[v1] Fix lazy mmu mode

[PATCH v1 3/4] sparc/mm: Avoid calling arch_enter/leave_lazy_mmu() in set_ptes

Posted by Ryan Roberts 8 months ago

With commit 1a10a44dfc1d ("sparc64: implement the new page table range
API") set_ptes was added to the sparc architecture. The implementation
included calling arch_enter/leave_lazy_mmu() calls.

The patch removes the usage of arch_enter/leave_lazy_mmu() since this
implies nesting of lazy mmu regions which is not supported. Without this
fix, lazy mmu mode is effectively disabled because we exit the mode
after the first set_ptes:

remap_pte_range()
  -> arch_enter_lazy_mmu()
  -> set_ptes()
      -> arch_enter_lazy_mmu()
      -> arch_leave_lazy_mmu()
  -> arch_leave_lazy_mmu()

Powerpc suffered the same problem and fixed it in a corresponding way
with commit 47b8def9358c ("powerpc/mm: Avoid calling
arch_enter/leave_lazy_mmu() in set_ptes").

Fixes: 1a10a44dfc1d ("sparc64: implement the new page table range API")
Signed-off-by: Ryan Roberts <ryan.roberts@arm.com>
---
 arch/sparc/include/asm/pgtable_64.h | 2 --
 1 file changed, 2 deletions(-)

diff --git a/arch/sparc/include/asm/pgtable_64.h b/arch/sparc/include/asm/pgtable_64.h
index 2b7f358762c1..dc28f2c4eee3 100644
--- a/arch/sparc/include/asm/pgtable_64.h
+++ b/arch/sparc/include/asm/pgtable_64.h
@@ -936,7 +936,6 @@ static inline void __set_pte_at(struct mm_struct *mm, unsigned long addr,
 static inline void set_ptes(struct mm_struct *mm, unsigned long addr,
 		pte_t *ptep, pte_t pte, unsigned int nr)
 {
-	arch_enter_lazy_mmu_mode();
 	for (;;) {
 		__set_pte_at(mm, addr, ptep, pte, 0);
 		if (--nr == 0)
@@ -945,7 +944,6 @@ static inline void set_ptes(struct mm_struct *mm, unsigned long addr,
 		pte_val(pte) += PAGE_SIZE;
 		addr += PAGE_SIZE;
 	}
-	arch_leave_lazy_mmu_mode();
 }
 #define set_ptes set_ptes
 
-- 
2.43.0

Re: [PATCH v1 3/4] sparc/mm: Avoid calling arch_enter/leave_lazy_mmu() in set_ptes

Posted by Andreas Larsson 8 months ago

On 2025-03-02 15:55, Ryan Roberts wrote:
> With commit 1a10a44dfc1d ("sparc64: implement the new page table range
> API") set_ptes was added to the sparc architecture. The implementation
> included calling arch_enter/leave_lazy_mmu() calls.
> 
> The patch removes the usage of arch_enter/leave_lazy_mmu() since this
> implies nesting of lazy mmu regions which is not supported. Without this
> fix, lazy mmu mode is effectively disabled because we exit the mode
> after the first set_ptes:
> 
> remap_pte_range()
>   -> arch_enter_lazy_mmu()
>   -> set_ptes()
>       -> arch_enter_lazy_mmu()
>       -> arch_leave_lazy_mmu()
>   -> arch_leave_lazy_mmu()
> 
> Powerpc suffered the same problem and fixed it in a corresponding way
> with commit 47b8def9358c ("powerpc/mm: Avoid calling
> arch_enter/leave_lazy_mmu() in set_ptes").
> 
> Fixes: 1a10a44dfc1d ("sparc64: implement the new page table range API")
> Signed-off-by: Ryan Roberts <ryan.roberts@arm.com>
> ---
>  arch/sparc/include/asm/pgtable_64.h | 2 --
>  1 file changed, 2 deletions(-)
> 
> diff --git a/arch/sparc/include/asm/pgtable_64.h b/arch/sparc/include/asm/pgtable_64.h
> index 2b7f358762c1..dc28f2c4eee3 100644
> --- a/arch/sparc/include/asm/pgtable_64.h
> +++ b/arch/sparc/include/asm/pgtable_64.h
> @@ -936,7 +936,6 @@ static inline void __set_pte_at(struct mm_struct *mm, unsigned long addr,
>  static inline void set_ptes(struct mm_struct *mm, unsigned long addr,
>  		pte_t *ptep, pte_t pte, unsigned int nr)
>  {
> -	arch_enter_lazy_mmu_mode();
>  	for (;;) {
>  		__set_pte_at(mm, addr, ptep, pte, 0);
>  		if (--nr == 0)
> @@ -945,7 +944,6 @@ static inline void set_ptes(struct mm_struct *mm, unsigned long addr,
>  		pte_val(pte) += PAGE_SIZE;
>  		addr += PAGE_SIZE;
>  	}
> -	arch_leave_lazy_mmu_mode();
>  }
>  #define set_ptes set_ptes

Acked-by: Andreas Larsson <andreas@gaisler.com>

Thanks,
Andreas

Re: [PATCH v1 3/4] sparc/mm: Avoid calling arch_enter/leave_lazy_mmu() in set_ptes

Posted by David Hildenbrand 8 months ago

On 02.03.25 15:55, Ryan Roberts wrote:
> With commit 1a10a44dfc1d ("sparc64: implement the new page table range
> API") set_ptes was added to the sparc architecture. The implementation
> included calling arch_enter/leave_lazy_mmu() calls.
> 
> The patch removes the usage of arch_enter/leave_lazy_mmu() since this
> implies nesting of lazy mmu regions which is not supported. Without this
> fix, lazy mmu mode is effectively disabled because we exit the mode
> after the first set_ptes:
> 
> remap_pte_range()
>    -> arch_enter_lazy_mmu()
>    -> set_ptes()
>        -> arch_enter_lazy_mmu()
>        -> arch_leave_lazy_mmu()
>    -> arch_leave_lazy_mmu()
> 
> Powerpc suffered the same problem and fixed it in a corresponding way
> with commit 47b8def9358c ("powerpc/mm: Avoid calling
> arch_enter/leave_lazy_mmu() in set_ptes").
> 
> Fixes: 1a10a44dfc1d ("sparc64: implement the new page table range API")
> Signed-off-by: Ryan Roberts <ryan.roberts@arm.com>
> ---
>   arch/sparc/include/asm/pgtable_64.h | 2 --
>   1 file changed, 2 deletions(-)
> 
> diff --git a/arch/sparc/include/asm/pgtable_64.h b/arch/sparc/include/asm/pgtable_64.h
> index 2b7f358762c1..dc28f2c4eee3 100644
> --- a/arch/sparc/include/asm/pgtable_64.h
> +++ b/arch/sparc/include/asm/pgtable_64.h
> @@ -936,7 +936,6 @@ static inline void __set_pte_at(struct mm_struct *mm, unsigned long addr,
>   static inline void set_ptes(struct mm_struct *mm, unsigned long addr,
>   		pte_t *ptep, pte_t pte, unsigned int nr)
>   {
> -	arch_enter_lazy_mmu_mode();
>   	for (;;) {
>   		__set_pte_at(mm, addr, ptep, pte, 0);
>   		if (--nr == 0)
> @@ -945,7 +944,6 @@ static inline void set_ptes(struct mm_struct *mm, unsigned long addr,
>   		pte_val(pte) += PAGE_SIZE;
>   		addr += PAGE_SIZE;
>   	}
> -	arch_leave_lazy_mmu_mode();
>   }
>   #define set_ptes set_ptes
>   

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

-- 
Cheers,

David / dhildenb