Skip to contentSkip to conversation

mm: slab/slub: use page->list consistently instead of page->lru

PATCH In review Dave Hansen 4 participants 6 messages Dec 11, 2013 – Dec 12, 2013

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/2 mm: slab/slub: use page->list consistently instead of page->lru

From: Dave Hansen <[email protected]>

'struct page' has two list_head fields: 'lru' and 'list'. Conveniently, they are unioned together. This means that code can use them interchangably, which gets horribly confusing like with this nugget from slab.c:

Quoted 3 lines
	list_del(&page->lru);
	if (page->active == cachep->num)
		list_add(&page->list, &n->slabs_full);

This patch makes the slab and slub code use page->list universally instead of mixing ->list and ->lru.

It also adds some comments to attempt to keep new users from picking up uses of ->list.

So, the new rule is: page->list is what the slabs use. page->lru is for everybody else. This is a pretty arbitrary rule, but we need _something_. Maybe we should just axe the ->list one and make the sl?bs use ->lru.

---

 linux.git-davehans/include/linux/mm_types.h |    5 ++
 linux.git-davehans/mm/slab.c                |   50 ++++++++++++++--------------
 2 files changed, 29 insertions(+), 26 deletions(-)
include/linux/mm_types.h renamed from include/linux/mm_types.h~make-slab-use-page-lru-vs-list-consistently +4 -1
diff -puN include/linux/mm_types.h~make-slab-use-page-lru-vs-list-consistently include/linux/mm_types.h
@@ -123,6 +123,8 @@ struct page {
 	union {
 		struct list_head lru;	/* Pageout list, eg. active_list
 					 * protected by zone->lru_lock !
+					 * Can be used as a generic list
+					 * by the page owner.
 					 */
 		struct {		/* slub per cpu partial pages */
 			struct page *next;	/* Next partial slab */
@@ -135,7 +137,8 @@ struct page {
 #endif
 		};
 
-		struct list_head list;	/* slobs list of pages */
+		struct list_head list;	/* sl[aou]bs list of pages.
+					 * do not use outside of slabs */
 		struct slab *slab_page; /* slab fields */
 		struct rcu_head rcu_head;	/* Used by SLAB
 						 * when destroying via RCU
mm/slab.c renamed from mm/slab.c~make-slab-use-page-lru-vs-list-consistently +25 -25

170 lines not on this page · raw message

_

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

#1 lore raw

PATCH 2/2 mm: blk-mq: uses page->list incorrectly

From: Dave Hansen <[email protected]>

'struct page' has two list_head fields: 'lru' and 'list'. Conveniently, they are unioned together. This means that code can use them interchangably, which gets horribly confusing.

The blk-mq made the logical decision to try to use page->list. But, that field was actually introduced just for the slub code. ->lru is the right field to use outside of slab/slub.

---

 linux.git-davehans/block/blk-mq.c |    6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
block/blk-mq.c renamed from block/blk-mq.c~blk-mq-uses-page-list-incorrectly +3 -3
diff -puN block/blk-mq.c~blk-mq-uses-page-list-incorrectly block/blk-mq.c
@@ -1087,8 +1087,8 @@ static void blk_mq_free_rq_map(struct bl
 	struct page *page;
 
 	while (!list_empty(&hctx->page_list)) {
-		page = list_first_entry(&hctx->page_list, struct page, list);
-		list_del_init(&page->list);
+		page = list_first_entry(&hctx->page_list, struct page, lru);
+		list_del_init(&page->lru);
 		__free_pages(page, page->private);
 	}
 
@@ -1152,7 +1152,7 @@ static int blk_mq_init_rq_map(struct blk
 			break;
 
 		page->private = this_order;
-		list_add_tail(&page->list, &hctx->page_list);
+		list_add_tail(&page->lru, &hctx->page_list);
 
 		p = page_address(page);
 		entries_per_page = order_to_size(this_order) / rq_size;

_

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

Acked-by on v1 2/2
#2 lore raw

Re: [PATCH 2/2] mm: blk-mq: uses page->list incorrectly

Quoted 7 lines · Dave Hansen

From: Dave Hansen <[email protected]>

'struct page' has two list_head fields: 'lru' and 'list'. Conveniently, they are unioned together. This means that code can use them interchangably, which gets horribly confusing.

Quoted 5 lines

The blk-mq made the logical decision to try to use page->list. But, that field was actually introduced just for the slub code. ->lru is the right field to use outside of slab/slub.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

Acked-by on v1 2/2
#3 lore raw

RE: [PATCH 2/2] mm: blk-mq: uses page->list incorrectly

In reply to Dave Hansen
Quoted 7 lines · Dave Hansen

From: Dave Hansen <[email protected]>

'struct page' has two list_head fields: 'lru' and 'list'. Conveniently, they are unioned together. This means that code can use them interchangably, which gets horribly confusing.

Quoted 5 lines

The blk-mq made the logical decision to try to use page->list. But, that field was actually introduced just for the slub code. ->lru is the right field to use outside of slab/slub.

Looks good to me.

for both.

Signature · 1 line
 Kirill A. Shutemov
#4 lore raw
In reply to Dave Hansen
Quoted 14 lines · Dave Hansen

From: Dave Hansen <[email protected]>

'struct page' has two list_head fields: 'lru' and 'list'. Conveniently, they are unioned together. This means that code can use them interchangably, which gets horribly confusing like with this nugget from slab.c:

Quoted 3 lines
	list_del(&page->lru);
	if (page->active == cachep->num)
		list_add(&page->list, &n->slabs_full);

This patch makes the slab and slub code use page->list universally instead of mixing ->list and ->lru.

Quoted 8 lines

It also adds some comments to attempt to keep new users from picking up uses of ->list.

So, the new rule is: page->list is what the slabs use. page->lru is for everybody else. This is a pretty arbitrary rule, but we need _something_. Maybe we should just axe the ->list one and make the sl?bs use ->lru.

I'd recommend this suggestion, I don't see why the slab allocators can't use a page->lru field to maintain their lists of slab pages and it makes the code much cleaner. Anybody hacking thise code will know it's not really a lru and we're just reusing a field from struct page without adding unnecessary complexity.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

Acked-by on v1 1/2
#5 lore raw
In reply to Dave Hansen
Quoted 4 lines · Dave Hansen

'struct page' has two list_head fields: 'lru' and 'list'. Conveniently, they are unioned together. This means that code can use them interchangably, which gets horribly confusing like with this nugget from slab.c:

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>