container_of(node->array, ..., i_pages) just to access i_pages again
is an incredibly roundabout way of accessing node->array itself.
Simplify it.

Signed-off-by: Pedro Falcato <pfalcato@suse.de>
---
 mm/workingset.c | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)

diff --git a/mm/workingset.c b/mm/workingset.c
index 4841ae8af41113797378846f08336cd7c5757bd5..6e7f4cb1b9a7807e9288955f180a5b6cffab1a40 100644
--- a/mm/workingset.c
+++ b/mm/workingset.c
@@ -612,7 +612,6 @@ struct list_lru shadow_nodes;
 
 void workingset_update_node(struct xa_node *node)
 {
-	struct address_space *mapping;
 	struct page *page = virt_to_page(node);
 
 	/*
@@ -623,8 +622,7 @@ void workingset_update_node(struct xa_node *node)
 	 * already where they should be. The list_empty() test is safe
 	 * as node->private_list is protected by the i_pages lock.
 	 */
-	mapping = container_of(node->array, struct address_space, i_pages);
-	lockdep_assert_held(&mapping->i_pages.xa_lock);
+	lockdep_assert_held(&node->array->xa_lock);
 
 	if (node->count && node->count == node->nr_values) {
 		if (list_empty(&node->private_list)) {

---
base-commit: 856ddd6c69a1a1935f32de8411550d997bcbe6f5
change-id: 20250421-workingset-simplify-eef38019a78c

Best regards,
-- 
Pedro Falcato <pfalcato@suse.de>

On Mon, Apr 21, 2025 at 06:16:28PM +0100, Pedro Falcato wrote:
> container_of(node->array, ..., i_pages) just to access i_pages again
> is an incredibly roundabout way of accessing node->array itself.
> Simplify it.

Agreed. The other side is shadow_lru_isolate() which goes also goes
through mapping->i_pages. That's probably how it came to be. But your
patch doesn't make that relationship any harder to understand.

> Signed-off-by: Pedro Falcato <pfalcato@suse.de>

Acked-by: Johannes Weiner <hannes@cmpxchg.org>

Thanks

On Mon, Apr 21, 2025 at 06:16:28PM +0100, Pedro Falcato wrote:
> container_of(node->array, ..., i_pages) just to access i_pages again
> is an incredibly roundabout way of accessing node->array itself.
> Simplify it.
> 
> Signed-off-by: Pedro Falcato <pfalcato@suse.de>
> ---
>  mm/workingset.c | 4 +---
>  1 file changed, 1 insertion(+), 3 deletions(-)
> 
> diff --git a/mm/workingset.c b/mm/workingset.c
> index 4841ae8af41113797378846f08336cd7c5757bd5..6e7f4cb1b9a7807e9288955f180a5b6cffab1a40 100644
> --- a/mm/workingset.c
> +++ b/mm/workingset.c
> @@ -612,7 +612,6 @@ struct list_lru shadow_nodes;
>  
>  void workingset_update_node(struct xa_node *node)
>  {
> -	struct address_space *mapping;
>  	struct page *page = virt_to_page(node);
>  
>  	/*
> @@ -623,8 +622,7 @@ void workingset_update_node(struct xa_node *node)
>  	 * already where they should be. The list_empty() test is safe
>  	 * as node->private_list is protected by the i_pages lock.
>  	 */
> -	mapping = container_of(node->array, struct address_space, i_pages);
> -	lockdep_assert_held(&mapping->i_pages.xa_lock);
> +	lockdep_assert_held(&node->array->xa_lock);
>  
>  	if (node->count && node->count == node->nr_values) {
>  		if (list_empty(&node->private_list)) {
> 

Actually, not sure if this is wanted given the original code is a little more
explicit on what the lock is. +CC the original author

If people think this is worse, just drop the patch, I don't really care - was just
checking out the function for other purposes and found this bit a little confusing.

-- 
Pedro

On Mon, Apr 21, 2025 at 06:39:53PM +0100, Pedro Falcato wrote:
> Actually, not sure if this is wanted given the original code is a little more
> explicit on what the lock is. +CC the original author
> 
> If people think this is worse, just drop the patch, I don't really care - was just
> checking out the function for other purposes and found this bit a little confusing.

I think it's clearer.

Reviewed-by: Matthew Wilcox (Oracle) <willy@infradead.org>