Skip to contentSkip to conversation

ksm: abstract the function try_to_get_old_rmap_item

PATCH Superseded [email protected] 2 participants 4 messages Dec 30, 2022 – Feb 13, 2023

Subsystems (2)

v5 of 10 Superseded Latest is v10 →

lore
Share & tools
lore
#0 lore raw

PATCH v5 1/6 ksm: abstract the function try_to_get_old_rmap_item

From: xu xin <[email protected]>

A new function try_to_get_old_rmap_item is abstracted from get_next_rmap_item. This function will be reused by the subsequent patches about counting ksm_zero_pages.

The patch improves the readability and reusability of KSM code.

Cc David Hildenbrand <[email protected]>
Cc Claudio Imbrenda <[email protected]>
Cc Xuexin Jiang <[email protected]>
---
 mm/ksm.c | 25 +++++++++++++++++++------
 1 file changed, 19 insertions(+), 6 deletions(-)
mm/ksm.c +19 -6
index 83e2f74ae7da..5b0a7343ff4a 100644
@@ -2214,23 +2214,36 @@ static void cmp_and_merge_page(struct page *page, struct ksm_rmap_item *rmap_ite
 	}
 }
 
-static struct ksm_rmap_item *get_next_rmap_item(struct ksm_mm_slot *mm_slot,
-					    struct ksm_rmap_item **rmap_list,
-					    unsigned long addr)
+static struct ksm_rmap_item *try_to_get_old_rmap_item(unsigned long addr,
+					 struct ksm_rmap_item **rmap_list)
 {
-	struct ksm_rmap_item *rmap_item;
-
 	while (*rmap_list) {
-		rmap_item = *rmap_list;
+		struct ksm_rmap_item *rmap_item = *rmap_list;
 		if ((rmap_item->address & PAGE_MASK) == addr)
 			return rmap_item;
 		if (rmap_item->address > addr)
 			break;
 		*rmap_list = rmap_item->rmap_list;
+		/* Running here indicates it's vma has been UNMERGEABLE */
 		remove_rmap_item_from_tree(rmap_item);
 		free_rmap_item(rmap_item);
 	}
 
+	return NULL;
+}
+
+static struct ksm_rmap_item *get_next_rmap_item(struct ksm_mm_slot *mm_slot,
+					    struct ksm_rmap_item **rmap_list,
+					    unsigned long addr)
+{
+	struct ksm_rmap_item *rmap_item;
+
+	/* lookup if we have a old rmap_item matching the addr*/
+	rmap_item = try_to_get_old_rmap_item(addr, rmap_list);
+	if (rmap_item)
+		return rmap_item;
+
+	/* Need to allocate a new rmap_item */
 	rmap_item = alloc_rmap_item();
 	if (rmap_item) {
 		/* It has already been zeroed */
Signature · 1 line
2.15.2
#1 lore raw
Quoted 27 lines · On 30.12.22 02:12, [email protected] wrote:

From: xu xin <[email protected]>

A new function try_to_get_old_rmap_item is abstracted from get_next_rmap_item. This function will be reused by the subsequent patches about counting ksm_zero_pages.

The patch improves the readability and reusability of KSM code.

Cc David Hildenbrand <[email protected]>
Cc Claudio Imbrenda <[email protected]>
Cc Xuexin Jiang <[email protected]>
---
  mm/ksm.c | 25 +++++++++++++++++++------
  1 file changed, 19 insertions(+), 6 deletions(-)
mm/ksm.c +0 -1
index 83e2f74ae7da..5b0a7343ff4a 100644
@@ -2214,23 +2214,36 @@ static void cmp_and_merge_page(struct page *page, struct ksm_rmap_item *rmap_ite
  	}
  }
 
-static struct ksm_rmap_item *get_next_rmap_item(struct ksm_mm_slot *mm_slot,

mm/ksm.c static void cmp_and_merge_page(struct page *page, struct ksm_rmap_item *rmap_ite · line 2221

Quoted 10 lines
mm/ksm.c +3 -5
-					    struct ksm_rmap_item **rmap_list,
-					    unsigned long addr)
+static struct ksm_rmap_item *try_to_get_old_rmap_item(unsigned long addr,
+					 struct ksm_rmap_item **rmap_list)
  {
-	struct ksm_rmap_item *rmap_item;
-
  	while (*rmap_list) {
-		rmap_item = *rmap_list;
+		struct ksm_rmap_item *rmap_item = *rmap_list;

Empty line missing.

mm/ksm.c static void cmp_and_merge_page(struct page *page, struct ksm_rmap_item *rmap_ite · line 2227

Quoted 6 lines
mm/ksm.c +1 -0
  		if ((rmap_item->address & PAGE_MASK) == addr)
  			return rmap_item;
  		if (rmap_item->address > addr)
  			break;
  		*rmap_list = rmap_item->rmap_list;
+		/* Running here indicates it's vma has been UNMERGEABLE */

"If we end up here, the VMA is UNMERGEABLE."

Although I am not sure if that is true?

Quoted 4 lines
mm/ksm.c +0 -0
  		remove_rmap_item_from_tree(rmap_item);
  		free_rmap_item(rmap_item);
  	}
 

mm/ksm.c static void cmp_and_merge_page(struct page *page, struct ksm_rmap_item *rmap_ite · line 2241

Quoted 10 lines
mm/ksm.c +10 -0
+	return NULL;
+}
+
+static struct ksm_rmap_item *get_next_rmap_item(struct ksm_mm_slot *mm_slot,
+					    struct ksm_rmap_item **rmap_list,
+					    unsigned long addr)
+{
+	struct ksm_rmap_item *rmap_item;
+
+	/* lookup if we have a old rmap_item matching the addr*/

I suggest dropping that comment, "try_to_get_old_rmap_item()" is expressive enough.

mm/ksm.c static void cmp_and_merge_page(struct page *page, struct ksm_rmap_item *rmap_ite · line 2246

Quoted 5 lines
mm/ksm.c +5 -0
+	rmap_item = try_to_get_old_rmap_item(addr, rmap_list);
+	if (rmap_item)
+		return rmap_item;
+
+	/* Need to allocate a new rmap_item */

I suggest dropping that comment for the same reason.

Quoted 3 lines
mm/ksm.c +0 -0
  	rmap_item = alloc_rmap_item();
  	if (rmap_item) {
  		/* It has already been zeroed */
Signature · 3 lines
Thanks,

David / dhildenb
#3 lore raw
Quoted 1 line · On 04.02.23 07:43, [email protected] wrote:

Thanks for these suggestions, we will take them into patch v6.

Please don't remove all context from your replies. Makes it really hard to follow-up :)

Further, there seems to be an issue with threading and your mails. For example, your cover letter [1] does not link the other patches like [2], for example, because the patches don't have "in-reply-to:" tags. Try using "git send-email", that should take care of threading automatically.

[1] https://lore.kernel.org/all/[email protected]/
[2] https://lore.kernel.org/all/[email protected]/

Signature · 3 lines
Thanks,

David / dhildenb