[PATCH mm-hotfixes-unstable] mm: Pass page directly instead of using folio_page

[PATCH mm-hotfixes-unstable] mm: Pass page directly instead of using folio_page

[Dev Jain]

In commit_anon_folio_batch(), we iterate over all pages pointed to by the
PTE batch. Therefore we need to know the first page of the batch;
currently we derive that via folio_page(folio, 0), but, that takes us
to the first (head) page of the folio instead - our PTE batch may lie
in the middle of the folio, leading to incorrectness.

Bite the bullet and throw away the micro-optimization of reusing the
folio in favour of code simplicity. Derive the page and the folio in
change_pte_range, and pass the page too to commit_anon_folio_batch to
fix the aforementioned issue.

Reported-by: syzbot+57bcc752f0df8bb1365c@syzkaller.appspotmail.com
Fixes: cac1db8c3aad ("mm: optimize mprotect() by PTE batching")
Signed-off-by: Dev Jain <dev.jain@arm.com>
---
 mm/mprotect.c | 23 ++++++++++-------------
 1 file changed, 10 insertions(+), 13 deletions(-)

diff --git a/mm/mprotect.c b/mm/mprotect.c
index 78bded7acf79..113b48985834 100644
--- a/mm/mprotect.c
+++ b/mm/mprotect.c
@@ -120,9 +120,8 @@ static int mprotect_folio_pte_batch(struct folio *folio, pte_t *ptep,
 
 static bool prot_numa_skip(struct vm_area_struct *vma, unsigned long addr,
 			   pte_t oldpte, pte_t *pte, int target_node,
-			   struct folio **foliop)
+			   struct folio *folio)
 {
-	struct folio *folio = NULL;
 	bool ret = true;
 	bool toptier;
 	int nid;
@@ -131,7 +130,6 @@ static bool prot_numa_skip(struct vm_area_struct *vma, unsigned long addr,
 	if (pte_protnone(oldpte))
 		goto skip;
 
-	folio = vm_normal_folio(vma, addr, oldpte);
 	if (!folio)
 		goto skip;
 
@@ -173,7 +171,6 @@ static bool prot_numa_skip(struct vm_area_struct *vma, unsigned long addr,
 		folio_xchg_access_time(folio, jiffies_to_msecs(jiffies));
 
 skip:
-	*foliop = folio;
 	return ret;
 }
 
@@ -231,10 +228,9 @@ static int page_anon_exclusive_sub_batch(int start_idx, int max_len,
  * retrieve sub-batches.
  */
 static void commit_anon_folio_batch(struct vm_area_struct *vma,
-		struct folio *folio, unsigned long addr, pte_t *ptep,
+		struct folio *folio, struct page *first_page, unsigned long addr, pte_t *ptep,
 		pte_t oldpte, pte_t ptent, int nr_ptes, struct mmu_gather *tlb)
 {
-	struct page *first_page = folio_page(folio, 0);
 	bool expected_anon_exclusive;
 	int sub_batch_idx = 0;
 	int len;
@@ -251,7 +247,7 @@ static void commit_anon_folio_batch(struct vm_area_struct *vma,
 }
 
 static void set_write_prot_commit_flush_ptes(struct vm_area_struct *vma,
-		struct folio *folio, unsigned long addr, pte_t *ptep,
+		struct folio *folio, struct page *page, unsigned long addr, pte_t *ptep,
 		pte_t oldpte, pte_t ptent, int nr_ptes, struct mmu_gather *tlb)
 {
 	bool set_write;
@@ -270,7 +266,7 @@ static void set_write_prot_commit_flush_ptes(struct vm_area_struct *vma,
 				       /* idx = */ 0, set_write, tlb);
 		return;
 	}
-	commit_anon_folio_batch(vma, folio, addr, ptep, oldpte, ptent, nr_ptes, tlb);
+	commit_anon_folio_batch(vma, folio, page, addr, ptep, oldpte, ptent, nr_ptes, tlb);
 }
 
 static long change_pte_range(struct mmu_gather *tlb,
@@ -305,15 +301,19 @@ static long change_pte_range(struct mmu_gather *tlb,
 			const fpb_t flags = FPB_RESPECT_SOFT_DIRTY | FPB_RESPECT_WRITE;
 			int max_nr_ptes = (end - addr) >> PAGE_SHIFT;
 			struct folio *folio = NULL;
+			struct page *page;
 			pte_t ptent;
 
+			page = vm_normal_page(vma, addr, oldpte);
+			if (page)
+				folio = page_folio(page);
 			/*
 			 * Avoid trapping faults against the zero or KSM
 			 * pages. See similar comment in change_huge_pmd.
 			 */
 			if (prot_numa) {
 				int ret = prot_numa_skip(vma, addr, oldpte, pte,
-							 target_node, &folio);
+							 target_node, folio);
 				if (ret) {
 
 					/* determine batch to skip */
@@ -323,9 +323,6 @@ static long change_pte_range(struct mmu_gather *tlb,
 				}
 			}
 
-			if (!folio)
-				folio = vm_normal_folio(vma, addr, oldpte);
-
 			nr_ptes = mprotect_folio_pte_batch(folio, pte, oldpte, max_nr_ptes, flags);
 
 			oldpte = modify_prot_start_ptes(vma, addr, pte, nr_ptes);
@@ -351,7 +348,7 @@ static long change_pte_range(struct mmu_gather *tlb,
 			 */
 			if ((cp_flags & MM_CP_TRY_CHANGE_WRITABLE) &&
 			     !pte_write(ptent))
-				set_write_prot_commit_flush_ptes(vma, folio,
+				set_write_prot_commit_flush_ptes(vma, folio, page,
 				addr, pte, oldpte, ptent, nr_ptes, tlb);
 			else
 				prot_commit_flush_ptes(vma, addr, pte, oldpte, ptent,
-- 
2.30.2

Re: [PATCH mm-hotfixes-unstable] mm: Pass page directly instead of using folio_page

[Lorenzo Stoakes]

On Wed, Aug 06, 2025 at 08:26:11PM +0530, Dev Jain wrote:
> In commit_anon_folio_batch(), we iterate over all pages pointed to by the
> PTE batch. Therefore we need to know the first page of the batch;
> currently we derive that via folio_page(folio, 0), but, that takes us
> to the first (head) page of the folio instead - our PTE batch may lie
> in the middle of the folio, leading to incorrectness.
>
> Bite the bullet and throw away the micro-optimization of reusing the
> folio in favour of code simplicity. Derive the page and the folio in
> change_pte_range, and pass the page too to commit_anon_folio_batch to
> fix the aforementioned issue.
>
> Reported-by: syzbot+57bcc752f0df8bb1365c@syzkaller.appspotmail.com
> Fixes: cac1db8c3aad ("mm: optimize mprotect() by PTE batching")
> Signed-off-by: Dev Jain <dev.jain@arm.com>

This looks reasonable, fixes the problem and compiles/works on my machine so:

Reviewed-by: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>

This badly needs refactoring as 13 parameters being passed to a function is
ridiculous, but we can do that later.

Let's get this in as a hotfix asap.

> ---
>  mm/mprotect.c | 23 ++++++++++-------------
>  1 file changed, 10 insertions(+), 13 deletions(-)
>
> diff --git a/mm/mprotect.c b/mm/mprotect.c
> index 78bded7acf79..113b48985834 100644
> --- a/mm/mprotect.c
> +++ b/mm/mprotect.c
> @@ -120,9 +120,8 @@ static int mprotect_folio_pte_batch(struct folio *folio, pte_t *ptep,
>
>  static bool prot_numa_skip(struct vm_area_struct *vma, unsigned long addr,
>  			   pte_t oldpte, pte_t *pte, int target_node,
> -			   struct folio **foliop)
> +			   struct folio *folio)
>  {
> -	struct folio *folio = NULL;
>  	bool ret = true;
>  	bool toptier;
>  	int nid;
> @@ -131,7 +130,6 @@ static bool prot_numa_skip(struct vm_area_struct *vma, unsigned long addr,
>  	if (pte_protnone(oldpte))
>  		goto skip;
>
> -	folio = vm_normal_folio(vma, addr, oldpte);
>  	if (!folio)
>  		goto skip;
>
> @@ -173,7 +171,6 @@ static bool prot_numa_skip(struct vm_area_struct *vma, unsigned long addr,
>  		folio_xchg_access_time(folio, jiffies_to_msecs(jiffies));
>
>  skip:
> -	*foliop = folio;
>  	return ret;
>  }
>
> @@ -231,10 +228,9 @@ static int page_anon_exclusive_sub_batch(int start_idx, int max_len,
>   * retrieve sub-batches.
>   */
>  static void commit_anon_folio_batch(struct vm_area_struct *vma,
> -		struct folio *folio, unsigned long addr, pte_t *ptep,
> +		struct folio *folio, struct page *first_page, unsigned long addr, pte_t *ptep,
>  		pte_t oldpte, pte_t ptent, int nr_ptes, struct mmu_gather *tlb)
>  {
> -	struct page *first_page = folio_page(folio, 0);
>  	bool expected_anon_exclusive;
>  	int sub_batch_idx = 0;
>  	int len;
> @@ -251,7 +247,7 @@ static void commit_anon_folio_batch(struct vm_area_struct *vma,
>  }
>
>  static void set_write_prot_commit_flush_ptes(struct vm_area_struct *vma,
> -		struct folio *folio, unsigned long addr, pte_t *ptep,
> +		struct folio *folio, struct page *page, unsigned long addr, pte_t *ptep,
>  		pte_t oldpte, pte_t ptent, int nr_ptes, struct mmu_gather *tlb)
>  {
>  	bool set_write;
> @@ -270,7 +266,7 @@ static void set_write_prot_commit_flush_ptes(struct vm_area_struct *vma,
>  				       /* idx = */ 0, set_write, tlb);
>  		return;
>  	}
> -	commit_anon_folio_batch(vma, folio, addr, ptep, oldpte, ptent, nr_ptes, tlb);
> +	commit_anon_folio_batch(vma, folio, page, addr, ptep, oldpte, ptent, nr_ptes, tlb);
>  }
>
>  static long change_pte_range(struct mmu_gather *tlb,
> @@ -305,15 +301,19 @@ static long change_pte_range(struct mmu_gather *tlb,
>  			const fpb_t flags = FPB_RESPECT_SOFT_DIRTY | FPB_RESPECT_WRITE;
>  			int max_nr_ptes = (end - addr) >> PAGE_SHIFT;
>  			struct folio *folio = NULL;
> +			struct page *page;
>  			pte_t ptent;
>
> +			page = vm_normal_page(vma, addr, oldpte);
> +			if (page)
> +				folio = page_folio(page);
>  			/*
>  			 * Avoid trapping faults against the zero or KSM
>  			 * pages. See similar comment in change_huge_pmd.
>  			 */
>  			if (prot_numa) {
>  				int ret = prot_numa_skip(vma, addr, oldpte, pte,
> -							 target_node, &folio);
> +							 target_node, folio);
>  				if (ret) {
>
>  					/* determine batch to skip */
> @@ -323,9 +323,6 @@ static long change_pte_range(struct mmu_gather *tlb,
>  				}
>  			}
>
> -			if (!folio)
> -				folio = vm_normal_folio(vma, addr, oldpte);
> -
>  			nr_ptes = mprotect_folio_pte_batch(folio, pte, oldpte, max_nr_ptes, flags);
>
>  			oldpte = modify_prot_start_ptes(vma, addr, pte, nr_ptes);
> @@ -351,7 +348,7 @@ static long change_pte_range(struct mmu_gather *tlb,
>  			 */
>  			if ((cp_flags & MM_CP_TRY_CHANGE_WRITABLE) &&
>  			     !pte_write(ptent))
> -				set_write_prot_commit_flush_ptes(vma, folio,
> +				set_write_prot_commit_flush_ptes(vma, folio, page,
>  				addr, pte, oldpte, ptent, nr_ptes, tlb);
>  			else
>  				prot_commit_flush_ptes(vma, addr, pte, oldpte, ptent,
> --
> 2.30.2
>

Re: [PATCH mm-hotfixes-unstable] mm: Pass page directly instead of using folio_page

[Dev Jain]

On 06/08/25 10:10 pm, Lorenzo Stoakes wrote:
> On Wed, Aug 06, 2025 at 08:26:11PM +0530, Dev Jain wrote:
>> In commit_anon_folio_batch(), we iterate over all pages pointed to by the
>> PTE batch. Therefore we need to know the first page of the batch;
>> currently we derive that via folio_page(folio, 0), but, that takes us
>> to the first (head) page of the folio instead - our PTE batch may lie
>> in the middle of the folio, leading to incorrectness.
>>
>> Bite the bullet and throw away the micro-optimization of reusing the
>> folio in favour of code simplicity. Derive the page and the folio in
>> change_pte_range, and pass the page too to commit_anon_folio_batch to
>> fix the aforementioned issue.
>>
>> Reported-by: syzbot+57bcc752f0df8bb1365c@syzkaller.appspotmail.com
>> Fixes: cac1db8c3aad ("mm: optimize mprotect() by PTE batching")
>> Signed-off-by: Dev Jain <dev.jain@arm.com>
> This looks reasonable, fixes the problem and compiles/works on my machine so:
>
> Reviewed-by: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>

Thanks.

>
> This badly needs refactoring as 13 parameters being passed to a function is

Which function are you talking about?

> ridiculous, but we can do that later.
>
> Let's get this in as a hotfix asap.
>
>> ---
>>   mm/mprotect.c | 23 ++++++++++-------------
>>   1 file changed, 10 insertions(+), 13 deletions(-)
>>
>> diff --git a/mm/mprotect.c b/mm/mprotect.c
>> index 78bded7acf79..113b48985834 100644
>> --- a/mm/mprotect.c
>> +++ b/mm/mprotect.c
>> @@ -120,9 +120,8 @@ static int mprotect_folio_pte_batch(struct folio *folio, pte_t *ptep,
>>
>>   static bool prot_numa_skip(struct vm_area_struct *vma, unsigned long addr,
>>   			   pte_t oldpte, pte_t *pte, int target_node,
>> -			   struct folio **foliop)
>> +			   struct folio *folio)
>>   {
>> -	struct folio *folio = NULL;
>>   	bool ret = true;
>>   	bool toptier;
>>   	int nid;
>> @@ -131,7 +130,6 @@ static bool prot_numa_skip(struct vm_area_struct *vma, unsigned long addr,
>>   	if (pte_protnone(oldpte))
>>   		goto skip;
>>
>> -	folio = vm_normal_folio(vma, addr, oldpte);
>>   	if (!folio)
>>   		goto skip;
>>
>> @@ -173,7 +171,6 @@ static bool prot_numa_skip(struct vm_area_struct *vma, unsigned long addr,
>>   		folio_xchg_access_time(folio, jiffies_to_msecs(jiffies));
>>
>>   skip:
>> -	*foliop = folio;
>>   	return ret;
>>   }
>>
>> @@ -231,10 +228,9 @@ static int page_anon_exclusive_sub_batch(int start_idx, int max_len,
>>    * retrieve sub-batches.
>>    */
>>   static void commit_anon_folio_batch(struct vm_area_struct *vma,
>> -		struct folio *folio, unsigned long addr, pte_t *ptep,
>> +		struct folio *folio, struct page *first_page, unsigned long addr, pte_t *ptep,
>>   		pte_t oldpte, pte_t ptent, int nr_ptes, struct mmu_gather *tlb)
>>   {
>> -	struct page *first_page = folio_page(folio, 0);
>>   	bool expected_anon_exclusive;
>>   	int sub_batch_idx = 0;
>>   	int len;
>> @@ -251,7 +247,7 @@ static void commit_anon_folio_batch(struct vm_area_struct *vma,
>>   }
>>
>>   static void set_write_prot_commit_flush_ptes(struct vm_area_struct *vma,
>> -		struct folio *folio, unsigned long addr, pte_t *ptep,
>> +		struct folio *folio, struct page *page, unsigned long addr, pte_t *ptep,
>>   		pte_t oldpte, pte_t ptent, int nr_ptes, struct mmu_gather *tlb)
>>   {
>>   	bool set_write;
>> @@ -270,7 +266,7 @@ static void set_write_prot_commit_flush_ptes(struct vm_area_struct *vma,
>>   				       /* idx = */ 0, set_write, tlb);
>>   		return;
>>   	}
>> -	commit_anon_folio_batch(vma, folio, addr, ptep, oldpte, ptent, nr_ptes, tlb);
>> +	commit_anon_folio_batch(vma, folio, page, addr, ptep, oldpte, ptent, nr_ptes, tlb);
>>   }
>>
>>   static long change_pte_range(struct mmu_gather *tlb,
>> @@ -305,15 +301,19 @@ static long change_pte_range(struct mmu_gather *tlb,
>>   			const fpb_t flags = FPB_RESPECT_SOFT_DIRTY | FPB_RESPECT_WRITE;
>>   			int max_nr_ptes = (end - addr) >> PAGE_SHIFT;
>>   			struct folio *folio = NULL;
>> +			struct page *page;
>>   			pte_t ptent;
>>
>> +			page = vm_normal_page(vma, addr, oldpte);
>> +			if (page)
>> +				folio = page_folio(page);
>>   			/*
>>   			 * Avoid trapping faults against the zero or KSM
>>   			 * pages. See similar comment in change_huge_pmd.
>>   			 */
>>   			if (prot_numa) {
>>   				int ret = prot_numa_skip(vma, addr, oldpte, pte,
>> -							 target_node, &folio);
>> +							 target_node, folio);
>>   				if (ret) {
>>
>>   					/* determine batch to skip */
>> @@ -323,9 +323,6 @@ static long change_pte_range(struct mmu_gather *tlb,
>>   				}
>>   			}
>>
>> -			if (!folio)
>> -				folio = vm_normal_folio(vma, addr, oldpte);
>> -
>>   			nr_ptes = mprotect_folio_pte_batch(folio, pte, oldpte, max_nr_ptes, flags);
>>
>>   			oldpte = modify_prot_start_ptes(vma, addr, pte, nr_ptes);
>> @@ -351,7 +348,7 @@ static long change_pte_range(struct mmu_gather *tlb,
>>   			 */
>>   			if ((cp_flags & MM_CP_TRY_CHANGE_WRITABLE) &&
>>   			     !pte_write(ptent))
>> -				set_write_prot_commit_flush_ptes(vma, folio,
>> +				set_write_prot_commit_flush_ptes(vma, folio, page,
>>   				addr, pte, oldpte, ptent, nr_ptes, tlb);
>>   			else
>>   				prot_commit_flush_ptes(vma, addr, pte, oldpte, ptent,
>> --
>> 2.30.2
>>

Re: [PATCH mm-hotfixes-unstable] mm: Pass page directly instead of using folio_page

[Lorenzo Stoakes]

On Wed, Aug 06, 2025 at 10:49:46PM +0530, Dev Jain wrote:
>
> On 06/08/25 10:10 pm, Lorenzo Stoakes wrote:
> > On Wed, Aug 06, 2025 at 08:26:11PM +0530, Dev Jain wrote:
> > > In commit_anon_folio_batch(), we iterate over all pages pointed to by the
> > > PTE batch. Therefore we need to know the first page of the batch;
> > > currently we derive that via folio_page(folio, 0), but, that takes us
> > > to the first (head) page of the folio instead - our PTE batch may lie
> > > in the middle of the folio, leading to incorrectness.
> > >
> > > Bite the bullet and throw away the micro-optimization of reusing the
> > > folio in favour of code simplicity. Derive the page and the folio in
> > > change_pte_range, and pass the page too to commit_anon_folio_batch to
> > > fix the aforementioned issue.
> > >
> > > Reported-by: syzbot+57bcc752f0df8bb1365c@syzkaller.appspotmail.com
> > > Fixes: cac1db8c3aad ("mm: optimize mprotect() by PTE batching")
> > > Signed-off-by: Dev Jain <dev.jain@arm.com>
> > This looks reasonable, fixes the problem and compiles/works on my machine so:
> >
> > Reviewed-by: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
>
> Thanks.
>
> >
> > This badly needs refactoring as 13 parameters being passed to a function is
>
> Which function are you talking about?

Misread, 9, but it's still silly. mprotect in general needs rework. I may submit
a patch for this at some point...

Anyway let's just get this hotfix in asap.

>
> > ridiculous, but we can do that later.
> >
> > Let's get this in as a hotfix asap.

Re: [PATCH mm-hotfixes-unstable] mm: Pass page directly instead of using folio_page

[David Hildenbrand]

On 06.08.25 16:56, Dev Jain wrote:
> In commit_anon_folio_batch(), we iterate over all pages pointed to by the
> PTE batch. Therefore we need to know the first page of the batch;
> currently we derive that via folio_page(folio, 0), but, that takes us
> to the first (head) page of the folio instead - our PTE batch may lie
> in the middle of the folio, leading to incorrectness.
> 
> Bite the bullet and throw away the micro-optimization of reusing the
> folio in favour of code simplicity.

Huh? We are still reusing the folio. There is a single 
vm_normal_page()+page_folio() lookup in that code.

Maybe what you meant is "Simplify the code by moving the folio lookup 
out of prot_numa_skip() such that we only have a single page+folio lookup."

> Derive the page and the folio in
> change_pte_range, and pass the page too to commit_anon_folio_batch to
> fix the aforementioned issue.
> 
> Reported-by: syzbot+57bcc752f0df8bb1365c@syzkaller.appspotmail.com
> Fixes: cac1db8c3aad ("mm: optimize mprotect() by PTE batching")

Fell free to add a

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

> Signed-off-by: Dev Jain <dev.jain@arm.com>

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

-- 
Cheers,

David / dhildenb