diff mbox series

[v5,4/9] mm/migrate: clear __GFP_RECLAIM to make the migration callback consistent with regular THP allocations

Message ID 1594622517-20681-5-git-send-email-iamjoonsoo.kim@lge.com (mailing list archive)
State New, archived
Headers show
Series clean-up the migration target allocation functions | expand

Commit Message

Joonsoo Kim July 13, 2020, 6:41 a.m. UTC 
From: Joonsoo Kim <iamjoonsoo.kim@lge.com>

new_page_nodemask is a migration callback and it tries to use a common
gfp flags for the target page allocation whether it is a base page or a
THP. The later only adds GFP_TRANSHUGE to the given mask. This results
in the allocation being slightly more aggressive than necessary because
the resulting gfp mask will contain also __GFP_RECLAIM_KSWAPD. THP
allocations usually exclude this flag to reduce over eager background
reclaim during a high THP allocation load which has been seen during
large mmaps initialization. There is no indication that this is a
problem for migration as well but theoretically the same might happen
when migrating large mappings to a different node. Make the migration
callback consistent with regular THP allocations.

Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
---
 mm/migrate.c | 5 +++++
 1 file changed, 5 insertions(+)

Comments

Michal Hocko July 13, 2020, 7:36 a.m. UTC | #1 
On Mon 13-07-20 15:41:52, Joonsoo Kim wrote:
> From: Joonsoo Kim <iamjoonsoo.kim@lge.com>
> 
> new_page_nodemask is a migration callback and it tries to use a common
> gfp flags for the target page allocation whether it is a base page or a
> THP. The later only adds GFP_TRANSHUGE to the given mask. This results
> in the allocation being slightly more aggressive than necessary because
> the resulting gfp mask will contain also __GFP_RECLAIM_KSWAPD. THP
> allocations usually exclude this flag to reduce over eager background
> reclaim during a high THP allocation load which has been seen during
> large mmaps initialization. There is no indication that this is a
> problem for migration as well but theoretically the same might happen
> when migrating large mappings to a different node. Make the migration
> callback consistent with regular THP allocations.
> 
> Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>

Acked-by: Michal Hocko <mhocko@suse.com>

> ---
>  mm/migrate.c | 5 +++++
>  1 file changed, 5 insertions(+)
> 
> diff --git a/mm/migrate.c b/mm/migrate.c
> index 3b3d918..1cfc965 100644
> --- a/mm/migrate.c
> +++ b/mm/migrate.c
> @@ -1547,6 +1547,11 @@ struct page *new_page_nodemask(struct page *page,
>  	}
>  
>  	if (PageTransHuge(page)) {
> +		/*
> +		 * clear __GFP_RECALIM to make the migration callback
> +		 * consistent with regular THP allocations.
> +		 */
> +		gfp_mask &= ~__GFP_RECLAIM;
>  		gfp_mask |= GFP_TRANSHUGE;
>  		order = HPAGE_PMD_ORDER;
>  	}
> -- 
> 2.7.4
>
Vlastimil Babka July 13, 2020, 7:52 a.m. UTC | #2 
On 7/13/20 8:41 AM, js1304@gmail.com wrote:
> From: Joonsoo Kim <iamjoonsoo.kim@lge.com>
> 
> new_page_nodemask is a migration callback and it tries to use a common
> gfp flags for the target page allocation whether it is a base page or a
> THP. The later only adds GFP_TRANSHUGE to the given mask. This results
> in the allocation being slightly more aggressive than necessary because
> the resulting gfp mask will contain also __GFP_RECLAIM_KSWAPD. THP
> allocations usually exclude this flag to reduce over eager background
> reclaim during a high THP allocation load which has been seen during
> large mmaps initialization. There is no indication that this is a
> problem for migration as well but theoretically the same might happen
> when migrating large mappings to a different node. Make the migration
> callback consistent with regular THP allocations.
> 
> Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>

Acked-by: Vlastimil Babka <vbabka@suse.cz>

Thanks!

Typo below (I assume Andrew will fix it)

> ---
>  mm/migrate.c | 5 +++++
>  1 file changed, 5 insertions(+)
> 
> diff --git a/mm/migrate.c b/mm/migrate.c
> index 3b3d918..1cfc965 100644
> --- a/mm/migrate.c
> +++ b/mm/migrate.c
> @@ -1547,6 +1547,11 @@ struct page *new_page_nodemask(struct page *page,
>  	}
>  
>  	if (PageTransHuge(page)) {
> +		/*
> +		 * clear __GFP_RECALIM to make the migration callback

                         __GFP_RECLAIM

> +		 * consistent with regular THP allocations.
> +		 */
> +		gfp_mask &= ~__GFP_RECLAIM;
>  		gfp_mask |= GFP_TRANSHUGE;
>  		order = HPAGE_PMD_ORDER;
>  	}
>
Joonsoo Kim July 14, 2020, 12:44 a.m. UTC | #3 
On Mon, Jul 13, 2020 at 09:52:20AM +0200, Vlastimil Babka wrote:
> On 7/13/20 8:41 AM, js1304@gmail.com wrote:
> > From: Joonsoo Kim <iamjoonsoo.kim@lge.com>
> > 
> > new_page_nodemask is a migration callback and it tries to use a common
> > gfp flags for the target page allocation whether it is a base page or a
> > THP. The later only adds GFP_TRANSHUGE to the given mask. This results
> > in the allocation being slightly more aggressive than necessary because
> > the resulting gfp mask will contain also __GFP_RECLAIM_KSWAPD. THP
> > allocations usually exclude this flag to reduce over eager background
> > reclaim during a high THP allocation load which has been seen during
> > large mmaps initialization. There is no indication that this is a
> > problem for migration as well but theoretically the same might happen
> > when migrating large mappings to a different node. Make the migration
> > callback consistent with regular THP allocations.
> > 
> > Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
> 
> Acked-by: Vlastimil Babka <vbabka@suse.cz>
> 
> Thanks!
> 
> Typo below (I assume Andrew will fix it)
> 
> > ---
> >  mm/migrate.c | 5 +++++
> >  1 file changed, 5 insertions(+)
> > 
> > diff --git a/mm/migrate.c b/mm/migrate.c
> > index 3b3d918..1cfc965 100644
> > --- a/mm/migrate.c
> > +++ b/mm/migrate.c
> > @@ -1547,6 +1547,11 @@ struct page *new_page_nodemask(struct page *page,
> >  	}
> >  
> >  	if (PageTransHuge(page)) {
> > +		/*
> > +		 * clear __GFP_RECALIM to make the migration callback
> 
>                          __GFP_RECLAIM
> 

Okay. Here goes a fixed version.

Thanks!


---------------------->8-----------------------------
From 6273f02fd8b8ef066c10c4a8ba54ea9efe6e70cd Mon Sep 17 00:00:00 2001
From: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Date: Mon, 6 Jul 2020 14:34:04 +0900
Subject: [PATCH v5 4/9] mm/migrate: clear __GFP_RECLAIM to make the migration
 callback consistent with regular THP allocations

new_page_nodemask is a migration callback and it tries to use a common
gfp flags for the target page allocation whether it is a base page or a
THP. The later only adds GFP_TRANSHUGE to the given mask. This results
in the allocation being slightly more aggressive than necessary because
the resulting gfp mask will contain also __GFP_RECLAIM_KSWAPD. THP
allocations usually exclude this flag to reduce over eager background
reclaim during a high THP allocation load which has been seen during
large mmaps initialization. There is no indication that this is a
problem for migration as well but theoretically the same might happen
when migrating large mappings to a different node. Make the migration
callback consistent with regular THP allocations.

Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
---
 mm/migrate.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/mm/migrate.c b/mm/migrate.c
index 3b3d918..faabb2e 100644
--- a/mm/migrate.c
+++ b/mm/migrate.c
@@ -1547,6 +1547,11 @@ struct page *new_page_nodemask(struct page *page,
 	}
 
 	if (PageTransHuge(page)) {
+		/*
+		 * clear __GFP_RECLAIM to make the migration callback
+		 * consistent with regular THP allocations.
+		 */
+		gfp_mask &= ~__GFP_RECLAIM;
 		gfp_mask |= GFP_TRANSHUGE;
 		order = HPAGE_PMD_ORDER;
 	}
diff mbox series

Patch

diff --git a/mm/migrate.c b/mm/migrate.c
index 3b3d918..1cfc965 100644
--- a/mm/migrate.c
+++ b/mm/migrate.c
@@ -1547,6 +1547,11 @@  struct page *new_page_nodemask(struct page *page,
 	}
 
 	if (PageTransHuge(page)) {
+		/*
+		 * clear __GFP_RECALIM to make the migration callback
+		 * consistent with regular THP allocations.
+		 */
+		gfp_mask &= ~__GFP_RECLAIM;
 		gfp_mask |= GFP_TRANSHUGE;
 		order = HPAGE_PMD_ORDER;
 	}