Re: [PATCH v13 06/22] x86/virt/tdx: Add SEAMCALL error printing for module initialization

[Huang, Kai]

On Fri, 2023-09-08 at 09:31 -0700, Dave Hansen wrote:
> On 8/25/23 05:14, Kai Huang wrote:
> > +#define SEAMCALL_PRERR(__seamcall_func, __fn, __args, __seamcall_err_func)	\
> > +({										\
> > +	u64 ___sret = __SEAMCALL_PRERR(__seamcall_func, __fn, __args,		\
> > +			__seamcall_err_func, pr_err);				\
> > +	int ___ret;								\
> > +										\
> > +	switch (___sret) {							\
> > +	case TDX_SUCCESS:							\
> > +		___ret = 0;							\
> > +		break;								\
> > +	case TDX_SEAMCALL_VMFAILINVALID:					\
> > +		pr_err("SEAMCALL failed: TDX module not loaded.\n");		\
> > +		___ret = -ENODEV;						\
> > +		break;								\
> > +	case TDX_SEAMCALL_GP:							\
> > +		pr_err("SEAMCALL failed: TDX disabled by BIOS.\n");		\
> > +		___ret = -EOPNOTSUPP;						\
> > +		break;								\
> > +	case TDX_SEAMCALL_UD:							\
> > +		pr_err("SEAMCALL failed: CPU not in VMX operation.\n");		\
> > +		___ret = -EACCES;						\
> > +		break;								\
> > +	default:								\
> > +		___ret = -EIO;							\
> > +	}									\
> > +	___ret;									\
> > +})
> 
> I have no clue where all of this came from or why it is necessary or why
> it has to be macros.  I'm just utterly confused.
> 
> I was really hoping to be able to run through this set and get it ready
> to be merged.  But it seems to still be seeing a *LOT* of change.
> Should I wait another few weeks for this to settle down again?

Those changes are due to SEAMCALL API change from the TDCALL/VMCALL/SEAMCALL
assembly change patchset.  I'll work internally to make this stable asap (I
tried before but was suggested to sent out to community for feedback).

Also I would appreciate if you could take a look at patch 18/19 (which are
separated small patches for better review) and patch 20 (reset PAMT in kexec).

Thanks in advance!