mm/hugetlb: Let unmap_hugepage_range() and

[PATCH v3 3/4] mm/hugetlb: Refactor __unmap_hugepage_range() to take folio instead of page

Posted by nifan.cxl@gmail.com 7 months, 3 weeks ago

From: Fan Ni <fan.ni@samsung.com>

The function __unmap_hugepage_range() has two kinds of users:
1) unmap_hugepage_range(), which passes in the head page of a folio.
   Since unmap_hugepage_range() already takes folio and there are no other
   uses of the folio struct in the function, it is natural for
   __unmap_hugepage_range() to take folio also.
2) All other uses, which pass in NULL pointer.

In both cases, we can pass in folio. Refactor __unmap_hugepage_range() to
take folio.

Signed-off-by: Fan Ni <fan.ni@samsung.com>
---
 include/linux/hugetlb.h |  4 ++--
 mm/hugetlb.c            | 10 +++++-----
 2 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/include/linux/hugetlb.h b/include/linux/hugetlb.h
index 83d85cbb4284..3a07a60c8cd9 100644
--- a/include/linux/hugetlb.h
+++ b/include/linux/hugetlb.h
@@ -133,7 +133,7 @@ void unmap_hugepage_range(struct vm_area_struct *,
 void __unmap_hugepage_range(struct mmu_gather *tlb,
 			  struct vm_area_struct *vma,
 			  unsigned long start, unsigned long end,
-			  struct page *ref_page, zap_flags_t zap_flags);
+			  struct folio *, zap_flags_t zap_flags);
 void hugetlb_report_meminfo(struct seq_file *);
 int hugetlb_report_node_meminfo(char *buf, int len, int nid);
 void hugetlb_show_meminfo_node(int nid);
@@ -452,7 +452,7 @@ static inline long hugetlb_change_protection(
 
 static inline void __unmap_hugepage_range(struct mmu_gather *tlb,
 			struct vm_area_struct *vma, unsigned long start,
-			unsigned long end, struct page *ref_page,
+			unsigned long end, struct folio *folio,
 			zap_flags_t zap_flags)
 {
 	BUG();
diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index 7601e3d344bc..6696206d556e 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -5808,7 +5808,7 @@ int move_hugetlb_page_tables(struct vm_area_struct *vma,
 
 void __unmap_hugepage_range(struct mmu_gather *tlb, struct vm_area_struct *vma,
 			    unsigned long start, unsigned long end,
-			    struct page *ref_page, zap_flags_t zap_flags)
+			    struct folio *folio, zap_flags_t zap_flags)
 {
 	struct mm_struct *mm = vma->vm_mm;
 	unsigned long address;
@@ -5885,8 +5885,8 @@ void __unmap_hugepage_range(struct mmu_gather *tlb, struct vm_area_struct *vma,
 		 * page is being unmapped, not a range. Ensure the page we
 		 * are about to unmap is the actual page of interest.
 		 */
-		if (ref_page) {
-			if (page != ref_page) {
+		if (folio) {
+			if (page_folio(page) != folio) {
 				spin_unlock(ptl);
 				continue;
 			}
@@ -5952,7 +5952,7 @@ void __unmap_hugepage_range(struct mmu_gather *tlb, struct vm_area_struct *vma,
 		/*
 		 * Bail out after unmapping reference page if supplied
 		 */
-		if (ref_page)
+		if (folio)
 			break;
 	}
 	tlb_end_vma(tlb, vma);
@@ -6027,7 +6027,7 @@ void unmap_hugepage_range(struct vm_area_struct *vma, unsigned long start,
 	tlb_gather_mmu(&tlb, vma->vm_mm);
 
 	__unmap_hugepage_range(&tlb, vma, start, end,
-			       &folio->page, zap_flags);
+			       folio, zap_flags);
 
 	mmu_notifier_invalidate_range_end(&range);
 	tlb_finish_mmu(&tlb);
-- 
2.47.2

Re: [PATCH v3 3/4] mm/hugetlb: Refactor __unmap_hugepage_range() to take folio instead of page

Posted by Oscar Salvador 7 months, 3 weeks ago

On Mon, Apr 28, 2025 at 10:11:46AM -0700, nifan.cxl@gmail.com wrote:
> From: Fan Ni <fan.ni@samsung.com>
> 
> The function __unmap_hugepage_range() has two kinds of users:
> 1) unmap_hugepage_range(), which passes in the head page of a folio.
>    Since unmap_hugepage_range() already takes folio and there are no other
>    uses of the folio struct in the function, it is natural for
>    __unmap_hugepage_range() to take folio also.
> 2) All other uses, which pass in NULL pointer.
> 
> In both cases, we can pass in folio. Refactor __unmap_hugepage_range() to
> take folio.
> 
> Signed-off-by: Fan Ni <fan.ni@samsung.com>

Reviewed-by: Oscar Salvador <osalvador@suse.de>

But:

>  void __unmap_hugepage_range(struct mmu_gather *tlb, struct vm_area_struct *vma,
>  			    unsigned long start, unsigned long end,
> -			    struct page *ref_page, zap_flags_t zap_flags)
> +			    struct folio *folio, zap_flags_t zap_flags)

I think we are kinda losing information here. ref_ was a good hint
and...

>  	struct mm_struct *mm = vma->vm_mm;
>  	unsigned long address;
> @@ -5885,8 +5885,8 @@ void __unmap_hugepage_range(struct mmu_gather *tlb, struct vm_area_struct *vma,
>  		 * page is being unmapped, not a range. Ensure the page we
>  		 * are about to unmap is the actual page of interest.
>  		 */
> -		if (ref_page) {
> -			if (page != ref_page) {
> +		if (folio) {
> +			if (page_folio(page) != folio) {

You have to update the comment above, since we are not passing a
reference page anymore but a folio.

 

-- 
Oscar Salvador
SUSE Labs

Re: [PATCH v3 3/4] mm/hugetlb: Refactor __unmap_hugepage_range() to take folio instead of page

Posted by Fan Ni 7 months, 2 weeks ago

On Wed, Apr 30, 2025 at 09:58:35AM +0200, Oscar Salvador wrote:
> On Mon, Apr 28, 2025 at 10:11:46AM -0700, nifan.cxl@gmail.com wrote:
> > From: Fan Ni <fan.ni@samsung.com>
> > 
> > The function __unmap_hugepage_range() has two kinds of users:
> > 1) unmap_hugepage_range(), which passes in the head page of a folio.
> >    Since unmap_hugepage_range() already takes folio and there are no other
> >    uses of the folio struct in the function, it is natural for
> >    __unmap_hugepage_range() to take folio also.
> > 2) All other uses, which pass in NULL pointer.
> > 
> > In both cases, we can pass in folio. Refactor __unmap_hugepage_range() to
> > take folio.
> > 
> > Signed-off-by: Fan Ni <fan.ni@samsung.com>
> 
> Reviewed-by: Oscar Salvador <osalvador@suse.de>
> 
> But:
> 
> >  void __unmap_hugepage_range(struct mmu_gather *tlb, struct vm_area_struct *vma,
> >  			    unsigned long start, unsigned long end,
> > -			    struct page *ref_page, zap_flags_t zap_flags)
> > +			    struct folio *folio, zap_flags_t zap_flags)
> 
> I think we are kinda losing information here. ref_ was a good hint
> and...

Hi Oscar,

Thanks for the feedback.
Since the sugguested change here is minor and does not affect the
function, and we do not have a aligned opinion here.
https://lore.kernel.org/linux-mm/b23ef51b-1284-4168-8157-432c3e045788@redhat.com/
I will leave it as it is until there are more pushes for the change.


> 
> >  	struct mm_struct *mm = vma->vm_mm;
> >  	unsigned long address;
> > @@ -5885,8 +5885,8 @@ void __unmap_hugepage_range(struct mmu_gather *tlb, struct vm_area_struct *vma,
> >  		 * page is being unmapped, not a range. Ensure the page we
> >  		 * are about to unmap is the actual page of interest.
> >  		 */
> > -		if (ref_page) {
> > -			if (page != ref_page) {
> > +		if (folio) {
> > +			if (page_folio(page) != folio) {
> 
> You have to update the comment above, since we are not passing a
> reference page anymore but a folio.

Will update in the next version. Thanks.

Fan

> 
>  
> 
> -- 
> Oscar Salvador
> SUSE Labs
> 

-- 
Fan Ni

Re: [PATCH v3 3/4] mm/hugetlb: Refactor __unmap_hugepage_range() to take folio instead of page

Posted by David Hildenbrand 7 months, 2 weeks ago

On 04.05.25 23:35, Fan Ni wrote:
> On Wed, Apr 30, 2025 at 09:58:35AM +0200, Oscar Salvador wrote:
>> On Mon, Apr 28, 2025 at 10:11:46AM -0700, nifan.cxl@gmail.com wrote:
>>> From: Fan Ni <fan.ni@samsung.com>
>>>
>>> The function __unmap_hugepage_range() has two kinds of users:
>>> 1) unmap_hugepage_range(), which passes in the head page of a folio.
>>>     Since unmap_hugepage_range() already takes folio and there are no other
>>>     uses of the folio struct in the function, it is natural for
>>>     __unmap_hugepage_range() to take folio also.
>>> 2) All other uses, which pass in NULL pointer.
>>>
>>> In both cases, we can pass in folio. Refactor __unmap_hugepage_range() to
>>> take folio.
>>>
>>> Signed-off-by: Fan Ni <fan.ni@samsung.com>
>>
>> Reviewed-by: Oscar Salvador <osalvador@suse.de>
>>
>> But:
>>
>>>   void __unmap_hugepage_range(struct mmu_gather *tlb, struct vm_area_struct *vma,
>>>   			    unsigned long start, unsigned long end,
>>> -			    struct page *ref_page, zap_flags_t zap_flags)
>>> +			    struct folio *folio, zap_flags_t zap_flags)
>>
>> I think we are kinda losing information here. ref_ was a good hint
>> and...
> 
> Hi Oscar,
> 
> Thanks for the feedback.
> Since the sugguested change here is minor and does not affect the
> function, and we do not have a aligned opinion here.
> https://lore.kernel.org/linux-mm/b23ef51b-1284-4168-8157-432c3e045788@redhat.com/
> I will leave it as it is until there are more pushes for the change.

I don't think we're losing any information. :) Especially if nowhere 
else in the kernel we use the term "ref_page" or "ref_folio". And things 
like put_ref_page(), free_unref_folios() ... talk about dropping 
references not "this is the reference folio".

"folio_to_unmap" might be clearer, but then, I think we should just use 
a single folio variable in that function.

This function might benefit from kerneldoc, though :)

-- 
Cheers,

David / dhildenb

Re: [PATCH v3 3/4] mm/hugetlb: Refactor __unmap_hugepage_range() to take folio instead of page

Posted by David Hildenbrand 7 months, 3 weeks ago

On 28.04.25 19:11, nifan.cxl@gmail.com wrote:
> From: Fan Ni <fan.ni@samsung.com>
> 
> The function __unmap_hugepage_range() has two kinds of users:
> 1) unmap_hugepage_range(), which passes in the head page of a folio.
>     Since unmap_hugepage_range() already takes folio and there are no other
>     uses of the folio struct in the function, it is natural for
>     __unmap_hugepage_range() to take folio also.
> 2) All other uses, which pass in NULL pointer.
> 
> In both cases, we can pass in folio. Refactor __unmap_hugepage_range() to
> take folio.
> 
> Signed-off-by: Fan Ni <fan.ni@samsung.com>
> ---

Acked-by: David Hildenbrand <david@redhat.com>

-- 
Cheers,

David / dhildenb