2024-03-29 06:59:53

by Baolin Wang

Subject: [PATCH] mm: huge_memory: add the missing folio_test_pmd_mappable() for THP split statistics

Now the mTHP can also be split or added into the deferred list, so add
folio_test_pmd_mappable() validation for PMD mapped THP, to avoid confusion
with PMD mapped THP related statistics.

Signed-off-by: Baolin Wang <[email protected]>
---
mm/huge_memory.c | 6 ++++--
1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/mm/huge_memory.c b/mm/huge_memory.c
index 1683de78c313..3ca9282a0dc9 100644
--- a/mm/huge_memory.c
+++ b/mm/huge_memory.c
@@ -3109,7 +3109,8 @@ int split_huge_page_to_list_to_order(struct page *page, struct list_head *list,
i_mmap_unlock_read(mapping);
out:
xas_destroy(&xas);
- count_vm_event(!ret ? THP_SPLIT_PAGE : THP_SPLIT_PAGE_FAILED);
+ if (folio_test_pmd_mappable(folio))
+ count_vm_event(!ret ? THP_SPLIT_PAGE : THP_SPLIT_PAGE_FAILED);
return ret;
}

@@ -3171,7 +3172,8 @@ void deferred_split_folio(struct folio *folio)

spin_lock_irqsave(&ds_queue->split_queue_lock, flags);
if (list_empty(&folio->_deferred_list)) {
- count_vm_event(THP_DEFERRED_SPLIT_PAGE);
+ if (folio_test_pmd_mappable(folio))
+ count_vm_event(THP_DEFERRED_SPLIT_PAGE);
list_add_tail(&folio->_deferred_list, &ds_queue->split_queue);
ds_queue->split_queue_len++;
#ifdef CONFIG_MEMCG
--
2.39.3

2024-04-04 18:45:34

by David Hildenbrand

Subject: Re: [PATCH] mm: huge_memory: add the missing folio_test_pmd_mappable() for THP split statistics

On 29.03.24 07:59, Baolin Wang wrote:
> Now the mTHP can also be split or added into the deferred list, so add
> folio_test_pmd_mappable() validation for PMD mapped THP, to avoid confusion
> with PMD mapped THP related statistics.
>
> Signed-off-by: Baolin Wang <[email protected]>
> ---
> mm/huge_memory.c | 6 ++++--
> 1 file changed, 4 insertions(+), 2 deletions(-)
>
> diff --git a/mm/huge_memory.c b/mm/huge_memory.c
> index 1683de78c313..3ca9282a0dc9 100644
> --- a/mm/huge_memory.c
> +++ b/mm/huge_memory.c
> @@ -3109,7 +3109,8 @@ int split_huge_page_to_list_to_order(struct page *page, struct list_head *list,
> i_mmap_unlock_read(mapping);
> out:
> xas_destroy(&xas);
> - count_vm_event(!ret ? THP_SPLIT_PAGE : THP_SPLIT_PAGE_FAILED);
> + if (folio_test_pmd_mappable(folio))
> + count_vm_event(!ret ? THP_SPLIT_PAGE : THP_SPLIT_PAGE_FAILED);
> return ret;
> }
>
> @@ -3171,7 +3172,8 @@ void deferred_split_folio(struct folio *folio)
>
> spin_lock_irqsave(&ds_queue->split_queue_lock, flags);
> if (list_empty(&folio->_deferred_list)) {
> - count_vm_event(THP_DEFERRED_SPLIT_PAGE);
> + if (folio_test_pmd_mappable(folio))
> + count_vm_event(THP_DEFERRED_SPLIT_PAGE);
> list_add_tail(&folio->_deferred_list, &ds_queue->split_queue);
> ds_queue->split_queue_len++;
> #ifdef CONFIG_MEMCG

Acked-by: David Hildenbrand <[email protected]>

--
Cheers,

David / dhildenb

2024-04-22 06:17:43

by Lance Yang

Subject: Re: [PATCH] mm: huge_memory: add the missing folio_test_pmd_mappable() for THP split statistics

Hey Baolin,

Maybe I spotted a bug and made a change to this patch as follows.

diff --git a/mm/huge_memory.c b/mm/huge_memory.c
index 824eff9211db..f3c4e08c6d30 100644
--- a/mm/huge_memory.c
+++ b/mm/huge_memory.c
@@ -2994,6 +2994,7 @@ int split_huge_page_to_list_to_order(struct page *page, struct list_head *list,
XA_STATE_ORDER(xas, &folio->mapping->i_pages, folio->index, new_order);
struct anon_vma *anon_vma = NULL;
struct address_space *mapping = NULL;
+ int order = folio_order(folio);
int extra_pins, ret;
pgoff_t end;
bool is_hzp;
@@ -3172,7 +3173,7 @@ int split_huge_page_to_list_to_order(struct page *page, struct list_head *list,
i_mmap_unlock_read(mapping);
out:
xas_destroy(&xas);
- if (folio_test_pmd_mappable(folio))

IIUC, if we split the folio successfully, it won’t be large here.

+ if (order >= HPAGE_PMD_ORDER)
count_vm_event(!ret ? THP_SPLIT_PAGE : THP_SPLIT_PAGE_FAILED);
return ret;
}

Thanks,
Lance

2024-04-22 06:49:47

by Baolin Wang

Subject: Re: [PATCH] mm: huge_memory: add the missing folio_test_pmd_mappable() for THP split statistics

On 2024/4/22 14:17, Lance Yang wrote:
> Hey Baolin,
>
> Maybe I spotted a bug and made a change to this patch as follows.
>
> diff --git a/mm/huge_memory.c b/mm/huge_memory.c
> index 824eff9211db..f3c4e08c6d30 100644
> --- a/mm/huge_memory.c
> +++ b/mm/huge_memory.c
> @@ -2994,6 +2994,7 @@ int split_huge_page_to_list_to_order(struct page *page, struct list_head *list,
> XA_STATE_ORDER(xas, &folio->mapping->i_pages, folio->index, new_order);
> struct anon_vma *anon_vma = NULL;
> struct address_space *mapping = NULL;
> + int order = folio_order(folio);
> int extra_pins, ret;
> pgoff_t end;
> bool is_hzp;
> @@ -3172,7 +3173,7 @@ int split_huge_page_to_list_to_order(struct page *page, struct list_head *list,
> i_mmap_unlock_read(mapping);
> out:
> xas_destroy(&xas);
> - if (folio_test_pmd_mappable(folio))
>
> IIUC, if we split the folio successfully, it won’t be large here.

Yes, good catch. I did a blindly copy-paste.

Will fix in next version. Thanks for reviewing.

> + if (order >= HPAGE_PMD_ORDER)
> count_vm_event(!ret ? THP_SPLIT_PAGE : THP_SPLIT_PAGE_FAILED);
> return ret;
> }
>
> Thanks,
> Lance