[PATCH RFC v4 11/13] mm/vmscan: Move count_vm_event(DROP_SLAB) into drop_slab()

[PATCH RFC v4 11/13] mm/vmscan: Move count_vm_event(DROP_SLAB) into drop_slab()

[David Hildenbrand]

Let's count within the function itself, so every invocation (of future
users) will be counted.

Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: linux-fsdevel@vger.kernel.org
Signed-off-by: David Hildenbrand <david@redhat.com>
---
 fs/drop_caches.c | 4 +---
 mm/vmscan.c      | 1 +
 2 files changed, 2 insertions(+), 3 deletions(-)

diff --git a/fs/drop_caches.c b/fs/drop_caches.c
index d31b6c72b476..a042da782fcd 100644
--- a/fs/drop_caches.c
+++ b/fs/drop_caches.c
@@ -61,10 +61,8 @@ int drop_caches_sysctl_handler(struct ctl_table *table, int write,
 			iterate_supers(drop_pagecache_sb, NULL);
 			count_vm_event(DROP_PAGECACHE);
 		}
-		if (sysctl_drop_caches & 2) {
+		if (sysctl_drop_caches & 2)
 			drop_slab();
-			count_vm_event(DROP_SLAB);
-		}
 		if (!stfu) {
 			pr_info("%s (%d): drop_caches: %d\n",
 				current->comm, task_pid_nr(current),
diff --git a/mm/vmscan.c b/mm/vmscan.c
index 5a6445e86328..c3e53502a84a 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -726,6 +726,7 @@ void drop_slab(void)
 
 	for_each_online_node(nid)
 		drop_slab_node(nid);
+	count_vm_event(DROP_SLAB);
 }
 
 static inline int is_page_cache_freeable(struct page *page)
-- 
2.23.0

[PATCH RFC v4 13/13] virtio-mem: Drop slab objects when unplug continues to fail

[David Hildenbrand]

Start dropping slab objects after 30 minutes and repeat every 30 minutes
in case we can't unplug more memory using alloc_contig_range().
Log messages and make it configurable. Enable dropping slab objects as
default (especially, reclaimable slab objects that are not movable).

In the future, we might want to implement+use drop_slab_range(), which
will also come in handy for other users (e.g., offlining, gigantic huge
pages).

Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Jason Wang <jasowang@redhat.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Igor Mammedov <imammedo@redhat.com>
Cc: Dave Young <dyoung@redhat.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Pavel Tatashin <pasha.tatashin@soleen.com>
Cc: Stefan Hajnoczi <stefanha@redhat.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: linux-fsdevel@vger.kernel.org
Signed-off-by: David Hildenbrand <david@redhat.com>
---
 drivers/virtio/virtio_mem.c | 30 ++++++++++++++++++++++++++++++
 1 file changed, 30 insertions(+)

diff --git a/drivers/virtio/virtio_mem.c b/drivers/virtio/virtio_mem.c
index 3a57434f92ed..8f25f7453a08 100644
--- a/drivers/virtio/virtio_mem.c
+++ b/drivers/virtio/virtio_mem.c
@@ -25,6 +25,11 @@ static bool unplug_online = true;
 module_param(unplug_online, bool, 0644);
 MODULE_PARM_DESC(unplug_online, "Try to unplug online memory");
 
+static bool drop_slab_objects = true;
+module_param(drop_slab_objects, bool, 0644);
+MODULE_PARM_DESC(drop_slab_objects,
+		 "Drop slab objects when unplug continues to fail");
+
 enum virtio_mem_mb_state {
 	/* Unplugged, not added to Linux. Can be reused later. */
 	VIRTIO_MEM_MB_STATE_UNUSED = 0,
@@ -1384,6 +1389,7 @@ static int virtio_mem_mb_unplug_any_sb_online(struct virtio_mem *vm,
 static int virtio_mem_unplug_request(struct virtio_mem *vm, uint64_t diff)
 {
 	uint64_t nb_sb = diff / vm->subblock_size;
+	bool retried = false;
 	unsigned long mb_id;
 	int rc;
 
@@ -1421,6 +1427,7 @@ static int virtio_mem_unplug_request(struct virtio_mem *vm, uint64_t diff)
 		return 0;
 	}
 
+retry_locked:
 	/* Try to unplug subblocks of partially plugged online blocks. */
 	virtio_mem_for_each_mb_state(vm, mb_id,
 				     VIRTIO_MEM_MB_STATE_ONLINE_PARTIAL) {
@@ -1445,6 +1452,29 @@ static int virtio_mem_unplug_request(struct virtio_mem *vm, uint64_t diff)
 	}
 
 	mutex_unlock(&vm->hotplug_mutex);
+
+	/*
+	 * If we can't unplug the requested amount of memory for a long time,
+	 * start freeing up memory in caches. This might harm performance,
+	 * is configurable, and we log a message. Retry imemdiately a second
+	 * time - then wait another VIRTIO_MEM_RETRY_TIMER_MAX_MS.
+	 */
+	if (nb_sb && !retried && drop_slab_objects &&
+	    vm->retry_timer_ms == VIRTIO_MEM_RETRY_TIMER_MAX_MS) {
+		if (vm->nid == NUMA_NO_NODE) {
+			dev_info(&vm->vdev->dev, "dropping all slab objects\n");
+			drop_slab();
+		} else {
+			dev_info(&vm->vdev->dev,
+				 "dropping all slab objects on node=%d\n",
+				 vm->nid);
+			drop_slab_node(vm->nid);
+		}
+		retried = true;
+		mutex_lock(&vm->hotplug_mutex);
+		goto retry_locked;
+	}
+
 	return nb_sb ? -EBUSY : 0;
 out_unlock:
 	mutex_unlock(&vm->hotplug_mutex);
-- 
2.23.0

Re: [PATCH RFC v4 11/13] mm/vmscan: Move count_vm_event(DROP_SLAB) into drop_slab()

[Michal Hocko]

On Thu 12-12-19 18:11:35, David Hildenbrand wrote:
> Let's count within the function itself, so every invocation (of future
> users) will be counted.
> 
> Cc: Alexander Viro <viro@zeniv.linux.org.uk>
> Cc: Andrew Morton <akpm@linux-foundation.org>
> Cc: linux-fsdevel@vger.kernel.org
> Signed-off-by: David Hildenbrand <david@redhat.com>

Slight inconsistency with the page cache droppint but nothing earth
shattering.

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

> ---
>  fs/drop_caches.c | 4 +---
>  mm/vmscan.c      | 1 +
>  2 files changed, 2 insertions(+), 3 deletions(-)
> 
> diff --git a/fs/drop_caches.c b/fs/drop_caches.c
> index d31b6c72b476..a042da782fcd 100644
> --- a/fs/drop_caches.c
> +++ b/fs/drop_caches.c
> @@ -61,10 +61,8 @@ int drop_caches_sysctl_handler(struct ctl_table *table, int write,
>  			iterate_supers(drop_pagecache_sb, NULL);
>  			count_vm_event(DROP_PAGECACHE);
>  		}
> -		if (sysctl_drop_caches & 2) {
> +		if (sysctl_drop_caches & 2)
>  			drop_slab();
> -			count_vm_event(DROP_SLAB);
> -		}
>  		if (!stfu) {
>  			pr_info("%s (%d): drop_caches: %d\n",
>  				current->comm, task_pid_nr(current),
> diff --git a/mm/vmscan.c b/mm/vmscan.c
> index 5a6445e86328..c3e53502a84a 100644
> --- a/mm/vmscan.c
> +++ b/mm/vmscan.c
> @@ -726,6 +726,7 @@ void drop_slab(void)
>  
>  	for_each_online_node(nid)
>  		drop_slab_node(nid);
> +	count_vm_event(DROP_SLAB);
>  }
>  
>  static inline int is_page_cache_freeable(struct page *page)
> -- 
> 2.23.0

-- 
Michal Hocko
SUSE Labs