[PATCH v2 2/6] mm/rmap: use huge_ptep_get() in try_to_unmap_one()

[David Hildenbrand (Arm)]

On 7/2/26 11:35, Muchun Song wrote:
> 
> 
>> On Jul 2, 2026, at 17:08, Dev Jain <dev.jain at arm.com> wrote:
>>
>>
>>
>> On 02/07/26 2:17 pm, Muchun Song wrote:
>>>
>>>
>>>
>>> Maybe I didn't express my thoughts clearly in the first version, let me
>>> explain in more detail.
>>>
>>> We should define this stub as a no-op for !CONFIG_HUGETLB_PAGE (like
>>> set_huge_pte_at, that is why I mentioned 5d4af6195c87c6 for your reference
>>> in your previous version). Currently, you've added a declaration, but the
>>> function itself doesn't actually exist, which seems quite strange to me.
>>
>> https://lore.kernel.org/all/a4fe8ba6-2ecd-4bb9-95a9-27f9f1e87d2e@kernel.org/
>>
>> David suggested this. Honestly I quite like David's suggestion, what do you
>> think?
> 
> Thanks for pointing that out, I missed it earlier. That said, looking at
> hugetlb.h, it already contains quite a few no-op stubs. To keep things
> consistent, I'd personally prefer a stub here. Since David suggested this,
> I’d love to hear his thoughts on this as well.

It's the shortest possible way to enforce (through the linker!) that you
function is not called from wrong context.

The only alternative is using a stub with a BUILD_BUG() in it. Which will 4 5 LOC :)

-- 
Cheers,

David