Re: [PATCH 06/11] VMX/altp2m: add code to support EPTP switching and #VE.

[Andrew Cooper <andrew.cooper3@citrix.com>]

On 09/01/15 21:26, Ed White wrote:
> Implement and hook up the code to enable VMX support of VMFUNC and #VE.
>
> VMFUNC leaf 0 (EPTP switching) and #VE are emulated on hardware that
> doesn't support them.
>
> Signed-off-by: Ed White <edmund.h.white@intel.com>
> ---
>  xen/arch/x86/hvm/vmx/vmx.c | 138 +++++++++++++++++++++++++++++++++++++++++++++
>  1 file changed, 138 insertions(+)
>
> diff --git a/xen/arch/x86/hvm/vmx/vmx.c b/xen/arch/x86/hvm/vmx/vmx.c
> index 931709b..a0a2d02 100644
> --- a/xen/arch/x86/hvm/vmx/vmx.c
> +++ b/xen/arch/x86/hvm/vmx/vmx.c
> @@ -56,6 +56,7 @@
>  #include <asm/debugger.h>
>  #include <asm/apic.h>
>  #include <asm/hvm/nestedhvm.h>
> +#include <asm/hvm/altp2mhvm.h>
>  #include <asm/event.h>
>  #include <public/arch-x86/cpuid.h>
>  
> @@ -1718,6 +1719,91 @@ static void vmx_enable_msr_exit_interception(struct domain *d)
>                                           MSR_TYPE_W);
>  }
>  
> +static void vmx_vcpu_update_eptp(struct vcpu *v)
> +{
> +    struct domain *d = v->domain;
> +    struct p2m_domain *p2m = altp2mhvm_active(d) ?
> +        p2m_get_altp2m(v) : p2m_get_hostp2m(d);
> +    struct ept_data *ept = &p2m->ept;
> +
> +    ept->asr = pagetable_get_pfn(p2m_get_pagetable(p2m));
> +
> +    vmx_vmcs_enter(v);
> +
> +    __vmwrite(EPT_POINTER, ept_get_eptp(ept));
> +
> +    vmx_vmcs_exit(v);
> +}
> +
> +static void vmx_vcpu_update_vmfunc_ve(struct vcpu *v)
> +{
> +    struct domain *d = v->domain;
> +    u32 mask = SECONDARY_EXEC_ENABLE_VM_FUNCTIONS;
> +
> +    if ( !cpu_has_vmx_vmfunc )
> +        return;
> +
> +    if ( cpu_has_vmx_virt_exceptions )
> +        mask |= SECONDARY_EXEC_ENABLE_VIRT_EXCEPTIONS;
> +
> +    vmx_vmcs_enter(v);
> +
> +    if ( !d->is_dying && altp2mhvm_active(d) )
> +    {
> +        v->arch.hvm_vmx.secondary_exec_control |= mask;
> +        __vmwrite(VM_FUNCTION_CONTROL, VMX_VMFUNC_EPTP_SWITCHING);
> +        __vmwrite(EPTP_LIST_ADDR, virt_to_maddr(d->arch.altp2m_eptp));
> +
> +        if ( cpu_has_vmx_virt_exceptions )
> +        {
> +            p2m_type_t t;
> +            mfn_t mfn;
> +
> +            mfn = get_gfn_query_unlocked(d, vcpu_altp2mhvm(v).veinfo, &t);
> +            __vmwrite(VIRT_EXCEPTION_INFO, mfn_x(mfn) << PAGE_SHIFT);
> +        }
> +    }
> +    else
> +        v->arch.hvm_vmx.secondary_exec_control &= ~mask;
> +
> +    __vmwrite(SECONDARY_VM_EXEC_CONTROL,
> +        v->arch.hvm_vmx.secondary_exec_control);
> +
> +    vmx_vmcs_exit(v);
> +}
> +
> +static bool_t vmx_vcpu_emulate_ve(struct vcpu *v)
> +{
> +    bool_t rc = 0;
> +    ve_info_t *veinfo = vcpu_altp2mhvm(v).veinfo ?
> +        hvm_map_guest_frame_rw(vcpu_altp2mhvm(v).veinfo, 0) : NULL;
> +
> +    if ( !veinfo )
> +        return 0;
> +
> +    if ( veinfo->semaphore != 0 )
> +        goto out;
> +
> +    rc = 1;
> +
> +    veinfo->exit_reason = EXIT_REASON_EPT_VIOLATION;
> +    veinfo->semaphore = ~0l;
> +    veinfo->eptp_index = vcpu_altp2mhvm(v).p2midx;
> +
> +    vmx_vmcs_enter(v);
> +    __vmread(EXIT_QUALIFICATION, &veinfo->exit_qualification);
> +    __vmread(GUEST_LINEAR_ADDRESS, &veinfo->gla);
> +    __vmread(GUEST_PHYSICAL_ADDRESS, &veinfo->gpa);
> +    vmx_vmcs_exit(v);
> +
> +    hvm_inject_hw_exception(TRAP_virtualisation,
> +                            HVM_DELIVER_NO_ERROR_CODE);
> +
> +out:
> +    hvm_unmap_guest_frame(veinfo, 0);
> +    return rc;
> +}
> +
>  static struct hvm_function_table __initdata vmx_function_table = {
>      .name                 = "VMX",
>      .cpu_up_prepare       = vmx_cpu_up_prepare,
> @@ -1777,6 +1863,9 @@ static struct hvm_function_table __initdata vmx_function_table = {
>      .nhvm_hap_walk_L1_p2m = nvmx_hap_walk_L1_p2m,
>      .hypervisor_cpuid_leaf = vmx_hypervisor_cpuid_leaf,
>      .enable_msr_exit_interception = vmx_enable_msr_exit_interception,
> +    .ahvm_vcpu_update_eptp = vmx_vcpu_update_eptp,
> +    .ahvm_vcpu_update_vmfunc_ve = vmx_vcpu_update_vmfunc_ve,
> +    .ahvm_vcpu_emulate_ve = vmx_vcpu_emulate_ve,
>  };
>  
>  const struct hvm_function_table * __init start_vmx(void)
> @@ -2551,6 +2640,17 @@ static void vmx_vmexit_ud_intercept(struct cpu_user_regs *regs)
>          hvm_inject_hw_exception(TRAP_invalid_op, HVM_DELIVER_NO_ERROR_CODE);
>          break;
>      case X86EMUL_EXCEPTION:
> +        /* check for a VMFUNC that should be emulated */
> +        if ( !cpu_has_vmx_vmfunc && altp2mhvm_active(current->domain) &&
> +             ctxt.insn_buf_bytes >= 3 && ctxt.insn_buf[0] == 0x0f &&
> +             ctxt.insn_buf[1] == 0x01 && ctxt.insn_buf[2] == 0xd4 &&
> +             regs->eax == 0 &&
> +             p2m_switch_vcpu_altp2m_by_id(current, (uint16_t)regs->ecx) )
> +        {
> +            regs->eip += 3;
> +            return;
> +        }
> +

This is not appropriate.  You should extend the x86_emulate to decode
and act on the VMFUNC instruction.

>          if ( ctxt.exn_pending )
>              hvm_inject_trap(&ctxt.trap);
>          /* fall through */
> @@ -2698,6 +2798,40 @@ void vmx_vmexit_handler(struct cpu_user_regs *regs)
>  
>      /* Now enable interrupts so it's safe to take locks. */
>      local_irq_enable();
> + 
> +    /*
> +     * If the guest has the ability to switch EPTP without an exit,
> +     * figure out whether it has done so and update the altp2m data.
> +     */
> +    if ( altp2mhvm_active(v->domain) &&
> +        (v->arch.hvm_vmx.secondary_exec_control &
> +        SECONDARY_EXEC_ENABLE_VM_FUNCTIONS) )
> +    {
> +        unsigned long idx;
> +
> +        if ( v->arch.hvm_vmx.secondary_exec_control &
> +            SECONDARY_EXEC_ENABLE_VIRT_EXCEPTIONS )
> +            __vmread(EPTP_INDEX, &idx);
> +        else
> +        {
> +            unsigned long eptp;
> +
> +            __vmread(EPT_POINTER, &eptp);
> +
> +            if ( !p2m_find_altp2m_by_eptp(v->domain, eptp, &idx) )
> +            {
> +                gdprintk(XENLOG_ERR, "EPTP not found in alternate p2m list\n");
> +                domain_crash(v->domain);
> +            }
> +        }
> +
> +        if ( (uint16_t)idx != vcpu_altp2mhvm(v).p2midx )
> +        {
> +            cpumask_clear_cpu(v->vcpu_id, p2m_get_altp2m(v)->dirty_cpumask);
> +            vcpu_altp2mhvm(v).p2midx = (uint16_t)idx;
> +            cpumask_set_cpu(v->vcpu_id, p2m_get_altp2m(v)->dirty_cpumask);
> +        }
> +    }
>  
>      /* XXX: This looks ugly, but we need a mechanism to ensure
>       * any pending vmresume has really happened
> @@ -3041,6 +3175,10 @@ void vmx_vmexit_handler(struct cpu_user_regs *regs)
>              update_guest_eip();
>          break;
>  
> +    case EXIT_REASON_VMFUNC:
> +        vmx_vmexit_ud_intercept(regs);

This should be a dedicated function to handle vmfunc, which is invoked
from here and from the x86_emulate when emulating a 'VMFUNC' instruction.

~Andrew

> +        break;
> +
>      case EXIT_REASON_INVEPT:
>          if ( nvmx_handle_invept(regs) == X86EMUL_OKAY )
>              update_guest_eip();