[PATCH] mm/vmscan: reclaim every zone in per-node proactive reclaim

[PATCH] mm/vmscan: reclaim every zone in per-node proactive reclaim

[Gregory Price (Meta)]

user_proactive_reclaim() limits per-node reclaim to
gfp_zone(GFP_KERNEL), which is ZONE_NORMAL. Proactive reclaim serves no
allocation, but that limit makes shrink_node() skip every ZONE_MOVABLE
folio, so writing to /sys/devices/system/node/nodeN/reclaim never
reclaims a node's movable memory. A movable-only node, such as dax/kmem
or hotplugged memory onlined movable, reclaims nothing and fails with
-EAGAIN.

To reproduce, fill a movable-only node with clean page cache and write
200M to its nodeN/reclaim: the write fails with -EAGAIN and the page
cache stays. The same write to a ZONE_NORMAL node frees 200MB.

Scan every zone, as memory.reclaim does.

Fixes: b980077899ea4 ("mm: introduce per-node proactive reclaim interface")
Cc: stable@vger.kernel.org
Signed-off-by: Gregory Price (Meta) <gourry@gourry.net>
---
 mm/vmscan.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mm/vmscan.c b/mm/vmscan.c
index e82ba737c028d..90f7a7bf4e426 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -8180,7 +8180,7 @@ int user_proactive_reclaim(char *buf,
 		} else {
 			struct scan_control sc = {
 				.gfp_mask = current_gfp_context(gfp_mask),
-				.reclaim_idx = gfp_zone(gfp_mask),
+				.reclaim_idx = MAX_NR_ZONES - 1,
 				.proactive_swappiness = swappiness == -1 ? NULL : &swappiness,
 				.priority = DEF_PRIORITY,
 				.may_writepage = 1,
-- 
2.53.0-Meta

Re: [PATCH] mm/vmscan: reclaim every zone in per-node proactive reclaim

[Barry Song]

On Tue, Oct 20, 2026 at 2:50 AM Gregory Price (Meta) <gourry@gourry.net> wrote:
>
> user_proactive_reclaim() limits per-node reclaim to
> gfp_zone(GFP_KERNEL), which is ZONE_NORMAL. Proactive reclaim serves no
> allocation, but that limit makes shrink_node() skip every ZONE_MOVABLE
> folio, so writing to /sys/devices/system/node/nodeN/reclaim never
> reclaims a node's movable memory. A movable-only node, such as dax/kmem
> or hotplugged memory onlined movable, reclaims nothing and fails with
> -EAGAIN.
>
> To reproduce, fill a movable-only node with clean page cache and write
> 200M to its nodeN/reclaim: the write fails with -EAGAIN and the page
> cache stays. The same write to a ZONE_NORMAL node frees 200MB.
>
> Scan every zone, as memory.reclaim does.
>
> Fixes: b980077899ea4 ("mm: introduce per-node proactive reclaim interface")
> Cc: stable@vger.kernel.org
> Signed-off-by: Gregory Price (Meta) <gourry@gourry.net>
> ---

Makes a lot of sense to me.

Reviewed-by: Barry Song <baohua@kernel.org>

Re: [PATCH] mm/vmscan: reclaim every zone in per-node proactive reclaim

[Johannes Weiner]

On Mon, Oct 19, 2026 at 02:50:02PM -0400, Gregory Price (Meta) wrote:
> user_proactive_reclaim() limits per-node reclaim to
> gfp_zone(GFP_KERNEL), which is ZONE_NORMAL. Proactive reclaim serves no
> allocation, but that limit makes shrink_node() skip every ZONE_MOVABLE
> folio, so writing to /sys/devices/system/node/nodeN/reclaim never
> reclaims a node's movable memory. A movable-only node, such as dax/kmem
> or hotplugged memory onlined movable, reclaims nothing and fails with
> -EAGAIN.
> 
> To reproduce, fill a movable-only node with clean page cache and write
> 200M to its nodeN/reclaim: the write fails with -EAGAIN and the page
> cache stays. The same write to a ZONE_NORMAL node frees 200MB.
> 
> Scan every zone, as memory.reclaim does.
> 
> Fixes: b980077899ea4 ("mm: introduce per-node proactive reclaim interface")
> Cc: stable@vger.kernel.org
> Signed-off-by: Gregory Price (Meta) <gourry@gourry.net>

Reviewed-by: Johannes Weiner <hannes@cmpxchg.org>

Re: [PATCH] mm/vmscan: reclaim every zone in per-node proactive reclaim

[Harry Yoo]

On Mon, Oct 19, 2026 at 02:50:02PM -0400, Gregory Price (Meta) wrote:
> user_proactive_reclaim() limits per-node reclaim to
> gfp_zone(GFP_KERNEL), which is ZONE_NORMAL. Proactive reclaim serves no
> allocation, but that limit makes shrink_node() skip every ZONE_MOVABLE
> folio, so writing to /sys/devices/system/node/nodeN/reclaim never
> reclaims a node's movable memory. A movable-only node, such as dax/kmem
> or hotplugged memory onlined movable, reclaims nothing and fails with
> -EAGAIN.
> 
> To reproduce, fill a movable-only node with clean page cache and write
> 200M to its nodeN/reclaim: the write fails with -EAGAIN and the page
> cache stays. The same write to a ZONE_NORMAL node frees 200MB.
> 
> Scan every zone, as memory.reclaim does.
> 
> Fixes: b980077899ea4 ("mm: introduce per-node proactive reclaim interface")
> Cc: stable@vger.kernel.org
> Signed-off-by: Gregory Price (Meta) <gourry@gourry.net>
> ---

Looks good to me,
Reviewed-by: Harry Yoo (Meta) <harry@kernel.org>

-- 
Cheers,
Harry / Hyeonggon

Re: [PATCH] mm/vmscan: reclaim every zone in per-node proactive reclaim

[Shakeel Butt]

On Mon, Oct 19, 2026 at 02:50:02PM -0400, Gregory Price (Meta) wrote:
> user_proactive_reclaim() limits per-node reclaim to
> gfp_zone(GFP_KERNEL), which is ZONE_NORMAL. Proactive reclaim serves no
> allocation, but that limit makes shrink_node() skip every ZONE_MOVABLE
> folio, so writing to /sys/devices/system/node/nodeN/reclaim never
> reclaims a node's movable memory. A movable-only node, such as dax/kmem
> or hotplugged memory onlined movable, reclaims nothing and fails with
> -EAGAIN.
> 
> To reproduce, fill a movable-only node with clean page cache and write
> 200M to its nodeN/reclaim: the write fails with -EAGAIN and the page
> cache stays. The same write to a ZONE_NORMAL node frees 200MB.
> 
> Scan every zone, as memory.reclaim does.
> 
> Fixes: b980077899ea4 ("mm: introduce per-node proactive reclaim interface")
> Cc: stable@vger.kernel.org
> Signed-off-by: Gregory Price (Meta) <gourry@gourry.net>

Acked-by: Shakeel Butt <shakeel.butt@linux.dev>

Re: [PATCH] mm/vmscan: reclaim every zone in per-node proactive reclaim

[Yosry Ahmed]

On Mon, Oct 19, 2026 at 02:50:02PM -0400, Gregory Price (Meta) wrote:
> user_proactive_reclaim() limits per-node reclaim to
> gfp_zone(GFP_KERNEL), which is ZONE_NORMAL. Proactive reclaim serves no
> allocation, but that limit makes shrink_node() skip every ZONE_MOVABLE
> folio, so writing to /sys/devices/system/node/nodeN/reclaim never
> reclaims a node's movable memory. A movable-only node, such as dax/kmem
> or hotplugged memory onlined movable, reclaims nothing and fails with
> -EAGAIN.
> 
> To reproduce, fill a movable-only node with clean page cache and write
> 200M to its nodeN/reclaim: the write fails with -EAGAIN and the page
> cache stays. The same write to a ZONE_NORMAL node frees 200MB.
> 
> Scan every zone, as memory.reclaim does.
> 
> Fixes: b980077899ea4 ("mm: introduce per-node proactive reclaim interface")
> Cc: stable@vger.kernel.org
> Signed-off-by: Gregory Price (Meta) <gourry@gourry.net>

Reviewed-by: Yosry Ahmed <yosry@kernel.org>

Re: [PATCH] mm/vmscan: reclaim every zone in per-node proactive reclaim

[SJ Park]

On Mon, 19 Oct 2026 14:50:02 -0400 "Gregory Price (Meta)" <gourry@gourry.net> wrote:

> user_proactive_reclaim() limits per-node reclaim to
> gfp_zone(GFP_KERNEL), which is ZONE_NORMAL. Proactive reclaim serves no
> allocation, but that limit makes shrink_node() skip every ZONE_MOVABLE
> folio, so writing to /sys/devices/system/node/nodeN/reclaim never
> reclaims a node's movable memory. A movable-only node, such as dax/kmem
> or hotplugged memory onlined movable, reclaims nothing and fails with
> -EAGAIN.
> 
> To reproduce, fill a movable-only node with clean page cache and write
> 200M to its nodeN/reclaim: the write fails with -EAGAIN and the page
> cache stays. The same write to a ZONE_NORMAL node frees 200MB.
> 
> Scan every zone, as memory.reclaim does.

Both the issue description and the fix make sense to me.

> 
> Fixes: b980077899ea4 ("mm: introduce per-node proactive reclaim interface")
> Cc: stable@vger.kernel.org
> Signed-off-by: Gregory Price (Meta) <gourry@gourry.net>

Reviewed-by: SJ Park <sj@kernel.org>


Thanks,
SJ

[...]

Re: [PATCH] mm/vmscan: reclaim every zone in per-node proactive reclaim

[Balbir Singh]

On 10/20/26 5:50 AM, Gregory Price (Meta) wrote:
> user_proactive_reclaim() limits per-node reclaim to
> gfp_zone(GFP_KERNEL), which is ZONE_NORMAL. Proactive reclaim serves no
> allocation, but that limit makes shrink_node() skip every ZONE_MOVABLE
> folio, so writing to /sys/devices/system/node/nodeN/reclaim never
> reclaims a node's movable memory. A movable-only node, such as dax/kmem
> or hotplugged memory onlined movable, reclaims nothing and fails with
> -EAGAIN.
> 
> To reproduce, fill a movable-only node with clean page cache and write
> 200M to its nodeN/reclaim: the write fails with -EAGAIN and the page
> cache stays. The same write to a ZONE_NORMAL node frees 200MB.
> 
> Scan every zone, as memory.reclaim does.
> 
> Fixes: b980077899ea4 ("mm: introduce per-node proactive reclaim interface")
> Cc: stable@vger.kernel.org
> Signed-off-by: Gregory Price (Meta) <gourry@gourry.net>
> ---
>  mm/vmscan.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/mm/vmscan.c b/mm/vmscan.c
> index e82ba737c028d..90f7a7bf4e426 100644
> --- a/mm/vmscan.c
> +++ b/mm/vmscan.c
> @@ -8180,7 +8180,7 @@ int user_proactive_reclaim(char *buf,
>  		} else {
>  			struct scan_control sc = {
>  				.gfp_mask = current_gfp_context(gfp_mask),
> -				.reclaim_idx = gfp_zone(gfp_mask),
> +				.reclaim_idx = MAX_NR_ZONES - 1,
>  				.proactive_swappiness = swappiness == -1 ? NULL : &swappiness,
>  				.priority = DEF_PRIORITY,
>  				.may_writepage = 1,


I just noticed some other issues with this code, in terms of honouring
zone_reclaim_node (not related to this patch). I'll send out fixes, for
this patch.

Reviewed-by: Balbir Singh <balbirs@nvidia.com>

Re: [PATCH] mm/vmscan: reclaim every zone in per-node proactive reclaim

[Gregory Price]

On Tue, Oct 20, 2026 at 12:49:37PM +1100, Balbir Singh wrote:
> > diff --git a/mm/vmscan.c b/mm/vmscan.c
> > index e82ba737c028d..90f7a7bf4e426 100644
> > --- a/mm/vmscan.c
> > +++ b/mm/vmscan.c
> > @@ -8180,7 +8180,7 @@ int user_proactive_reclaim(char *buf,
> >  		} else {
> >  			struct scan_control sc = {
> >  				.gfp_mask = current_gfp_context(gfp_mask),
> > -				.reclaim_idx = gfp_zone(gfp_mask),
> > +				.reclaim_idx = MAX_NR_ZONES - 1,
> >  				.proactive_swappiness = swappiness == -1 ? NULL : &swappiness,
> >  				.priority = DEF_PRIORITY,
> >  				.may_writepage = 1,
> 
> 
> I just noticed some other issues with this code, in terms of honouring
> zone_reclaim_node (not related to this patch). I'll send out fixes, for
> this patch.

I'm shocked, I tell you, shocked!

These are all things I accidentally ran into by just uh... using the
interface.  I'm absolutely positive these interfaces are full of bugs.

~Gregory

Re: [PATCH] mm/vmscan: reclaim every zone in per-node proactive reclaim

[Michal Hocko]

On Mon 19-10-26 14:50:02, Gregory Price (Meta) wrote:
> user_proactive_reclaim() limits per-node reclaim to
> gfp_zone(GFP_KERNEL), which is ZONE_NORMAL. Proactive reclaim serves no
> allocation, but that limit makes shrink_node() skip every ZONE_MOVABLE
> folio, so writing to /sys/devices/system/node/nodeN/reclaim never
> reclaims a node's movable memory. A movable-only node, such as dax/kmem
> or hotplugged memory onlined movable, reclaims nothing and fails with
> -EAGAIN.
> 
> To reproduce, fill a movable-only node with clean page cache and write
> 200M to its nodeN/reclaim: the write fails with -EAGAIN and the page
> cache stays. The same write to a ZONE_NORMAL node frees 200MB.
> 
> Scan every zone, as memory.reclaim does.
> 
> Fixes: b980077899ea4 ("mm: introduce per-node proactive reclaim interface")
> Cc: stable@vger.kernel.org
> Signed-off-by: Gregory Price (Meta) <gourry@gourry.net>
> ---
>  mm/vmscan.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/mm/vmscan.c b/mm/vmscan.c
> index e82ba737c028d..90f7a7bf4e426 100644
> --- a/mm/vmscan.c
> +++ b/mm/vmscan.c
> @@ -8180,7 +8180,7 @@ int user_proactive_reclaim(char *buf,
>  		} else {
>  			struct scan_control sc = {
>  				.gfp_mask = current_gfp_context(gfp_mask),
> -				.reclaim_idx = gfp_zone(gfp_mask),
> +				.reclaim_idx = MAX_NR_ZONES - 1,
>  				.proactive_swappiness = swappiness == -1 ? NULL : &swappiness,
>  				.priority = DEF_PRIORITY,
>  				.may_writepage = 1,

Would it make more sense to 
diff --git a/mm/vmscan.c b/mm/vmscan.c
index f11491ee9ed5..e07db9514e59 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -7963,7 +7963,7 @@ int user_proactive_reclaim(char *buf,
 	int swappiness = -1;
 	char *old_buf, *start;
 	substring_t args[MAX_OPT_ARGS];
-	gfp_t gfp_mask = GFP_KERNEL;
+	gfp_t gfp_mask = GFP_HIGHUSER_MOVABLE;
 
 	if (!buf || (!memcg && !pgdat) || (memcg && pgdat))
 		return -EINVAL;

That would take care of try_to_free_mem_cgroup_pages too.
-- 
Michal Hocko
SUSE Labs

Re: [PATCH] mm/vmscan: reclaim every zone in per-node proactive reclaim

[Kunwu Chan]

On Mon, 19 Oct 2026 14:50:02 -0400 "Gregory Price (Meta)" <gourry@gourry.net> wrote:

> user_proactive_reclaim() limits per-node reclaim to
> gfp_zone(GFP_KERNEL), which is ZONE_NORMAL. Proactive reclaim serves no
> allocation, but that limit makes shrink_node() skip every ZONE_MOVABLE
> folio, so writing to /sys/devices/system/node/nodeN/reclaim never
> reclaims a node's movable memory. A movable-only node, such as dax/kmem
> or hotplugged memory onlined movable, reclaims nothing and fails with
> -EAGAIN.
> 
> To reproduce, fill a movable-only node with clean page cache and write
> 200M to its nodeN/reclaim: the write fails with -EAGAIN and the page
> cache stays. The same write to a ZONE_NORMAL node frees 200MB.
> 
> Scan every zone, as memory.reclaim does.
> 
> Fixes: b980077899ea4 ("mm: introduce per-node proactive reclaim interface")
> Cc: stable@vger.kernel.org
> Signed-off-by: Gregory Price (Meta) <gourry@gourry.net>
> ---
>  mm/vmscan.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/mm/vmscan.c b/mm/vmscan.c
> index e82ba737c028d..90f7a7bf4e426 100644
> --- a/mm/vmscan.c
> +++ b/mm/vmscan.c
> @@ -8180,7 +8180,7 @@ int user_proactive_reclaim(char *buf,
>  		} else {
>  			struct scan_control sc = {
>  				.gfp_mask = current_gfp_context(gfp_mask),
> -				.reclaim_idx = gfp_zone(gfp_mask),
> +				.reclaim_idx = MAX_NR_ZONES - 1,
>  				.proactive_swappiness = swappiness == -1 ? NULL : &swappiness,
>  				.priority = DEF_PRIORITY,
>  				.may_writepage = 1,
> -- 
> 2.53.0-Meta
> 
> 

reclaim_idx = MAX_NR_ZONES - 1 is the right fix — proactive reclaim
isn't serving an allocation, so gfp_zone(GFP_KERNEL) needlessly caps
the scan.  GFP_HIGHUSER_MOVABLE would also work for the node path
via gfp_zone(), but that's indirect.

Reviewed-by: Kunwu Chan <chentao@kylinos.cn>

Re: [PATCH] mm/vmscan: reclaim every zone in per-node proactive reclaim

[Davidlohr Bueso]

On Mon, 19 Oct 2026, Gregory Price (Meta) wrote:

>user_proactive_reclaim() limits per-node reclaim to
>gfp_zone(GFP_KERNEL), which is ZONE_NORMAL. Proactive reclaim serves no
>allocation, but that limit makes shrink_node() skip every ZONE_MOVABLE
>folio, so writing to /sys/devices/system/node/nodeN/reclaim never
>reclaims a node's movable memory. A movable-only node, such as dax/kmem
>or hotplugged memory onlined movable, reclaims nothing and fails with
>-EAGAIN.
>
>To reproduce, fill a movable-only node with clean page cache and write
>200M to its nodeN/reclaim: the write fails with -EAGAIN and the page
>cache stays. The same write to a ZONE_NORMAL node frees 200MB.
>
>Scan every zone, as memory.reclaim does.
>
>Fixes: b980077899ea4 ("mm: introduce per-node proactive reclaim interface")
>Cc: stable@vger.kernel.org
>Signed-off-by: Gregory Price (Meta) <gourry@gourry.net>

This makes sense - sorry for the trouble but at least glad to see
people using this interface.

Acked-by: Davidlohr Bueso <dave@stgolabs.net>

>---
> mm/vmscan.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
>diff --git a/mm/vmscan.c b/mm/vmscan.c
>index e82ba737c028d..90f7a7bf4e426 100644
>--- a/mm/vmscan.c
>+++ b/mm/vmscan.c
>@@ -8180,7 +8180,7 @@ int user_proactive_reclaim(char *buf,
> 		} else {
> 			struct scan_control sc = {
> 				.gfp_mask = current_gfp_context(gfp_mask),
>-				.reclaim_idx = gfp_zone(gfp_mask),
>+				.reclaim_idx = MAX_NR_ZONES - 1,
> 				.proactive_swappiness = swappiness == -1 ? NULL : &swappiness,
> 				.priority = DEF_PRIORITY,
> 				.may_writepage = 1,
>-- 
>2.53.0-Meta
>

Re: [PATCH] mm/vmscan: reclaim every zone in per-node proactive reclaim

[Gregory Price (Meta)]

On 2026-10-20 09:49 +0200, Michal Hocko wrote:
> 
> Would it make more sense to 
> diff --git a/mm/vmscan.c b/mm/vmscan.c
> index f11491ee9ed5..e07db9514e59 100644
> --- a/mm/vmscan.c
> +++ b/mm/vmscan.c
> @@ -7963,7 +7963,7 @@ int user_proactive_reclaim(char *buf,
>  	int swappiness = -1;
>  	char *old_buf, *start;
>  	substring_t args[MAX_OPT_ARGS];
> -	gfp_t gfp_mask = GFP_KERNEL;
> +	gfp_t gfp_mask = GFP_HIGHUSER_MOVABLE;
>  
>  	if (!buf || (!memcg && !pgdat) || (memcg && pgdat))
>  		return -EINVAL;
> 
> That would take care of try_to_free_mem_cgroup_pages too.

funny enough, try_to_free_mem_cgroup_pages does what my patch does:

struct scan_control sc = {
	.reclaim_idx = MAX_NR_ZONES - 1,


now, whether the gfp_mask should be GFP_KERNEL, that's a different
question i think.

(tl;dr i think my patch is still correct)

~Gregory

-- 
Gregory Price (Meta) <gourry@gourry.net>

Re: [PATCH] mm/vmscan: reclaim every zone in per-node proactive reclaim

[Michal Hocko]

On Tue 20-10-26 15:58:53, Gregory Price (Meta) wrote:
> On 2026-10-20 09:49 +0200, Michal Hocko wrote:
> > 
> > Would it make more sense to 
> > diff --git a/mm/vmscan.c b/mm/vmscan.c
> > index f11491ee9ed5..e07db9514e59 100644
> > --- a/mm/vmscan.c
> > +++ b/mm/vmscan.c
> > @@ -7963,7 +7963,7 @@ int user_proactive_reclaim(char *buf,
> >  	int swappiness = -1;
> >  	char *old_buf, *start;
> >  	substring_t args[MAX_OPT_ARGS];
> > -	gfp_t gfp_mask = GFP_KERNEL;
> > +	gfp_t gfp_mask = GFP_HIGHUSER_MOVABLE;
> >  
> >  	if (!buf || (!memcg && !pgdat) || (memcg && pgdat))
> >  		return -EINVAL;
> > 
> > That would take care of try_to_free_mem_cgroup_pages too.
> 
> funny enough, try_to_free_mem_cgroup_pages does what my patch does:
> 
> struct scan_control sc = {
> 	.reclaim_idx = MAX_NR_ZONES - 1,
> 
> 
> now, whether the gfp_mask should be GFP_KERNEL, that's a different
> question i think.
> 
> (tl;dr i think my patch is still correct)

Ohh, right, I forgot that memcg reclaim always reclaims all zones.
Tweaking the mask seems more consistent but not something I will insist
on. Up to you which way to go. In any case

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