Re: [PATCH v10 03/11] x86/cpufeatures: Add TDX Guest CPU feature

[Thomas Gleixner]

On Wed, Oct 13 2021 at 23:07, Borislav Petkov wrote:
> On Wed, Oct 13, 2021 at 10:44:37PM +0200, Thomas Gleixner wrote:
>> No. This is tasteless garbage, really.
>> 
>> tdx_early_init() is invoked from x86_64_start_kernel() very early in the
>> boot process __before__ is_tdx_guest() is invoked.
>> 
>> So why on earth is it requried to keep those conditionals and cpuid()
>> muck around after init?
>
> Yah, reportedly, they wanna parse cmdline options so it has to be after
> copy_bootdata() but copy_bootdata() has a cc_platform_has() call which
> ends up in is_tdx_guest() on Intel and there you have the catch 22
> because CPUID hasn't happened yet in tdx_early_init().

Seriously?

So this ends up in doing:

   use();
   init();

Can you spot what's wrong with that?

That's a clear violation of common sense and is simply not going to
happen. Why? If you think about deep defensive programming then use()
will look like this:

use()
{
        assert(initialized);
}

which is not something made up. It's a fundamental principle of
programming and some languages enforce that for very good reasons.

Just because it can be done in C is no justification.

What's wrong with:

x86_64_start_kernel()

    tdx_early_init();

    copy_bootdata();
    
    tdx_late_init();

Absolutely nothing. It's clear, simple and well defined.

Thanks,

        tglx