Skip to contentSkip to conversation

mm,hwpoison: Return -EBUSY when migration fails

PATCH Applied Oscar Salvador 4 participants 6 messages Dec 9, 2020

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

PATCH 1/1 mm,hwpoison: Return -EBUSY when migration fails

Currently, we return -EIO when we fail to migrate the page.

Migrations' failures are rather transient as they can happen due to several reasons, e.g: high page refcount bump, mapping->migrate_page failing etc.
All meaning that at that time the page could not be migrated, but that has nothing to do with an EIO error.

Let us return -EBUSY instead, as we do in case we failed to isolate the page.

While are it, let us remove the "ret" print as its value does not change.

---
 mm/memory-failure.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
mm/memory-failure.c +3 -3
index 428991e297e2..1942fb83ac64 100644
@@ -1849,11 +1849,11 @@ static int __soft_offline_page(struct page *page)
 			pr_info("soft offline: %#lx: %s migration failed %d, type %lx (%pGp)\n",
 				pfn, msg_page[huge], ret, page->flags, &page->flags);
 			if (ret > 0)
-				ret = -EIO;
+				ret = -EBUSY;
 		}
 	} else {
-		pr_info("soft offline: %#lx: %s isolation failed: %d, page count %d, type %lx (%pGp)\n",
-			pfn, msg_page[huge], ret, page_count(page), page->flags, &page->flags);
+		pr_info("soft offline: %#lx: %s isolation failed, page count %d, type %lx (%pGp)\n",
+			pfn, msg_page[huge], page_count(page), page->flags, &page->flags);
 		ret = -EBUSY;
 	}
 	return ret;
Signature · 1 line
2.26.2
#1 lore raw
Quoted 8 lines · Oscar Salvador

Currently, we return -EIO when we fail to migrate the page.

Migrations' failures are rather transient as they can happen due to several reasons, e.g: high page refcount bump, mapping->migrate_page failing etc.
All meaning that at that time the page could not be migrated, but that has nothing to do with an EIO error.

Quoted 6 lines

Let us return -EBUSY instead, as we do in case we failed to isolate the page.

While are it, let us remove the "ret" print as its value does not change.

#2 lore raw
In reply to Oscar Salvador
Quoted 23 lines · Oscar Salvador

Currently, we return -EIO when we fail to migrate the page.

Migrations' failures are rather transient as they can happen due to several reasons, e.g: high page refcount bump, mapping->migrate_page failing etc.
All meaning that at that time the page could not be migrated, but that has nothing to do with an EIO error.

Let us return -EBUSY instead, as we do in case we failed to isolate the page.

While are it, let us remove the "ret" print as its value does not change.

---
 mm/memory-failure.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
mm/memory-failure.c +0 -0
index 428991e297e2..1942fb83ac64 100644

mm/memory-failure.c static int __soft_offline_page(struct page *page) · line 1852

Quoted 5 lines
mm/memory-failure.c +1 -1
@@ -1849,11 +1849,11 @@ static int __soft_offline_page(struct page *page)
 			pr_info("soft offline: %#lx: %s migration failed %d, type %lx (%pGp)\n",
 				pfn, msg_page[huge], ret, page->flags, &page->flags);
 			if (ret > 0)
-				ret = -EIO;
+				ret = -EBUSY;

Do we expect callers to retry immediately? -EAGAIN might make also sense. But -EBUSY is an obvious improvement. Do we have callers relying on this behavior?

Signature · 3 lines
Thanks,

David / dhildenb
#3 lore raw
Quoted 4 lines · David Hildenbrand

On 09.12.20 10:28, Oscar Salvador wrote:
Do we expect callers to retry immediately? -EAGAIN might make also sense. But -EBUSY is an obvious improvement. Do we have callers relying on this behavior?

Not really, unless something LTP takes a look at the error code in retries in case EBUSY.

Take into account that most of the callers do not even really check the return code (GHES, RAS/CEC, etc.)

Signature · 2 lines
Oscar Salvador
SUSE L3
Acked-by on v1 1/1
#4 lore raw
In reply to Oscar Salvador
Quoted 8 lines · Oscar Salvador

Currently, we return -EIO when we fail to migrate the page.

Migrations' failures are rather transient as they can happen due to several reasons, e.g: high page refcount bump, mapping->migrate_page failing etc.
All meaning that at that time the page could not be migrated, but that has nothing to do with an EIO error.

Quoted 6 lines

Let us return -EBUSY instead, as we do in case we failed to isolate the page.

While are it, let us remove the "ret" print as its value does not change.

Technically this affects madvise(2) so let's cc linux-api. The manpage doesn't document error codes of MADV_HWPOISON and MADV_SOFT_OFFLINE (besides EPERM) though so nothing to adjust there. It's meant only for the hwpoison testing suite anyway.

Quoted 24 lines
---
 mm/memory-failure.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
mm/memory-failure.c +3 -3
index 428991e297e2..1942fb83ac64 100644
@@ -1849,11 +1849,11 @@ static int __soft_offline_page(struct page *page)
 			pr_info("soft offline: %#lx: %s migration failed %d, type %lx (%pGp)\n",
 				pfn, msg_page[huge], ret, page->flags, &page->flags);
 			if (ret > 0)
-				ret = -EIO;
+				ret = -EBUSY;
 		}
 	} else {
-		pr_info("soft offline: %#lx: %s isolation failed: %d, page count %d, type %lx (%pGp)\n",
-			pfn, msg_page[huge], ret, page_count(page), page->flags, &page->flags);
+		pr_info("soft offline: %#lx: %s isolation failed, page count %d, type %lx (%pGp)\n",
+			pfn, msg_page[huge], page_count(page), page->flags, &page->flags);
 		ret = -EBUSY;
 	}
 	return ret;
#5 lore raw
Quoted 14 lines · Vlastimil Babka
Quoted 13 lines · Oscar Salvador

Currently, we return -EIO when we fail to migrate the page.

Migrations' failures are rather transient as they can happen due to several reasons, e.g: high page refcount bump, mapping->migrate_page failing etc.
All meaning that at that time the page could not be migrated, but that has nothing to do with an EIO error.

Let us return -EBUSY instead, as we do in case we failed to isolate the page.

While are it, let us remove the "ret" print as its value does not change.

Quoted 8 lines
Quoted 1 line

Technically this affects madvise(2) so let's cc linux-api. The manpage doesn't document error codes of MADV_HWPOISON and MADV_SOFT_OFFLINE (besides EPERM) though so nothing to adjust there. It's meant only for the hwpoison testing suite anyway.

Well, not only for hwpoison testing suite.

RAS/CEC and GHES also use soft_offline_page by means of memory_failure_queue in case a page cec count goes beyond a certain thereshold, but they do not really check the return code.

Only madvise does.

Signature · 2 lines
Oscar Salvador
SUSE L3