Re: [PATCH v19 007/130] x86/virt/tdx: Export SEAMCALL functions

[Sean Christopherson]

On Mon, Apr 22, 2024, Rick P Edgecombe wrote:
> On Mon, 2024-04-22 at 14:46 +0300, kirill.shutemov@xxxxxxxxxxxxxxx wrote:
> > On Fri, Apr 19, 2024 at 08:04:26PM +0000, Edgecombe, Rick P wrote:
> > > On Fri, 2024-04-19 at 17:46 +0300, kirill.shutemov@xxxxxxxxxxxxxxx wrote:
> > > > 
> > > > > Side topic #3, the ud2 to induce panic should be out-of-line.
> > > > 
> > > > Yeah. I switched to the inline one while debugging one section mismatch
> > > > issue and forgot to switch back.
> > > 
> > > Sorry, why do we need to panic?
> > 
> > It panics in cases that should never occur if the TDX module is
> > functioning properly. For example, TDVMCALL itself should never fail,
> > although the leaf function could.
> 
> Panic should normally be for desperate situations when horrible things will
> likely happen if we continue, right? Why are we adding a panic when we didn't
> have one before? Is it a second change, or a side affect of the refactor?

The kernel already does panic() if TDCALL itself fails,

  static inline void tdcall(u64 fn, struct tdx_module_args *args)
  {
	if (__tdcall_ret(fn, args))
		panic("TDCALL %lld failed (Buggy TDX module!)\n", fn);
  }

  /* Called from __tdx_hypercall() for unrecoverable failure */
  noinstr void __noreturn __tdx_hypercall_failed(void)
  {
	instrumentation_begin();
	panic("TDVMCALL failed. TDX module bug?");
  }

it's just doesn in C code via panic(), not in asm via a bare ud2.