diff mbox series

[-V2,5/7] migrate_pages(): fix failure counting for THP on -ENOSYS

Message ID 20220711084948.274787-6-ying.huang@intel.com (mailing list archive)
State New
Headers show
Series [-V2,1/7] migrate: fix syscall move_pages() return value for failure | expand

Commit Message

Huang, Ying July 11, 2022, 8:49 a.m. UTC 
If THP or hugetlbfs page migration isn't supported, unmap_and_move()
or unmap_and_move_huge_page() will return -ENOSYS.  For THP, splitting
will be tried, but if splitting doesn't succeed, the THP will be left
in "from" list wrongly.  If some other pages are retried, the THP
migration failure will counted again.  This is fixed via moving the
failure THP from "from" to "ret_pages".

Another issue of the original code is that the unsupported failure
processing isn't consistent between THP and hugetlbfs page.  Make them
consistent in this patch to make the code easier to be understood too.

Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Fixes: 5984fabb6e82 ("mm: move_pages: report the number of non-attempted pages")
Reviewed-by: Baolin Wang <baolin.wang@linux.alibaba.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Yang Shi <shy828301@gmail.com>
---
 mm/migrate.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

Comments

Oscar Salvador July 11, 2022, 12:26 p.m. UTC | #1 
On Mon, Jul 11, 2022 at 04:49:46PM +0800, Huang Ying wrote:
> If THP or hugetlbfs page migration isn't supported, unmap_and_move()
> or unmap_and_move_huge_page() will return -ENOSYS.  For THP, splitting
> will be tried, but if splitting doesn't succeed, the THP will be left
> in "from" list wrongly.  If some other pages are retried, the THP
> migration failure will counted again.  This is fixed via moving the
> failure THP from "from" to "ret_pages".
> 
> Another issue of the original code is that the unsupported failure
> processing isn't consistent between THP and hugetlbfs page.  Make them
> consistent in this patch to make the code easier to be understood too.
> 
> Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
> Fixes: 5984fabb6e82 ("mm: move_pages: report the number of non-attempted pages")
> Reviewed-by: Baolin Wang <baolin.wang@linux.alibaba.com>
> Cc: Zi Yan <ziy@nvidia.com>
> Cc: Yang Shi <shy828301@gmail.com>
> ---
>  mm/migrate.c | 6 +++---
>  1 file changed, 3 insertions(+), 3 deletions(-)
> 
> diff --git a/mm/migrate.c b/mm/migrate.c
> index 4bceba143db0..8cce73b7c046 100644
> --- a/mm/migrate.c
> +++ b/mm/migrate.c
> @@ -1192,10 +1192,8 @@ static int unmap_and_move_huge_page(new_page_t get_new_page,
>  	 * tables or check whether the hugepage is pmd-based or not before
>  	 * kicking migration.
>  	 */
> -	if (!hugepage_migration_supported(page_hstate(hpage))) {
> -		list_move_tail(&hpage->lru, ret);
> +	if (!hugepage_migration_supported(page_hstate(hpage)))
>  		return -ENOSYS;
> -	}
>  
>  	if (page_count(hpage) == 1) {
>  		/* page was freed from under us. So we are done. */
> @@ -1392,6 +1390,7 @@ int migrate_pages(struct list_head *from, new_page_t get_new_page,
>  			 *		 page will be put back
>  			 *	-EAGAIN: stay on the from list
>  			 *	-ENOMEM: stay on the from list
> +			 *	-ENOSYS: stay on the from list
>  			 *	Other errno: put on ret_pages list then splice to
>  			 *		     from list
>  			 */
> @@ -1421,6 +1420,7 @@ int migrate_pages(struct list_head *from, new_page_t get_new_page,
>  				}
>  
>  				nr_failed_pages += nr_subpages;
> +				list_move_tail(&page->lru, &ret_pages);

I must be missing something, but migrate_pages() calls unmap_and_move_huge_page() 
with ret being ret_pages, so

list_move_tail(&hpage->lru, ret) == list_move_tail(&page->lru, &ret_pages)

Yet, you say "This is fixed via moving the failure THP from "from" to "ret_pages"".
/me confused.
Huang, Ying July 13, 2022, 8:30 a.m. UTC | #2 
Oscar Salvador <osalvador@suse.de> writes:

> On Mon, Jul 11, 2022 at 04:49:46PM +0800, Huang Ying wrote:
>> If THP or hugetlbfs page migration isn't supported, unmap_and_move()
>> or unmap_and_move_huge_page() will return -ENOSYS.  For THP, splitting
>> will be tried, but if splitting doesn't succeed, the THP will be left
>> in "from" list wrongly.  If some other pages are retried, the THP
>> migration failure will counted again.  This is fixed via moving the
>> failure THP from "from" to "ret_pages".
>> 
>> Another issue of the original code is that the unsupported failure
>> processing isn't consistent between THP and hugetlbfs page.  Make them
>> consistent in this patch to make the code easier to be understood too.
>> 
>> Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
>> Fixes: 5984fabb6e82 ("mm: move_pages: report the number of non-attempted pages")
>> Reviewed-by: Baolin Wang <baolin.wang@linux.alibaba.com>
>> Cc: Zi Yan <ziy@nvidia.com>
>> Cc: Yang Shi <shy828301@gmail.com>
>> ---
>>  mm/migrate.c | 6 +++---
>>  1 file changed, 3 insertions(+), 3 deletions(-)
>> 
>> diff --git a/mm/migrate.c b/mm/migrate.c
>> index 4bceba143db0..8cce73b7c046 100644
>> --- a/mm/migrate.c
>> +++ b/mm/migrate.c
>> @@ -1192,10 +1192,8 @@ static int unmap_and_move_huge_page(new_page_t get_new_page,
>>  	 * tables or check whether the hugepage is pmd-based or not before
>>  	 * kicking migration.
>>  	 */
>> -	if (!hugepage_migration_supported(page_hstate(hpage))) {
>> -		list_move_tail(&hpage->lru, ret);
>> +	if (!hugepage_migration_supported(page_hstate(hpage)))
>>  		return -ENOSYS;
>> -	}
>>  
>>  	if (page_count(hpage) == 1) {
>>  		/* page was freed from under us. So we are done. */
>> @@ -1392,6 +1390,7 @@ int migrate_pages(struct list_head *from, new_page_t get_new_page,
>>  			 *		 page will be put back
>>  			 *	-EAGAIN: stay on the from list
>>  			 *	-ENOMEM: stay on the from list
>> +			 *	-ENOSYS: stay on the from list
>>  			 *	Other errno: put on ret_pages list then splice to
>>  			 *		     from list
>>  			 */
>> @@ -1421,6 +1420,7 @@ int migrate_pages(struct list_head *from, new_page_t get_new_page,
>>  				}
>>  
>>  				nr_failed_pages += nr_subpages;
>> +				list_move_tail(&page->lru, &ret_pages);
>
> I must be missing something, but migrate_pages() calls unmap_and_move_huge_page() 
> with ret being ret_pages, so
>
> list_move_tail(&hpage->lru, ret) == list_move_tail(&page->lru, &ret_pages)
>
> Yet, you say "This is fixed via moving the failure THP from "from" to "ret_pages"".
> /me confused.

To make it consistent between hugetlb page and THP/normal page, I have
revised the unmap_and_move_huge_page() via deleting the list_move_tail()
there for ENOSYS.  After that, we move the hugetlb page and THP/normal
page in the same way in migrate_pages().

Best Regards,
Huang, Ying
diff mbox series

Patch

diff --git a/mm/migrate.c b/mm/migrate.c
index 4bceba143db0..8cce73b7c046 100644
--- a/mm/migrate.c
+++ b/mm/migrate.c
@@ -1192,10 +1192,8 @@  static int unmap_and_move_huge_page(new_page_t get_new_page,
 	 * tables or check whether the hugepage is pmd-based or not before
 	 * kicking migration.
 	 */
-	if (!hugepage_migration_supported(page_hstate(hpage))) {
-		list_move_tail(&hpage->lru, ret);
+	if (!hugepage_migration_supported(page_hstate(hpage)))
 		return -ENOSYS;
-	}
 
 	if (page_count(hpage) == 1) {
 		/* page was freed from under us. So we are done. */
@@ -1392,6 +1390,7 @@  int migrate_pages(struct list_head *from, new_page_t get_new_page,
 			 *		 page will be put back
 			 *	-EAGAIN: stay on the from list
 			 *	-ENOMEM: stay on the from list
+			 *	-ENOSYS: stay on the from list
 			 *	Other errno: put on ret_pages list then splice to
 			 *		     from list
 			 */
@@ -1421,6 +1420,7 @@  int migrate_pages(struct list_head *from, new_page_t get_new_page,
 				}
 
 				nr_failed_pages += nr_subpages;
+				list_move_tail(&page->lru, &ret_pages);
 				break;
 			case -ENOMEM:
 				/*