[PATCH] mm: move dummy_vm_ops out of a header

[PATCH] mm: move dummy_vm_ops out of a header

[Mateusz Guzik]

Otherwise the kernel ends up with multiple copies:
$ nm vmlinux | grep dummy_vm_ops
ffffffff81e4ea00 d dummy_vm_ops.2
ffffffff81e11760 d dummy_vm_ops.254
ffffffff81e406e0 d dummy_vm_ops.4
ffffffff81e3c780 d dummy_vm_ops.7

While here prefix it with vma_.

[if someone has better ideas where to put it, please move it]

Signed-off-by: Mateusz Guzik <mjguzik@gmail.com>
---
 include/linux/mm.h | 6 +++---
 mm/init-mm.c       | 2 ++
 2 files changed, 5 insertions(+), 3 deletions(-)

diff --git a/include/linux/mm.h b/include/linux/mm.h
index 406ab9ea818f..14898e76bbf1 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -756,17 +756,17 @@ static inline void vma_mark_detached(struct vm_area_struct *vma,
 
 #endif /* CONFIG_PER_VMA_LOCK */
 
+extern const struct vm_operations_struct vma_dummy_vm_ops;
+
 /*
  * WARNING: vma_init does not initialize vma->vm_lock.
  * Use vm_area_alloc()/vm_area_free() if vma needs locking.
  */
 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
 {
-	static const struct vm_operations_struct dummy_vm_ops = {};
-
 	memset(vma, 0, sizeof(*vma));
 	vma->vm_mm = mm;
-	vma->vm_ops = &dummy_vm_ops;
+	vma->vm_ops = &vma_dummy_vm_ops;
 	INIT_LIST_HEAD(&vma->anon_vma_chain);
 	vma_mark_detached(vma, false);
 	vma_numab_state_init(vma);
diff --git a/mm/init-mm.c b/mm/init-mm.c
index efa97b57acfd..cfd367822cdd 100644
--- a/mm/init-mm.c
+++ b/mm/init-mm.c
@@ -17,6 +17,8 @@
 #define INIT_MM_CONTEXT(name)
 #endif
 
+const struct vm_operations_struct vma_dummy_vm_ops;
+
 /*
  * For dynamically allocated mm_structs, there is a dynamically sized cpumask
  * at the end of the structure, the size of which depends on the maximum CPU
-- 
2.39.2

Re: [PATCH] mm: move dummy_vm_ops out of a header

[Matthew Wilcox]

On Mon, Aug 07, 2023 at 01:16:11AM +0200, Mateusz Guzik wrote:
> Otherwise the kernel ends up with multiple copies:
> $ nm vmlinux | grep dummy_vm_ops
> ffffffff81e4ea00 d dummy_vm_ops.2
> ffffffff81e11760 d dummy_vm_ops.254
> ffffffff81e406e0 d dummy_vm_ops.4
> ffffffff81e3c780 d dummy_vm_ops.7
> 
> While here prefix it with vma_.

It really shouldn't be prefixed with vma.  Other than that, I love this
patch.

> [if someone has better ideas where to put it, please move it]
> 
> Signed-off-by: Mateusz Guzik <mjguzik@gmail.com>
> ---
>  include/linux/mm.h | 6 +++---
>  mm/init-mm.c       | 2 ++
>  2 files changed, 5 insertions(+), 3 deletions(-)
> 
> diff --git a/include/linux/mm.h b/include/linux/mm.h
> index 406ab9ea818f..14898e76bbf1 100644
> --- a/include/linux/mm.h
> +++ b/include/linux/mm.h
> @@ -756,17 +756,17 @@ static inline void vma_mark_detached(struct vm_area_struct *vma,
>  
>  #endif /* CONFIG_PER_VMA_LOCK */
>  
> +extern const struct vm_operations_struct vma_dummy_vm_ops;
> +
>  /*
>   * WARNING: vma_init does not initialize vma->vm_lock.
>   * Use vm_area_alloc()/vm_area_free() if vma needs locking.
>   */
>  static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
>  {
> -	static const struct vm_operations_struct dummy_vm_ops = {};
> -
>  	memset(vma, 0, sizeof(*vma));
>  	vma->vm_mm = mm;
> -	vma->vm_ops = &dummy_vm_ops;
> +	vma->vm_ops = &vma_dummy_vm_ops;
>  	INIT_LIST_HEAD(&vma->anon_vma_chain);
>  	vma_mark_detached(vma, false);
>  	vma_numab_state_init(vma);
> diff --git a/mm/init-mm.c b/mm/init-mm.c
> index efa97b57acfd..cfd367822cdd 100644
> --- a/mm/init-mm.c
> +++ b/mm/init-mm.c
> @@ -17,6 +17,8 @@
>  #define INIT_MM_CONTEXT(name)
>  #endif
>  
> +const struct vm_operations_struct vma_dummy_vm_ops;
> +
>  /*
>   * For dynamically allocated mm_structs, there is a dynamically sized cpumask
>   * at the end of the structure, the size of which depends on the maximum CPU
> -- 
> 2.39.2
> 
> 

Re: [PATCH] mm: move dummy_vm_ops out of a header

[Mateusz Guzik]

On 8/7/23, Matthew Wilcox <willy@infradead.org> wrote:
> On Mon, Aug 07, 2023 at 01:16:11AM +0200, Mateusz Guzik wrote:
>> Otherwise the kernel ends up with multiple copies:
>> $ nm vmlinux | grep dummy_vm_ops
>> ffffffff81e4ea00 d dummy_vm_ops.2
>> ffffffff81e11760 d dummy_vm_ops.254
>> ffffffff81e406e0 d dummy_vm_ops.4
>> ffffffff81e3c780 d dummy_vm_ops.7
>>
>> While here prefix it with vma_.
>
> It really shouldn't be prefixed with vma.  Other than that, I love this
> patch.
>

I think an unprefixed global is iffy, but I'm not going to insist on
semi-cosmetics.

The one thing I expected people to complain about is the location of
the struct. The mm_init.c file was the least bad choice I found, but
maybe something else is preferred?
I'm just trying to avoid sending V3 for this patch after someone
claims dummy ops should land elsewhere. ;)

-- 
Mateusz Guzik <mjguzik gmail.com>