On 06/01/2025 14:23, Qi Zheng wrote:
> On 2025/1/6 20:36, Alexander Gordeev wrote:
>> On Mon, Jan 06, 2025 at 06:55:58PM +0800, Qi Zheng wrote:
>>>>> +static inline void pagetable_dtor(struct ptdesc *ptdesc)
>>>>> +{
>>>>> + struct folio *folio = ptdesc_folio(ptdesc);
>>>>> +
>>>>> + ptlock_free(ptdesc);
>>>>> + __folio_clear_pgtable(folio);
>>>>> + lruvec_stat_sub_folio(folio, NR_PAGETABLE);
>>>>> +}
>>>>> +
>>>>
>>>> If I am not mistaken, it is just pagetable_pte_dtor() rename.
>>>> What is the point in moving the code around?
>>>
>>> No, this is to unify pagetable_p*_dtor() into pagetable_dtor(), so
>>> that we can move pagetable_dtor() to __tlb_remove_table(), and then
>>> ptlock and PTE page can be freed together through RCU, which is
>>> also the main purpose of this patch series.
>>
>> I am only talking about this patch. pagetable_dtor() code above is
>> the same pagetable_pte_dtor() below - it is only the function name
>> that changed. So why to move the function body? Anyway, that is
>
> Ah, I just don't want to put pagetable_dtor() in between
> pagetable_pte_ctor() and ___pte_offset_map(), so I moved it above
> pagetable_pte_ctor(). No other special reason. 😉
I think inserting pagetable_dtor() there makes sense. I wouldn't say
that pagetable_pte_dtor() is being renamed to pagetable_dtor(), because
in fact this patch replaces all of pagetable_{pte,pmd,pud}_dtor() with
pagetable_dtor(), and it is arguably clearer to insert the latter higher
up in mm.h.
FWIW my follow-up series introduces a common __pagetable_dtor(),
inserted below pagetable_ctor() [1].
- Kevin
[1]
https://lore.kernel.org/linux-mm/20250103184415.2744423-2-kevin.brodsky@xxxxxxx/
