hypervisor: add API to query VM physaddr size

Add a function that returns the size of guest physical addresses. This
wraps KVM_CAP_ARM_VM_IPA_SIZE on aarch64 and CPUID to query host
physical address size on x86_64. This replaces the phy_max_address_bits
function from x86_64, and all callers are migrated to use the new Vm
function instead.

BUG=b:210727578
TEST=check on trogdor64 with debug prints

Change-Id: I3107fe6357fcf166b7ad0e2a7c55919705da3b0c
Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/crosvm/+/3364971
Reviewed-by: Dmitry Torokhov <dtor@chromium.org>
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Alexandre Courbot <acourbot@chromium.org>
Reviewed-by: Andrew Walbran <qwandor@google.com>
Commit-Queue: Daniel Verkamp <dverkamp@chromium.org>

aarch64/src/lib.rs

@@ -14,7 +14,7 @@ use devices::{
     Bus, BusDeviceObj, BusError, IrqChip, IrqChipAArch64, PciAddress, PciConfigMmio, PciDevice,
 };
 use hypervisor::{
-    DeviceKind, Hypervisor, HypervisorCap, ProtectionType, VcpuAArch64, VcpuFeature, VmAArch64,
+    DeviceKind, Hypervisor, HypervisorCap, ProtectionType, VcpuAArch64, VcpuFeature, Vm, VmAArch64,
 };
 use minijail::Minijail;
 use remain::sorted;
@@ -245,12 +245,8 @@ impl arch::LinuxArch for AArch64 {
         Ok(arch_memory_regions(components.memory_size))
     }
 
-    fn get_phys_max_addr() -> u64 {
+    fn create_system_allocator<V: Vm>(vm: &V) -> SystemAllocator {
-        u64::max_value()
+        Self::get_resource_allocator(vm.get_memory().memory_size())
-    }
-
-    fn create_system_allocator(guest_mem: &GuestMemory) -> SystemAllocator {
-        Self::get_resource_allocator(guest_mem.memory_size())
     }
 
     fn build_vm<V, Vcpu>(
@@ -507,8 +503,8 @@ impl arch::LinuxArch for AArch64 {
         })
     }
 
-    fn configure_vcpu(
+    fn configure_vcpu<V: Vm>(
-        _guest_mem: &GuestMemory,
+        _vm: &V,
         _hypervisor: &dyn Hypervisor,
         _irq_chip: &mut dyn IrqChipAArch64,
         _vcpu: &mut dyn VcpuAArch64,

arch/src/lib.rs

@@ -148,14 +148,12 @@ pub trait LinuxArch {
         components: &VmComponents,
     ) -> std::result::Result<Vec<(GuestAddress, u64)>, Self::Error>;
 
-    /// Creates a new `SystemAllocator` that fits the given `GuestMemory`'s layout.
+    /// Creates a new `SystemAllocator` that fits the given `Vm`'s memory layout.
     ///
     /// # Arguments
     ///
-    /// * `guest_mem` - The memory to be used as a template for the `SystemAllocator`.
+    /// * `vm` - The virtual machine to be used as a template for the `SystemAllocator`.
-    fn create_system_allocator(guest_mem: &GuestMemory) -> SystemAllocator;
+    fn create_system_allocator<V: Vm>(vm: &V) -> SystemAllocator;
-
-    fn get_phys_max_addr() -> u64;
 
     /// Takes `VmComponents` and generates a `RunnableLinuxVm`.
     ///
@@ -197,15 +195,15 @@ pub trait LinuxArch {
     ///
     /// # Arguments
     ///
-    /// * `guest_mem` - The memory to be used by the guest.
+    /// * `vm` - The virtual machine object.
     /// * `hypervisor` - The `Hypervisor` that created the vcpu.
     /// * `irq_chip` - The `IrqChip` associated with this vm.
     /// * `vcpu` - The VCPU object to configure.
     /// * `vcpu_id` - The id of the given `vcpu`.
     /// * `num_cpus` - Number of virtual CPUs the guest will have.
     /// * `has_bios` - Whether the `VmImage` is a `Bios` image
-    fn configure_vcpu(
+    fn configure_vcpu<V: Vm>(
-        guest_mem: &GuestMemory,
+        vm: &V,
         hypervisor: &dyn HypervisorArch,
         irq_chip: &mut dyn IrqChipArch,
         vcpu: &mut dyn VcpuArch,

hypervisor/src/kvm/aarch64.rs

@@ -40,6 +40,19 @@ impl Kvm {
         // Use the lower 8 bits representing the IPA space as the machine type
         Ok((ipa_size & KVM_VM_TYPE_ARM_IPA_SIZE_MASK) | protection_flag)
     }
+
+    /// Get the size of guest physical addresses (IPA) in bits.
+    pub fn get_guest_phys_addr_size(&self) -> u8 {
+        // Safe because we know self is a real kvm fd
+        let vm_ipa_size = match unsafe {
+            ioctl_with_val(self, KVM_CHECK_EXTENSION(), KVM_CAP_ARM_VM_IPA_SIZE.into())
+        } {
+            // Default physical address size is 40 bits if the extension is not supported.
+            ret if ret < 0 => 40,
+            ipa => ipa as u8,
+        };
+        vm_ipa_size
+    }
 }
 
 impl KvmVm {

hypervisor/src/kvm/mod.rs

@@ -453,6 +453,10 @@ impl Vm for KvmVm {
         }
     }
 
+    fn get_guest_phys_addr_size(&self) -> u8 {
+        self.kvm.get_guest_phys_addr_size()
+    }
+
     fn get_memory(&self) -> &GuestMemory {
         &self.guest_mem
     }

hypervisor/src/kvm/x86_64.rs

@@ -2,6 +2,8 @@
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
+use std::arch::x86_64::__cpuid;
+
 use base::IoctlNr;
 
 use libc::E2BIG;
@@ -74,6 +76,20 @@ impl Kvm {
             Err(Error::new(libc::EINVAL))
         }
     }
+
+    /// Get the size of guest physical addresses in bits.
+    pub fn get_guest_phys_addr_size(&self) -> u8 {
+        // Get host cpu max physical address bits.
+        // Assume the guest physical address size is the same as the host.
+        let highest_ext_function = unsafe { __cpuid(0x80000000) };
+        if highest_ext_function.eax >= 0x80000008 {
+            let addr_size = unsafe { __cpuid(0x80000008) };
+            // Low 8 bits of 0x80000008 leaf: host physical address size in bits.
+            addr_size.eax as u8
+        } else {
+            36
+        }
+    }
 }
 
 impl HypervisorX86_64 for Kvm {

hypervisor/src/lib.rs

@@ -52,6 +52,9 @@ pub trait Vm: Send {
     /// reflects the usable capabilities.
     fn check_capability(&self, c: VmCap) -> bool;
 
+    /// Get the guest physical address size in bits.
+    fn get_guest_phys_addr_size(&self) -> u8;
+
     /// Gets the guest-mapped memory for the Vm.
     fn get_memory(&self) -> &GuestMemory;
 

src/linux.rs

@@ -2071,7 +2071,7 @@ where
     }
 
     Arch::configure_vcpu(
-        vm.get_memory(),
+        &vm,
         vm.get_hypervisor(),
         irq_chip,
         &mut vcpu,
@@ -2778,7 +2778,7 @@ where
 
     let exit_evt = Event::new().context("failed to create event")?;
     let reset_evt = Event::new().context("failed to create event")?;
-    let mut sys_allocator = Arch::create_system_allocator(vm.get_memory());
+    let mut sys_allocator = Arch::create_system_allocator(&vm);
 
     // Allocate the ramoops region first. AArch64::build_vm() assumes this.
     let ramoops_region = match &components.pstore {
@@ -2789,7 +2789,7 @@ where
         None => None,
     };
 
-    let phys_max_addr = Arch::get_phys_max_addr();
+    let phys_max_addr = (1u64 << vm.get_guest_phys_addr_size()) - 1;
     let mut devices = create_devices(
         &cfg,
         &mut vm,

x86_64/src/cpuid.rs

@@ -214,19 +214,6 @@ pub fn setup_cpuid(
         .map_err(Error::SetSupportedCpusFailed)
 }
 
-/// get host cpu max physical address bits
-pub fn phy_max_address_bits() -> u32 {
-    let mut phys_bits: u32 = 36;
-
-    let highest_ext_function = unsafe { __cpuid(0x80000000) };
-    if highest_ext_function.eax >= 0x80000008 {
-        let addr_size = unsafe { __cpuid(0x80000008) };
-        phys_bits = addr_size.eax & 0xff;
-    }
-
-    phys_bits
-}
-
 #[cfg(test)]
 mod tests {
     use super::*;

x86_64/src/lib.rs

@@ -53,16 +53,14 @@ use std::sync::Arc;
 use crate::bootparam::boot_params;
 use acpi_tables::sdt::SDT;
 use acpi_tables::{aml, aml::Aml};
-use arch::{
+use arch::{get_serial_cmdline, GetSerialCmdlineError, RunnableLinuxVm, VmComponents, VmImage};
-    get_serial_cmdline, GetSerialCmdlineError, LinuxArch, RunnableLinuxVm, VmComponents, VmImage,
-};
 use base::Event;
 use devices::serial_device::{SerialHardware, SerialParameters};
 use devices::{
     BusDeviceObj, BusResumeDevice, IrqChip, IrqChipX86_64, PciAddress, PciConfigIo, PciConfigMmio,
     PciDevice,
 };
-use hypervisor::{HypervisorX86_64, ProtectionType, VcpuX86_64, VmX86_64};
+use hypervisor::{HypervisorX86_64, ProtectionType, VcpuX86_64, Vm, VmX86_64};
 use minijail::Minijail;
 use remain::sorted;
 use resources::SystemAllocator;
@@ -370,13 +368,10 @@ impl arch::LinuxArch for X8664arch {
         Ok(arch_memory_regions(components.memory_size, bios_size))
     }
 
-    fn get_phys_max_addr() -> u64 {
+    fn create_system_allocator<V: Vm>(vm: &V) -> SystemAllocator {
-        (1u64 << cpuid::phy_max_address_bits()) - 1
+        let guest_mem = vm.get_memory();
-    }
-
-    fn create_system_allocator(guest_mem: &GuestMemory) -> SystemAllocator {
         let high_mmio_start = Self::get_high_mmio_base(guest_mem);
-        let high_mmio_size = Self::get_high_mmio_size(guest_mem);
+        let high_mmio_size = Self::get_high_mmio_size(vm);
         SystemAllocator::builder()
             .add_io_addresses(0xc000, 0x1_0000)
             .add_low_mmio_addresses(END_ADDR_BEFORE_32BITS, PCI_MMIO_SIZE)
@@ -476,6 +471,7 @@ impl arch::LinuxArch for X8664arch {
         let max_bus = ((PCIE_CFG_MMIO_SIZE / 0x100000) - 1) as u8;
 
         let (acpi_dev_resource, bat_control) = Self::setup_acpi_devices(
+            &vm,
             &mem,
             &io_bus,
             system_allocator,
@@ -601,8 +597,8 @@ impl arch::LinuxArch for X8664arch {
         })
     }
 
-    fn configure_vcpu(
+    fn configure_vcpu<V: Vm>(
-        guest_mem: &GuestMemory,
+        vm: &V,
         hypervisor: &dyn HypervisorX86_64,
         irq_chip: &mut dyn IrqChipX86_64,
         vcpu: &mut dyn VcpuX86_64,
@@ -627,8 +623,9 @@ impl arch::LinuxArch for X8664arch {
             return Ok(());
         }
 
+        let guest_mem = vm.get_memory();
         let kernel_load_addr = GuestAddress(KERNEL_START_OFFSET);
-        regs::setup_msrs(vcpu, END_ADDR_BEFORE_32BITS).map_err(Error::SetupMsrs)?;
+        regs::setup_msrs(vm, vcpu, END_ADDR_BEFORE_32BITS).map_err(Error::SetupMsrs)?;
         let kernel_end = guest_mem
             .checked_offset(kernel_load_addr, KERNEL_64BIT_ENTRY_OFFSET)
             .ok_or(Error::KernelOffsetPastEnd)?;
@@ -1116,11 +1113,11 @@ impl X8664arch {
     ///
     /// # Arguments
     ///
-    /// * mem: The memory to be used by the guest
+    /// * `vm`: The virtual machine
-    fn get_high_mmio_size(mem: &GuestMemory) -> u64 {
+    fn get_high_mmio_size<V: Vm>(vm: &V) -> u64 {
-        let phys_mem_end = Self::get_phys_max_addr() + 1;
+        let phys_mem_end = 1u64 << vm.get_guest_phys_addr_size();
         let high_mmio_end = std::cmp::min(phys_mem_end, HIGH_MMIO_MAX_END);
-        high_mmio_end - Self::get_high_mmio_base(mem)
+        high_mmio_end - Self::get_high_mmio_base(vm.get_memory())
     }
 
     /// This returns a minimal kernel command for this architecture
@@ -1204,7 +1201,8 @@ impl X8664arch {
     /// * - `irq_chip` the IrqChip object for registering irq events
     /// * - `battery` indicate whether to create the battery
     /// * - `mmio_bus` the MMIO bus to add the devices to
-    fn setup_acpi_devices(
+    fn setup_acpi_devices<V: VmX86_64>(
+        vm: &V,
         mem: &GuestMemory,
         io_bus: &devices::Bus,
         resources: &mut SystemAllocator,
@@ -1264,7 +1262,7 @@ impl X8664arch {
                     aml::AddressSpaceCachable::NotCacheable,
                     true,
                     Self::get_high_mmio_base(mem),
-                    Self::get_high_mmio_size(mem),
+                    Self::get_high_mmio_size(vm),
                 ),
             ]),
         );

x86_64/src/regs.rs

@@ -5,7 +5,7 @@
 use std::{mem, result};
 
 use base::{self, warn};
-use hypervisor::{Fpu, Register, Regs, Sregs, VcpuX86_64};
+use hypervisor::{Fpu, Register, Regs, Sregs, VcpuX86_64, Vm};
 use remain::sorted;
 use thiserror::Error;
 use vm_memory::{GuestAddress, GuestMemory};
@@ -93,7 +93,12 @@ fn get_mtrr_pairs(base: u64, len: u64) -> Vec<(u64, u64)> {
     vecs
 }
 
-fn append_mtrr_entries(vpu: &dyn VcpuX86_64, pci_start: u64, entries: &mut Vec<Register>) {
+fn append_mtrr_entries(
+    vm: &dyn Vm,
+    vpu: &dyn VcpuX86_64,
+    pci_start: u64,
+    entries: &mut Vec<Register>,
+) {
     // Get VAR MTRR num from MSR_MTRRcap
     let mut msrs = vec![Register {
         id: crate::msr_index::MSR_MTRRcap,
@@ -117,7 +122,7 @@ fn append_mtrr_entries(vpu: &dyn VcpuX86_64, pci_start: u64, entries: &mut Vec<R
         return;
     }
 
-    let phys_mask: u64 = (1 << crate::cpuid::phy_max_address_bits()) - 1;
+    let phys_mask: u64 = (1 << vm.get_guest_phys_addr_size()) - 1;
     for (idx, (base, len)) in vecs.iter().enumerate() {
         let reg_idx = idx as u32 * 2;
         entries.push(Register {
@@ -137,7 +142,7 @@ fn append_mtrr_entries(vpu: &dyn VcpuX86_64, pci_start: u64, entries: &mut Vec<R
     });
 }
 
-fn create_msr_entries(vcpu: &dyn VcpuX86_64, pci_start: u64) -> Vec<Register> {
+fn create_msr_entries(vm: &dyn Vm, vcpu: &dyn VcpuX86_64, pci_start: u64) -> Vec<Register> {
     let mut entries = vec![
         Register {
             id: crate::msr_index::MSR_IA32_SYSENTER_CS,
@@ -182,7 +187,7 @@ fn create_msr_entries(vcpu: &dyn VcpuX86_64, pci_start: u64) -> Vec<Register> {
             value: crate::msr_index::MSR_IA32_MISC_ENABLE_FAST_STRING as u64,
         },
     ];
-    append_mtrr_entries(vcpu, pci_start, &mut entries);
+    append_mtrr_entries(vm, vcpu, pci_start, &mut entries);
     entries
 }
 
@@ -191,8 +196,8 @@ fn create_msr_entries(vcpu: &dyn VcpuX86_64, pci_start: u64) -> Vec<Register> {
 /// # Arguments
 ///
 /// * `vcpu` - Structure for the vcpu that holds the vcpu fd.
-pub fn setup_msrs(vcpu: &dyn VcpuX86_64, pci_start: u64) -> Result<()> {
+pub fn setup_msrs(vm: &dyn Vm, vcpu: &dyn VcpuX86_64, pci_start: u64) -> Result<()> {
-    let msrs = create_msr_entries(vcpu, pci_start);
+    let msrs = create_msr_entries(vm, vcpu, pci_start);
     vcpu.set_msrs(&msrs).map_err(Error::MsrIoctlFailed)
 }
 

x86_64/src/test_integration.rs

@@ -101,9 +101,8 @@ where
     let arch_mem_regions = arch_memory_regions(memory_size, None);
     let guest_mem = GuestMemory::new(&arch_mem_regions).unwrap();
 
-    let mut resources = X8664arch::create_system_allocator(&guest_mem);
-
     let (hyp, mut vm) = create_vm(guest_mem.clone());
+    let mut resources = X8664arch::create_system_allocator(&vm);
     let (irqchip_tube, device_tube) = Tube::pair().expect("failed to create irq tube");
 
     let mut irq_chip = create_irq_chip(vm.try_clone().expect("failed to clone vm"), 1, device_tube);
@@ -184,6 +183,7 @@ where
     let suspend_evt = Event::new().unwrap();
     let mut resume_notify_devices = Vec::new();
     let acpi_dev_resource = X8664arch::setup_acpi_devices(
+        &vm,
         &guest_mem,
         &io_bus,
         &mut resources,
@@ -246,7 +246,7 @@ where
                 .expect("failed to add vcpu to irqchip");
 
             setup_cpuid(&hyp, &irq_chip, &vcpu, 0, 1, false, false).unwrap();
-            setup_msrs(&vcpu, END_ADDR_BEFORE_32BITS).unwrap();
+            setup_msrs(&vm, &vcpu, END_ADDR_BEFORE_32BITS).unwrap();
 
             setup_regs(
                 &vcpu,