/arm/whpx/whpx-all.c |  5 +++--
 target/i386/hvf/hvf.c      | 11 ++++++++++
 10 files changed, 30 insertions(+), 70 deletions(-)
 delete mode 100644 target/arm/hvf-stub.c
accel/hvf/hvf-all.c +5 -2
index 0a4b498e83..8229ad8640 100644
@@ -17,6 +17,7 @@
 #include "system/hvf_int.h"
 #include "hw/core/cpu.h"
 #include "hw/boards.h"
+#include "target/arm/hvf_arm.h"
 #include "trace.h"
 
 bool hvf_allowed;
@@ -256,8 +257,10 @@ static int hvf_accel_init(AccelState *as, MachineState *ms)
     int pa_range = 36;
     MachineClass *mc = MACHINE_GET_CLASS(ms);
 
-    if (mc->hvf_get_physical_address_range) {
-        pa_range = mc->hvf_get_physical_address_range(ms);
+
+    if (mc->get_physical_address_range) {
+        pa_range = mc->get_physical_address_range(ms,
+            hvf_arch_get_default_ipa_bit_size(), hvf_arch_get_max_ipa_bit_size());
         if (pa_range < 0) {
             return -EINVAL;
         }
hw/arm/virt.c +4 -37

62 lines not on this page · raw message

include/hw/boards.h +2 -2
index dde8013bd7..fa3843a11b 100644
@@ -275,8 +275,8 @@ struct MachineClass {
     void (*reset)(MachineState *state, ResetType type);
     void (*wakeup)(MachineState *state);
     int (*kvm_type)(MachineState *machine, const char *arg);
-    int (*hvf_get_physical_address_range)(MachineState *machine);
-    int (*whpx_get_physical_address_range)(MachineState *machine);
+    int (*get_physical_address_range)(MachineState *machine,
+        int default_ipa_size, int max_ipa_size);
 
     BlockInterfaceType block_default_type;
     int units_per_default_bus;
include/system/hvf_int.h +2 -0
index a3b06a3e75..8b6447c238 100644
@@ -71,6 +71,8 @@ void assert_hvf_ok_impl(hv_return_t ret, const char *file, unsigned int line,
 const char *hvf_return_string(hv_return_t ret);
 int hvf_arch_init(void);
 hv_return_t hvf_arch_vm_create(MachineState *ms, uint32_t pa_range);
+uint32_t hvf_arch_get_default_ipa_bit_size(void);
+uint32_t hvf_arch_get_max_ipa_bit_size(void);
 int hvf_arch_init_vcpu(CPUState *cpu);
 void hvf_arch_vcpu_destroy(CPUState *cpu);
 int hvf_vcpu_exec(CPUState *);
target/arm/hvf-stub.c deleted +0 -20
deleted file mode 100644
index ff137267a0..0000000000
@@ -1,20 +0,0 @@
-/*
- * QEMU Hypervisor.framework (HVF) stubs for ARM
- *
- *  Copyright (c) Linaro
- *
- * SPDX-License-Identifier: GPL-2.0-or-later
- */
-
-#include "qemu/osdep.h"
-#include "hvf_arm.h"
-
-uint32_t hvf_arm_get_default_ipa_bit_size(void)
-{
-    g_assert_not_reached();
-}
-
-uint32_t hvf_arm_get_max_ipa_bit_size(void)
-{
-    g_assert_not_reached();
-}
target/arm/hvf/hvf.c +3 -3
index 47b0cd3a35..58135d1e18 100644
@@ -850,7 +850,7 @@ static uint64_t hvf_get_reg(CPUState *cpu, int rt)
 static void clamp_id_aa64mmfr0_parange_to_ipa_size(ARMISARegisters *isar)
 {
     uint32_t ipa_size = chosen_ipa_bit_size ?
-            chosen_ipa_bit_size : hvf_arm_get_max_ipa_bit_size();
+            chosen_ipa_bit_size : hvf_arch_get_max_ipa_bit_size();
     uint64_t id_aa64mmfr0;
 
     /* Clamp down the PARange to the IPA size the kernel supports. */
@@ -940,7 +940,7 @@ static bool hvf_arm_get_host_cpu_features(ARMHostCPUFeatures *ahcf)
     return r == HV_SUCCESS;
 }
 
-uint32_t hvf_arm_get_default_ipa_bit_size(void)
+uint32_t hvf_arch_get_default_ipa_bit_size(void)
 {
     uint32_t default_ipa_size;
     hv_return_t ret = hv_vm_config_get_default_ipa_size(&default_ipa_size);
@@ -949,7 +949,7 @@ uint32_t hvf_arm_get_default_ipa_bit_size(void)
     return default_ipa_size;
 }
 
-uint32_t hvf_arm_get_max_ipa_bit_size(void)
+uint32_t hvf_arch_get_max_ipa_bit_size(void)
 {
     uint32_t max_ipa_size;
     hv_return_t ret = hv_vm_config_get_max_ipa_size(&max_ipa_size);
target/arm/hvf_arm.h +0 -3
index ea82f2691d..5d19d82e5d 100644
@@ -22,7 +22,4 @@ void hvf_arm_init_debug(void);
 
 void hvf_arm_set_cpu_features_from_host(ARMCPU *cpu);
 
-uint32_t hvf_arm_get_default_ipa_bit_size(void);
-uint32_t hvf_arm_get_max_ipa_bit_size(void);
-
 #endif
target/arm/meson.build +0 -1
index e28bd3f8e2..f0f1b55e43 100644
@@ -32,7 +32,6 @@ arm_common_system_ss.add(files('cpu.c'))
 arm_common_system_ss.add(when: 'TARGET_AARCH64', if_false: files(
   'cpu32-stubs.c'))
 arm_common_system_ss.add(when: 'CONFIG_KVM', if_false: files('kvm-stub.c'))
-arm_common_system_ss.add(when: 'CONFIG_HVF', if_false: files('hvf-stub.c'))
 arm_common_system_ss.add(files(
   'arch_dump.c',
   'arm-powerctl.c',
target/arm/whpx/whpx-all.c +3 -2
index c18ac4ebc5..3dcccfe7fd 100644
@@ -806,8 +806,9 @@ int whpx_accel_init(AccelState *as, MachineState *ms)
         goto error;
     }
 
-    if (mc->whpx_get_physical_address_range) {
-        pa_range = mc->whpx_get_physical_address_range(ms);
+    if (mc->get_physical_address_range) {
+        pa_range = mc->get_physical_address_range(ms,
+            whpx_arm_get_ipa_bit_size(), whpx_arm_get_ipa_bit_size());
         if (pa_range < 0) {
             return -EINVAL;
         }
target/i386/hvf/hvf.c +11 -0
index 818b50419f..a6065da00b 100644
@@ -225,6 +225,17 @@ int hvf_arch_init(void)
     return 0;
 }
 
+/* 48-bit on all Intel Macs. Function currently unused. */
+uint32_t hvf_arch_get_default_ipa_bit_size(void)
+{
+    g_assert_not_reached();
+}
+
+uint32_t hvf_arch_get_max_ipa_bit_size(void)
+{
+    g_assert_not_reached();
+}
+
 hv_return_t hvf_arch_vm_create(MachineState *ms, uint32_t pa_range)
 {
     return hv_vm_create(HV_VM_DEFAULT);
Signature · 1 line
2.39.5 (Apple Git-154)
#21 lore raw

PATCH v5 15/18 whpx: arm64: implement -cpu host

In reply to Mohamed Mediouni

OpenProcessorKey and ReadRegU64 adapted from:
https://github.com/FEX-Emu/FEX/blob/e6de17e72ef03aa88ba14fa0ec13163061608c74/Source/Windows/Common/CPUFeatures.cpp#L62

---
 hw/arm/virt.c              |   2 +-
 target/arm/cpu64.c         |  19 +++++--
 target/arm/whpx/whpx-all.c | 110 +++++++++++++++++++++++++++++++++++++
 target/arm/whpx_arm.h      |   1 +
 4 files changed, 125 insertions(+), 7 deletions(-)
hw/arm/virt.c +1 -1
index c980f59e82..a23edeb2d2 100644
@@ -3217,7 +3217,7 @@ static void virt_machine_class_init(ObjectClass *oc, const void *data)
 #ifdef TARGET_AARCH64
         ARM_CPU_TYPE_NAME("cortex-a53"),
         ARM_CPU_TYPE_NAME("cortex-a57"),
-#if defined(CONFIG_KVM) || defined(CONFIG_HVF)
+#if defined(CONFIG_KVM) || defined(CONFIG_HVF) || defined(CONFIG_WHPX)
         ARM_CPU_TYPE_NAME("host"),
 #endif /* CONFIG_KVM || CONFIG_HVF */
 #endif /* TARGET_AARCH64 */
target/arm/cpu64.c +13 -6
index 26cf7e6dfa..3f00071081 100644
@@ -26,10 +26,13 @@
 #include "qemu/units.h"
 #include "system/kvm.h"
 #include "system/hvf.h"
+#include "system/whpx.h"
+#include "system/hw_accel.h"
 #include "system/qtest.h"
 #include "system/tcg.h"
 #include "kvm_arm.h"
 #include "hvf_arm.h"
+#include "whpx_arm.h"
 #include "qapi/visitor.h"
 #include "hw/qdev-properties.h"
 #include "internals.h"
@@ -522,7 +525,7 @@ void arm_cpu_pauth_finalize(ARMCPU *cpu, Error **errp)
     isar2 = FIELD_DP64(isar2, ID_AA64ISAR2, APA3, 0);
     isar2 = FIELD_DP64(isar2, ID_AA64ISAR2, GPA3, 0);
 
-    if (kvm_enabled() || hvf_enabled()) {
+    if (hwaccel_enabled()) {
         /*
          * Exit early if PAuth is enabled and fall through to disable it.
          * The algorithm selection properties are not present.
@@ -599,10 +602,10 @@ void aarch64_add_pauth_properties(Object *obj)
 
     /* Default to PAUTH on, with the architected algorithm on TCG. */
     qdev_property_add_static(DEVICE(obj), &arm_cpu_pauth_property);
-    if (kvm_enabled() || hvf_enabled()) {
+    if (hwaccel_enabled()) {
         /*
          * Mirror PAuth support from the probed sysregs back into the
-         * property for KVM or hvf. Is it just a bit backward? Yes it is!
+         * property for HW accel. Is it just a bit backward? Yes it is!
          * Note that prop_pauth is true whether the host CPU supports the
          * architected QARMA5 algorithm or the IMPDEF one. We don't
          * provide the separate pauth-impdef property for KVM or hvf,
@@ -773,6 +776,10 @@ static void aarch64_host_initfn(Object *obj)
     ARMCPU *cpu = ARM_CPU(obj);
     hvf_arm_set_cpu_features_from_host(cpu);
     aarch64_add_pauth_properties(obj);
+#elif defined(CONFIG_WHPX)
+    ARMCPU *cpu = ARM_CPU(obj);
+    whpx_arm_set_cpu_features_from_host(cpu);
+    aarch64_add_pauth_properties(obj);
 #else
     g_assert_not_reached();
 #endif
@@ -780,8 +787,8 @@ static void aarch64_host_initfn(Object *obj)
 
 static void aarch64_max_initfn(Object *obj)
 {
-    if (kvm_enabled() || hvf_enabled()) {
-        /* With KVM or HVF, '-cpu max' is identical to '-cpu host' */
+    if (hwaccel_enabled()) {
+        /* When hardware acceleration enabled, '-cpu max' is identical to '-cpu host' */
         aarch64_host_initfn(obj);
         return;
     }
@@ -800,7 +807,7 @@ static const ARMCPUInfo aarch64_cpus[] = {
     { .name = "cortex-a57",         .initfn = aarch64_a57_initfn },
     { .name = "cortex-a53",         .initfn = aarch64_a53_initfn },
     { .name = "max",                .initfn = aarch64_max_initfn },
-#if defined(CONFIG_KVM) || defined(CONFIG_HVF)
+#if defined(CONFIG_KVM) || defined(CONFIG_HVF) || defined(CONFIG_WHPX)
     { .name = "host",               .initfn = aarch64_host_initfn },
 #endif
 };
target/arm/whpx/whpx-all.c +110 -0

122 lines not on this page · raw message

target/arm/whpx_arm.h +1 -0
index de7406b66f..df65fd753c 100644
@@ -12,5 +12,6 @@
 #include "target/arm/cpu-qom.h"
 
 uint32_t whpx_arm_get_ipa_bit_size(void);
+void whpx_arm_set_cpu_features_from_host(ARMCPU *cpu);
 
 #endif
Signature · 1 line
2.39.5 (Apple Git-154)
#22 lore raw

Re: [PATCH v5 15/18] whpx: arm64: implement -cpu host

target/arm/whpx/whpx-all.c static void clamp_id_aa64mmfr0_parange_to_ipa_size(ARMISARegisters *isar) · line 763

Quoted 12 lines · Mohamed Mediouni

OpenProcessorKey and ReadRegU64 adapted from:
https://github.com/FEX-Emu/FEX/blob/e6de17e72ef03aa88ba14fa0ec13163061608c74/Source/Windows/Common/CPUFeatures.cpp#L62

target/arm/whpx/whpx-all.c +6 -0
+    /*
+     * MIDR_EL1 is not a global register on WHPX
+     * As such, read the CPU0 from the registry to get a consistent value.
+     * Otherwise, on heterogenous systems, you'll get variance between CPUs.
+     */
+    HKEY ProcessorKey = OpenProcessorKey();

Can you follow the QEMU coding style, please (here and elsewhere)? Variables and function names should be all lower case, and variable declarations go at the start of a C code block, not in the middle of one.

thanks
-- PMM

#23 lore raw

Re: [PATCH v5 15/18] whpx: arm64: implement -cpu host

Quoted 15 lines · Peter Maydell

target/arm/whpx/whpx-all.c static void clamp_id_aa64mmfr0_parange_to_ipa_size(ARMISARegisters *isar) · line 763

Quoted 12 lines · Mohamed Mediouni

OpenProcessorKey and ReadRegU64 adapted from:
https://github.com/FEX-Emu/FEX/blob/e6de17e72ef03aa88ba14fa0ec13163061608c74/Source/Windows/Common/CPUFeatures.cpp#L62

target/arm/whpx/whpx-all.c +6 -0
+    /*
+     * MIDR_EL1 is not a global register on WHPX
+     * As such, read the CPU0 from the registry to get a consistent value.
+     * Otherwise, on heterogenous systems, you'll get variance between CPUs.
+     */
+    HKEY ProcessorKey = OpenProcessorKey();
Quoted 5 lines

Can you follow the QEMU coding style, please (here and elsewhere)? Variables and function names should be all lower case, and variable declarations go at the start of a C code block, not in the middle of one.

In some cases, including in this function, I feel that the rule to declare variables at the start of a block is not really helpful, and is more related to legacy C than a real point nowadays. As well, it sometimes forces to reuse some variables between various sub blocks, which definitely can create bugs.

Anyway, I'm not discussing the existing QEMU coding style, but just asking if for the current context, is it really a problem to declare variable here?

Quoted 2 lines

thanks
-- PMM

#24 lore raw

Re: [PATCH v5 15/18] whpx: arm64: implement -cpu host

Quoted 15 lines · Pierrick Bouvier
Quoted 6 lines · Peter Maydell

On Fri, 8 Aug 2025 at 07:55, Mohamed Mediouni <[email protected]> wrote:
Can you follow the QEMU coding style, please (here and elsewhere)? Variables and function names should be all lower case,
and variable declarations go at the start of a C code
block, not in the middle of one.

In some cases, including in this function, I feel that the rule to declare variables at the start of a block is not really helpful, and is more related to legacy C than a real point nowadays.
As well, it sometimes forces to reuse some variables between various sub blocks, which definitely can create bugs.

Quoted 3 lines

Anyway, I'm not discussing the existing QEMU coding style, but just asking if for the current context, is it really a problem to declare variable here?

The point of a coding style is to aim for consistency. QEMU is pretty terrible at being consistent, but we should try. The rule about variables at start of block is not because some compilers fail to compile it, but because we think it's overall more readable that way.

thanks
-- PMM

#25 lore raw

Re: [PATCH v5 15/18] whpx: arm64: implement -cpu host

Quoted 21 lines · Peter Maydell
Quoted 15 lines · Pierrick Bouvier
Quoted 6 lines · Peter Maydell

On Fri, 8 Aug 2025 at 07:55, Mohamed Mediouni <[email protected]> wrote:
Can you follow the QEMU coding style, please (here and elsewhere)? Variables and function names should be all lower case,
and variable declarations go at the start of a C code
block, not in the middle of one.

In some cases, including in this function, I feel that the rule to declare variables at the start of a block is not really helpful, and is more related to legacy C than a real point nowadays.
As well, it sometimes forces to reuse some variables between various sub blocks, which definitely can create bugs.

Quoted 3 lines

Anyway, I'm not discussing the existing QEMU coding style, but just asking if for the current context, is it really a problem to declare variable here?

Quoted 5 lines

The point of a coding style is to aim for consistency. QEMU is pretty terrible at being consistent, but we should try. The rule about variables at start of block is not because some compilers fail to compile it, but because we think it's overall more readable that way.

There are also potential[1] functional problems with not declaring at the start of block, because if you have a "goto cleanup" which jumps over the line of the declaration, the variable will have undefined state when the 'cleanup:' block is running. This is something which is very subtle and easily missed when reading the code flow.

With regards,
Daniel

[1] I say "potential" because we unconditionally build with -ftrivial-auto-var-init=zero to mitigate this danger in practice.

Signature · 3 lines
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|
#26 lore raw

Re: [PATCH v5 15/18] whpx: arm64: implement -cpu host

Quoted 28 lines · Daniel P. Berrangé
Quoted 21 lines · Peter Maydell
Quoted 15 lines · Pierrick Bouvier
Quoted 6 lines · Peter Maydell

On Fri, 8 Aug 2025 at 07:55, Mohamed Mediouni <[email protected]> wrote:
Can you follow the QEMU coding style, please (here and elsewhere)? Variables and function names should be all lower case,
and variable declarations go at the start of a C code
block, not in the middle of one.

In some cases, including in this function, I feel that the rule to declare variables at the start of a block is not really helpful, and is more related to legacy C than a real point nowadays.
As well, it sometimes forces to reuse some variables between various sub blocks, which definitely can create bugs.

Quoted 3 lines

Anyway, I'm not discussing the existing QEMU coding style, but just asking if for the current context, is it really a problem to declare variable here?

Quoted 5 lines

The point of a coding style is to aim for consistency. QEMU is pretty terrible at being consistent, but we should try. The rule about variables at start of block is not because some compilers fail to compile it, but because we think it's overall more readable that way.

Quoted 7 lines

There are also potential[1] functional problems with not declaring at the start of block, because if you have a "goto cleanup" which jumps over the line of the declaration, the variable will have undefined state when the 'cleanup:' block is running. This is something which is very subtle and easily missed when reading the code flow.

This has nothing to do with where variables are declared, but where they are assigned. The same issue can happen whether or not it's declared at the start of a block.

I suspect we use -ftrivial-auto-var-init precisely because we force variables to be declared at start of the scope, i.e. where they don't have any value yet. So, instead of forcing an explicit initialization or rely on compiler warnings for uninitialized values, it was decided to initialize them to 0 by default.

If we declared them at the point where they have a defined semantic value, this problem would not exist anyway, out of the goto_cleanup situation, which has the same fundamental issue in both cases.

Quoted 6 lines

With regards,
Daniel

[1] I say "potential" because we unconditionally build with -ftrivial-auto-var-init=zero to mitigate this danger in practice.

#27 lore raw

Re: [PATCH v5 15/18] whpx: arm64: implement -cpu host

Quoted 47 lines · Pierrick Bouvier
Quoted 28 lines · Daniel P. Berrangé
Quoted 21 lines · Peter Maydell
Quoted 15 lines · Pierrick Bouvier
Quoted 6 lines · Peter Maydell

On Fri, 8 Aug 2025 at 07:55, Mohamed Mediouni <[email protected]> wrote:
Can you follow the QEMU coding style, please (here and elsewhere)? Variables and function names should be all lower case,
and variable declarations go at the start of a C code
block, not in the middle of one.

In some cases, including in this function, I feel that the rule to declare variables at the start of a block is not really helpful, and is more related to legacy C than a real point nowadays.
As well, it sometimes forces to reuse some variables between various sub blocks, which definitely can create bugs.

Quoted 3 lines

Anyway, I'm not discussing the existing QEMU coding style, but just asking if for the current context, is it really a problem to declare variable here?

Quoted 5 lines

The point of a coding style is to aim for consistency. QEMU is pretty terrible at being consistent, but we should try. The rule about variables at start of block is not because some compilers fail to compile it, but because we think it's overall more readable that way.

Quoted 7 lines

There are also potential[1] functional problems with not declaring at the start of block, because if you have a "goto cleanup" which jumps over the line of the declaration, the variable will have undefined state when the 'cleanup:' block is running. This is something which is very subtle and easily missed when reading the code flow.

This has nothing to do with where variables are declared, but where they are assigned. The same issue can happen whether or not it's declared at the start of a block.

I suspect we use -ftrivial-auto-var-init precisely because we force variables to be declared at start of the scope, i.e. where they don't have any value yet. So, instead of forcing an explicit initialization or rely on compiler warnings for uninitialized values, it was decided to initialize them to 0 by default.

Quoted 3 lines

If we declared them at the point where they have a defined semantic value, this problem would not exist anyway, out of the goto_cleanup situation, which has the same fundamental issue in both cases.

It really isn't the same issue when you compare

  void bar(void) {
    char *foo = NULL;

if (blah)
goto cleanup:

  cleanup:
    if (foo)
       ....
  }

vs

  void bar(void) {
    if (blah)
       goto cleanup:
    char *foo = NULL;

    ...some code...
  cleanup:
    if (foo)
       ....
  }

The late declaration of 'foo' is outright misleading to reviewers.

Its initialization at time of declaration gives the impression that 'foo' has well defined value in the 'cleanup' block, when that is not actually true. In big methods it is very easy to overlook an earlier 'goto' that jumps across a variable declaration and initialization.

Even if not all methods have this problem, the coding standards guide us into the habit of writing code that is immune from this kind of problem. That habit only forms reliably if we apply the coding standards unconditionally, rather than selectively.

With regards,
Daniel

Signature · 3 lines
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|
#28 lore raw

Re: [PATCH v5 15/18] whpx: arm64: implement -cpu host

Quoted 79 lines · Daniel P. Berrangé
Quoted 47 lines · Pierrick Bouvier
Quoted 28 lines · Daniel P. Berrangé
Quoted 21 lines · Peter Maydell
Quoted 15 lines · Pierrick Bouvier
Quoted 6 lines · Peter Maydell

On Fri, 8 Aug 2025 at 07:55, Mohamed Mediouni <[email protected]> wrote:
Can you follow the QEMU coding style, please (here and elsewhere)? Variables and function names should be all lower case,
and variable declarations go at the start of a C code
block, not in the middle of one.

In some cases, including in this function, I feel that the rule to declare variables at the start of a block is not really helpful, and is more related to legacy C than a real point nowadays.
As well, it sometimes forces to reuse some variables between various sub blocks, which definitely can create bugs.

Quoted 3 lines

Anyway, I'm not discussing the existing QEMU coding style, but just asking if for the current context, is it really a problem to declare variable here?

Quoted 5 lines

The point of a coding style is to aim for consistency. QEMU is pretty terrible at being consistent, but we should try. The rule about variables at start of block is not because some compilers fail to compile it, but because we think it's overall more readable that way.

Quoted 7 lines

There are also potential[1] functional problems with not declaring at the start of block, because if you have a "goto cleanup" which jumps over the line of the declaration, the variable will have undefined state when the 'cleanup:' block is running. This is something which is very subtle and easily missed when reading the code flow.

This has nothing to do with where variables are declared, but where they are assigned. The same issue can happen whether or not it's declared at the start of a block.

I suspect we use -ftrivial-auto-var-init precisely because we force variables to be declared at start of the scope, i.e. where they don't have any value yet. So, instead of forcing an explicit initialization or rely on compiler warnings for uninitialized values, it was decided to initialize them to 0 by default.

Quoted 3 lines

If we declared them at the point where they have a defined semantic value, this problem would not exist anyway, out of the goto_cleanup situation, which has the same fundamental issue in both cases.

It really isn't the same issue when you compare

   void bar(void) {
     char *foo = NULL;

if (blah)
goto cleanup:

   cleanup:
     if (foo)
        ....
   }

vs

   void bar(void) {
     if (blah)
        goto cleanup:
     char *foo = NULL;

     ...some code...
   cleanup:>      if (foo)
        ....
   }
Quoted 8 lines

The late declaration of 'foo' is outright misleading to reviewers.

Its initialization at time of declaration gives the impression that 'foo' has well defined value in the 'cleanup' block, when that is not actually true. In big methods it is very easy to overlook an earlier 'goto' that jumps across a variable declaration and initialization.

"Big" method is probably the issue there. If it's not possible to follow control flow in a given function, it's a strong hint there is a problem with its size, independently of any standard. And even though goto_cleanup is a legit pattern in C, I still don't get the argument about declaring variable far from their definition point in this case. It seems that we are trying to solve the consequence without really understanding the root cause issue.

Quoted 5 lines

Even if not all methods have this problem, the coding standards guide us into the habit of writing code that is immune from this kind of problem. That habit only forms reliably if we apply the coding standards unconditionally, rather than selectively.

That's right, but humanly enforced coding standard are usually a waste of time for everyone (reviewers and developers).

How many messages and exchanges on the mailing list could we save by using something like clang-format on the codebase, and force it to be "clean" as part of the CI? There would be no more discussion, as there would be only one single and objective source of truth.

Quoted 2 lines

With regards,
Daniel

#29 lore raw

Re: [PATCH v5 15/18] whpx: arm64: implement -cpu host

Quoted 89 lines · Pierrick Bouvier
Quoted 79 lines · Daniel P. Berrangé
Quoted 47 lines · Pierrick Bouvier
Quoted 28 lines · Daniel P. Berrangé
Quoted 21 lines · Peter Maydell
Quoted 15 lines · Pierrick Bouvier
Quoted 6 lines · Peter Maydell

On Fri, 8 Aug 2025 at 07:55, Mohamed Mediouni <[email protected]> wrote:
Can you follow the QEMU coding style, please (here and elsewhere)? Variables and function names should be all lower case,
and variable declarations go at the start of a C code
block, not in the middle of one.

In some cases, including in this function, I feel that the rule to declare variables at the start of a block is not really helpful, and is more related to legacy C than a real point nowadays.
As well, it sometimes forces to reuse some variables between various sub blocks, which definitely can create bugs.

Quoted 3 lines

Anyway, I'm not discussing the existing QEMU coding style, but just asking if for the current context, is it really a problem to declare variable here?

Quoted 5 lines

The point of a coding style is to aim for consistency. QEMU is pretty terrible at being consistent, but we should try. The rule about variables at start of block is not because some compilers fail to compile it, but because we think it's overall more readable that way.

Quoted 7 lines

There are also potential[1] functional problems with not declaring at the start of block, because if you have a "goto cleanup" which jumps over the line of the declaration, the variable will have undefined state when the 'cleanup:' block is running. This is something which is very subtle and easily missed when reading the code flow.

This has nothing to do with where variables are declared, but where they are assigned. The same issue can happen whether or not it's declared at the start of a block.

I suspect we use -ftrivial-auto-var-init precisely because we force variables to be declared at start of the scope, i.e. where they don't have any value yet. So, instead of forcing an explicit initialization or rely on compiler warnings for uninitialized values, it was decided to initialize them to 0 by default.

Quoted 3 lines

If we declared them at the point where they have a defined semantic value, this problem would not exist anyway, out of the goto_cleanup situation, which has the same fundamental issue in both cases.

It really isn't the same issue when you compare

   void bar(void) {
     char *foo = NULL;

if (blah)
goto cleanup:

   cleanup:
     if (foo)
        ....
   }

vs

   void bar(void) {
     if (blah)
        goto cleanup:
     char *foo = NULL;

     ...some code...
   cleanup:>      if (foo)
        ....
   }
Quoted 8 lines

The late declaration of 'foo' is outright misleading to reviewers.

Its initialization at time of declaration gives the impression that 'foo' has well defined value in the 'cleanup' block, when that is not actually true. In big methods it is very easy to overlook an earlier 'goto' that jumps across a variable declaration and initialization.

Quoted 3 lines

"Big" method is probably the issue there. If it's not possible to follow control flow in a given function, it's a strong hint there is a problem with its size, independently of any standard.

Certainly some methods are too big & deserve refactoring, but that's a non-trivial investment, and it isn't always a clearcut win to split code out into a bunch of arbitrarily short methods. You may solve the goto/initialization problem, but make other things harder as you often still have to fully page all the code into mind to understand it.

Quoted 8 lines
Quoted 5 lines

Even if not all methods have this problem, the coding standards guide us into the habit of writing code that is immune from this kind of problem. That habit only forms reliably if we apply the coding standards unconditionally, rather than selectively.

That's right, but humanly enforced coding standard are usually a waste of time for everyone (reviewers and developers).

Human enforced standards are absolutely better than a free-for-all. Over time contributors will gain familiarity with the project standards and largely comply without enforcement being required. If contributors repeatedly ignore coding standards, it will disincentivise reviewers from looking at their patches.

Quoted 4 lines

How many messages and exchanges on the mailing list could we save by using something like clang-format on the codebase, and force it to be "clean" as part of the CI? There would be no more discussion, as there would be only one single and objective source of truth.

I would really love if it we could apply clang-format to everything, but that has a non-trivial impact on maint when done on a large pre-existing codebase like QEMU. Cherry-picking to upstream stable or distros would be immensely painful, verging on impossible, after a bulk reformat. For any new codebase I'd go for clang-format every time.

With regards,
Daniel

Signature · 3 lines
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|
#30 lore raw

Re: [PATCH v5 15/18] whpx: arm64: implement -cpu host

Quoted 94 lines · Daniel P. Berrangé
Quoted 89 lines · Pierrick Bouvier
Quoted 79 lines · Daniel P. Berrangé
Quoted 47 lines · Pierrick Bouvier
Quoted 28 lines · Daniel P. Berrangé
Quoted 21 lines · Peter Maydell
Quoted 15 lines · Pierrick Bouvier
Quoted 6 lines · Peter Maydell

On Fri, 8 Aug 2025 at 07:55, Mohamed Mediouni <[email protected]> wrote:
Can you follow the QEMU coding style, please (here and elsewhere)? Variables and function names should be all lower case,
and variable declarations go at the start of a C code
block, not in the middle of one.

In some cases, including in this function, I feel that the rule to declare variables at the start of a block is not really helpful, and is more related to legacy C than a real point nowadays.
As well, it sometimes forces to reuse some variables between various sub blocks, which definitely can create bugs.

Quoted 3 lines

Anyway, I'm not discussing the existing QEMU coding style, but just asking if for the current context, is it really a problem to declare variable here?

Quoted 5 lines

The point of a coding style is to aim for consistency. QEMU is pretty terrible at being consistent, but we should try. The rule about variables at start of block is not because some compilers fail to compile it, but because we think it's overall more readable that way.

Quoted 7 lines

There are also potential[1] functional problems with not declaring at the start of block, because if you have a "goto cleanup" which jumps over the line of the declaration, the variable will have undefined state when the 'cleanup:' block is running. This is something which is very subtle and easily missed when reading the code flow.

This has nothing to do with where variables are declared, but where they are assigned. The same issue can happen whether or not it's declared at the start of a block.

I suspect we use -ftrivial-auto-var-init precisely because we force variables to be declared at start of the scope, i.e. where they don't have any value yet. So, instead of forcing an explicit initialization or rely on compiler warnings for uninitialized values, it was decided to initialize them to 0 by default.

Quoted 3 lines

If we declared them at the point where they have a defined semantic value, this problem would not exist anyway, out of the goto_cleanup situation, which has the same fundamental issue in both cases.

It really isn't the same issue when you compare

    void bar(void) {
      char *foo = NULL;

      if (blah)
         goto cleanup:

    cleanup:
      if (foo)
         ....
    }

vs

    void bar(void) {
      if (blah)
         goto cleanup:

      char *foo = NULL;

      ...some code...

    cleanup:>      if (foo)
         ....
    }
Quoted 8 lines

The late declaration of 'foo' is outright misleading to reviewers.

Its initialization at time of declaration gives the impression that 'foo' has well defined value in the 'cleanup' block, when that is not actually true. In big methods it is very easy to overlook an earlier 'goto' that jumps across a variable declaration and initialization.

Quoted 3 lines

"Big" method is probably the issue there. If it's not possible to follow control flow in a given function, it's a strong hint there is a problem with its size, independently of any standard.

Quoted 6 lines

Certainly some methods are too big & deserve refactoring, but that's a non-trivial investment, and it isn't always a clearcut win to split code out into a bunch of arbitrarily short methods. You may solve the goto/initialization problem, but make other things harder as you often still have to fully page all the code into mind to understand it.

I am still looking for an example of where breaking down a big function in smaller logical chunks has reduced the readability for anyone, but I never met one so far in my professional life. Usually it comes with the additional benefit that you need to *name* things explicitely, which is usually better than add a comment about them.
The point of breaking down code is explicitely to remove the need to keep things in mind and assume functions do what they are named for.

I respect the difference about tastes concerning readability and code structure, and I know the context and era from which QEMU codebase comes from.

However, arguing that variables should be at the start of a block because of a potential goto_cleanup situation is not a good argument.

Quoted 6 lines
Quoted 6 lines
Quoted 5 lines

Even if not all methods have this problem, the coding standards guide us into the habit of writing code that is immune from this kind of problem. That habit only forms reliably if we apply the coding standards unconditionally, rather than selectively.

Quoted 9 lines
Quoted 2 lines

That's right, but humanly enforced coding standard are usually a waste of time for everyone (reviewers and developers).

Human enforced standards are absolutely better than a free-for-all. Over time contributors will gain familiarity with the project standards and largely comply without enforcement being required. If contributors repeatedly ignore coding standards, it will disincentivise reviewers from looking at their patches.

Sure.
As well, incessant pushbacks and nitpicking from those same reviewers can disincentivise people to send any patch. But maybe the whole point is simply to keep people out of their lawn, or reduce the amount of patches they need to process daily, who knows.

Quoted 11 lines
Quoted 4 lines

How many messages and exchanges on the mailing list could we save by using something like clang-format on the codebase, and force it to be "clean" as part of the CI? There would be no more discussion, as there would be only one single and objective source of truth.

I would really love if it we could apply clang-format to everything, but that has a non-trivial impact on maint when done on a large pre-existing codebase like QEMU. Cherry-picking to upstream stable or distros would be immensely painful, verging on impossible, after a bulk reformat. For any new codebase I'd go for clang-format every time.

Maybe we could organize a conversation about this, because the benefits are worth making cherry-picking a little bit harder. In this case, all the community benefits from this, while blocking this penalizes everyone except the stable maintenance part and downstream forks. As well, it would be a once in a lifetime price to pay.

Quoted 2 lines

With regards,
Daniel

#31 lore raw

PATCH v5 16/18 target/arm: whpx: instantiate GIC early

In reply to Mohamed Mediouni

While figuring out a better spot for it, put it in whpx_accel_init.

Needs to be done before WHvSetupPartition.

---
 target/arm/whpx/whpx-all.c | 23 +++++++++++++++++++++++
 1 file changed, 23 insertions(+)
target/arm/whpx/whpx-all.c +23 -0
index b33473dcd3..6b839bd558 100644
@@ -980,6 +980,29 @@ int whpx_accel_init(AccelState *as, MachineState *ms)
 
     memset(&prop, 0, sizeof(WHV_PARTITION_PROPERTY));
 
+    WHV_ARM64_IC_PARAMETERS ic_params = {
+        .EmulationMode = WHvArm64IcEmulationModeGicV3,
+        .GicV3Parameters = {
+            .GicdBaseAddress = 0x08000000,
+            .GitsTranslaterBaseAddress = 0x08080000,
+            .GicLpiIntIdBits = 0,
+            .GicPpiPerformanceMonitorsInterrupt = VIRTUAL_PMU_IRQ,
+            .GicPpiOverflowInterruptFromCntv = ARCH_TIMER_VIRT_IRQ
+        }
+    };
+    prop.Arm64IcParameters = ic_params;
+
+    hr = whp_dispatch.WHvSetPartitionProperty(
+            whpx->partition,
+            WHvPartitionPropertyCodeArm64IcParameters,
+            &prop,
+            sizeof(WHV_PARTITION_PROPERTY));
+    if (FAILED(hr)) {
+        error_report("WHPX: Failed to enable GICv3 interrupt controller, hr=%08lx", hr);
+        ret = -EINVAL;
+        goto error;
+    }
+
     hr = whp_dispatch.WHvSetupPartition(whpx->partition);
     if (FAILED(hr)) {
         error_report("WHPX: Failed to setup partition, hr=%08lx", hr);
Signature · 1 line
2.39.5 (Apple Git-154)
#32 lore raw

PATCH v5 17/18 whpx: arm64: gicv3: add migration blocker

In reply to Mohamed Mediouni

GICv3 state save-restore is currently not implemented yet.

---
 hw/intc/arm_gicv3_whpx.c | 10 ++++++++++
 1 file changed, 10 insertions(+)
hw/intc/arm_gicv3_whpx.c +10 -0
index 35dc5ac531..700b2a1a2f 100644
@@ -17,6 +17,7 @@
 #include "system/whpx-internal.h"
 #include "gicv3_internal.h"
 #include "vgic_common.h"
+#include "migration/blocker.h"
 #include "qom/object.h"
 #include "target/arm/cpregs.h"
 
@@ -228,6 +229,15 @@ static void whpx_gicv3_realize(DeviceState *dev, Error **errp)
         error_setg(errp, "Nested virtualisation not currently supported by WHPX.");
         return;
     }
+
+    Error *whpx_migration_blocker = NULL;
+
+    error_setg(&whpx_migration_blocker,
+        "Live migration disabled because GIC state save/restore not supported on WHPX");
+    if (migrate_add_blocker(&whpx_migration_blocker, errp)) {
+        error_free(whpx_migration_blocker);
+        return;
+    }
 }
 
 static void whpx_gicv3_class_init(ObjectClass *klass, const void *data)
Signature · 1 line
2.39.5 (Apple Git-154)
Reviewed-by on v5 17/18
#33 lore raw

Re: [PATCH v5 17/18] whpx: arm64: gicv3: add migration blocker

Quoted 25 lines · Mohamed Mediouni

GICv3 state save-restore is currently not implemented yet.

---
  hw/intc/arm_gicv3_whpx.c | 10 ++++++++++
  1 file changed, 10 insertions(+)
hw/intc/arm_gicv3_whpx.c +3 -0
index 35dc5ac531..700b2a1a2f 100644
@@ -17,6 +17,7 @@
  #include "system/whpx-internal.h"
  #include "gicv3_internal.h"
  #include "vgic_common.h"
+#include "migration/blocker.h"
  #include "qom/object.h"
  #include "target/arm/cpregs.h"
 
@@ -228,6 +229,15 @@ static void whpx_gicv3_realize(DeviceState *dev, Error **errp)
          error_setg(errp, "Nested virtualisation not currently supported by WHPX.");
          return;
      }
+
+    Error *whpx_migration_blocker = NULL;

hw/intc/arm_gicv3_whpx.c static void whpx_gicv3_realize(DeviceState *dev, Error **errp) · line 243

Quoted 10 lines
hw/intc/arm_gicv3_whpx.c +7 -0
+
+    error_setg(&whpx_migration_blocker,
+        "Live migration disabled because GIC state save/restore not supported on WHPX");
+    if (migrate_add_blocker(&whpx_migration_blocker, errp)) {
+        error_free(whpx_migration_blocker);
+        return;
+    }
  }
 
  static void whpx_gicv3_class_init(ObjectClass *klass, const void *data)
#34 lore raw

Re: [PATCH v5 17/18] whpx: arm64: gicv3: add migration blocker

In reply to Mohamed Mediouni
Quoted 23 lines · Mohamed Mediouni

GICv3 state save-restore is currently not implemented yet.

---
 hw/intc/arm_gicv3_whpx.c | 10 ++++++++++
 1 file changed, 10 insertions(+)
hw/intc/arm_gicv3_whpx.c +1 -0
index 35dc5ac531..700b2a1a2f 100644
@@ -17,6 +17,7 @@
 #include "system/whpx-internal.h"
 #include "gicv3_internal.h"
 #include "vgic_common.h"
+#include "migration/blocker.h"
 #include "qom/object.h"
 #include "target/arm/cpregs.h"
 
@@ -228,6 +229,15 @@ static void whpx_gicv3_realize(DeviceState *dev, Error **errp)
         error_setg(errp, "Nested virtualisation not currently supported by WHPX.");
         return;

hw/intc/arm_gicv3_whpx.c static void whpx_gicv3_realize(DeviceState *dev, Error **errp) · line 240

Quoted 10 lines
hw/intc/arm_gicv3_whpx.c +9 -0
     }
+
+    Error *whpx_migration_blocker = NULL;
+
+    error_setg(&whpx_migration_blocker,
+        "Live migration disabled because GIC state save/restore not supported on WHPX");
+    if (migrate_add_blocker(&whpx_migration_blocker, errp)) {
+        error_free(whpx_migration_blocker);
+        return;
+    }

If this is the last patch in the series then something else in the series is appearing too early in it. The patchset should put in all the handling for WHPX first, and then as the last thing it does, enable it so that you can configure/enable it. There shouldn't be a point in the middle of the commit series where you can enable and build it but it's still missing pieces.

thanks
-- PMM

#35 lore raw

PATCH v5 18/18 MAINTAINERS: Add myself as a maintainer for WHPX

In reply to Mohamed Mediouni

And add arm64 files.

---
 MAINTAINERS | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
MAINTAINERS +4 -1
index 070ba2e9cb..0a1f6e620a 100644
@@ -539,11 +539,14 @@ F: accel/stubs/hvf-stub.c
 F: include/system/hvf.h
 F: include/system/hvf_int.h
 
-WHPX CPUs
+WHPX
 M: Sunil Muthuswamy <[email protected]>
+M: Mohamed Mediouni <[email protected]>
 S: Supported
 F: accel/whpx/
 F: target/i386/whpx/
+F: target/arm/whpx_arm.h
+F: target/arm/whpx/
 F: accel/stubs/whpx-stub.c
 F: include/system/whpx.h
 F: include/system/whpx-accel-ops.h
Signature · 1 line
2.39.5 (Apple Git-154)
Reviewed-by on v5 18/18
#36 lore raw

Re: [PATCH v5 18/18] MAINTAINERS: Add myself as a maintainer for WHPX

Quoted 16 lines · Mohamed Mediouni

And add arm64 files.

---
  MAINTAINERS | 5 ++++-
  1 file changed, 4 insertions(+), 1 deletion(-)
MAINTAINERS +0 -0
index 070ba2e9cb..0a1f6e620a 100644
@@ -539,11 +539,14 @@ F: accel/stubs/hvf-stub.c
  F: include/system/hvf.h
  F: include/system/hvf_int.h

MAINTAINERS F: accel/stubs/hvf-stub.c · line 549

Quoted 10 lines
MAINTAINERS +4 -1
 
-WHPX CPUs
+WHPX
  M: Sunil Muthuswamy <[email protected]>
+M: Mohamed Mediouni <[email protected]>
  S: Supported
  F: accel/whpx/
  F: target/i386/whpx/
+F: target/arm/whpx_arm.h
+F: target/arm/whpx/

(you might want to have a common section with both of you,
a x86 one for Sunil and a ARM one for you -- adding there
L: [email protected])

MAINTAINERS F: accel/stubs/hvf-stub.c · line 552

Quoted 3 lines
MAINTAINERS +0 -0
  F: accel/stubs/whpx-stub.c
  F: include/system/whpx.h
  F: include/system/whpx-accel-ops.h
Tested-by on the v5 series
#37 lore raw
In reply to Mohamed Mediouni

Am 8. August 2025 06:54:01 UTC schrieb Mohamed Mediouni <[email protected]>:

Quoted 108 lines

Link to branch: https://github.com/mediouni-m/qemu whpx (tag for this submission: whpx-v5)

Missing features:
- PSCI state sync with Hyper-V: notably breaks reboots when multiple cores are enabled
- Interrupt controller save-restore
- SVE register sync

To do:
- Switching ITS to off by default when using HVF w/ vGIC or WHPX (will probably do when both the series are merged)
- After that, perhaps exit if ITS is explicitly enabled.

Note:

"accel/system: Introduce hwaccel_enabled() helper" taken from the mailing list, added here as part of this series to make it compilable as a whole.

"hw/arm: virt: add GICv2m for the case when ITS is not available" present in both the HVF vGIC and this series.

Updates since v4:
- Taking into account review comments
- Add migration blocker in the vGICv3 code due to missing interrupt controller save/restore
- Debug register sync

Updates since v3:
- Disabling SVE on WHPX
- Taking into account review comments incl:

- fixing x86 support
- reduce the amount of __x86_64__ checks in common code to the minimum (winhvemulation) which can be reduced even further down the road.
- generalize get_physical_address_range into something common between hvf and whpx

Updates since v2:
- Fixed up a rebase screwup for whpx-internal.h
- Fixed ID_AA64ISAR1_EL1 and ID_AA64ISAR2_EL1 feature probe for -cpu host
- Switched to ID_AA64PFR1_EL1/ID_AA64DFR0_EL1 instead of their non-AA64 variant

Updates since v1:
- Shutdowns and reboots
- MPIDR_EL1 register sync
- Fixing GICD_TYPER_LPIS value
- IPA size clamping
- -cpu host now implemented

Mohamed Mediouni (17):
hw/arm: virt: add GICv2m for the case when ITS is not available
whpx: Move around files before introducing AArch64 support
whpx: reshuffle common code
whpx: ifdef out winhvemulation on non-x86_64
whpx: common: add WHPX_INTERCEPT_DEBUG_TRAPS define
hw, target, accel: whpx: change apic_in_platform to kernel_irqchip
whpx: interrupt controller support
whpx: add arm64 support
whpx: copy over memory management logic from hvf target/arm: cpu: mark WHPX as supporting PSCI 1.1
hw/arm: virt: cleanly fail on attempt to use the platform vGIC
together with ITS
whpx: arm64: clamp down IPA size
hw/arm, accel/hvf, whpx: unify get_physical_address_range between WHPX
and HVF
whpx: arm64: implement -cpu host
target/arm: whpx: instantiate GIC early
whpx: arm64: gicv3: add migration blocker
MAINTAINERS: Add myself as a maintainer for WHPX

Philippe Mathieu-Daudé (1):
accel/system: Introduce hwaccel_enabled() helper

MAINTAINERS                                   |    9 +-
accel/hvf/hvf-all.c                           |    7 +-
accel/meson.build                             |    1 +
accel/whpx/meson.build                        |    7 +
{target/i386 => accel}/whpx/whpx-accel-ops.c  |    6 +-
accel/whpx/whpx-common.c                      |  666 +++++++++++
hw/arm/virt-acpi-build.c                      |    4 +-
hw/arm/virt.c                                 |   34 +-
hw/i386/x86-cpu.c                             |    4 +-
hw/intc/arm_gicv3_common.c                    |    3 +
hw/intc/arm_gicv3_whpx.c                      |  272 +++++
hw/intc/meson.build                           |    1 +
include/hw/arm/virt.h                         |    2 +
include/hw/boards.h                           |    3 +-
include/hw/intc/arm_gicv3_common.h            |    3 +
include/system/hvf_int.h                      |    2 +
include/system/hw_accel.h                     |   13 +
.../whpx => include/system}/whpx-accel-ops.h  |    4 +-
include/system/whpx-all.h                     |   20 +
include/system/whpx-common.h                  |   26 +
.../whpx => include/system}/whpx-internal.h   |   14 +-
include/system/whpx.h                         |    4 +-
meson.build                                   |   21 +-
target/arm/cpu.c                              |    3 +-
target/arm/cpu64.c                            |   19 +-
target/arm/hvf-stub.c                         |   20 -
target/arm/hvf/hvf.c                          |    6 +-
target/arm/hvf_arm.h                          |    3 -
target/arm/meson.build                        |    2 +-
target/arm/whpx/meson.build                   |    5 +
target/arm/whpx/whpx-all.c                    | 1025 +++++++++++++++++
target/arm/whpx/whpx-stub.c                   |   15 +
target/arm/whpx_arm.h                         |   17 +
target/i386/cpu-apic.c                        |    2 +-
target/i386/hvf/hvf.c                         |   11 +
target/i386/whpx/meson.build                  |    1 -
target/i386/whpx/whpx-all.c                   |  569 +--------
target/i386/whpx/whpx-apic.c                  |    2 +-
38 files changed, 2207 insertions(+), 619 deletions(-)
Quoted 14 lines

create mode 100644 accel/whpx/meson.build
rename {target/i386 => accel}/whpx/whpx-accel-ops.c (96%) create mode 100644 accel/whpx/whpx-common.c
create mode 100644 hw/intc/arm_gicv3_whpx.c
rename {target/i386/whpx => include/system}/whpx-accel-ops.h (92%) create mode 100644 include/system/whpx-all.h
create mode 100644 include/system/whpx-common.h
rename {target/i386/whpx => include/system}/whpx-internal.h (96%) delete mode 100644 target/arm/hvf-stub.c
create mode 100644 target/arm/whpx/meson.build
create mode 100644 target/arm/whpx/whpx-all.c
create mode 100644 target/arm/whpx/whpx-stub.c
create mode 100644 target/arm/whpx_arm.h

Ran this branch for a couple of days on x86_64 without any issues. So, for the x86_64 refactorings: