Re: [PATCH 5/5] x86: don't build unused entry code when !PV32

[Jan Beulich <jbeulich@xxxxxxxx>]

On 01.04.2021 16:01, Roger Pau Monné wrote:
> On Wed, Nov 25, 2020 at 09:51:33AM +0100, Jan Beulich wrote:
>> @@ -230,6 +230,13 @@ ENTRY(cstar_enter)
>>          sti
>>  
>>          movq  STACK_CPUINFO_FIELD(current_vcpu)(%rbx), %rbx
>> +
>> +#ifndef CONFIG_PV32
>> +
>> +        jmp   switch_to_kernel
>> +
>> +#else
>> +
>>          movq  VCPU_domain(%rbx),%rcx
>>          cmpb  $0,DOMAIN_is_32bit_pv(%rcx)
>>          je    switch_to_kernel
>> @@ -393,3 +400,5 @@ compat_crash_page_fault:
>>          jmp   .Lft14
>>  .previous
>>          _ASM_EXTABLE(.Lft14, .Lfx14)
>> +
>> +#endif /* CONFIG_PV32 */
> 
> Seeing this chunk, would it make sense to move the cstar_enter part
> relevant to !is_32bit_pv into the common entry.S and leave the rest
> here as compat_cstar_enter or some such?
> 
> AFAICT we only need a tiny part of the compat stuff when !CONFIG_PV32,
> so I think we could make the hole compat/entry.S depend on
> CONFIG_PV32.

To be honest I was meaning to see whether this would work out reasonably
well in a separate, follow-on change, as the code movement might make it
harder to judge on the change here compared to the #ifdef insertions.
But maybe I was wrong and should give this a try right away.

Jan