Re: [PATCH 3/4] x86/tdx: Relax SEPT_VE_DISABLE check for debug TD

[Sathyanarayanan Kuppuswamy]

On 12/9/22 5:25 AM, Kirill A. Shutemov wrote:
> SEPT_VE_DISABLE check is required to keep the TD protected from VMM
> attacks, but it makes harder to debug guest kernel bugs. If guest
> touches unaccepted memory the TD will get terminated without any
> traces on what has happened.
> 
> Relax the SEPT_VE_DISABLE check to warning on debug TD and panic() in
> the #VE handler on EPT-violation on private memory. It will produce
> useful backtrace.
> 
> Signed-off-by: Kirill A. Shutemov <kirill.shutemov@xxxxxxxxxxxxxxx>
> ---
>  arch/x86/coco/tdx/tdx.c | 14 ++++++++++++--
>  1 file changed, 12 insertions(+), 2 deletions(-)
> 
> diff --git a/arch/x86/coco/tdx/tdx.c b/arch/x86/coco/tdx/tdx.c
> index 8ad04d101270..0e47846ff8ff 100644
> --- a/arch/x86/coco/tdx/tdx.c
> +++ b/arch/x86/coco/tdx/tdx.c
> @@ -38,6 +38,7 @@
>  #define VE_GET_PORT_NUM(e)	((e) >> 16)
>  #define VE_IS_IO_STRING(e)	((e) & BIT(4))
>  
> +#define ATTR_DEBUG		BIT(0)
>  #define ATTR_SEPT_VE_DISABLE	BIT(28)
>  
>  /* TDX Module call error codes */
> @@ -207,8 +208,15 @@ static void tdx_parse_tdinfo(u64 *cc_mask)
>  	 * TD-private memory.  Only VMM-shared memory (MMIO) will #VE.
>  	 */
>  	td_attr = out.rdx;
> -	if (!(td_attr & ATTR_SEPT_VE_DISABLE))
> -		tdx_panic("TD misconfiguration: SEPT_VE_DISABLE attribute must be set.");
> +	if (!(td_attr & ATTR_SEPT_VE_DISABLE)) {
> +		const char *msg = "TD misconfiguration: SEPT_VE_DISABLE attribute must be set.";
> +
> +		/* Relax SEPT_VE_DISABLE check for debug TD. */
> +		if (td_attr & ATTR_DEBUG)
> +			pr_warn("%s\n", msg);
> +		else
> +			tdx_panic(msg);
> +	}
>  }
>  
>  /*
> @@ -682,6 +690,8 @@ static int virt_exception_kernel(struct pt_regs *regs, struct ve_info *ve)
>  	case EXIT_REASON_CPUID:
>  		return handle_cpuid(regs, ve);
>  	case EXIT_REASON_EPT_VIOLATION:
> +		if (ve->gpa != cc_mkdec(ve->gpa))
> +			panic("Unexpected EPT-violation on private memory.");

Why add this change part of TD debug check? Should this be a separate patch?

>  		return handle_mmio(regs, ve);
>  	case EXIT_REASON_IO_INSTRUCTION:
>  		return handle_io(regs, ve);

-- 
Sathyanarayanan Kuppuswamy
Linux Kernel Developer