net/hyperv: Add flow control based on hi/low watermark

In the existing code, we only stop queue when the ringbuffer is full,
so the current packet has to be dropped or retried from upper layer.

This patch stops the tx queue when available ringbuffer is below
the low watermark. So the ringbuffer still has small amount of space
available for the current packet. This will reduce the overhead of
retries on sending.

Signed-off-by: Haiyang Zhang <haiyangz@microsoft.com>
Reviewed-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

4 files changed, 69 insertions, 36 deletions

diff --git a/drivers/hv/ring_buffer.c b/drivers/hv/ring_buffer.c
index 8af25a097d75..7233c88f01b8 100644
--- a/drivers/hv/ring_buffer.c
+++ b/drivers/hv/ring_buffer.c
@@ -30,37 +30,6 @@
 #include "hyperv_vmbus.h"
 
 
-/* #defines */
-
-
-/* Amount of space to write to */
-#define BYTES_AVAIL_TO_WRITE(r, w, z) \
-        ((w) >= (r)) ? ((z) - ((w) - (r))) : ((r) - (w))
-
-
-/*
- *
- * hv_get_ringbuffer_availbytes()
- *
- * Get number of bytes available to read and to write to
- * for the specified ring buffer
- */
-static inline void
-hv_get_ringbuffer_availbytes(struct hv_ring_buffer_info *rbi,
-                          u32 *read, u32 *write)
-{
-        u32 read_loc, write_loc;
-
-        smp_read_barrier_depends();
-
-        /* Capture the read/write indices before they changed */
-        read_loc = rbi->ring_buffer->read_index;
-        write_loc = rbi->ring_buffer->write_index;
-
-        *write = BYTES_AVAIL_TO_WRITE(read_loc, write_loc, rbi->ring_datasize);
-        *read = rbi->ring_datasize - *write;
-}
-
 /*
  * hv_get_next_write_location()
  *
diff --git a/drivers/net/hyperv/netvsc.c b/drivers/net/hyperv/netvsc.c
index d025c83cd12a..8b919471472f 100644
--- a/drivers/net/hyperv/netvsc.c
+++ b/drivers/net/hyperv/netvsc.c
@@ -428,6 +428,24 @@ int netvsc_device_remove(struct hv_device *device)
         return 0;
 }
 
+
+#define RING_AVAIL_PERCENT_HIWATER 20
+#define RING_AVAIL_PERCENT_LOWATER 10
+
+/*
+ * Get the percentage of available bytes to write in the ring.
+ * The return value is in range from 0 to 100.
+ */
+static inline u32 hv_ringbuf_avail_percent(
+                struct hv_ring_buffer_info *ring_info)
+{
+        u32 avail_read, avail_write;
+
+        hv_get_ringbuffer_availbytes(ring_info, &avail_read, &avail_write);
+
+        return avail_write * 100 / ring_info->ring_datasize;
+}
+
 static void netvsc_send_completion(struct hv_device *device,
                                    struct vmpacket_descriptor *packet)
 {
@@ -455,6 +473,8 @@ static void netvsc_send_completion(struct hv_device *device,
                 complete(&net_device->channel_init_wait);
         } else if (nvsp_packet->hdr.msg_type ==
                    NVSP_MSG1_TYPE_SEND_RNDIS_PKT_COMPLETE) {
+                int num_outstanding_sends;
+
                 /* Get the send context */
                 nvsc_packet = (struct hv_netvsc_packet *)(unsigned long)
                         packet->trans_id;
@@ -463,10 +483,14 @@ static void netvsc_send_completion(struct hv_device *device,
                 nvsc_packet->completion.send.send_completion(
                         nvsc_packet->completion.send.send_completion_ctx);
 
-                atomic_dec(&net_device->num_outstanding_sends);
+                num_outstanding_sends =
+                        atomic_dec_return(&net_device->num_outstanding_sends);
 
-                if (netif_queue_stopped(ndev) && !net_device->start_remove)
-                        netif_wake_queue(ndev);
+                if (netif_queue_stopped(ndev) && !net_device->start_remove &&
+                        (hv_ringbuf_avail_percent(&device->channel->outbound)
+                        > RING_AVAIL_PERCENT_HIWATER ||
+                        num_outstanding_sends < 1))
+                                netif_wake_queue(ndev);
         } else {
                 netdev_err(ndev, "Unknown send completion packet type- "
                            "%d received!!\n", nvsp_packet->hdr.msg_type);
@@ -519,10 +543,19 @@ int netvsc_send(struct hv_device *device,
 
         if (ret == 0) {
                 atomic_inc(&net_device->num_outstanding_sends);
+                if (hv_ringbuf_avail_percent(&device->channel->outbound) <
+                        RING_AVAIL_PERCENT_LOWATER) {
+                        netif_stop_queue(ndev);
+                        if (atomic_read(&net_device->
+                                num_outstanding_sends) < 1)
+                                netif_wake_queue(ndev);
+                }
         } else if (ret == -EAGAIN) {
                 netif_stop_queue(ndev);
-                if (atomic_read(&net_device->num_outstanding_sends) < 1)
+                if (atomic_read(&net_device->num_outstanding_sends) < 1) {
                         netif_wake_queue(ndev);
+                        ret = -ENOSPC;
+                }
         } else {
                 netdev_err(ndev, "Unable to send packet %p ret %d\n",
                            packet, ret);
diff --git a/drivers/net/hyperv/netvsc_drv.c b/drivers/net/hyperv/netvsc_drv.c
index dd294783b5c5..a0cc12786be4 100644
--- a/drivers/net/hyperv/netvsc_drv.c
+++ b/drivers/net/hyperv/netvsc_drv.c
@@ -224,9 +224,13 @@ static int netvsc_start_xmit(struct sk_buff *skb, struct net_device *net)
                 net->stats.tx_packets++;
         } else {
                 kfree(packet);
+                if (ret != -EAGAIN) {
+                        dev_kfree_skb_any(skb);
+                        net->stats.tx_dropped++;
+                }
         }
 
-        return ret ? NETDEV_TX_BUSY : NETDEV_TX_OK;
+        return (ret == -EAGAIN) ? NETDEV_TX_BUSY : NETDEV_TX_OK;
 }
 
 /*
diff --git a/include/linux/hyperv.h b/include/linux/hyperv.h
index 5852545e6bba..6af8738ae7e9 100644
--- a/include/linux/hyperv.h
+++ b/include/linux/hyperv.h
@@ -274,6 +274,33 @@ struct hv_ring_buffer_debug_info {
         u32 bytes_avail_towrite;
 };
 
+
+/*
+ *
+ * hv_get_ringbuffer_availbytes()
+ *
+ * Get number of bytes available to read and to write to
+ * for the specified ring buffer
+ */
+static inline void
+hv_get_ringbuffer_availbytes(struct hv_ring_buffer_info *rbi,
+                          u32 *read, u32 *write)
+{
+        u32 read_loc, write_loc, dsize;
+
+        smp_read_barrier_depends();
+
+        /* Capture the read/write indices before they changed */
+        read_loc = rbi->ring_buffer->read_index;
+        write_loc = rbi->ring_buffer->write_index;
+        dsize = rbi->ring_datasize;
+
+        *write = write_loc >= read_loc ? dsize - (write_loc - read_loc) :
+                read_loc - write_loc;
+        *read = dsize - *write;
+}
+
+
 /*
  * We use the same version numbering for all Hyper-V modules.
  *