On Mon, 2024-02-26 at 00:25 -0800, isaku.yamahata@xxxxxxxxx wrote:
> --- a/arch/x86/kvm/vmx/tdx.h
> +++ b/arch/x86/kvm/vmx/tdx.h
> @@ -3,6 +3,9 @@
> #define __KVM_X86_TDX_H
>
> #ifdef CONFIG_INTEL_TDX_HOST
> +
> +#include "tdx_ops.h"
> +
> struct kvm_tdx {
> struct kvm kvm;
> /* TDX specific members follow. */
I am consistently hitting build error for the middle patches in our
internal tree, mostly because of this madness of header file inclusion.
I found the above inclusion of "tdx_ops.h" in "tdx.h" just out of blue.
We have
- "tdx_arch.h"
- "tdx_errno.h"
- "tdx_ops.h"
- "tdx.h"
The first two can be included by the "tdx.h", so that we can have a rule
for C files to just include "tdx.h", i.e., the C files should never need
to include the first two explicitly.
The "tdx_ops.h" is a little bit confusing. I _think_ the purpose of it is
to only contain SEAMCALL wrappers. But I am not sure whether it can be
included by any C file directly.
Based on above code change, I _think_ the intention is to also embed it to
"tdx.h", so the C files should just include "tdx.h".
But based on Sean's comments, the SEAMCALL wrappers will be changed to
take 'struct kvm_tdx *' and 'struct vcpu_tdx *', so they need the
declaration of those structures which are in "tdx.h".
I think we can just make a rule that, "tdx_ops.h" should never be directly
included by any C file, instead, we include "tdx_ops.h" into "tdx.h"
somewhere after declaration of 'struct kvm_tdx' and 'struct vcpu_tdx'.
And such inclusion should happen when the "tdx_ops.h" is introduced.
Am I missing anything?
