Re: [V2 PATCH] x86/tdx: add VIRT_CPUID2 virtualization if REDUCE_VE was not successful

[Dave Hansen <dave.hansen@intel.com>]

On 4/30/25 08:09, Zhiquan Li wrote:
> On 2025/4/30 21:44, Dave Hansen wrote:
>>> Currently the most customer's complaints come from the CPUID leaf 0x2
>>> not virtualization, and most of access come from user space.  Is it
>>> appropriate for such behavior directly cause a guest kernel panic?
>> If a VMM doesn't properly configure topology properly and allows
>> REDUCE_VE, then panic. They obviously got something wrong because that
>> configuration doesn't make any sense.
> OK, I agree with you.  Is it better to simplify the logic like this:
> 
>     static void reduce_unnecessary_ve(void)
>     {
>         u64 configured;
> 
>         /* Has the VMM provided a valid topology configuration? */
>         tdg_vm_rd(TDCS_TOPOLOGY_ENUM_CONFIGURED, &configured);
> 
>         if (!configured)
>             panic("VMM did not configure X2APIC_IDs properly\n");
> 
>         tdg_vm_wr(TDCS_TD_CTLS, TD_CTLS_REDUCE_VE, TD_CTLS_REDUCE_VE);
>     }
> 
> Since merely fall back to enable ENUM_TOPOLOGY isn't enough, the guest
> still suffering the CPUID leaf 0x2 not virtualization regression, like
> the glibc bug.  Full REDUCE_VE is really expected.

Either that or fix the TDX module in some way to untangle the mess.