Merge branch 'netvsc-NAPI'

Stephen Hemminger says:

====================
NAPI support for Hyper-V

These patches enable NAPI, GRO and napi_alloc_skb for Hyper-V netvsc
driver.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

drivers/hv/ring_buffer.c

@@ -32,6 +32,8 @@
 
 #include "hyperv_vmbus.h"
 
+#define VMBUS_PKT_TRAILER	8
+
 /*
  * When we write to the ring buffer, check if the host needs to
  * be signaled. Here is the details of this protocol:
@@ -336,6 +338,12 @@ int hv_ringbuffer_write(struct vmbus_channel *channel,
 	return 0;
 }
 
+static inline void
+init_cached_read_index(struct hv_ring_buffer_info *rbi)
+{
+	rbi->cached_read_index = rbi->ring_buffer->read_index;
+}
+
 int hv_ringbuffer_read(struct vmbus_channel *channel,
 		       void *buffer, u32 buflen, u32 *buffer_actual_len,
 		       u64 *requestid, bool raw)
@@ -366,7 +374,8 @@ int hv_ringbuffer_read(struct vmbus_channel *channel,
 		return ret;
 	}
 
-	init_cached_read_index(channel);
+	init_cached_read_index(inring_info);
+
 	next_read_location = hv_get_next_read_location(inring_info);
 	next_read_location = hv_copyfrom_ringbuffer(inring_info, &desc,
 						    sizeof(desc),
@@ -410,3 +419,86 @@ int hv_ringbuffer_read(struct vmbus_channel *channel,
 
 	return ret;
 }
+
+/*
+ * Determine number of bytes available in ring buffer after
+ * the current iterator (priv_read_index) location.
+ *
+ * This is similar to hv_get_bytes_to_read but with private
+ * read index instead.
+ */
+static u32 hv_pkt_iter_avail(const struct hv_ring_buffer_info *rbi)
+{
+	u32 priv_read_loc = rbi->priv_read_index;
+	u32 write_loc = READ_ONCE(rbi->ring_buffer->write_index);
+
+	if (write_loc >= priv_read_loc)
+		return write_loc - priv_read_loc;
+	else
+		return (rbi->ring_datasize - priv_read_loc) + write_loc;
+}
+
+/*
+ * Get first vmbus packet from ring buffer after read_index
+ *
+ * If ring buffer is empty, returns NULL and no other action needed.
+ */
+struct vmpacket_descriptor *hv_pkt_iter_first(struct vmbus_channel *channel)
+{
+	struct hv_ring_buffer_info *rbi = &channel->inbound;
+
+	/* set state for later hv_signal_on_read() */
+	init_cached_read_index(rbi);
+
+	if (hv_pkt_iter_avail(rbi) < sizeof(struct vmpacket_descriptor))
+		return NULL;
+
+	return hv_get_ring_buffer(rbi) + rbi->priv_read_index;
+}
+EXPORT_SYMBOL_GPL(hv_pkt_iter_first);
+
+/*
+ * Get next vmbus packet from ring buffer.
+ *
+ * Advances the current location (priv_read_index) and checks for more
+ * data. If the end of the ring buffer is reached, then return NULL.
+ */
+struct vmpacket_descriptor *
+__hv_pkt_iter_next(struct vmbus_channel *channel,
+		   const struct vmpacket_descriptor *desc)
+{
+	struct hv_ring_buffer_info *rbi = &channel->inbound;
+	u32 packetlen = desc->len8 << 3;
+	u32 dsize = rbi->ring_datasize;
+
+	/* bump offset to next potential packet */
+	rbi->priv_read_index += packetlen + VMBUS_PKT_TRAILER;
+	if (rbi->priv_read_index >= dsize)
+		rbi->priv_read_index -= dsize;
+
+	/* more data? */
+	if (hv_pkt_iter_avail(rbi) < sizeof(struct vmpacket_descriptor))
+		return NULL;
+	else
+		return hv_get_ring_buffer(rbi) + rbi->priv_read_index;
+}
+EXPORT_SYMBOL_GPL(__hv_pkt_iter_next);
+
+/*
+ * Update host ring buffer after iterating over packets.
+ */
+void hv_pkt_iter_close(struct vmbus_channel *channel)
+{
+	struct hv_ring_buffer_info *rbi = &channel->inbound;
+
+	/*
+	 * Make sure all reads are done before we update the read index since
+	 * the writer may start writing to the read area once the read index
+	 * is updated.
+	 */
+	virt_rmb();
+	rbi->ring_buffer->read_index = rbi->priv_read_index;
+
+	hv_signal_on_read(channel);
+}
+EXPORT_SYMBOL_GPL(hv_pkt_iter_close);

drivers/net/hyperv/hyperv_net.h

@@ -196,6 +196,7 @@ int netvsc_recv_callback(struct net_device *net,
 			 const struct ndis_tcp_ip_checksum_info *csum_info,
 			 const struct ndis_pkt_8021q_info *vlan);
 void netvsc_channel_cb(void *context);
+int netvsc_poll(struct napi_struct *napi, int budget);
 int rndis_filter_open(struct netvsc_device *nvdev);
 int rndis_filter_close(struct netvsc_device *nvdev);
 int rndis_filter_device_add(struct hv_device *dev,
@@ -720,6 +721,7 @@ struct net_device_context {
 /* Per channel data */
 struct netvsc_channel {
 	struct vmbus_channel *channel;
+	struct napi_struct napi;
 	struct multi_send_data msd;
 	struct multi_recv_comp mrc;
 	atomic_t queue_sends;

drivers/net/hyperv/netvsc_drv.c

@@ -589,13 +589,14 @@ void netvsc_linkstatus_callback(struct hv_device *device_obj,
 }
 
 static struct sk_buff *netvsc_alloc_recv_skb(struct net_device *net,
+					     struct napi_struct *napi,
 					     const struct ndis_tcp_ip_checksum_info *csum_info,
 					     const struct ndis_pkt_8021q_info *vlan,
 					     void *data, u32 buflen)
 {
 	struct sk_buff *skb;
 
-	skb = netdev_alloc_skb_ip_align(net, buflen);
+	skb = napi_alloc_skb(napi, buflen);
 	if (!skb)
 		return skb;
 
@@ -642,11 +643,11 @@ int netvsc_recv_callback(struct net_device *net,
 {
 	struct net_device_context *net_device_ctx = netdev_priv(net);
 	struct netvsc_device *net_device = net_device_ctx->nvdev;
+	u16 q_idx = channel->offermsg.offer.sub_channel_index;
+	struct netvsc_channel *nvchan = &net_device->chan_table[q_idx];
 	struct net_device *vf_netdev;
 	struct sk_buff *skb;
 	struct netvsc_stats *rx_stats;
-	u16 q_idx = channel->offermsg.offer.sub_channel_index;
-
 
 	if (net->reg_state != NETREG_REGISTERED)
 		return NVSP_STAT_FAIL;
@@ -664,7 +665,8 @@ int netvsc_recv_callback(struct net_device *net,
 		net = vf_netdev;
 
 	/* Allocate a skb - TODO direct I/O to pages? */
-	skb = netvsc_alloc_recv_skb(net, csum_info, vlan, data, len);
+	skb = netvsc_alloc_recv_skb(net, &nvchan->napi,
+				    csum_info, vlan, data, len);
 	if (unlikely(!skb)) {
 		++net->stats.rx_dropped;
 		rcu_read_unlock();
@@ -679,7 +681,7 @@ int netvsc_recv_callback(struct net_device *net,
 	 * on the synthetic device because modifying the VF device
 	 * statistics will not work correctly.
 	 */
-	rx_stats = &net_device->chan_table[q_idx].rx_stats;
+	rx_stats = &nvchan->rx_stats;
 	u64_stats_update_begin(&rx_stats->syncp);
 	rx_stats->packets++;
 	rx_stats->bytes += len;
@@ -690,12 +692,7 @@ int netvsc_recv_callback(struct net_device *net,
 		++rx_stats->multicast;
 	u64_stats_update_end(&rx_stats->syncp);
 
-	/*
-	 * Pass the skb back up. Network stack will deallocate the skb when it
-	 * is done.
-	 * TODO - use NAPI?
-	 */
-	netif_receive_skb(skb);
+	napi_gro_receive(&nvchan->napi, skb);
 	rcu_read_unlock();
 
 	return 0;

drivers/net/hyperv/rndis_filter.c

@@ -1012,6 +1012,8 @@ static void netvsc_sc_open(struct vmbus_channel *new_sc)
 	if (ret == 0)
 		nvscdev->chan_table[chn_index].channel = new_sc;
 
+	napi_enable(&nvscdev->chan_table[chn_index].napi);
+
 	spin_lock_irqsave(&nvscdev->sc_lock, flags);
 	nvscdev->num_sc_offered--;
 	spin_unlock_irqrestore(&nvscdev->sc_lock, flags);

include/linux/hyperv.h

@@ -1504,14 +1504,6 @@ static inline  void hv_signal_on_read(struct vmbus_channel *channel)
 	return;
 }
 
-static inline void
-init_cached_read_index(struct vmbus_channel *channel)
-{
-	struct hv_ring_buffer_info *rbi = &channel->inbound;
-
-	rbi->cached_read_index = rbi->ring_buffer->read_index;
-}
-
 /*
  * Mask off host interrupt callback notifications
  */
@@ -1545,76 +1537,48 @@ static inline u32 hv_end_read(struct hv_ring_buffer_info *rbi)
 /*
  * An API to support in-place processing of incoming VMBUS packets.
  */
-#define VMBUS_PKT_TRAILER	8
 
-static inline struct vmpacket_descriptor *
-get_next_pkt_raw(struct vmbus_channel *channel)
+/* Get data payload associated with descriptor */
+static inline void *hv_pkt_data(const struct vmpacket_descriptor *desc)
 {
-	struct hv_ring_buffer_info *ring_info = &channel->inbound;
-	u32 priv_read_loc = ring_info->priv_read_index;
-	void *ring_buffer = hv_get_ring_buffer(ring_info);
-	u32 dsize = ring_info->ring_datasize;
-	/*
-	 * delta is the difference between what is available to read and
-	 * what was already consumed in place. We commit read index after
-	 * the whole batch is processed.
-	 */
-	u32 delta = priv_read_loc >= ring_info->ring_buffer->read_index ?
-		priv_read_loc - ring_info->ring_buffer->read_index :
-		(dsize - ring_info->ring_buffer->read_index) + priv_read_loc;
-	u32 bytes_avail_toread = (hv_get_bytes_to_read(ring_info) - delta);
-
-	if (bytes_avail_toread < sizeof(struct vmpacket_descriptor))
-		return NULL;
-
-	return ring_buffer + priv_read_loc;
+	return (void *)((unsigned long)desc + (desc->offset8 << 3));
 }
 
-/*
- * A helper function to step through packets "in-place"
- * This API is to be called after each successful call
- * get_next_pkt_raw().
- */
-static inline void put_pkt_raw(struct vmbus_channel *channel,
-				struct vmpacket_descriptor *desc)
+/* Get data size associated with descriptor */
+static inline u32 hv_pkt_datalen(const struct vmpacket_descriptor *desc)
 {
-	struct hv_ring_buffer_info *ring_info = &channel->inbound;
-	u32 packetlen = desc->len8 << 3;
-	u32 dsize = ring_info->ring_datasize;
-
-	/*
-	 * Include the packet trailer.
-	 */
-	ring_info->priv_read_index += packetlen + VMBUS_PKT_TRAILER;
-	ring_info->priv_read_index %= dsize;
+	return (desc->len8 << 3) - (desc->offset8 << 3);
 }
 
+
+struct vmpacket_descriptor *
+hv_pkt_iter_first(struct vmbus_channel *channel);
+
+struct vmpacket_descriptor *
+__hv_pkt_iter_next(struct vmbus_channel *channel,
+		   const struct vmpacket_descriptor *pkt);
+
+void hv_pkt_iter_close(struct vmbus_channel *channel);
+
 /*
- * This call commits the read index and potentially signals the host.
- * Here is the pattern for using the "in-place" consumption APIs:
- *
- * init_cached_read_index();
- *
- * while (get_next_pkt_raw() {
- *	process the packet "in-place";
- *	put_pkt_raw();
- * }
- * if (packets processed in place)
- *	commit_rd_index();
+ * Get next packet descriptor from iterator
+ * If at end of list, return NULL and update host.
  */
-static inline void commit_rd_index(struct vmbus_channel *channel)
+static inline struct vmpacket_descriptor *
+hv_pkt_iter_next(struct vmbus_channel *channel,
+		 const struct vmpacket_descriptor *pkt)
 {
-	struct hv_ring_buffer_info *ring_info = &channel->inbound;
-	/*
-	 * Make sure all reads are done before we update the read index since
-	 * the writer may start writing to the read area once the read index
-	 * is updated.
-	 */
-	virt_rmb();
-	ring_info->ring_buffer->read_index = ring_info->priv_read_index;
+	struct vmpacket_descriptor *nxt;
+
+	nxt = __hv_pkt_iter_next(channel, pkt);
+	if (!nxt)
+		hv_pkt_iter_close(channel);
 
-	hv_signal_on_read(channel);
+	return nxt;
 }
 
+#define foreach_vmbus_pkt(pkt, channel) \
+	for (pkt = hv_pkt_iter_first(channel); pkt; \
+	    pkt = hv_pkt_iter_next(channel, pkt))
 
 #endif /* _HYPERV_H */