[PATCH 17/18] xen/pvcalls: implement write

[Stefano Stabellini]

When the other end notifies us that there is data to be written
(pvcalls_back_conn_event), add the relative sock_mapping to the ioworker
list, increment the io and write counters, and schedule the ioworker.

Implement the write function called by ioworker by reading the data from
the data ring, writing it to the socket by calling inet_sendmsg.

Set out_error on error.

Signed-off-by: Stefano Stabellini <stefano@xxxxxxxxxxx>
CC: boris.ostrovsky@xxxxxxxxxx
CC: jgross@xxxxxxxx
---
 drivers/xen/pvcalls-back.c | 80 +++++++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 79 insertions(+), 1 deletion(-)

diff --git a/drivers/xen/pvcalls-back.c b/drivers/xen/pvcalls-back.c
index 0f715a8..2de43c3 100644
--- a/drivers/xen/pvcalls-back.c
+++ b/drivers/xen/pvcalls-back.c
@@ -179,7 +179,67 @@ static void pvcalls_conn_back_read(unsigned long opaque)
 
 static int pvcalls_conn_back_write(struct sock_mapping *map)
 {
-	return 0;
+	struct pvcalls_data_intf *intf = map->ring;
+	struct pvcalls_data *data = &map->data;
+	struct msghdr msg;
+	struct kvec vec[2];
+	RING_IDX cons, prod, size, ring_size;
+	int ret;
+
+	cons = intf->out_cons;
+	prod = intf->out_prod;
+	/* read the indexes before dealing with the data */
+	virt_mb();
+
+	ring_size = XEN_FLEX_RING_SIZE(map->ring_order);
+	size = pvcalls_queued(prod, cons, ring_size);
+	if (size == 0)
+		return 0;
+
+	memset(&msg, 0, sizeof(msg));
+	msg.msg_flags |= MSG_DONTWAIT;
+	msg.msg_iter.type = ITER_KVEC|READ;
+	msg.msg_iter.count = size;
+	if (pvcalls_mask(prod, ring_size) > pvcalls_mask(cons, ring_size)) {
+		vec[0].iov_base = data->out + pvcalls_mask(cons, ring_size);
+		vec[0].iov_len = size;
+		msg.msg_iter.kvec = vec;
+		msg.msg_iter.nr_segs = 1;
+	} else {
+		vec[0].iov_base = data->out + pvcalls_mask(cons, ring_size);
+		vec[0].iov_len = XEN_FLEX_RING_SIZE(ring_size) -
+			pvcalls_mask(cons, ring_size);
+		vec[1].iov_base = data->out;
+		vec[1].iov_len = size - vec[0].iov_len;
+		msg.msg_iter.kvec = vec;
+		msg.msg_iter.nr_segs = 2;
+	}
+
+	atomic_set(&map->write, 0);
+	ret = inet_sendmsg(map->sock, &msg, size);
+	if (ret == -EAGAIN || ret < size) {
+		atomic_inc(&map->write);
+		atomic_inc(&pvcalls_back_global.ioworkers[map->data_worker].io);
+	}
+	if (ret == -EAGAIN)
+		return ret;
+
+	/* write the data, then update the indexes */
+	virt_wmb();
+	if (ret < 0) {
+		intf->out_error = ret;
+	} else {
+		intf->out_error = 0;
+		intf->out_cons = cons + ret;
+		prod = intf->out_prod;
+	}
+	/* update the indexes, then notify the other end */
+	virt_wmb();
+	if (prod != cons + ret)
+		atomic_inc(&map->write);
+	notify_remote_via_irq(map->irq);
+
+	return ret;
 }
 
 static void pvcalls_back_ioworker(struct work_struct *work)
@@ -914,6 +974,24 @@ static irqreturn_t pvcalls_back_event(int irq, void *dev_id)
 
 static irqreturn_t pvcalls_back_conn_event(int irq, void *sock_map)
 {
+	struct sock_mapping *map = sock_map;
+	struct pvcalls_ioworker *iow;
+	unsigned long flags;
+
+	if (map == NULL || map->sock == NULL || map->sock->sk == NULL ||
+		map->sock->sk->sk_user_data != map)
+		return IRQ_HANDLED;
+
+	iow = &pvcalls_back_global.ioworkers[map->data_worker];
+	spin_lock_irqsave(&iow->lock, flags);
+	atomic_inc(&map->write);
+	if (list_empty(&map->queue))
+		list_add_tail(&map->queue, &iow->wqs);
+	spin_unlock_irqrestore(&iow->lock, flags);
+	atomic_inc(&iow->io);
+	queue_work_on(map->data_worker, pvcalls_back_global.wq,
+		&iow->register_work);
+
 	return IRQ_HANDLED;
 }
 
-- 
1.9.1