[PATCH 0/2] mm: remove page_rmapping()

[PATCH 0/2] mm: remove page_rmapping()

[Peng Zhang]

From: ZhangPeng <zhangpeng362@huawei.com>

This minor patch series remove page_rmapping() and convert the last user
fault_dirty_shared_page() to use a folio.

ZhangPeng (2):
  mm: remove page_rmapping()
  mm: use a folio in fault_dirty_shared_page()

 include/linux/mm.h |  1 -
 mm/memory.c        | 10 +++++-----
 mm/util.c          |  6 ------
 3 files changed, 5 insertions(+), 12 deletions(-)

-- 
2.25.1

[PATCH 1/2] mm: remove page_rmapping()

[Peng Zhang]

From: ZhangPeng <zhangpeng362@huawei.com>

After converting the last user to folio_raw_mapping(), we can safely
remove the function.

Signed-off-by: ZhangPeng <zhangpeng362@huawei.com>
---
 include/linux/mm.h | 1 -
 mm/memory.c        | 2 +-
 mm/util.c          | 6 ------
 3 files changed, 1 insertion(+), 8 deletions(-)

diff --git a/include/linux/mm.h b/include/linux/mm.h
index 9ecb8b9c07f6..8c7eba7370b2 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -2129,7 +2129,6 @@ static inline void *folio_address(const struct folio *folio)
 	return page_address(&folio->page);
 }
 
-extern void *page_rmapping(struct page *page);
 extern pgoff_t __page_file_index(struct page *page);
 
 /*
diff --git a/mm/memory.c b/mm/memory.c
index 21fab2727209..6921df44a99f 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -2979,7 +2979,7 @@ static vm_fault_t fault_dirty_shared_page(struct vm_fault *vmf)
 	 * pinned by vma->vm_file's reference.  We rely on unlock_page()'s
 	 * release semantics to prevent the compiler from undoing this copying.
 	 */
-	mapping = page_rmapping(page);
+	mapping = folio_raw_mapping(page_folio(page));
 	unlock_page(page);
 
 	if (!page_mkwrite)
diff --git a/mm/util.c b/mm/util.c
index dd12b9531ac4..5e9305189c3f 100644
--- a/mm/util.c
+++ b/mm/util.c
@@ -734,12 +734,6 @@ void *vcalloc(size_t n, size_t size)
 }
 EXPORT_SYMBOL(vcalloc);
 
-/* Neutral page->mapping pointer to address_space or anon_vma or other */
-void *page_rmapping(struct page *page)
-{
-	return folio_raw_mapping(page_folio(page));
-}
-
 struct anon_vma *folio_anon_vma(struct folio *folio)
 {
 	unsigned long mapping = (unsigned long)folio->mapping;
-- 
2.25.1

Re: [PATCH 1/2] mm: remove page_rmapping()

[Sidhartha Kumar]

On 6/30/23 5:13 AM, Peng Zhang wrote:
> From: ZhangPeng <zhangpeng362@huawei.com>
> 
> After converting the last user to folio_raw_mapping(), we can safely
> remove the function.
> 
> Signed-off-by: ZhangPeng <zhangpeng362@huawei.com>
> ---
>   include/linux/mm.h | 1 -
>   mm/memory.c        | 2 +-
>   mm/util.c          | 6 ------
>   3 files changed, 1 insertion(+), 8 deletions(-)
> 
> diff --git a/include/linux/mm.h b/include/linux/mm.h
> index 9ecb8b9c07f6..8c7eba7370b2 100644
> --- a/include/linux/mm.h
> +++ b/include/linux/mm.h
> @@ -2129,7 +2129,6 @@ static inline void *folio_address(const struct folio *folio)
>   	return page_address(&folio->page);
>   }
>   
> -extern void *page_rmapping(struct page *page);
>   extern pgoff_t __page_file_index(struct page *page);
>   
>   /*
> diff --git a/mm/memory.c b/mm/memory.c
> index 21fab2727209..6921df44a99f 100644
> --- a/mm/memory.c
> +++ b/mm/memory.c
> @@ -2979,7 +2979,7 @@ static vm_fault_t fault_dirty_shared_page(struct vm_fault *vmf)
>   	 * pinned by vma->vm_file's reference.  We rely on unlock_page()'s
>   	 * release semantics to prevent the compiler from undoing this copying.
>   	 */
> -	mapping = page_rmapping(page);
> +	mapping = folio_raw_mapping(page_folio(page));
>   	unlock_page(page);
>   
>   	if (!page_mkwrite)
> diff --git a/mm/util.c b/mm/util.c
> index dd12b9531ac4..5e9305189c3f 100644
> --- a/mm/util.c
> +++ b/mm/util.c
> @@ -734,12 +734,6 @@ void *vcalloc(size_t n, size_t size)
>   }
>   EXPORT_SYMBOL(vcalloc);
>   
> -/* Neutral page->mapping pointer to address_space or anon_vma or other */
> -void *page_rmapping(struct page *page)
> -{
> -	return folio_raw_mapping(page_folio(page));
> -}
> -
>   struct anon_vma *folio_anon_vma(struct folio *folio)
>   {
>   	unsigned long mapping = (unsigned long)folio->mapping;
Reviewed-by: Sidhartha Kumar <sidhartha.kumar@oracle.com>

Re: [PATCH 1/2] mm: remove page_rmapping()

[Sidhartha Kumar]

On 6/30/23 5:13 AM, Peng Zhang wrote:
> From: ZhangPeng <zhangpeng362@huawei.com>
> 
> After converting the last user to folio_raw_mapping(), we can safely
> remove the function.
> 
> Signed-off-by: ZhangPeng <zhangpeng362@huawei.com>
> ---
>   include/linux/mm.h | 1 -
>   mm/memory.c        | 2 +-
>   mm/util.c          | 6 ------
>   3 files changed, 1 insertion(+), 8 deletions(-)
> 
> diff --git a/include/linux/mm.h b/include/linux/mm.h
> index 9ecb8b9c07f6..8c7eba7370b2 100644
> --- a/include/linux/mm.h
> +++ b/include/linux/mm.h
> @@ -2129,7 +2129,6 @@ static inline void *folio_address(const struct folio *folio)
>   	return page_address(&folio->page);
>   }
>   
> -extern void *page_rmapping(struct page *page);
>   extern pgoff_t __page_file_index(struct page *page);
>   
>   /*
> diff --git a/mm/memory.c b/mm/memory.c
> index 21fab2727209..6921df44a99f 100644
> --- a/mm/memory.c
> +++ b/mm/memory.c
> @@ -2979,7 +2979,7 @@ static vm_fault_t fault_dirty_shared_page(struct vm_fault *vmf)
>   	 * pinned by vma->vm_file's reference.  We rely on unlock_page()'s
>   	 * release semantics to prevent the compiler from undoing this copying.
>   	 */


	struct page *page = vmf->page;
	
	...

	dirtied = set_page_dirty(page);
	VM_BUG_ON_PAGE(PageAnon(page), page);

Can this earlier part of the function be converted to folios so a folio 
can be passed directly in folio_raw_mapping()?
> -	mapping = page_rmapping(page);
> +	mapping = folio_raw_mapping(page_folio(page));
>   	unlock_page(page);
>   
>   	if (!page_mkwrite)
> diff --git a/mm/util.c b/mm/util.c
> index dd12b9531ac4..5e9305189c3f 100644
> --- a/mm/util.c
> +++ b/mm/util.c
> @@ -734,12 +734,6 @@ void *vcalloc(size_t n, size_t size)
>   }
>   EXPORT_SYMBOL(vcalloc);
>   
> -/* Neutral page->mapping pointer to address_space or anon_vma or other */
> -void *page_rmapping(struct page *page)
> -{
> -	return folio_raw_mapping(page_folio(page));
> -}
> -
>   struct anon_vma *folio_anon_vma(struct folio *folio)
>   {
>   	unsigned long mapping = (unsigned long)folio->mapping;

Re: [PATCH 1/2] mm: remove page_rmapping()

[zhangpeng (AS)]

On 2023/7/1 0:03, Sidhartha Kumar wrote:

> On 6/30/23 5:13 AM, Peng Zhang wrote:
>> From: ZhangPeng <zhangpeng362@huawei.com>
>>
>> After converting the last user to folio_raw_mapping(), we can safely
>> remove the function.
>>
>> Signed-off-by: ZhangPeng <zhangpeng362@huawei.com>
>> ---
>>   include/linux/mm.h | 1 -
>>   mm/memory.c        | 2 +-
>>   mm/util.c          | 6 ------
>>   3 files changed, 1 insertion(+), 8 deletions(-)
>>
>> diff --git a/include/linux/mm.h b/include/linux/mm.h
>> index 9ecb8b9c07f6..8c7eba7370b2 100644
>> --- a/include/linux/mm.h
>> +++ b/include/linux/mm.h
>> @@ -2129,7 +2129,6 @@ static inline void *folio_address(const struct 
>> folio *folio)
>>       return page_address(&folio->page);
>>   }
>>   -extern void *page_rmapping(struct page *page);
>>   extern pgoff_t __page_file_index(struct page *page);
>>     /*
>> diff --git a/mm/memory.c b/mm/memory.c
>> index 21fab2727209..6921df44a99f 100644
>> --- a/mm/memory.c
>> +++ b/mm/memory.c
>> @@ -2979,7 +2979,7 @@ static vm_fault_t 
>> fault_dirty_shared_page(struct vm_fault *vmf)
>>        * pinned by vma->vm_file's reference.  We rely on unlock_page()'s
>>        * release semantics to prevent the compiler from undoing this 
>> copying.
>>        */
>
>
>     struct page *page = vmf->page;
>
>     ...
>
>     dirtied = set_page_dirty(page);
>     VM_BUG_ON_PAGE(PageAnon(page), page);
>
> Can this earlier part of the function be converted to folios so a 
> folio can be passed directly in folio_raw_mapping()?

Thanks for your feedback.

I convert the last user to use a folio in patch 2. I'll reverse the order of the two patches.

Best Regards,
Peng

>> -    mapping = page_rmapping(page);
>> +    mapping = folio_raw_mapping(page_folio(page));
>>       unlock_page(page);
>>         if (!page_mkwrite)
>> diff --git a/mm/util.c b/mm/util.c
>> index dd12b9531ac4..5e9305189c3f 100644
>> --- a/mm/util.c
>> +++ b/mm/util.c
>> @@ -734,12 +734,6 @@ void *vcalloc(size_t n, size_t size)
>>   }
>>   EXPORT_SYMBOL(vcalloc);
>>   -/* Neutral page->mapping pointer to address_space or anon_vma or 
>> other */
>> -void *page_rmapping(struct page *page)
>> -{
>> -    return folio_raw_mapping(page_folio(page));
>> -}
>> -
>>   struct anon_vma *folio_anon_vma(struct folio *folio)
>>   {
>>       unsigned long mapping = (unsigned long)folio->mapping;
>

[PATCH 2/2] mm: use a folio in fault_dirty_shared_page()

[Peng Zhang]

From: ZhangPeng <zhangpeng362@huawei.com>

We can replace four implicit calls to compound_head() with one by using
folio.

Signed-off-by: ZhangPeng <zhangpeng362@huawei.com>
---
 mm/memory.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/mm/memory.c b/mm/memory.c
index 6921df44a99f..73b03706451c 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -2967,20 +2967,20 @@ static vm_fault_t fault_dirty_shared_page(struct vm_fault *vmf)
 {
 	struct vm_area_struct *vma = vmf->vma;
 	struct address_space *mapping;
-	struct page *page = vmf->page;
+	struct folio *folio = page_folio(vmf->page);
 	bool dirtied;
 	bool page_mkwrite = vma->vm_ops && vma->vm_ops->page_mkwrite;
 
-	dirtied = set_page_dirty(page);
-	VM_BUG_ON_PAGE(PageAnon(page), page);
+	dirtied = folio_mark_dirty(folio);
+	VM_BUG_ON_FOLIO(folio_test_anon(folio), folio);
 	/*
 	 * Take a local copy of the address_space - page.mapping may be zeroed
 	 * by truncate after unlock_page().   The address_space itself remains
 	 * pinned by vma->vm_file's reference.  We rely on unlock_page()'s
 	 * release semantics to prevent the compiler from undoing this copying.
 	 */
-	mapping = folio_raw_mapping(page_folio(page));
-	unlock_page(page);
+	mapping = folio_raw_mapping(folio);
+	folio_unlock(folio);
 
 	if (!page_mkwrite)
 		file_update_time(vma->vm_file);
-- 
2.25.1

Re: [PATCH 2/2] mm: use a folio in fault_dirty_shared_page()

[Sidhartha Kumar]

On 6/30/23 5:13 AM, Peng Zhang wrote:
> From: ZhangPeng <zhangpeng362@huawei.com>
> 
> We can replace four implicit calls to compound_head() with one by using
> folio.
> 
> Signed-off-by: ZhangPeng <zhangpeng362@huawei.com>
> ---
>   mm/memory.c | 10 +++++-----
>   1 file changed, 5 insertions(+), 5 deletions(-)
> 
> diff --git a/mm/memory.c b/mm/memory.c
> index 6921df44a99f..73b03706451c 100644
> --- a/mm/memory.c
> +++ b/mm/memory.c
> @@ -2967,20 +2967,20 @@ static vm_fault_t fault_dirty_shared_page(struct vm_fault *vmf)
>   {
>   	struct vm_area_struct *vma = vmf->vma;
>   	struct address_space *mapping;
> -	struct page *page = vmf->page;
> +	struct folio *folio = page_folio(vmf->page);
>   	bool dirtied;
>   	bool page_mkwrite = vma->vm_ops && vma->vm_ops->page_mkwrite;
>   
> -	dirtied = set_page_dirty(page);
> -	VM_BUG_ON_PAGE(PageAnon(page), page);
> +	dirtied = folio_mark_dirty(folio);
> +	VM_BUG_ON_FOLIO(folio_test_anon(folio), folio);
>   	/*
>   	 * Take a local copy of the address_space - page.mapping may be zeroed
>   	 * by truncate after unlock_page().   The address_space itself remains
>   	 * pinned by vma->vm_file's reference.  We rely on unlock_page()'s
>   	 * release semantics to prevent the compiler from undoing this copying.
>   	 */
> -	mapping = folio_raw_mapping(page_folio(page));
> -	unlock_page(page);
> +	mapping = folio_raw_mapping(folio);
> +	folio_unlock(folio);
>   
>   	if (!page_mkwrite)
>   		file_update_time(vma->vm_file);
Reviewed-by: Sidhartha Kumar <sidhartha.kumar@oracle.com>

Re: [PATCH 0/2] mm: remove page_rmapping()

[Matthew Wilcox]

On Fri, Jun 30, 2023 at 08:13:08PM +0800, Peng Zhang wrote:
> From: ZhangPeng <zhangpeng362@huawei.com>
> 
> This minor patch series remove page_rmapping() and convert the last user
> fault_dirty_shared_page() to use a folio.

This patch series works better if you do it in the other order.
First convert the user, then remove the now-unused page_rmapping().

It's almost always a mistake to start with "I want to remove all the
users of API x".  You should be thinking "Here is function y that still
uses pages; I shall convert it to use folios".

Re: [PATCH 0/2] mm: remove page_rmapping()

[zhangpeng (AS)]

On 2023/7/1 5:39, Matthew Wilcox wrote:

> On Fri, Jun 30, 2023 at 08:13:08PM +0800, Peng Zhang wrote:
>> From: ZhangPeng <zhangpeng362@huawei.com>
>>
>> This minor patch series remove page_rmapping() and convert the last user
>> fault_dirty_shared_page() to use a folio.
> This patch series works better if you do it in the other order.
> First convert the user, then remove the now-unused page_rmapping().
>
> It's almost always a mistake to start with "I want to remove all the
> users of API x".  You should be thinking "Here is function y that still
> uses pages; I shall convert it to use folios".

Thanks for your feedback.

I'll reverse the order of the two patches. Convert user first, then remove the now-unused page_rmapping().

Best Regards,
Peng