x86: add_branch_prefix_frag_p() vs PadLock insns

[Jan Beulich]

On 26.09.2025 23:14, H.J. Lu wrote:
> On Fri, Sep 26, 2025 at 6:29 PM Jan Beulich <jbeulich@suse.com> wrote:
>>
>> On 26.09.2025 12:21, H.J. Lu wrote:
>>> On Fri, Sep 26, 2025, 6:16 PM Jan Beulich <jbeulich@suse.com> wrote:
>>>> why is it that add_branch_prefix_frag_p() special-cases PadLock insns?
>>>> They're by far not the only ones including prefixes in their opcodes.
>>>> IOW - I wonder if the conditional there is either insufficient or
>>>> excessive.
>>>
>>> If I remember correctly, this is to mitigate
>>> an Intel processor issue.  No Intel processor
>>> support such instructions.
>>
>> Yes. But then why was the PadLock check added there right when that code
>> was introduced? (It has meanwhile been extended, but if that's really
>> needed for some reason, it likely wasn't extended enough.)
> 
> This command-line mitigation can be applied to any assembly
> codes including padlock which will be skipped.

Well. Once again I'm in trouble seeing how this answers, even just remotely,
my question. Please read my original message again, and then I'd appreciate
an answer which at least gives a me a rough hint in which direction to alter
things. The way the code is right now simply cannot be right.

Jan