pvh/arch-x86_64: Initialize vCPU regs for PVH

Set the initial values of the KVM vCPU registers as specified in
the PVH boot ABI:

https://xenbits.xen.org/docs/unstable/misc/pvh.html

Signed-off-by: Colin Percival <[email protected]>
Co-authored-by: Alejandro Jimenez <[email protected]>

src/arch/src/x86_64/gdt.rs

@@ -1,3 +1,5 @@
+// Copyright © 2020, Oracle and/or its affiliates.
+//
 // Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved.
 // SPDX-License-Identifier: Apache-2.0
 //
@@ -24,8 +26,34 @@ fn get_base(entry: u64) -> u64 {
         | (((entry) & 0x0000_0000_FFFF_0000) >> 16)
 }
 
+// Extract the segment limit from the GDT segment descriptor.
+//
+// In a segment descriptor, the limit field is 20 bits, so it can directly describe
+// a range from 0 to 0xFFFFF (1 MB). When G flag is set (4-KByte page granularity) it
+// scales the value in the limit field by a factor of 2^12 (4 Kbytes), making the effective
+// limit range from 0xFFF (4 KBytes) to 0xFFFF_FFFF (4 GBytes).
+//
+// However, the limit field in the VMCS definition is a 32 bit field, and the limit value is not
+// automatically scaled using the G flag. This means that for a desired range of 4GB for a
+// given segment, its limit must be specified as 0xFFFF_FFFF. Therefore the method of obtaining
+// the limit from the GDT entry is not sufficient, since it only provides 20 bits when 32 bits
+// are necessary. Fortunately, we can check if the G flag is set when extracting the limit since
+// the full GDT entry is passed as an argument, and perform the scaling of the limit value to
+// return the full 32 bit value.
+//
+// The scaling mentioned above is required when using PVH boot, since the guest boots in protected
+// (32-bit) mode and must be able to access the entire 32-bit address space. It does not cause
+// issues for the case of direct boot to 64-bit (long) mode, since in 64-bit mode the processor does
+// not perform runtime limit checking on code or data segments.
 fn get_limit(entry: u64) -> u32 {
-    ((((entry) & 0x000F_0000_0000_0000) >> 32) | ((entry) & 0x0000_0000_0000_FFFF)) as u32
+    let limit: u32 =
+        ((((entry) & 0x000F_0000_0000_0000) >> 32) | ((entry) & 0x0000_0000_0000_FFFF)) as u32;
+
+    // Perform manual limit scaling if G flag is set
+    match get_g(entry) {
+        0 => limit,
+        _ => (limit << 12) | 0xFFF, // G flag is either 0 or 1
+    }
 }
 
 fn get_g(entry: u64) -> u8 {
@@ -109,7 +137,7 @@ mod tests {
         assert_eq!(0xB, seg.type_);
         // base and limit
         assert_eq!(0x10_0000, seg.base);
-        assert_eq!(0xfffff, seg.limit);
+        assert_eq!(0xffff_ffff, seg.limit);
         assert_eq!(0x0, seg.unusable);
     }
 }

src/arch/src/x86_64/layout.rs

@@ -27,5 +27,8 @@ pub const IRQ_MAX: u32 = 23;
 /// Address for the TSS setup.
 pub const KVM_TSS_ADDRESS: u64 = 0xfffb_d000;
 
+/// Address of the hvm_start_info struct used in PVH boot
+pub const PVH_INFO_START: u64 = 0x6000;
+
 /// The 'zero page', a.k.a linux kernel bootparams.
 pub const ZERO_PAGE_START: u64 = 0x7000;

src/arch/src/x86_64/regs.rs

@@ -1,3 +1,4 @@
+// Copyright © 2020, Oracle and/or its affiliates.
 // Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved.
 // SPDX-License-Identifier: Apache-2.0
 //
@@ -11,6 +12,7 @@ use kvm_bindings::{kvm_fpu, kvm_regs, kvm_sregs};
 use kvm_ioctls::VcpuFd;
 use vm_memory::{Address, Bytes, GuestAddress, GuestMemory, GuestMemoryMmap};
 
+use super::super::{BootProtocol, EntryPoint};
 use super::gdt::{gdt_entry, kvm_segment_from_gdt};
 
 // Initial pagetables.
@@ -100,20 +102,33 @@ impl fmt::Display for SetupRegistersError {
 /// # Errors
 ///
 /// When [`kvm_ioctls::ioctls::vcpu::VcpuFd::set_regs`] errors.
-pub fn setup_regs(vcpu: &VcpuFd, boot_ip: u64) -> std::result::Result<(), SetupRegistersError> {
-    let regs: kvm_regs = kvm_regs {
-        rflags: 0x0000_0000_0000_0002u64,
-        rip: boot_ip,
-        // Frame pointer. It gets a snapshot of the stack pointer (rsp) so that when adjustments are
-        // made to rsp (i.e. reserving space for local variables or pushing values on to the stack),
-        // local variables and function parameters are still accessible from a constant offset from
-        // rbp.
-        rsp: super::layout::BOOT_STACK_POINTER,
-        // Starting stack pointer.
-        rbp: super::layout::BOOT_STACK_POINTER,
-        // Must point to zero page address per Linux ABI. This is x86_64 specific.
-        rsi: super::layout::ZERO_PAGE_START,
-        ..Default::default()
+pub fn setup_regs(
+    vcpu: &VcpuFd,
+    entry_point: EntryPoint,
+) -> std::result::Result<(), SetupRegistersError> {
+    let regs: kvm_regs = match entry_point.protocol {
+        BootProtocol::PvhBoot => kvm_regs {
+            // Configure regs as required by PVH boot protocol.
+            rflags: 0x0000_0000_0000_0002u64,
+            rbx: super::layout::PVH_INFO_START,
+            rip: entry_point.entry_addr.raw_value(),
+            ..Default::default()
+        },
+        BootProtocol::LinuxBoot => kvm_regs {
+            // Configure regs as required by Linux 64-bit boot protocol.
+            rflags: 0x0000_0000_0000_0002u64,
+            rip: entry_point.entry_addr.raw_value(),
+            // Frame pointer. It gets a snapshot of the stack pointer (rsp) so that when adjustments
+            // are made to rsp (i.e. reserving space for local variables or pushing
+            // values on to the stack), local variables and function parameters are
+            // still accessible from a constant offset from rbp.
+            rsp: super::layout::BOOT_STACK_POINTER,
+            // Starting stack pointer.
+            rbp: super::layout::BOOT_STACK_POINTER,
+            // Must point to zero page address per Linux ABI. This is x86_64 specific.
+            rsi: super::layout::ZERO_PAGE_START,
+            ..Default::default()
+        },
     };
 
     vcpu.set_regs(&regs).map_err(SetupRegistersError)
@@ -142,6 +157,7 @@ pub enum SetupSpecialRegistersError {
 ///
 /// * `mem` - The memory that will be passed to the guest.
 /// * `vcpu` - Structure for the VCPU that holds the VCPU's fd.
+/// * `boot_prot` - The boot protocol being used.
 ///
 /// # Errors
 ///
@@ -153,14 +169,18 @@ pub enum SetupSpecialRegistersError {
 pub fn setup_sregs(
     mem: &GuestMemoryMmap,
     vcpu: &VcpuFd,
+    boot_prot: BootProtocol,
 ) -> std::result::Result<(), SetupSpecialRegistersError> {
     let mut sregs: kvm_sregs = vcpu
         .get_sregs()
         .map_err(SetupSpecialRegistersError::GetSpecialRegisters)?;
 
-    configure_segments_and_sregs(mem, &mut sregs)
+    configure_segments_and_sregs(mem, &mut sregs, boot_prot)
         .map_err(SetupSpecialRegistersError::ConfigureSegmentsAndSpecialRegisters)?;
-    setup_page_tables(mem, &mut sregs).map_err(SetupSpecialRegistersError::SetupPageTables)?; // TODO(dgreid) - Can this be done once per system instead?
+    if let BootProtocol::LinuxBoot = boot_prot {
+        setup_page_tables(mem, &mut sregs).map_err(SetupSpecialRegistersError::SetupPageTables)?;
+        // TODO(dgreid) - Can this be done once per system instead?
+    }
 
     vcpu.set_sregs(&sregs)
         .map_err(SetupSpecialRegistersError::SetSpecialRegisters)
@@ -175,6 +195,7 @@ const EFER_LMA: u64 = 0x400;
 const EFER_LME: u64 = 0x100;
 
 const X86_CR0_PE: u64 = 0x1;
+const X86_CR0_ET: u64 = 0x10;
 const X86_CR0_PG: u64 = 0x8000_0000;
 const X86_CR4_PAE: u64 = 0x20;
 
@@ -198,13 +219,31 @@ fn write_idt_value(val: u64, guest_mem: &GuestMemoryMmap) -> Result<()> {
         .map_err(|_| Error::WriteIDT)
 }
 
-fn configure_segments_and_sregs(mem: &GuestMemoryMmap, sregs: &mut kvm_sregs) -> Result<()> {
-    let gdt_table: [u64; BOOT_GDT_MAX] = [
-        gdt_entry(0, 0, 0),            // NULL
-        gdt_entry(0xa09b, 0, 0xfffff), // CODE
-        gdt_entry(0xc093, 0, 0xfffff), // DATA
-        gdt_entry(0x808b, 0, 0xfffff), // TSS
-    ];
+fn configure_segments_and_sregs(
+    mem: &GuestMemoryMmap,
+    sregs: &mut kvm_sregs,
+    boot_prot: BootProtocol,
+) -> Result<()> {
+    let gdt_table: [u64; BOOT_GDT_MAX] = match boot_prot {
+        BootProtocol::PvhBoot => {
+            // Configure GDT entries as specified by PVH boot protocol
+            [
+                gdt_entry(0, 0, 0),                // NULL
+                gdt_entry(0xc09b, 0, 0xffff_ffff), // CODE
+                gdt_entry(0xc093, 0, 0xffff_ffff), // DATA
+                gdt_entry(0x008b, 0, 0x67),        // TSS
+            ]
+        }
+        BootProtocol::LinuxBoot => {
+            // Configure GDT entries as specified by Linux 64bit boot protocol
+            [
+                gdt_entry(0, 0, 0),            // NULL
+                gdt_entry(0xa09b, 0, 0xfffff), // CODE
+                gdt_entry(0xc093, 0, 0xfffff), // DATA
+                gdt_entry(0x808b, 0, 0xfffff), // TSS
+            ]
+        }
+    };
 
     let code_seg = kvm_segment_from_gdt(gdt_table[1], 1);
     let data_seg = kvm_segment_from_gdt(gdt_table[2], 2);
@@ -227,9 +266,17 @@ fn configure_segments_and_sregs(mem: &GuestMemoryMmap, sregs: &mut kvm_sregs) ->
     sregs.ss = data_seg;
     sregs.tr = tss_seg;
 
-    // 64-bit protected mode
-    sregs.cr0 |= X86_CR0_PE;
-    sregs.efer |= EFER_LME | EFER_LMA;
+    match boot_prot {
+        BootProtocol::PvhBoot => {
+            sregs.cr0 = X86_CR0_PE | X86_CR0_ET;
+            sregs.cr4 = 0;
+        }
+        BootProtocol::LinuxBoot => {
+            // 64-bit protected mode
+            sregs.cr0 |= X86_CR0_PE;
+            sregs.efer |= EFER_LME | EFER_LMA;
+        }
+    }
 
     Ok(())
 }
@@ -287,24 +334,45 @@ mod tests {
         gm.read_obj(read_addr).unwrap()
     }
 
-    fn validate_segments_and_sregs(gm: &GuestMemoryMmap, sregs: &kvm_sregs) {
+    fn validate_segments_and_sregs(
+        gm: &GuestMemoryMmap,
+        sregs: &kvm_sregs,
+        boot_prot: BootProtocol,
+    ) {
+        if let BootProtocol::LinuxBoot = boot_prot {
+            assert_eq!(0xaf_9b00_0000_ffff, read_u64(gm, BOOT_GDT_OFFSET + 8));
+            assert_eq!(0xcf_9300_0000_ffff, read_u64(gm, BOOT_GDT_OFFSET + 16));
+            assert_eq!(0x8f_8b00_0000_ffff, read_u64(gm, BOOT_GDT_OFFSET + 24));
+
+            assert_eq!(0xffff_ffff, sregs.tr.limit);
+
+            assert!(sregs.cr0 & X86_CR0_PE != 0);
+            assert!(sregs.efer & EFER_LME != 0 && sregs.efer & EFER_LMA != 0);
+        } else {
+            // Validate values that are specific to PVH boot protocol
+            assert_eq!(0xcf_9b00_0000_ffff, read_u64(gm, BOOT_GDT_OFFSET + 8));
+            assert_eq!(0xcf_9300_0000_ffff, read_u64(gm, BOOT_GDT_OFFSET + 16));
+            assert_eq!(0x00_8b00_0000_0067, read_u64(gm, BOOT_GDT_OFFSET + 24));
+
+            assert_eq!(0x67, sregs.tr.limit);
+            assert_eq!(0, sregs.tr.g);
+
+            assert!(sregs.cr0 & X86_CR0_PE != 0 && sregs.cr0 & X86_CR0_ET != 0);
+            assert_eq!(0, sregs.cr4);
+        }
+
+        // Common settings for both PVH and Linux boot protocol
         assert_eq!(0x0, read_u64(gm, BOOT_GDT_OFFSET));
-        assert_eq!(0xaf_9b00_0000_ffff, read_u64(gm, BOOT_GDT_OFFSET + 8));
-        assert_eq!(0xcf_9300_0000_ffff, read_u64(gm, BOOT_GDT_OFFSET + 16));
-        assert_eq!(0x8f_8b00_0000_ffff, read_u64(gm, BOOT_GDT_OFFSET + 24));
         assert_eq!(0x0, read_u64(gm, BOOT_IDT_OFFSET));
 
         assert_eq!(0, sregs.cs.base);
-        assert_eq!(0xfffff, sregs.ds.limit);
+        assert_eq!(0xffff_ffff, sregs.ds.limit);
         assert_eq!(0x10, sregs.es.selector);
         assert_eq!(1, sregs.fs.present);
         assert_eq!(1, sregs.gs.g);
         assert_eq!(0, sregs.ss.avl);
         assert_eq!(0, sregs.tr.base);
-        assert_eq!(0xfffff, sregs.tr.limit);
         assert_eq!(0, sregs.tr.avl);
-        assert!(sregs.cr0 & X86_CR0_PE != 0);
-        assert!(sregs.efer & EFER_LME != 0 && sregs.efer & EFER_LMA != 0);
     }
 
     fn validate_page_tables(gm: &GuestMemoryMmap, sregs: &kvm_sregs) {
@@ -356,7 +424,12 @@ mod tests {
             ..Default::default()
         };
 
-        setup_regs(&vcpu, expected_regs.rip).unwrap();
+        let entry_point: EntryPoint = EntryPoint {
+            entry_addr: GuestAddress(expected_regs.rip),
+            protocol: BootProtocol::LinuxBoot,
+        };
+
+        setup_regs(&vcpu, entry_point).unwrap();
 
         let actual_regs: kvm_regs = vcpu.get_regs().unwrap();
         assert_eq!(actual_regs, expected_regs);
@@ -369,16 +442,22 @@ mod tests {
         let vcpu = vm.create_vcpu(0).unwrap();
         let gm = create_guest_mem(None);
 
-        assert!(vcpu.set_sregs(&Default::default()).is_ok());
-        setup_sregs(&gm, &vcpu).unwrap();
-
-        let mut sregs: kvm_sregs = vcpu.get_sregs().unwrap();
-        // for AMD KVM_GET_SREGS returns g = 0 for each kvm_segment.
-        // We set it to 1, otherwise the test will fail.
-        sregs.gs.g = 1;
-
-        validate_segments_and_sregs(&gm, &sregs);
-        validate_page_tables(&gm, &sregs);
+        [BootProtocol::LinuxBoot, BootProtocol::PvhBoot]
+            .iter()
+            .for_each(|boot_prot| {
+                assert!(vcpu.set_sregs(&Default::default()).is_ok());
+                setup_sregs(&gm, &vcpu, *boot_prot).unwrap();
+
+                let mut sregs: kvm_sregs = vcpu.get_sregs().unwrap();
+                // for AMD KVM_GET_SREGS returns g = 0 for each kvm_segment.
+                // We set it to 1, otherwise the test will fail.
+                sregs.gs.g = 1;
+
+                validate_segments_and_sregs(&gm, &sregs, *boot_prot);
+                if let BootProtocol::LinuxBoot = *boot_prot {
+                    validate_page_tables(&gm, &sregs);
+                }
+            });
     }
 
     #[test]
@@ -423,9 +502,13 @@ mod tests {
     fn test_configure_segments_and_sregs() {
         let mut sregs: kvm_sregs = Default::default();
         let gm = create_guest_mem(None);
-        configure_segments_and_sregs(&gm, &mut sregs).unwrap();
+        configure_segments_and_sregs(&gm, &mut sregs, BootProtocol::LinuxBoot).unwrap();
+
+        validate_segments_and_sregs(&gm, &sregs, BootProtocol::LinuxBoot);
+
+        configure_segments_and_sregs(&gm, &mut sregs, BootProtocol::PvhBoot).unwrap();
 
-        validate_segments_and_sregs(&gm, &sregs);
+        validate_segments_and_sregs(&gm, &sregs, BootProtocol::PvhBoot);
     }
 
     #[test]

src/vmm/src/builder.rs

@@ -385,7 +385,7 @@ pub fn build_microvm_for_boot(
         &vmm,
         vcpus.as_mut(),
         vcpu_config,
-        entry_point.entry_addr,
+        entry_point,
         &initrd,
         boot_cmdline,
     )?;
@@ -849,7 +849,7 @@ pub fn configure_system_for_boot(
     vmm: &Vmm,
     vcpus: &mut [Vcpu],
     vcpu_config: VcpuConfig,
-    entry_addr: GuestAddress,
+    entry_point: EntryPoint,
     initrd: &Option<InitrdConfig>,
     boot_cmdline: LoaderKernelCmdline,
 ) -> std::result::Result<(), StartMicrovmError> {
@@ -860,7 +860,7 @@ pub fn configure_system_for_boot(
             vcpu.kvm_vcpu
                 .configure(
                     vmm.guest_memory(),
-                    entry_addr,
+                    entry_point,
                     &vcpu_config,
                     vmm.vm.supported_cpuid().clone(),
                 )
@@ -893,7 +893,7 @@ pub fn configure_system_for_boot(
     {
         for vcpu in vcpus.iter_mut() {
             vcpu.kvm_vcpu
-                .configure(vmm.guest_memory(), entry_addr)
+                .configure(vmm.guest_memory(), entry_point.entry_addr)
                 .map_err(Error::VcpuConfigure)
                 .map_err(Internal)?;
         }