vmm: acpi: provide FADT PM1a event/control blocks for nested Hyper-V

A Windows guest that launches nested Hyper-V (for example to run WSL2)
fails to start its hypervisor on cloud-hypervisor's HW-reduced-ACPI
FADT. hvloader's hypervisor-launch path (0x18000f01c -> 0x180015628 ->
0x180015788) registers every legacy PM register block via 0x1800158dc
and rejects any block whose GAS address is 0 with status 8
(STATUS_INVALID_DEVICE_REQUEST). hvix64 then never launches and
HypervisorPresent stays False. The HW-reduced FADT leaves those blocks
zero.

Emit valid PM1a event/control blocks (I/O ports, lengths and X_GAS) in
the FADT and reserve those ports in the I/O allocator so nothing else
claims them. The HW-reduced guest OS ignores the legacy ports; only
hvloader's ACPI validation reads them.

These blocks are only useful to a guest that itself runs an enlightened
hypervisor, so emit them only when both guest nesting and the Hyper-V
enlightenments are enabled (--cpu nested=on,kvm_hyperv=on).

Signed-off-by: doge <me@crackerben.com>

vmm/src/acpi.rs

@@ -5,6 +5,7 @@
 use std::time::Instant;
 
 use acpi_tables::Aml;
+use acpi_tables::gas::{AccessSize, AddressSpace, GAS};
 use acpi_tables::rsdp::Rsdp;
 #[cfg(target_arch = "aarch64")]
 use acpi_tables::sdt::GenericAddress;
@@ -274,7 +275,11 @@ pub fn create_dsdt_table(
 
 const FACP_DSDT_OFFSET: usize = 140;
 
-fn create_facp_table(dsdt_offset: GuestAddress, device_manager: &DeviceManager) -> Sdt {
+fn create_facp_table(
+    dsdt_offset: GuestAddress,
+    device_manager: &DeviceManager,
+    legacy_acpi_pm1a: bool,
+) -> Sdt {
     trace_scoped!("create_facp_table");
 
     // Revision 6 of the ACPI FADT table is 276 bytes long
@@ -326,6 +331,50 @@ fn create_facp_table(dsdt_offset: GuestAddress, device_manager: &DeviceManager)
     // Hypervisor Vendor Identity
     facp.write_bytes(268, b"CLOUDHYP");
 
+    // Windows' nested-Hyper-V hvloader rejects a HW-reduced FADT whose PM1a GAS is
+    // zero; point the blocks at unused ACPI I/O ports (conforming guests ignore them).
+    if legacy_acpi_pm1a {
+        const PM1A_EVT_PORT: u16 = 0x60c;
+        facp.write(56, PM1A_EVT_PORT as u32); // PM1a_EVT_BLK (0x38)
+        facp.write(88, 4u8); // PM1_EVT_LEN (0x58)
+        facp.write_bytes(
+            148, // X_PM1a_EVT_BLK (0x94)
+            GAS::new(
+                AddressSpace::SystemIo,
+                32,
+                0,
+                AccessSize::WordAccess,
+                PM1A_EVT_PORT.into(),
+            )
+            .as_bytes(),
+        );
+
+        const PM1A_CNT_PORT: u16 = 0x610;
+        facp.write(64, PM1A_CNT_PORT as u32); // PM1a_CNT_BLK (0x40)
+        facp.write(89, 2u8); // PM1_CNT_LEN (0x59)
+        facp.write_bytes(
+            172, // X_PM1a_CNT_BLK (0xac)
+            GAS::new(
+                AddressSpace::SystemIo,
+                16,
+                0,
+                AccessSize::WordAccess,
+                PM1A_CNT_PORT.into(),
+            )
+            .as_bytes(),
+        );
+
+        let mut allocator = device_manager.allocator().lock().unwrap();
+        for (port, len) in [(PM1A_EVT_PORT, 4u64), (PM1A_CNT_PORT, 2u64)] {
+            if allocator
+                .allocate_io_addresses(Some(GuestAddress(port.into())), len, None)
+                .is_none()
+            {
+                warn!("Could not reserve PM1a I/O port {port:#x} advertised in the FADT");
+            }
+        }
+    }
+
     facp.update_checksum();
 
     facp
@@ -855,7 +904,12 @@ fn create_acpi_tables_internal(
     tables_bytes.extend_from_slice(dsdt.as_slice());
 
     // FACP aka FADT
-    let facp = create_facp_table(dsdt_addr, device_manager);
+    // The legacy PM1a blocks are x86-only; there is nothing to emit on other arches.
+    #[cfg(target_arch = "x86_64")]
+    let legacy_acpi_pm1a = cpu_manager.nested() && cpu_manager.kvm_hyperv();
+    #[cfg(not(target_arch = "x86_64"))]
+    let legacy_acpi_pm1a = false;
+    let facp = create_facp_table(dsdt_addr, device_manager, legacy_acpi_pm1a);
     let facp_addr = dsdt_addr.checked_add(dsdt.len() as u64).unwrap();
     tables_bytes.extend_from_slice(facp.as_slice());
     xsdt_table_pointers.push(facp_addr.0);
@@ -1130,7 +1184,12 @@ pub fn create_acpi_tables_tdx(
     )];
 
     // FACP aka FADT
-    tables.push(create_facp_table(GuestAddress(0), device_manager));
+    let legacy_acpi_pm1a = cpu_manager.nested() && cpu_manager.kvm_hyperv();
+    tables.push(create_facp_table(
+        GuestAddress(0),
+        device_manager,
+        legacy_acpi_pm1a,
+    ));
 
     // MADT
     tables.push(cpu_manager.create_madt());

vmm/src/cpu.rs

@@ -1089,6 +1089,16 @@ impl CpuManager {
         Ok(())
     }
 
+    #[cfg(target_arch = "x86_64")]
+    pub fn nested(&self) -> bool {
+        self.config.nested
+    }
+
+    #[cfg(target_arch = "x86_64")]
+    pub fn kvm_hyperv(&self) -> bool {
+        self.config.kvm_hyperv
+    }
+
     /// Only create new vCPUs if there aren't any inactive ones to reuse
     fn create_vcpus(
         &mut self,