Re: [Xen-devel] [PATCH 3/5] x86emul: support RTM instructions

[Andrew Cooper <andrew.cooper3@xxxxxxxxxx>]

On 08/09/16 14:44, Jan Beulich wrote:
> Minimal emulation: XBEGIN aborts right away, hence
> - XABORT is just a no-op,
> - XEND always raises #GP,
> - XTEST always signals neither RTM nor HLE are active.
>
> Signed-off-by: Jan Beulich <jbeulich@xxxxxxxx>
>
> --- a/xen/arch/x86/x86_emulate/x86_emulate.c
> +++ b/xen/arch/x86/x86_emulate/x86_emulate.c
> @@ -1172,6 +1172,8 @@ static bool_t vcpu_has(
>  #define vcpu_has_clflush() vcpu_has(       1, EDX, 19, ctxt, ops)
>  #define vcpu_has_lzcnt() vcpu_has(0x80000001, ECX,  5, ctxt, ops)
>  #define vcpu_has_bmi1()  vcpu_has(0x00000007, EBX,  3, ctxt, ops)
> +#define vcpu_has_hle()   vcpu_has(0x00000007, EBX,  4, ctxt, ops)
> +#define vcpu_has_rtm()   vcpu_has(0x00000007, EBX, 11, ctxt, ops)
>  
>  #define vcpu_must_have(leaf, reg, bit) \
>      generate_exception_if(!vcpu_has(leaf, reg, bit, ctxt, ops), EXC_UD, -1)
> @@ -2852,7 +2854,18 @@ x86_emulate(
>          lock_prefix = 1;
>          break;
>  
> -    case 0xc6 ... 0xc7: /* mov (sole member of Grp11) */
> +    case 0xc6: /* Grp11: mov / xabort */
> +    case 0xc7: /* Grp11: mov / xbegin */
> +        if ( modrm == 0xf8 && vcpu_has_rtm() )
> +        {
> +            if ( b & 1 )
> +            {
> +                jmp_rel((int32_t)src.val);

This should be based on op_bytes.  There are two forms, one with a rel16
jump and one with rel32, and I don't see this being accounted for
anywhere else.

> +                _regs.eax = 0;
> +            }
> +            dst.type = OP_NONE;

The XABORT instruction should explicitly set bit.

Incidentally, what is supposed to happen if we branch into the middle of
an RTM region?

> +            break;
> +        }
>          generate_exception_if((modrm_reg & 7) != 0, EXC_UD, -1);
>      case 0x88 ... 0x8b: /* mov */
>      case 0xa0 ... 0xa1: /* mov mem.offs,{%al,%ax,%eax,%rax} */
> @@ -4246,6 +4259,17 @@ x86_emulate(
>                  goto done;
>              goto no_writeback;
>  
> +        case 0xd5: /* xend */
> +            generate_exception_if(vcpu_has_rtm() && !vex.pfx, EXC_GP, 0);
> +            break;
> +
> +        case 0xd6: /* xtest */
> +            if ( (!vcpu_has_rtm() && !vcpu_has_hle()) || vex.pfx )
> +                break;

Shouldn't this raise #UD explicitly?  I can't spot anything which does
if we break out.

~Andrew

> +            /* Neither HLE nor RTM can be active when we get here. */
> +            _regs.eflags |= EFLG_ZF;
> +            goto no_writeback;
> +
>          case 0xdf: /* invlpga */
>              generate_exception_if(!in_protmode(ctxt, ops), EXC_UD, -1);
>              generate_exception_if(!mode_ring0(), EXC_GP, 0);
>
>
>


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