Re: [Xen-devel] [PATCH v2 08/10] nEPT: handle invept instruction from L1 VMM

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 19.12.12 at 20:44, Xiantao Zhang <xiantao.zhang@xxxxxxxxx> wrote:
> --- a/xen/arch/x86/hvm/vmx/vmx.c
> +++ b/xen/arch/x86/hvm/vmx/vmx.c
> @@ -2572,11 +2572,13 @@ void vmx_vmexit_handler(struct cpu_user_regs *regs)
>          if ( nvmx_handle_vmresume(regs) == X86EMUL_OKAY )
>              update_guest_eip();
>          break;
> -
> +    case EXIT_REASON_INVEPT:
> +        if ( nvmx_handle_invept(regs) == X86EMUL_OKAY )
> +            update_guest_eip();
> +        break;

In (potentially going to become) long switch statements, please
don't drop the blank lines between individual cases - instead of
dropping the line here, you wold want to insert another one
below the new separately handled case.

>      case EXIT_REASON_MWAIT_INSTRUCTION:
>      case EXIT_REASON_MONITOR_INSTRUCTION:
>      case EXIT_REASON_GETSEC:
> -    case EXIT_REASON_INVEPT:
>      case EXIT_REASON_INVVPID:
>          /*
>           * We should never exit on GETSEC because CR4.SMXE is always 0 when
> --- a/xen/arch/x86/hvm/vmx/vvmx.c
> +++ b/xen/arch/x86/hvm/vmx/vvmx.c
> @@ -1356,6 +1356,45 @@ int nvmx_handle_vmwrite(struct cpu_user_regs *regs)
>      return X86EMUL_OKAY;
>  }
>  
> +int nvmx_handle_invept(struct cpu_user_regs *regs)
> +{
> +    struct vmx_inst_decoded decode;
> +    unsigned long eptp;
> +    u64 inv_type;
> +
> +    if ( !cpu_has_vmx_ept )
> +        return X86EMUL_EXCEPTION;
> +
> +    if ( decode_vmx_inst(regs, &decode, &eptp, 0)
> +             != X86EMUL_OKAY )
> +        return X86EMUL_EXCEPTION;
> +
> +    inv_type = reg_read(regs, decode.reg2);
> +    gdprintk(XENLOG_DEBUG,"inv_type:%ld, eptp:%lx\n", inv_type, eptp);

An unconditional printk() on an operation potentially happening
quite frequently? Even with XENLOG_DEBUG this is not acceptable
imo.

> +
> +    switch ( inv_type ) {
> +    case INVEPT_SINGLE_CONTEXT:
> +        {
> +            struct p2m_domain *p2m = vcpu_nestedhvm(current).nv_p2m;
> +            if ( p2m )
> +            {
> +                 p2m_flush(current, p2m);

Despite your comment in 00/10, there still is a whitespace issues
at least here (didn't look that closely elsewhere).

> +                ept_sync_domain(p2m);
> +            }
> +        }
> +        break;
> +    case INVEPT_ALL_CONTEXT:
> +        p2m_flush_nestedp2m(current->domain);
> +        __invept(INVEPT_ALL_CONTEXT, 0, 0);
> +        break;
> +    default:
> +        return X86EMUL_EXCEPTION;
> +    }
> +    vmreturn(regs, VMSUCCEED);
> +    return X86EMUL_OKAY;
> +}
> +
> +
>  #define __emul_value(enable1, default1) \
>      ((enable1 | default1) << 32 | (default1))
>  
> --- a/xen/arch/x86/mm/p2m.c
> +++ b/xen/arch/x86/mm/p2m.c
> @@ -1465,7 +1465,7 @@ p2m_flush_table(struct p2m_domain *p2m)
>  void
>  p2m_flush(struct vcpu *v, struct p2m_domain *p2m)
>  {
> -    ASSERT(v->domain == p2m->domain);
> +    ASSERT(p2m && v->domain == p2m->domain);

How is this change related to the rest of the patch?

Jan

>      vcpu_nestedhvm(v).nv_p2m = NULL;
>      p2m_flush_table(p2m);
>      hvm_asid_flush_vcpu(v);



_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel