Skip to contentSkip to conversation

mm/memory_hotplug: cleanup do_migrate_range()

PATCH In review SeongJae Park 3 participants 8 messages Feb 15, 2023 – Feb 16, 2023

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

v2 cover mm/memory_hotplug: cleanup do_migrate_range()

Make do_migrate_range() return value mechanism simple.

Changes from v1
(https://lore.kernel.org/linux-mm/[email protected]/)
- Simply return nothing from do_migrate_range() (David Hildenbrand)
- Add a cleanup for 'ret' variable

SeongJae Park (2):
mm/memory_hotplug: return nothing from do_migrate_range() mm/memory_hotplug: cleanup 'ret' variable usage in do_migrate_range()

 mm/memory_hotplug.c | 8 ++------
 1 file changed, 2 insertions(+), 6 deletions(-)
Signature · 1 line
2.25.1
#1 lore raw

PATCH v2 1/2 mm/memory_hotplug: return nothing from do_migrate_range()

Return value mechanism of do_migrate_range() is not very simple, while no caller of the function checks the return value. Make the function return nothing to be more simple.

---
 mm/memory_hotplug.c | 5 +----
 1 file changed, 1 insertion(+), 4 deletions(-)
mm/memory_hotplug.c +1 -4
index bcb0dc41c2f2..6c615ba1a5c7 100644
@@ -1620,8 +1620,7 @@ static int scan_movable_pages(unsigned long start, unsigned long end,
 	return 0;
 }
 
-static int
-do_migrate_range(unsigned long start_pfn, unsigned long end_pfn)
+static void do_migrate_range(unsigned long start_pfn, unsigned long end_pfn)
 {
 	unsigned long pfn;
 	struct page *page, *head;
@@ -1721,8 +1720,6 @@ do_migrate_range(unsigned long start_pfn, unsigned long end_pfn)
 			putback_movable_pages(&source);
 		}
 	}
-
-	return ret;
 }
 
 static int __init cmdline_parse_movable_node(char *p)
Signature · 1 line
2.25.1
#6 lore raw

PATCH v2 2/2 mm/memory_hotplug: cleanup 'ret' variable usage in do_migrate_range()

In reply to SeongJae Park

Because do_migrate_range() is returning nothing, some 'ret' variable usages are unnecessary. Remove unnecessary usage and reduce its scope.

---
 mm/memory_hotplug.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)
mm/memory_hotplug.c +1 -2
index 6c615ba1a5c7..6df3072d11df 100644
@@ -1624,7 +1624,6 @@ static void do_migrate_range(unsigned long start_pfn, unsigned long end_pfn)
 {
 	unsigned long pfn;
 	struct page *page, *head;
-	int ret = 0;
 	LIST_HEAD(source);
 	static DEFINE_RATELIMIT_STATE(migrate_rs, DEFAULT_RATELIMIT_INTERVAL,
 				      DEFAULT_RATELIMIT_BURST);
@@ -1678,7 +1677,6 @@ static void do_migrate_range(unsigned long start_pfn, unsigned long end_pfn)
 						    page_is_file_lru(page));
 
 		} else {
-			ret = -EBUSY;
 			if (__ratelimit(&migrate_rs)) {
 				pr_warn("failed to isolate pfn %lx\n", pfn);
 				dump_page(page, "isolation failed");
@@ -1692,6 +1690,7 @@ static void do_migrate_range(unsigned long start_pfn, unsigned long end_pfn)
 			.nmask = &nmask,
 			.gfp_mask = GFP_USER | __GFP_MOVABLE | __GFP_RETRY_MAYFAIL,
 		};
+		int ret;
 
 		/*
 		 * We have checked that migration range is on a single zone so
Signature · 1 line
2.25.1
Acked-by on v2 1/2
#2 lore raw

Re: [PATCH v2 1/2] mm/memory_hotplug: return nothing from do_migrate_range()

In reply to SeongJae Park
Quoted 25 lines · SeongJae Park

Return value mechanism of do_migrate_range() is not very simple, while no caller of the function checks the return value. Make the function return nothing to be more simple.

---
  mm/memory_hotplug.c | 5 +----
  1 file changed, 1 insertion(+), 4 deletions(-)
mm/memory_hotplug.c +1 -2
index bcb0dc41c2f2..6c615ba1a5c7 100644
@@ -1620,8 +1620,7 @@ static int scan_movable_pages(unsigned long start, unsigned long end,
  	return 0;
  }
 
-static int
-do_migrate_range(unsigned long start_pfn, unsigned long end_pfn)
+static void do_migrate_range(unsigned long start_pfn, unsigned long end_pfn)
  {
  	unsigned long pfn;
  	struct page *page, *head;

mm/memory_hotplug.c do_migrate_range(unsigned long start_pfn, unsigned long end_pfn) · line 1723

Quoted 6 lines
mm/memory_hotplug.c +0 -2
@@ -1721,8 +1720,6 @@ do_migrate_range(unsigned long start_pfn, unsigned long end_pfn)
  			putback_movable_pages(&source);
  		}
  	}
-
-	return ret;
  }

I think this patch should also stop initializing ret = 0 inside that function.

With that

Signature · 3 lines
Thanks,

David / dhildenb
#7 lore raw

Re: [PATCH v2 2/2] mm/memory_hotplug: cleanup 'ret' variable usage in do_migrate_range()

In reply to SeongJae Park
Quoted 29 lines · SeongJae Park

Because do_migrate_range() is returning nothing, some 'ret' variable usages are unnecessary. Remove unnecessary usage and reduce its scope.

---
  mm/memory_hotplug.c | 3 +--
  1 file changed, 1 insertion(+), 2 deletions(-)
mm/memory_hotplug.c +0 -2
index 6c615ba1a5c7..6df3072d11df 100644
@@ -1624,7 +1624,6 @@ static void do_migrate_range(unsigned long start_pfn, unsigned long end_pfn)
  {
  	unsigned long pfn;
  	struct page *page, *head;
-	int ret = 0;
  	LIST_HEAD(source);
  	static DEFINE_RATELIMIT_STATE(migrate_rs, DEFAULT_RATELIMIT_INTERVAL,
  				      DEFAULT_RATELIMIT_BURST);
@@ -1678,7 +1677,6 @@ static void do_migrate_range(unsigned long start_pfn, unsigned long end_pfn)
  						    page_is_file_lru(page));
 
  		} else {
-			ret = -EBUSY;
  			if (__ratelimit(&migrate_rs)) {
  				pr_warn("failed to isolate pfn %lx\n", pfn);
  				dump_page(page, "isolation failed");

mm/memory_hotplug.c static void do_migrate_range(unsigned long start_pfn, unsigned long end_pfn) · line 1696

Quoted 7 lines
mm/memory_hotplug.c +1 -0
@@ -1692,6 +1690,7 @@ static void do_migrate_range(unsigned long start_pfn, unsigned long end_pfn)
  			.nmask = &nmask,
  			.gfp_mask = GFP_USER | __GFP_MOVABLE | __GFP_RETRY_MAYFAIL,
  		};
+		int ret;
 
  		/*
  		 * We have checked that migration range is on a single zone so

Please squash that into the previous patch, just calling it "mm/memory_hotplug: cleanup return value handling in do_migrate_range()" or similar.

Signature · 3 lines
Thanks,

David / dhildenb
Reviewed-by on v2 1/2
#4 lore raw

Re: [PATCH v2 1/2] mm/memory_hotplug: return nothing from do_migrate_range()

In reply to SeongJae Park
Quoted 25 lines · SeongJae Park

Return value mechanism of do_migrate_range() is not very simple, while no caller of the function checks the return value. Make the function return nothing to be more simple.

---
  mm/memory_hotplug.c | 5 +----
  1 file changed, 1 insertion(+), 4 deletions(-)
mm/memory_hotplug.c +1 -2
index bcb0dc41c2f2..6c615ba1a5c7 100644
@@ -1620,8 +1620,7 @@ static int scan_movable_pages(unsigned long start, unsigned long end,
  	return 0;
  }
 
-static int
-do_migrate_range(unsigned long start_pfn, unsigned long end_pfn)
+static void do_migrate_range(unsigned long start_pfn, unsigned long end_pfn)
  {
  	unsigned long pfn;
  	struct page *page, *head;

mm/memory_hotplug.c do_migrate_range(unsigned long start_pfn, unsigned long end_pfn) · line 1725

Quoted 5 lines
mm/memory_hotplug.c +0 -2
@@ -1721,8 +1720,6 @@ do_migrate_range(unsigned long start_pfn, unsigned long end_pfn)
  			putback_movable_pages(&source);
  		}
  	}
-
-	return ret;

Nit: while we are at it, can we also remove the "TODO" comment in offline_pages()?

ret = scan_movable_pages(pfn, end_pfn, &pfn);

if (!ret) {
	/*
	 * TODO: fatal migration failures should bail
	 * out
	 */
	do_migrate_range(pfn, end_pfn);
}

With David's comments:

#3 lore raw

Re: [PATCH v2 1/2] mm/memory_hotplug: return nothing from do_migrate_range()

In reply to David Hildenbrand

Hi David,

Quoted 19 lines · On Thu, 16 Feb 2023 10:46:59 +0100 David Hildenbrand <[email protected]> wrote:
Quoted 18 lines · SeongJae Park

Return value mechanism of do_migrate_range() is not very simple, while no caller of the function checks the return value. Make the function return nothing to be more simple.

---
  mm/memory_hotplug.c | 5 +----
  1 file changed, 1 insertion(+), 4 deletions(-)
mm/memory_hotplug.c +0 -0
index bcb0dc41c2f2..6c615ba1a5c7 100644
@@ -1620,8 +1620,7 @@ static int scan_movable_pages(unsigned long start, unsigned long end,
  	return 0;
  }
 
Quoted 16 lines

mm/memory_hotplug.c do_migrate_range(unsigned long start_pfn, unsigned long end_pfn) · line 1723

Quoted 13 lines
mm/memory_hotplug.c +1 -4
-static int
-do_migrate_range(unsigned long start_pfn, unsigned long end_pfn)
+static void do_migrate_range(unsigned long start_pfn, unsigned long end_pfn)
  {
  	unsigned long pfn;
  	struct page *page, *head;
@@ -1721,8 +1720,6 @@ do_migrate_range(unsigned long start_pfn, unsigned long end_pfn)
  			putback_movable_pages(&source);
  		}
  	}
-
-	return ret;
  }

I think this patch should also stop initializing ret = 0 inside that function.

The second patch does it, and I read your reply to the patch asking squashing it in this one. I will send the new version soon.

Quoted 4 lines

With that

Thanks you.

Thanks,
SJ

Quoted 7 lines
Signature · 4 lines
-- 
Thanks,

David / dhildenb
#5 lore raw

Re: [PATCH v2 1/2] mm/memory_hotplug: return nothing from do_migrate_range()

In reply to Baolin Wang

Hi Baolin,

Quoted 46 lines · On Thu, 16 Feb 2023 18:53:34 +0800 Baolin Wang <[email protected]>…
Quoted 25 lines · SeongJae Park

Return value mechanism of do_migrate_range() is not very simple, while no caller of the function checks the return value. Make the function return nothing to be more simple.

---
  mm/memory_hotplug.c | 5 +----
  1 file changed, 1 insertion(+), 4 deletions(-)
mm/memory_hotplug.c +1 -2
index bcb0dc41c2f2..6c615ba1a5c7 100644
@@ -1620,8 +1620,7 @@ static int scan_movable_pages(unsigned long start, unsigned long end,
  	return 0;
  }
 
-static int
-do_migrate_range(unsigned long start_pfn, unsigned long end_pfn)
+static void do_migrate_range(unsigned long start_pfn, unsigned long end_pfn)
  {
  	unsigned long pfn;
  	struct page *page, *head;

mm/memory_hotplug.c do_migrate_range(unsigned long start_pfn, unsigned long end_pfn) · line 1725

Quoted 5 lines
mm/memory_hotplug.c +0 -2
@@ -1721,8 +1720,6 @@ do_migrate_range(unsigned long start_pfn, unsigned long end_pfn)
  			putback_movable_pages(&source);
  		}
  	}
-
-	return ret;

Nit: while we are at it, can we also remove the "TODO" comment in offline_pages()?

ret = scan_movable_pages(pfn, end_pfn, &pfn);

if (!ret) {
	/*
	 * TODO: fatal migration failures should bail
	 * out
	 */
	do_migrate_range(pfn, end_pfn);
}
Quoted 2 lines

With David's comments:

As David mentioned we're just still unclear what issues are fatal[1], I was thinking we didn't forgive the future work, so left the comment. If anyone has different opinion or I'm getting something wrong, please let me know.

[1] https://lore.kernel.org/linux-mm/[email protected]/

Thanks,
SJ