* [PATCH v2] mm: huge_memory: convert split_huge_pages_all() to use a folio
@ 2022-12-30 9:30 Kefeng Wang
2022-12-30 21:45 ` Andrew Morton
0 siblings, 1 reply; 5+ messages in thread
From: Kefeng Wang @ 2022-12-30 9:30 UTC (permalink / raw)
To: Andrew Morton, linux-mm; +Cc: linux-kernel, willy, Kefeng Wang
Straightforwardly convert split_huge_pages_all() to use a folio.
Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
---
- adjust indenting of code pointed by Matthew
mm/huge_memory.c | 25 +++++++++++++++++--------
1 file changed, 17 insertions(+), 8 deletions(-)
diff --git a/mm/huge_memory.c b/mm/huge_memory.c
index 266c4b557946..46e86293935b 100644
--- a/mm/huge_memory.c
+++ b/mm/huge_memory.c
@@ -2932,6 +2932,7 @@ static void split_huge_pages_all(void)
{
struct zone *zone;
struct page *page;
+ struct folio *folio;
unsigned long pfn, max_zone_pfn;
unsigned long total = 0, split = 0;
@@ -2944,24 +2945,32 @@ static void split_huge_pages_all(void)
int nr_pages;
page = pfn_to_online_page(pfn);
- if (!page || !get_page_unless_zero(page))
+ if (!page || PageTail(page))
+ continue;
+ folio = page_folio(page);
+ if (!folio_try_get(folio))
continue;
- if (zone != page_zone(page))
+ if (unlikely(page_folio(page) != folio))
+ goto next;
+
+ if (zone != folio_zone(folio))
goto next;
- if (!PageHead(page) || PageHuge(page) || !PageLRU(page))
+ if (!folio_test_large(folio) ||
+ folio_test_hugetlb(folio) ||
+ !folio_test_lru(folio))
goto next;
total++;
- lock_page(page);
- nr_pages = thp_nr_pages(page);
- if (!split_huge_page(page))
+ folio_lock(folio);
+ nr_pages = folio_nr_pages(folio);
+ if (!split_folio(folio))
split++;
pfn += nr_pages - 1;
- unlock_page(page);
+ folio_unlock(folio);
next:
- put_page(page);
+ folio_put(folio);
cond_resched();
}
}
--
2.35.3
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH v2] mm: huge_memory: convert split_huge_pages_all() to use a folio
2022-12-30 9:30 [PATCH v2] mm: huge_memory: convert split_huge_pages_all() to use a folio Kefeng Wang
@ 2022-12-30 21:45 ` Andrew Morton
2023-01-03 1:58 ` Kefeng Wang
0 siblings, 1 reply; 5+ messages in thread
From: Andrew Morton @ 2022-12-30 21:45 UTC (permalink / raw)
To: Kefeng Wang; +Cc: linux-mm, linux-kernel, willy
On Fri, 30 Dec 2022 17:30:20 +0800 Kefeng Wang <wangkefeng.wang@huawei.com> wrote:
> Straightforwardly convert split_huge_pages_all() to use a folio.
>
> ...
>
> --- a/mm/huge_memory.c
> +++ b/mm/huge_memory.c
> @@ -2932,6 +2932,7 @@ static void split_huge_pages_all(void)
> {
> struct zone *zone;
> struct page *page;
> + struct folio *folio;
> unsigned long pfn, max_zone_pfn;
> unsigned long total = 0, split = 0;
>
> @@ -2944,24 +2945,32 @@ static void split_huge_pages_all(void)
> int nr_pages;
>
> page = pfn_to_online_page(pfn);
> - if (!page || !get_page_unless_zero(page))
> + if (!page || PageTail(page))
> + continue;
> + folio = page_folio(page);
> + if (!folio_try_get(folio))
> continue;
>
> - if (zone != page_zone(page))
> + if (unlikely(page_folio(page) != folio))
> + goto next;
> +
> + if (zone != folio_zone(folio))
> goto next;
I'm still not understanding the above hunk. Why is the
"page_folio(page) != folio" check added? Should it be commented?
> - if (!PageHead(page) || PageHuge(page) || !PageLRU(page))
> + if (!folio_test_large(folio) ||
> + folio_test_hugetlb(folio) ||
> + !folio_test_lru(folio))
> goto next;
>
> total++;
> - lock_page(page);
> - nr_pages = thp_nr_pages(page);
> - if (!split_huge_page(page))
> + folio_lock(folio);
> + nr_pages = folio_nr_pages(folio);
> + if (!split_folio(folio))
> split++;
> pfn += nr_pages - 1;
> - unlock_page(page);
> + folio_unlock(folio);
> next:
> - put_page(page);
> + folio_put(folio);
> cond_resched();
> }
> }
> --
> 2.35.3
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH v2] mm: huge_memory: convert split_huge_pages_all() to use a folio
2022-12-30 21:45 ` Andrew Morton
@ 2023-01-03 1:58 ` Kefeng Wang
2023-01-06 3:06 ` Andrew Morton
0 siblings, 1 reply; 5+ messages in thread
From: Kefeng Wang @ 2023-01-03 1:58 UTC (permalink / raw)
To: Andrew Morton; +Cc: linux-mm, linux-kernel, willy
On 2022/12/31 5:45, Andrew Morton wrote:
> On Fri, 30 Dec 2022 17:30:20 +0800 Kefeng Wang <wangkefeng.wang@huawei.com> wrote:
>
>> Straightforwardly convert split_huge_pages_all() to use a folio.
>>
>> ...
>>
>> --- a/mm/huge_memory.c
>> +++ b/mm/huge_memory.c
>> @@ -2932,6 +2932,7 @@ static void split_huge_pages_all(void)
>> {
>> struct zone *zone;
>> struct page *page;
>> + struct folio *folio;
>> unsigned long pfn, max_zone_pfn;
>> unsigned long total = 0, split = 0;
>>
>> @@ -2944,24 +2945,32 @@ static void split_huge_pages_all(void)
>> int nr_pages;
>>
>> page = pfn_to_online_page(pfn);
>> - if (!page || !get_page_unless_zero(page))
>> + if (!page || PageTail(page))
>> + continue;
>> + folio = page_folio(page);
>> + if (!folio_try_get(folio))
>> continue;
>>
>> - if (zone != page_zone(page))
>> + if (unlikely(page_folio(page) != folio))
>> + goto next;
>> +
>> + if (zone != folio_zone(folio))
>> goto next;
>
> I'm still not understanding the above hunk. Why is the
> "page_folio(page) != folio" check added? Should it be commented?
There is a comment in try_get_folio(), is it enough?
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH v2] mm: huge_memory: convert split_huge_pages_all() to use a folio
2023-01-03 1:58 ` Kefeng Wang
@ 2023-01-06 3:06 ` Andrew Morton
2023-01-06 6:23 ` Kefeng Wang
0 siblings, 1 reply; 5+ messages in thread
From: Andrew Morton @ 2023-01-06 3:06 UTC (permalink / raw)
To: Kefeng Wang; +Cc: linux-mm, linux-kernel, willy
On Tue, 3 Jan 2023 09:58:36 +0800 Kefeng Wang <wangkefeng.wang@huawei.com> wrote:
>
>
> On 2022/12/31 5:45, Andrew Morton wrote:
> > On Fri, 30 Dec 2022 17:30:20 +0800 Kefeng Wang <wangkefeng.wang@huawei.com> wrote:
> >
> >> Straightforwardly convert split_huge_pages_all() to use a folio.
> >>
> >> ...
> >>
> >> --- a/mm/huge_memory.c
> >> +++ b/mm/huge_memory.c
> >> @@ -2932,6 +2932,7 @@ static void split_huge_pages_all(void)
> >> {
> >> struct zone *zone;
> >> struct page *page;
> >> + struct folio *folio;
> >> unsigned long pfn, max_zone_pfn;
> >> unsigned long total = 0, split = 0;
> >>
> >> @@ -2944,24 +2945,32 @@ static void split_huge_pages_all(void)
> >> int nr_pages;
> >>
> >> page = pfn_to_online_page(pfn);
> >> - if (!page || !get_page_unless_zero(page))
> >> + if (!page || PageTail(page))
> >> + continue;
> >> + folio = page_folio(page);
> >> + if (!folio_try_get(folio))
> >> continue;
> >>
> >> - if (zone != page_zone(page))
> >> + if (unlikely(page_folio(page) != folio))
> >> + goto next;
> >> +
> >> + if (zone != folio_zone(folio))
> >> goto next;
> >
> > I'm still not understanding the above hunk. Why is the
> > "page_folio(page) != folio" check added? Should it be commented?
>
> There is a comment in try_get_folio(), is it enough?
Nobody would think to look at a private function in gup.c.
I suggest this folio_try_get() rule be documented at the
folio_try_get() implementation site.
Most callers do perform this check, but not the ones in vmscan.c?
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH v2] mm: huge_memory: convert split_huge_pages_all() to use a folio
2023-01-06 3:06 ` Andrew Morton
@ 2023-01-06 6:23 ` Kefeng Wang
0 siblings, 0 replies; 5+ messages in thread
From: Kefeng Wang @ 2023-01-06 6:23 UTC (permalink / raw)
To: Andrew Morton; +Cc: linux-mm, linux-kernel, willy
On 2023/1/6 11:06, Andrew Morton wrote:
> On Tue, 3 Jan 2023 09:58:36 +0800 Kefeng Wang <wangkefeng.wang@huawei.com> wrote:
>
>>
>>
>> On 2022/12/31 5:45, Andrew Morton wrote:
>>> On Fri, 30 Dec 2022 17:30:20 +0800 Kefeng Wang <wangkefeng.wang@huawei.com> wrote:
>>>
>>>> Straightforwardly convert split_huge_pages_all() to use a folio.
>>>>
>>>> ...
>>>>
>>>> --- a/mm/huge_memory.c
>>>> +++ b/mm/huge_memory.c
>>>> @@ -2932,6 +2932,7 @@ static void split_huge_pages_all(void)
>>>> {
>>>> struct zone *zone;
>>>> struct page *page;
>>>> + struct folio *folio;
>>>> unsigned long pfn, max_zone_pfn;
>>>> unsigned long total = 0, split = 0;
>>>>
>>>> @@ -2944,24 +2945,32 @@ static void split_huge_pages_all(void)
>>>> int nr_pages;
>>>>
>>>> page = pfn_to_online_page(pfn);
>>>> - if (!page || !get_page_unless_zero(page))
>>>> + if (!page || PageTail(page))
>>>> + continue;
>>>> + folio = page_folio(page);
>>>> + if (!folio_try_get(folio))
>>>> continue;
>>>>
>>>> - if (zone != page_zone(page))
>>>> + if (unlikely(page_folio(page) != folio))
>>>> + goto next;
>>>> +
>>>> + if (zone != folio_zone(folio))
>>>> goto next;
>>>
>>> I'm still not understanding the above hunk. Why is the
>>> "page_folio(page) != folio" check added? Should it be commented?
>>
>> There is a comment in try_get_folio(), is it enough?
>
> Nobody would think to look at a private function in gup.c.
Sorry，I mean that comment is to explain why adding a new recheck.
>
> I suggest this folio_try_get() rule be documented at the
> folio_try_get() implementation site.
>
> Most callers do perform this check, but not the ones in vmscan.c ?
Hope Matthew could give some advise, thanks
^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2023-01-06 6:23 UTC | newest]
Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-12-30 9:30 [PATCH v2] mm: huge_memory: convert split_huge_pages_all() to use a folio Kefeng Wang
2022-12-30 21:45 ` Andrew Morton
2023-01-03 1:58 ` Kefeng Wang
2023-01-06 3:06 ` Andrew Morton
2023-01-06 6:23 ` Kefeng Wang
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox