Re: [PATCH v5 07/21] x86/fpu: Provide fpu_enable_guest_xfd_features() for KVM

[Yang Zhong <yang.zhong@intel.com>]

On Wed, Jan 05, 2022 at 02:06:40PM +0100, Paolo Bonzini wrote:
> On 1/5/22 13:35, Yang Zhong wrote:
> >+int fpu_enable_guest_xfd_features(struct fpu_guest *guest_fpu, u64 xfeatures)
> >+{
> >+	lockdep_assert_preemption_enabled();
> >+
> 
> The old fpu_update_guest_perm_features(guest_fpu) is equivalent to
> 
> 	fpu_enable_guest_xfd_features(guest_fpu, guest_fpu->perm);
> 
> Missing doc comment:
> 
> /*
>  * fpu_enable_guest_xfd_features - Enable xfeatures according to guest perm
>  * @guest_fpu:         Pointer to the guest FPU container
>  * @xfeatures:         Features requested by guest CPUID
>  *
>  * Enable all dynamic xfeatures according to guest perm and requested CPUID.
>  * Invoked if the caller wants to conservatively expand fpstate buffer instead
>  * of waiting until XCR0 or XFD MSR is written.
>  *
>  * Return: 0 on success, error code otherwise
>  */
> 
> Also, the check for 32-bit is slightly imprecise:
> 
> 	/* Dynamic xfeatures are not supported with 32-bit kernels. */
> 	if (!IS_ENABLED(CONFIG_X86_64))
> -		return 0;
> +		return -EINVAL;
> 
> since we only get here with xfeatures != 0 (if it compiles, just removing
> the IS_ENABLED check altogether would be even better).  With these changes,
> 

  Paolo, I did 32 bit kernel build tests

  (1). w/ IS_ENABLED(CONFIG_X86_64)

      if (!IS_ENABLED(CONFIG_X86_64))
      	return -EINVAL; 
    
     This 32 bit kernel can successfully build.

  (2). remove IS_ENABLED(CONFIG_X86_64)

      The 32 bit kernel build is failed, and the error as:

      ld: arch/x86/kernel/fpu/core.o: in function `fpu_enable_guest_xfd_features':
/root/amx/v5/kvm/arch/x86/kernel/fpu/core.c:278: undefined reference to `__xfd_enable_feature'

      Seems we need define 32 bit API for __xfd_enable_feature().

   I also tried (1) in desktop, but I could not reboot the machine after installed 32 bit kernel.

   thanks a lot!

   Yang 





> Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
> 
> Thanks,
> 
> Paolo