drbd: introduce drbd_recv_header_maybe_unplug

Recently, drbd_recv_header() was changed to potentially
implicitly "unplug" the backend device(s), in case there
is currently nothing to receive.

Be more explicit about it: re-introduce the original drbd_recv_header(),
and introduce a new drbd_recv_header_maybe_unplug() for use by the
receiver "main loop".

Using explicit plugging via blk_start_plug(); blk_finish_plug();
really helps the io-scheduler of the backend with merging requests.

Wrap the receiver "main loop" with such a plug.
Also catch unplug events on the Primary,
and try to propagate.

This is performance relevant.  Without this, if the receiving side does
not merge requests, number of IOPS on the peer can me significantly
higher than IOPS on the Primary, and can easily become the bottleneck.

Together, both changes should help to reduce the number of IOPS
as seen on the backend of the receiving side, by increasing
the chance of merging mergable requests, without trading latency
for more throughput.

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

1 files changed, 55 insertions, 0 deletions

diff --git a/drivers/block/drbd/drbd_req.c b/drivers/block/drbd/drbd_req.c
index 447c975f5481..5cf43f13e7eb 100644
--- a/drivers/block/drbd/drbd_req.c
+++ b/drivers/block/drbd/drbd_req.c
@@ -1279,6 +1279,56 @@ static bool may_do_writes(struct drbd_device *device)
 	return s.disk == D_UP_TO_DATE || s.pdsk == D_UP_TO_DATE;
 }
 
+struct drbd_plug_cb {
+	struct blk_plug_cb cb;
+	struct drbd_request *most_recent_req;
+	/* do we need more? */
+};
+
+static void drbd_unplug(struct blk_plug_cb *cb, bool from_schedule)
+{
+	struct drbd_plug_cb *plug = container_of(cb, struct drbd_plug_cb, cb);
+	struct drbd_resource *resource = plug->cb.data;
+	struct drbd_request *req = plug->most_recent_req;
+
+	if (!req)
+		return;
+
+	spin_lock_irq(&resource->req_lock);
+	/* In case the sender did not process it yet, raise the flag to
+	 * have it followed with P_UNPLUG_REMOTE just after. */
+	req->rq_state |= RQ_UNPLUG;
+	/* but also queue a generic unplug */
+	drbd_queue_unplug(req->device);
+	spin_unlock_irq(&resource->req_lock);
+	kref_put(&req->kref, drbd_req_destroy);
+}
+
+static struct drbd_plug_cb* drbd_check_plugged(struct drbd_resource *resource)
+{
+	/* A lot of text to say
+	 * return (struct drbd_plug_cb*)blk_check_plugged(); */
+	struct drbd_plug_cb *plug;
+	struct blk_plug_cb *cb = blk_check_plugged(drbd_unplug, resource, sizeof(*plug));
+
+	if (cb)
+		plug = container_of(cb, struct drbd_plug_cb, cb);
+	else
+		plug = NULL;
+	return plug;
+}
+
+static void drbd_update_plug(struct drbd_plug_cb *plug, struct drbd_request *req)
+{
+	struct drbd_request *tmp = plug->most_recent_req;
+	/* Will be sent to some peer.
+	 * Remember to tag it with UNPLUG_REMOTE on unplug */
+	kref_get(&req->kref);
+	plug->most_recent_req = req;
+	if (tmp)
+		kref_put(&tmp->kref, drbd_req_destroy);
+}
+
 static void drbd_send_and_submit(struct drbd_device *device, struct drbd_request *req)
 {
 	struct drbd_resource *resource = device->resource;
@@ -1287,6 +1337,8 @@ static void drbd_send_and_submit(struct drbd_device *device, struct drbd_request
 	bool no_remote = false;
 	bool submit_private_bio = false;
 
+	struct drbd_plug_cb *plug = drbd_check_plugged(resource);
+
 	spin_lock_irq(&resource->req_lock);
 	if (rw == WRITE) {
 		/* This may temporarily give up the req_lock,
@@ -1351,6 +1403,9 @@ static void drbd_send_and_submit(struct drbd_device *device, struct drbd_request
 			no_remote = true;
 	}
 
+	if (plug != NULL && no_remote == false)
+		drbd_update_plug(plug, req);
+
 	/* If it took the fast path in drbd_request_prepare, add it here.
 	 * The slow path has added it already. */
 	if (list_empty(&req->req_pending_master_completion))