From 0b307ebd6834012c483b8ee9801164123a54f79a Mon Sep 17 00:00:00 2001
From: stephen hemminger <stephen@networkplumber.org>
Date: Tue, 24 Jan 2017 13:05:58 -0800
Subject: netvsc: remove no longer needed receive staging buffers

The ring buffer mapping now handles the wraparound case
inside get_next_pkt_raw. Therefore it is not necessary to have an
additional special receive staging buffer.

See commit 1562edaed8c164ca5199 ("Drivers: hv: ring_buffer: count on
wrap around mappings")

Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
 drivers/net/hyperv/netvsc.c | 83 ++++++---------------------------------------
 1 file changed, 11 insertions(+), 72 deletions(-)

(limited to 'drivers/net/hyperv/netvsc.c')

diff --git a/drivers/net/hyperv/netvsc.c b/drivers/net/hyperv/netvsc.c
index 5a1cc089acb7..5e90c7fb1bd2 100644
--- a/drivers/net/hyperv/netvsc.c
+++ b/drivers/net/hyperv/netvsc.c
@@ -67,12 +67,6 @@ static struct netvsc_device *alloc_net_device(void)
 	if (!net_device)
 		return NULL;
 
-	net_device->cb_buffer = kzalloc(NETVSC_PACKET_SIZE, GFP_KERNEL);
-	if (!net_device->cb_buffer) {
-		kfree(net_device);
-		return NULL;
-	}
-
 	net_device->mrc[0].buf = vzalloc(NETVSC_RECVSLOT_MAX *
 					 sizeof(struct recv_comp_data));
 
@@ -93,7 +87,6 @@ static void free_netvsc_device(struct netvsc_device *nvdev)
 	for (i = 0; i < VRSS_CHANNEL_MAX; i++)
 		vfree(nvdev->mrc[i].buf);
 
-	kfree(nvdev->cb_buffer);
 	kfree(nvdev);
 }
 
@@ -584,7 +577,6 @@ void netvsc_device_remove(struct hv_device *device)
 	vmbus_close(device->channel);
 
 	/* Release all resources */
-	vfree(net_device->sub_cb_buf);
 	free_netvsc_device(net_device);
 }
 
@@ -1271,16 +1263,11 @@ static void netvsc_process_raw_pkt(struct hv_device *device,
 
 void netvsc_channel_cb(void *context)
 {
-	int ret;
-	struct vmbus_channel *channel = (struct vmbus_channel *)context;
+	struct vmbus_channel *channel = context;
 	u16 q_idx = channel->offermsg.offer.sub_channel_index;
 	struct hv_device *device;
 	struct netvsc_device *net_device;
-	u32 bytes_recvd;
-	u64 request_id;
 	struct vmpacket_descriptor *desc;
-	unsigned char *buffer;
-	int bufferlen = NETVSC_PACKET_SIZE;
 	struct net_device *ndev;
 	bool need_to_commit = false;
 
@@ -1292,65 +1279,19 @@ void netvsc_channel_cb(void *context)
 	net_device = get_inbound_net_device(device);
 	if (!net_device)
 		return;
+
 	ndev = hv_get_drvdata(device);
-	buffer = get_per_channel_state(channel);
-
-	do {
-		desc = get_next_pkt_raw(channel);
-		if (desc != NULL) {
-			netvsc_process_raw_pkt(device,
-					       channel,
-					       net_device,
-					       ndev,
-					       desc->trans_id,
-					       desc);
-
-			put_pkt_raw(channel, desc);
-			need_to_commit = true;
-			continue;
-		}
-		if (need_to_commit) {
-			need_to_commit = false;
-			commit_rd_index(channel);
-		}
 
-		ret = vmbus_recvpacket_raw(channel, buffer, bufferlen,
-					   &bytes_recvd, &request_id);
-		if (ret == 0) {
-			if (bytes_recvd > 0) {
-				desc = (struct vmpacket_descriptor *)buffer;
-				netvsc_process_raw_pkt(device,
-						       channel,
-						       net_device,
-						       ndev,
-						       request_id,
-						       desc);
-			} else {
-				/*
-				 * We are done for this pass.
-				 */
-				break;
-			}
-
-		} else if (ret == -ENOBUFS) {
-			if (bufferlen > NETVSC_PACKET_SIZE)
-				kfree(buffer);
-			/* Handle large packet */
-			buffer = kmalloc(bytes_recvd, GFP_ATOMIC);
-			if (buffer == NULL) {
-				/* Try again next time around */
-				netdev_err(ndev,
-					   "unable to allocate buffer of size "
-					   "(%d)!!\n", bytes_recvd);
-				break;
-			}
-
-			bufferlen = bytes_recvd;
-		}
-	} while (1);
+	while ((desc = get_next_pkt_raw(channel)) != NULL) {
+		netvsc_process_raw_pkt(device, channel, net_device,
+				       ndev, desc->trans_id, desc);
 
-	if (bufferlen > NETVSC_PACKET_SIZE)
-		kfree(buffer);
+		put_pkt_raw(channel, desc);
+		need_to_commit = true;
+	}
+
+	if (need_to_commit)
+		commit_rd_index(channel);
 
 	netvsc_chk_recv_comp(net_device, channel, q_idx);
 }
@@ -1374,8 +1315,6 @@ int netvsc_device_add(struct hv_device *device, void *additional_info)
 
 	net_device->ring_size = ring_size;
 
-	set_per_channel_state(device->channel, net_device->cb_buffer);
-
 	/* Open the channel */
 	ret = vmbus_open(device->channel, ring_size * PAGE_SIZE,
 			 ring_size * PAGE_SIZE, NULL, 0,
-- 
cgit 


From b8b835a89b2f7a7fe681983dfe5c489cb9ad9500 Mon Sep 17 00:00:00 2001
From: stephen hemminger <stephen@networkplumber.org>
Date: Tue, 24 Jan 2017 13:06:07 -0800
Subject: netvsc: group all per-channel state together

Put all the per-channel state together in one data struct.

Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
 drivers/net/hyperv/hyperv_net.h   | 14 +++++---
 drivers/net/hyperv/netvsc.c       | 73 +++++++++++++++++++--------------------
 drivers/net/hyperv/netvsc_drv.c   |  5 ++-
 drivers/net/hyperv/rndis_filter.c |  6 ++--
 4 files changed, 51 insertions(+), 47 deletions(-)

(limited to 'drivers/net/hyperv/netvsc.c')

diff --git a/drivers/net/hyperv/hyperv_net.h b/drivers/net/hyperv/hyperv_net.h
index db11f7ab67a8..fb73caad0965 100644
--- a/drivers/net/hyperv/hyperv_net.h
+++ b/drivers/net/hyperv/hyperv_net.h
@@ -714,6 +714,14 @@ struct net_device_context {
 	u32 vf_serial;
 };
 
+/* Per channel data */
+struct netvsc_channel {
+	struct vmbus_channel *channel;
+	struct multi_send_data msd;
+	struct multi_recv_comp mrc;
+	atomic_t queue_sends;
+};
+
 /* Per netvsc device */
 struct netvsc_device {
 	u32 nvsp_version;
@@ -744,27 +752,25 @@ struct netvsc_device {
 
 	struct nvsp_message revoke_packet;
 
-	struct vmbus_channel *chn_table[VRSS_CHANNEL_MAX];
 	u32 send_table[VRSS_SEND_TAB_SIZE];
 	u32 max_chn;
 	u32 num_chn;
 	spinlock_t sc_lock; /* Protects num_sc_offered variable */
 	u32 num_sc_offered;
-	atomic_t queue_sends[VRSS_CHANNEL_MAX];
 
 	/* Holds rndis device info */
 	void *extension;
 
 	int ring_size;
 
-	struct multi_send_data msd[VRSS_CHANNEL_MAX];
 	u32 max_pkt; /* max number of pkt in one send, e.g. 8 */
 	u32 pkt_align; /* alignment bytes, e.g. 8 */
 
-	struct multi_recv_comp mrc[VRSS_CHANNEL_MAX];
 	atomic_t num_outstanding_recvs;
 
 	atomic_t open_cnt;
+
+	struct netvsc_channel chan_table[VRSS_CHANNEL_MAX];
 };
 
 static inline struct netvsc_device *
diff --git a/drivers/net/hyperv/netvsc.c b/drivers/net/hyperv/netvsc.c
index 5e90c7fb1bd2..4dd2a1f2da11 100644
--- a/drivers/net/hyperv/netvsc.c
+++ b/drivers/net/hyperv/netvsc.c
@@ -67,8 +67,8 @@ static struct netvsc_device *alloc_net_device(void)
 	if (!net_device)
 		return NULL;
 
-	net_device->mrc[0].buf = vzalloc(NETVSC_RECVSLOT_MAX *
-					 sizeof(struct recv_comp_data));
+	net_device->chan_table[0].mrc.buf
+		= vzalloc(NETVSC_RECVSLOT_MAX * sizeof(struct recv_comp_data));
 
 	init_waitqueue_head(&net_device->wait_drain);
 	net_device->destroy = false;
@@ -85,7 +85,7 @@ static void free_netvsc_device(struct netvsc_device *nvdev)
 	int i;
 
 	for (i = 0; i < VRSS_CHANNEL_MAX; i++)
-		vfree(nvdev->mrc[i].buf);
+		vfree(nvdev->chan_table[i].mrc.buf);
 
 	kfree(nvdev);
 }
@@ -632,7 +632,9 @@ static void netvsc_send_tx_complete(struct netvsc_device *net_device,
 
 	num_outstanding_sends =
 		atomic_dec_return(&net_device->num_outstanding_sends);
-	queue_sends = atomic_dec_return(&net_device->queue_sends[q_idx]);
+
+	queue_sends =
+		atomic_dec_return(&net_device->chan_table[q_idx].queue_sends);
 
 	if (net_device->destroy && num_outstanding_sends == 0)
 		wake_up(&net_device->wait_drain);
@@ -757,9 +759,11 @@ static inline int netvsc_send_pkt(
 	struct sk_buff *skb)
 {
 	struct nvsp_message nvmsg;
-	u16 q_idx = packet->q_idx;
-	struct vmbus_channel *out_channel = net_device->chn_table[q_idx];
+	struct netvsc_channel *nvchan
+		= &net_device->chan_table[packet->q_idx];
+	struct vmbus_channel *out_channel = nvchan->channel;
 	struct net_device *ndev = hv_get_drvdata(device);
+	struct netdev_queue *txq = netdev_get_tx_queue(ndev, packet->q_idx);
 	u64 req_id;
 	int ret;
 	struct hv_page_buffer *pgbuf;
@@ -820,22 +824,18 @@ static inline int netvsc_send_pkt(
 
 	if (ret == 0) {
 		atomic_inc(&net_device->num_outstanding_sends);
-		atomic_inc(&net_device->queue_sends[q_idx]);
+		atomic_inc_return(&nvchan->queue_sends);
 
 		if (ring_avail < RING_AVAIL_PERCENT_LOWATER) {
-			netif_tx_stop_queue(netdev_get_tx_queue(ndev, q_idx));
+			netif_tx_stop_queue(txq);
 
-			if (atomic_read(&net_device->
-				queue_sends[q_idx]) < 1)
-				netif_tx_wake_queue(netdev_get_tx_queue(
-						    ndev, q_idx));
+			if (atomic_read(&nvchan->queue_sends) < 1)
+				netif_tx_wake_queue(txq);
 		}
 	} else if (ret == -EAGAIN) {
-		netif_tx_stop_queue(netdev_get_tx_queue(
-				    ndev, q_idx));
-		if (atomic_read(&net_device->queue_sends[q_idx]) < 1) {
-			netif_tx_wake_queue(netdev_get_tx_queue(
-					    ndev, q_idx));
+		netif_tx_stop_queue(txq);
+		if (atomic_read(&nvchan->queue_sends) < 1) {
+			netif_tx_wake_queue(txq);
 			ret = -ENOSPC;
 		}
 	} else {
@@ -866,8 +866,7 @@ int netvsc_send(struct hv_device *device,
 {
 	struct netvsc_device *net_device;
 	int ret = 0;
-	struct vmbus_channel *out_channel;
-	u16 q_idx = packet->q_idx;
+	struct netvsc_channel *nvchan;
 	u32 pktlen = packet->total_data_buflen, msd_len = 0;
 	unsigned int section_index = NETVSC_INVALID_INDEX;
 	struct multi_send_data *msdp;
@@ -887,8 +886,7 @@ int netvsc_send(struct hv_device *device,
 	if (!net_device->send_section_map)
 		return -EAGAIN;
 
-	out_channel = net_device->chn_table[q_idx];
-
+	nvchan = &net_device->chan_table[packet->q_idx];
 	packet->send_buf_index = NETVSC_INVALID_INDEX;
 	packet->cp_partial = false;
 
@@ -900,9 +898,8 @@ int netvsc_send(struct hv_device *device,
 		goto send_now;
 	}
 
-	msdp = &net_device->msd[q_idx];
-
 	/* batch packets in send buffer if possible */
+	msdp = &nvchan->msd;
 	if (msdp->pkt)
 		msd_len = msdp->pkt->total_data_buflen;
 
@@ -1003,8 +1000,9 @@ static int netvsc_send_recv_completion(struct vmbus_channel *channel,
 static inline void count_recv_comp_slot(struct netvsc_device *nvdev, u16 q_idx,
 					u32 *filled, u32 *avail)
 {
-	u32 first = nvdev->mrc[q_idx].first;
-	u32 next = nvdev->mrc[q_idx].next;
+	struct multi_recv_comp *mrc = &nvdev->chan_table[q_idx].mrc;
+	u32 first = mrc->first;
+	u32 next = mrc->next;
 
 	*filled = (first > next) ? NETVSC_RECVSLOT_MAX - first + next :
 		  next - first;
@@ -1016,26 +1014,26 @@ static inline void count_recv_comp_slot(struct netvsc_device *nvdev, u16 q_idx,
 static inline struct recv_comp_data *read_recv_comp_slot(struct netvsc_device
 							 *nvdev, u16 q_idx)
 {
+	struct multi_recv_comp *mrc = &nvdev->chan_table[q_idx].mrc;
 	u32 filled, avail;
 
-	if (!nvdev->mrc[q_idx].buf)
+	if (unlikely(!mrc->buf))
 		return NULL;
 
 	count_recv_comp_slot(nvdev, q_idx, &filled, &avail);
 	if (!filled)
 		return NULL;
 
-	return nvdev->mrc[q_idx].buf + nvdev->mrc[q_idx].first *
-	       sizeof(struct recv_comp_data);
+	return mrc->buf + mrc->first * sizeof(struct recv_comp_data);
 }
 
 /* Put the first filled slot back to available pool */
 static inline void put_recv_comp_slot(struct netvsc_device *nvdev, u16 q_idx)
 {
+	struct multi_recv_comp *mrc = &nvdev->chan_table[q_idx].mrc;
 	int num_recv;
 
-	nvdev->mrc[q_idx].first = (nvdev->mrc[q_idx].first + 1) %
-				  NETVSC_RECVSLOT_MAX;
+	mrc->first = (mrc->first + 1) % NETVSC_RECVSLOT_MAX;
 
 	num_recv = atomic_dec_return(&nvdev->num_outstanding_recvs);
 
@@ -1070,13 +1068,14 @@ static void netvsc_chk_recv_comp(struct netvsc_device *nvdev,
 static inline struct recv_comp_data *get_recv_comp_slot(
 	struct netvsc_device *nvdev, struct vmbus_channel *channel, u16 q_idx)
 {
+	struct multi_recv_comp *mrc = &nvdev->chan_table[q_idx].mrc;
 	u32 filled, avail, next;
 	struct recv_comp_data *rcd;
 
-	if (!nvdev->recv_section)
+	if (unlikely(!nvdev->recv_section))
 		return NULL;
 
-	if (!nvdev->mrc[q_idx].buf)
+	if (unlikely(!mrc->buf))
 		return NULL;
 
 	if (atomic_read(&nvdev->num_outstanding_recvs) >
@@ -1087,9 +1086,9 @@ static inline struct recv_comp_data *get_recv_comp_slot(
 	if (!avail)
 		return NULL;
 
-	next = nvdev->mrc[q_idx].next;
-	rcd = nvdev->mrc[q_idx].buf + next * sizeof(struct recv_comp_data);
-	nvdev->mrc[q_idx].next = (next + 1) % NETVSC_RECVSLOT_MAX;
+	next = mrc->next;
+	rcd = mrc->buf + next * sizeof(struct recv_comp_data);
+	mrc->next = (next + 1) % NETVSC_RECVSLOT_MAX;
 
 	atomic_inc(&nvdev->num_outstanding_recvs);
 
@@ -1159,7 +1158,7 @@ static void netvsc_receive(struct netvsc_device *net_device,
 					      channel);
 	}
 
-	if (!net_device->mrc[q_idx].buf) {
+	if (!net_device->chan_table[q_idx].mrc.buf) {
 		ret = netvsc_send_recv_completion(channel,
 						  vmxferpage_packet->d.trans_id,
 						  status);
@@ -1333,7 +1332,7 @@ int netvsc_device_add(struct hv_device *device, void *additional_info)
 	 * opened.
 	 */
 	for (i = 0; i < VRSS_CHANNEL_MAX; i++)
-		net_device->chn_table[i] = device->channel;
+		net_device->chan_table[i].channel = device->channel;
 
 	/* Writing nvdev pointer unlocks netvsc_send(), make sure chn_table is
 	 * populated.
diff --git a/drivers/net/hyperv/netvsc_drv.c b/drivers/net/hyperv/netvsc_drv.c
index f96b44eee270..da5863641703 100644
--- a/drivers/net/hyperv/netvsc_drv.c
+++ b/drivers/net/hyperv/netvsc_drv.c
@@ -135,7 +135,7 @@ static int netvsc_close(struct net_device *net)
 	while (true) {
 		aread = 0;
 		for (i = 0; i < nvdev->num_chn; i++) {
-			chn = nvdev->chn_table[i];
+			chn = nvdev->chan_table[i].channel;
 			if (!chn)
 				continue;
 
@@ -225,7 +225,7 @@ static u16 netvsc_select_queue(struct net_device *ndev, struct sk_buff *skb,
 		q_idx = new_idx;
 	}
 
-	if (unlikely(!nvsc_dev->chn_table[q_idx]))
+	if (unlikely(!nvsc_dev->chan_table[q_idx].channel))
 		q_idx = 0;
 
 	return q_idx;
@@ -545,7 +545,6 @@ no_memory:
 	++net_device_ctx->eth_stats.tx_no_memory;
 	goto drop;
 }
-
 /*
  * netvsc_linkstatus_callback - Link up/down notification
  */
diff --git a/drivers/net/hyperv/rndis_filter.c b/drivers/net/hyperv/rndis_filter.c
index fc98d2a16ddd..0df02c9808e8 100644
--- a/drivers/net/hyperv/rndis_filter.c
+++ b/drivers/net/hyperv/rndis_filter.c
@@ -1012,15 +1012,15 @@ static void netvsc_sc_open(struct vmbus_channel *new_sc)
 	if (chn_index >= nvscdev->num_chn)
 		return;
 
-	nvscdev->mrc[chn_index].buf = vzalloc(NETVSC_RECVSLOT_MAX *
-					      sizeof(struct recv_comp_data));
+	nvscdev->chan_table[chn_index].mrc.buf
+		= vzalloc(NETVSC_RECVSLOT_MAX * sizeof(struct recv_comp_data));
 
 	ret = vmbus_open(new_sc, nvscdev->ring_size * PAGE_SIZE,
 			 nvscdev->ring_size * PAGE_SIZE, NULL, 0,
 			 netvsc_channel_cb, new_sc);
 
 	if (ret == 0)
-		nvscdev->chn_table[chn_index] = new_sc;
+		nvscdev->chan_table[chn_index].channel = new_sc;
 
 	spin_lock_irqsave(&nvscdev->sc_lock, flags);
 	nvscdev->num_sc_offered--;
-- 
cgit 


From dc54a08cd3620e6457382c0cd0c8f03513dd749a Mon Sep 17 00:00:00 2001
From: stephen hemminger <stephen@networkplumber.org>
Date: Tue, 24 Jan 2017 13:06:08 -0800
Subject: netvsc: optimize receive path

Do manual optimizations of receive path:
  - remove checks for impossible conditions (but keep checks
    for bad data from host)
  - pass argument down, rather than having callee recompute what
    is already known
  - remove indirection about receive buffer datalength
  - remove dependence on VLAN_TAG_PRESENCE
  - use _hot/_cold and likely/unlikely

Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
 drivers/net/hyperv/hyperv_net.h   | 21 +++++----
 drivers/net/hyperv/netvsc.c       | 74 ++++++++++++-----------------
 drivers/net/hyperv/netvsc_drv.c   | 32 ++++++-------
 drivers/net/hyperv/rndis_filter.c | 99 +++++++++++++++------------------------
 4 files changed, 93 insertions(+), 133 deletions(-)

(limited to 'drivers/net/hyperv/netvsc.c')

diff --git a/drivers/net/hyperv/hyperv_net.h b/drivers/net/hyperv/hyperv_net.h
index fb73caad0965..4b91e37c85ca 100644
--- a/drivers/net/hyperv/hyperv_net.h
+++ b/drivers/net/hyperv/hyperv_net.h
@@ -119,6 +119,7 @@ struct ndis_recv_scale_param { /* NDIS_RECEIVE_SCALE_PARAMETERS */
 
 /* Fwd declaration */
 struct ndis_tcp_ip_checksum_info;
+struct ndis_pkt_8021q_info;
 
 /*
  * Represent netvsc packet which contains 1 RNDIS and 1 ethernet frame
@@ -186,12 +187,11 @@ int netvsc_send(struct hv_device *device,
 		struct sk_buff *skb);
 void netvsc_linkstatus_callback(struct hv_device *device_obj,
 				struct rndis_message *resp);
-int netvsc_recv_callback(struct hv_device *device_obj,
-			struct hv_netvsc_packet *packet,
-			void **data,
-			struct ndis_tcp_ip_checksum_info *csum_info,
-			struct vmbus_channel *channel,
-			u16 vlan_tci);
+int netvsc_recv_callback(struct net_device *net,
+			 struct vmbus_channel *channel,
+			 void  *data, u32 len,
+			 const struct ndis_tcp_ip_checksum_info *csum_info,
+			 const struct ndis_pkt_8021q_info *vlan);
 void netvsc_channel_cb(void *context);
 int rndis_filter_open(struct netvsc_device *nvdev);
 int rndis_filter_close(struct netvsc_device *nvdev);
@@ -200,10 +200,11 @@ int rndis_filter_device_add(struct hv_device *dev,
 void rndis_filter_device_remove(struct hv_device *dev);
 int rndis_filter_set_rss_param(struct rndis_device *rdev,
 			       const u8 *key, int num_queue);
-int rndis_filter_receive(struct hv_device *dev,
-			struct hv_netvsc_packet *pkt,
-			void **data,
-			struct vmbus_channel *channel);
+int rndis_filter_receive(struct net_device *ndev,
+			 struct netvsc_device *net_dev,
+			 struct hv_device *dev,
+			 struct vmbus_channel *channel,
+			 void *data, u32 buflen);
 
 int rndis_filter_set_packet_filter(struct rndis_device *dev, u32 new_filter);
 int rndis_filter_set_device_mac(struct net_device *ndev, char *mac);
diff --git a/drivers/net/hyperv/netvsc.c b/drivers/net/hyperv/netvsc.c
index 4dd2a1f2da11..80eecb4f9456 100644
--- a/drivers/net/hyperv/netvsc.c
+++ b/drivers/net/hyperv/netvsc.c
@@ -1095,51 +1095,35 @@ static inline struct recv_comp_data *get_recv_comp_slot(
 	return rcd;
 }
 
-static void netvsc_receive(struct netvsc_device *net_device,
-			struct vmbus_channel *channel,
-			struct hv_device *device,
-			struct vmpacket_descriptor *packet)
+static void netvsc_receive(struct net_device *ndev,
+		   struct netvsc_device *net_device,
+		   struct net_device_context *net_device_ctx,
+		   struct hv_device *device,
+		   struct vmbus_channel *channel,
+		   struct vmtransfer_page_packet_header *vmxferpage_packet,
+		   struct nvsp_message *nvsp)
 {
-	struct vmtransfer_page_packet_header *vmxferpage_packet;
-	struct nvsp_message *nvsp_packet;
-	struct hv_netvsc_packet nv_pkt;
-	struct hv_netvsc_packet *netvsc_packet = &nv_pkt;
+	char *recv_buf = net_device->recv_buf;
 	u32 status = NVSP_STAT_SUCCESS;
 	int i;
 	int count = 0;
-	struct net_device *ndev = hv_get_drvdata(device);
-	void *data;
 	int ret;
 	struct recv_comp_data *rcd;
 	u16 q_idx = channel->offermsg.offer.sub_channel_index;
 
-	/*
-	 * All inbound packets other than send completion should be xfer page
-	 * packet
-	 */
-	if (packet->type != VM_PKT_DATA_USING_XFER_PAGES) {
-		netdev_err(ndev, "Unknown packet type received - %d\n",
-			   packet->type);
-		return;
-	}
-
-	nvsp_packet = (struct nvsp_message *)((unsigned long)packet +
-			(packet->offset8 << 3));
-
 	/* Make sure this is a valid nvsp packet */
-	if (nvsp_packet->hdr.msg_type !=
-	    NVSP_MSG1_TYPE_SEND_RNDIS_PKT) {
-		netdev_err(ndev, "Unknown nvsp packet type received-"
-			" %d\n", nvsp_packet->hdr.msg_type);
+	if (unlikely(nvsp->hdr.msg_type != NVSP_MSG1_TYPE_SEND_RNDIS_PKT)) {
+		netif_err(net_device_ctx, rx_err, ndev,
+			  "Unknown nvsp packet type received %u\n",
+			  nvsp->hdr.msg_type);
 		return;
 	}
 
-	vmxferpage_packet = (struct vmtransfer_page_packet_header *)packet;
-
-	if (vmxferpage_packet->xfer_pageset_id != NETVSC_RECEIVE_BUFFER_ID) {
-		netdev_err(ndev, "Invalid xfer page set id - "
-			   "expecting %x got %x\n", NETVSC_RECEIVE_BUFFER_ID,
-			   vmxferpage_packet->xfer_pageset_id);
+	if (unlikely(vmxferpage_packet->xfer_pageset_id != NETVSC_RECEIVE_BUFFER_ID)) {
+		netif_err(net_device_ctx, rx_err, ndev,
+			  "Invalid xfer page set id - expecting %x got %x\n",
+			  NETVSC_RECEIVE_BUFFER_ID,
+			  vmxferpage_packet->xfer_pageset_id);
 		return;
 	}
 
@@ -1147,15 +1131,13 @@ static void netvsc_receive(struct netvsc_device *net_device,
 
 	/* Each range represents 1 RNDIS pkt that contains 1 ethernet frame */
 	for (i = 0; i < count; i++) {
-		/* Initialize the netvsc packet */
-		data = (void *)((unsigned long)net_device->
-			recv_buf + vmxferpage_packet->ranges[i].byte_offset);
-		netvsc_packet->total_data_buflen =
-					vmxferpage_packet->ranges[i].byte_count;
+		void *data = recv_buf
+			+ vmxferpage_packet->ranges[i].byte_offset;
+		u32 buflen = vmxferpage_packet->ranges[i].byte_count;
 
 		/* Pass it to the upper layer */
-		status = rndis_filter_receive(device, netvsc_packet, &data,
-					      channel);
+		status = rndis_filter_receive(ndev, net_device, device,
+					      channel, data, buflen);
 	}
 
 	if (!net_device->chan_table[q_idx].mrc.buf) {
@@ -1234,11 +1216,10 @@ static void netvsc_process_raw_pkt(struct hv_device *device,
 				   u64 request_id,
 				   struct vmpacket_descriptor *desc)
 {
-	struct nvsp_message *nvmsg;
 	struct net_device_context *net_device_ctx = netdev_priv(ndev);
-
-	nvmsg = (struct nvsp_message *)((unsigned long)
-		desc + (desc->offset8 << 3));
+	struct nvsp_message *nvmsg
+		= (struct nvsp_message *)((unsigned long)desc
+					  + (desc->offset8 << 3));
 
 	switch (desc->type) {
 	case VM_PKT_COMP:
@@ -1246,7 +1227,10 @@ static void netvsc_process_raw_pkt(struct hv_device *device,
 		break;
 
 	case VM_PKT_DATA_USING_XFER_PAGES:
-		netvsc_receive(net_device, channel, device, desc);
+		netvsc_receive(ndev, net_device, net_device_ctx,
+			       device, channel,
+			       (struct vmtransfer_page_packet_header *)desc,
+			       nvmsg);
 		break;
 
 	case VM_PKT_DATA_INBAND:
diff --git a/drivers/net/hyperv/netvsc_drv.c b/drivers/net/hyperv/netvsc_drv.c
index da5863641703..4bc1fdbc8cd7 100644
--- a/drivers/net/hyperv/netvsc_drv.c
+++ b/drivers/net/hyperv/netvsc_drv.c
@@ -596,13 +596,13 @@ void netvsc_linkstatus_callback(struct hv_device *device_obj,
 }
 
 static struct sk_buff *netvsc_alloc_recv_skb(struct net_device *net,
-				struct hv_netvsc_packet *packet,
-				struct ndis_tcp_ip_checksum_info *csum_info,
-				void *data, u16 vlan_tci)
+					     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, packet->total_data_buflen);
+	skb = netdev_alloc_skb_ip_align(net, buflen);
 	if (!skb)
 		return skb;
 
@@ -610,8 +610,7 @@ static struct sk_buff *netvsc_alloc_recv_skb(struct net_device *net,
 	 * Copy to skb. This copy is needed here since the memory pointed by
 	 * hv_netvsc_packet cannot be deallocated
 	 */
-	memcpy(skb_put(skb, packet->total_data_buflen), data,
-	       packet->total_data_buflen);
+	memcpy(skb_put(skb, buflen), data, buflen);
 
 	skb->protocol = eth_type_trans(skb, net);
 
@@ -628,9 +627,12 @@ static struct sk_buff *netvsc_alloc_recv_skb(struct net_device *net,
 			skb->ip_summed = CHECKSUM_UNNECESSARY;
 	}
 
-	if (vlan_tci & VLAN_TAG_PRESENT)
+	if (vlan) {
+		u16 vlan_tci = vlan->vlanid | (vlan->pri << VLAN_PRIO_SHIFT);
+
 		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q),
 				       vlan_tci);
+	}
 
 	return skb;
 }
@@ -639,14 +641,12 @@ static struct sk_buff *netvsc_alloc_recv_skb(struct net_device *net,
  * netvsc_recv_callback -  Callback when we receive a packet from the
  * "wire" on the specified device.
  */
-int netvsc_recv_callback(struct hv_device *device_obj,
-				struct hv_netvsc_packet *packet,
-				void **data,
-				struct ndis_tcp_ip_checksum_info *csum_info,
-				struct vmbus_channel *channel,
-				u16 vlan_tci)
+int netvsc_recv_callback(struct net_device *net,
+			 struct vmbus_channel *channel,
+			 void  *data, u32 len,
+			 const struct ndis_tcp_ip_checksum_info *csum_info,
+			 const struct ndis_pkt_8021q_info *vlan)
 {
-	struct net_device *net = hv_get_drvdata(device_obj);
 	struct net_device_context *net_device_ctx = netdev_priv(net);
 	struct net_device *vf_netdev;
 	struct sk_buff *skb;
@@ -668,7 +668,7 @@ int netvsc_recv_callback(struct hv_device *device_obj,
 		net = vf_netdev;
 
 	/* Allocate a skb - TODO direct I/O to pages? */
-	skb = netvsc_alloc_recv_skb(net, packet, csum_info, *data, vlan_tci);
+	skb = netvsc_alloc_recv_skb(net, csum_info, vlan, data, len);
 	if (unlikely(!skb)) {
 		++net->stats.rx_dropped;
 		rcu_read_unlock();
@@ -687,7 +687,7 @@ int netvsc_recv_callback(struct hv_device *device_obj,
 	rx_stats = this_cpu_ptr(net_device_ctx->rx_stats);
 	u64_stats_update_begin(&rx_stats->syncp);
 	rx_stats->packets++;
-	rx_stats->bytes += packet->total_data_buflen;
+	rx_stats->bytes += len;
 
 	if (skb->pkt_type == PACKET_BROADCAST)
 		++rx_stats->broadcast;
diff --git a/drivers/net/hyperv/rndis_filter.c b/drivers/net/hyperv/rndis_filter.c
index 0df02c9808e8..decce4f8fda8 100644
--- a/drivers/net/hyperv/rndis_filter.c
+++ b/drivers/net/hyperv/rndis_filter.c
@@ -132,7 +132,7 @@ static void put_rndis_request(struct rndis_device *dev,
 }
 
 static void dump_rndis_message(struct hv_device *hv_dev,
-			struct rndis_message *rndis_msg)
+			       const struct rndis_message *rndis_msg)
 {
 	struct net_device *netdev = hv_get_drvdata(hv_dev);
 
@@ -347,102 +347,78 @@ static inline void *rndis_get_ppi(struct rndis_packet *rpkt, u32 type)
 	return NULL;
 }
 
-static int rndis_filter_receive_data(struct rndis_device *dev,
-				   struct rndis_message *msg,
-				   struct hv_netvsc_packet *pkt,
-				   void **data,
-				   struct vmbus_channel *channel)
+static int rndis_filter_receive_data(struct net_device *ndev,
+				     struct rndis_device *dev,
+				     struct rndis_message *msg,
+				     struct vmbus_channel *channel,
+				     void *data, u32 data_buflen)
 {
-	struct rndis_packet *rndis_pkt;
+	struct rndis_packet *rndis_pkt = &msg->msg.pkt;
+	const struct ndis_tcp_ip_checksum_info *csum_info;
+	const struct ndis_pkt_8021q_info *vlan;
 	u32 data_offset;
-	struct ndis_pkt_8021q_info *vlan;
-	struct ndis_tcp_ip_checksum_info *csum_info;
-	u16 vlan_tci = 0;
-	struct net_device_context *net_device_ctx = netdev_priv(dev->ndev);
-
-	rndis_pkt = &msg->msg.pkt;
 
 	/* Remove the rndis header and pass it back up the stack */
 	data_offset = RNDIS_HEADER_SIZE + rndis_pkt->data_offset;
 
-	pkt->total_data_buflen -= data_offset;
+	data_buflen -= data_offset;
 
 	/*
 	 * Make sure we got a valid RNDIS message, now total_data_buflen
 	 * should be the data packet size plus the trailer padding size
 	 */
-	if (pkt->total_data_buflen < rndis_pkt->data_len) {
+	if (unlikely(data_buflen < rndis_pkt->data_len)) {
 		netdev_err(dev->ndev, "rndis message buffer "
 			   "overflow detected (got %u, min %u)"
 			   "...dropping this message!\n",
-			   pkt->total_data_buflen, rndis_pkt->data_len);
+			   data_buflen, rndis_pkt->data_len);
 		return NVSP_STAT_FAIL;
 	}
 
+	vlan = rndis_get_ppi(rndis_pkt, IEEE_8021Q_INFO);
+
 	/*
 	 * Remove the rndis trailer padding from rndis packet message
 	 * rndis_pkt->data_len tell us the real data length, we only copy
 	 * the data packet to the stack, without the rndis trailer padding
 	 */
-	pkt->total_data_buflen = rndis_pkt->data_len;
-	*data = (void *)((unsigned long)(*data) + data_offset);
-
-	vlan = rndis_get_ppi(rndis_pkt, IEEE_8021Q_INFO);
-	if (vlan) {
-		vlan_tci = VLAN_TAG_PRESENT | vlan->vlanid |
-			(vlan->pri << VLAN_PRIO_SHIFT);
-	}
-
+	data = (void *)((unsigned long)data + data_offset);
 	csum_info = rndis_get_ppi(rndis_pkt, TCPIP_CHKSUM_PKTINFO);
-	return netvsc_recv_callback(net_device_ctx->device_ctx, pkt, data,
-				    csum_info, channel, vlan_tci);
+	return netvsc_recv_callback(ndev, channel,
+				    data, rndis_pkt->data_len,
+				    csum_info, vlan);
 }
 
-int rndis_filter_receive(struct hv_device *dev,
-				struct hv_netvsc_packet	*pkt,
-				void **data,
-				struct vmbus_channel *channel)
+int rndis_filter_receive(struct net_device *ndev,
+			 struct netvsc_device *net_dev,
+			 struct hv_device *dev,
+			 struct vmbus_channel *channel,
+			 void *data, u32 buflen)
 {
-	struct net_device *ndev = hv_get_drvdata(dev);
 	struct net_device_context *net_device_ctx = netdev_priv(ndev);
-	struct netvsc_device *net_dev = net_device_ctx->nvdev;
-	struct rndis_device *rndis_dev;
-	struct rndis_message *rndis_msg;
-	int ret = 0;
-
-	if (!net_dev) {
-		ret = NVSP_STAT_FAIL;
-		goto exit;
-	}
+	struct rndis_device *rndis_dev = net_dev->extension;
+	struct rndis_message *rndis_msg = data;
 
 	/* Make sure the rndis device state is initialized */
-	if (!net_dev->extension) {
-		netdev_err(ndev, "got rndis message but no rndis device - "
-			  "dropping this message!\n");
-		ret = NVSP_STAT_FAIL;
-		goto exit;
+	if (unlikely(!rndis_dev)) {
+		netif_err(net_device_ctx, rx_err, ndev,
+			  "got rndis message but no rndis device!\n");
+		return NVSP_STAT_FAIL;
 	}
 
-	rndis_dev = (struct rndis_device *)net_dev->extension;
-	if (rndis_dev->state == RNDIS_DEV_UNINITIALIZED) {
-		netdev_err(ndev, "got rndis message but rndis device "
-			   "uninitialized...dropping this message!\n");
-		ret = NVSP_STAT_FAIL;
-		goto exit;
+	if (unlikely(rndis_dev->state == RNDIS_DEV_UNINITIALIZED)) {
+		netif_err(net_device_ctx, rx_err, ndev,
+			  "got rndis message uninitialized\n");
+		return NVSP_STAT_FAIL;
 	}
 
-	rndis_msg = *data;
-
-	if (netif_msg_rx_err(net_device_ctx))
+	if (netif_msg_rx_status(net_device_ctx))
 		dump_rndis_message(dev, rndis_msg);
 
 	switch (rndis_msg->ndis_msg_type) {
 	case RNDIS_MSG_PACKET:
-		/* data msg */
-		ret = rndis_filter_receive_data(rndis_dev, rndis_msg, pkt,
-						data, channel);
-		break;
-
+		return rndis_filter_receive_data(ndev, rndis_dev, rndis_msg,
+						 channel, data, buflen);
 	case RNDIS_MSG_INIT_C:
 	case RNDIS_MSG_QUERY_C:
 	case RNDIS_MSG_SET_C:
@@ -462,8 +438,7 @@ int rndis_filter_receive(struct hv_device *dev,
 		break;
 	}
 
-exit:
-	return ret;
+	return 0;
 }
 
 static int rndis_filter_query_device(struct rndis_device *dev, u32 oid,
-- 
cgit 


From 2c7f83ca713fa0ac5c4698e4134b09a355f60263 Mon Sep 17 00:00:00 2001
From: stephen hemminger <stephen@networkplumber.org>
Date: Tue, 24 Jan 2017 13:06:09 -0800
Subject: netvsc: don't pass void * to internal device_add

All the caller's/callee's know that the format of the device_add
parameter is a netvsc_device_info struct.

Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
 drivers/net/hyperv/hyperv_net.h   | 5 +++--
 drivers/net/hyperv/netvsc.c       | 6 +++---
 drivers/net/hyperv/rndis_filter.c | 5 ++---
 3 files changed, 8 insertions(+), 8 deletions(-)

(limited to 'drivers/net/hyperv/netvsc.c')

diff --git a/drivers/net/hyperv/hyperv_net.h b/drivers/net/hyperv/hyperv_net.h
index 4b91e37c85ca..28cbd6a2ecf8 100644
--- a/drivers/net/hyperv/hyperv_net.h
+++ b/drivers/net/hyperv/hyperv_net.h
@@ -178,7 +178,8 @@ struct rndis_device {
 /* Interface */
 struct rndis_message;
 struct netvsc_device;
-int netvsc_device_add(struct hv_device *device, void *additional_info);
+int netvsc_device_add(struct hv_device *device,
+		      const struct netvsc_device_info *info);
 void netvsc_device_remove(struct hv_device *device);
 int netvsc_send(struct hv_device *device,
 		struct hv_netvsc_packet *packet,
@@ -196,7 +197,7 @@ void netvsc_channel_cb(void *context);
 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,
-			void *additional_info);
+			    struct netvsc_device_info *info);
 void rndis_filter_device_remove(struct hv_device *dev);
 int rndis_filter_set_rss_param(struct rndis_device *rdev,
 			       const u8 *key, int num_queue);
diff --git a/drivers/net/hyperv/netvsc.c b/drivers/net/hyperv/netvsc.c
index 80eecb4f9456..359e7ef7040b 100644
--- a/drivers/net/hyperv/netvsc.c
+++ b/drivers/net/hyperv/netvsc.c
@@ -1283,11 +1283,11 @@ void netvsc_channel_cb(void *context)
  * netvsc_device_add - Callback when the device belonging to this
  * driver is added
  */
-int netvsc_device_add(struct hv_device *device, void *additional_info)
+int netvsc_device_add(struct hv_device *device,
+		      const struct netvsc_device_info *device_info)
 {
 	int i, ret = 0;
-	int ring_size =
-	((struct netvsc_device_info *)additional_info)->ring_size;
+	int ring_size = device_info->ring_size;
 	struct netvsc_device *net_device;
 	struct net_device *ndev = hv_get_drvdata(device);
 	struct net_device_context *net_device_ctx = netdev_priv(ndev);
diff --git a/drivers/net/hyperv/rndis_filter.c b/drivers/net/hyperv/rndis_filter.c
index decce4f8fda8..e3b29f35366c 100644
--- a/drivers/net/hyperv/rndis_filter.c
+++ b/drivers/net/hyperv/rndis_filter.c
@@ -1005,13 +1005,12 @@ static void netvsc_sc_open(struct vmbus_channel *new_sc)
 }
 
 int rndis_filter_device_add(struct hv_device *dev,
-			    void *additional_info)
+			    struct netvsc_device_info *device_info)
 {
 	struct net_device *net = hv_get_drvdata(dev);
 	struct net_device_context *net_device_ctx = netdev_priv(net);
 	struct netvsc_device *net_device;
 	struct rndis_device *rndis_device;
-	struct netvsc_device_info *device_info = additional_info;
 	struct ndis_offload hwcaps;
 	struct ndis_offload_params offloads;
 	struct nvsp_message *init_packet;
@@ -1035,7 +1034,7 @@ int rndis_filter_device_add(struct hv_device *dev,
 	 * NOTE! Once the channel is created, we may get a receive callback
 	 * (RndisFilterOnReceive()) before this call is completed
 	 */
-	ret = netvsc_device_add(dev, additional_info);
+	ret = netvsc_device_add(dev, device_info);
 	if (ret != 0) {
 		kfree(rndis_device);
 		return ret;
-- 
cgit 


From 46b4f7f5d1f7410de48128540ef2d1aab913a619 Mon Sep 17 00:00:00 2001
From: stephen hemminger <stephen@networkplumber.org>
Date: Tue, 24 Jan 2017 13:06:11 -0800
Subject: netvsc: eliminate per-device outstanding send counter

Since now keep track of per-queue outstanding sends, we can avoid
one atomic update by removing no longer needed per-device atomic.

Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
 drivers/net/hyperv/hyperv_net.h   |  1 -
 drivers/net/hyperv/netvsc.c       | 44 ++++++++++++++-------------------------
 drivers/net/hyperv/rndis_filter.c | 21 ++++++++++++++++---
 3 files changed, 34 insertions(+), 32 deletions(-)

(limited to 'drivers/net/hyperv/netvsc.c')

diff --git a/drivers/net/hyperv/hyperv_net.h b/drivers/net/hyperv/hyperv_net.h
index 757205c9cb93..fec365241f37 100644
--- a/drivers/net/hyperv/hyperv_net.h
+++ b/drivers/net/hyperv/hyperv_net.h
@@ -729,7 +729,6 @@ struct netvsc_channel {
 struct netvsc_device {
 	u32 nvsp_version;
 
-	atomic_t num_outstanding_sends;
 	wait_queue_head_t wait_drain;
 	bool destroy;
 
diff --git a/drivers/net/hyperv/netvsc.c b/drivers/net/hyperv/netvsc.c
index 359e7ef7040b..bd055146f098 100644
--- a/drivers/net/hyperv/netvsc.c
+++ b/drivers/net/hyperv/netvsc.c
@@ -90,29 +90,23 @@ static void free_netvsc_device(struct netvsc_device *nvdev)
 	kfree(nvdev);
 }
 
-static struct netvsc_device *get_outbound_net_device(struct hv_device *device)
-{
-	struct netvsc_device *net_device = hv_device_to_netvsc_device(device);
 
-	if (net_device && net_device->destroy)
-		net_device = NULL;
+static inline bool netvsc_channel_idle(const struct netvsc_device *net_device,
+				       u16 q_idx)
+{
+	const struct netvsc_channel *nvchan = &net_device->chan_table[q_idx];
 
-	return net_device;
+	return atomic_read(&net_device->num_outstanding_recvs) == 0 &&
+		atomic_read(&nvchan->queue_sends) == 0;
 }
 
-static struct netvsc_device *get_inbound_net_device(struct hv_device *device)
+static struct netvsc_device *get_outbound_net_device(struct hv_device *device)
 {
 	struct netvsc_device *net_device = hv_device_to_netvsc_device(device);
 
-	if (!net_device)
-		goto get_in_err;
-
-	if (net_device->destroy &&
-	    atomic_read(&net_device->num_outstanding_sends) == 0 &&
-	    atomic_read(&net_device->num_outstanding_recvs) == 0)
+	if (net_device && net_device->destroy)
 		net_device = NULL;
 
-get_in_err:
 	return net_device;
 }
 
@@ -612,7 +606,6 @@ static void netvsc_send_tx_complete(struct netvsc_device *net_device,
 	struct net_device *ndev = hv_get_drvdata(device);
 	struct net_device_context *net_device_ctx = netdev_priv(ndev);
 	struct vmbus_channel *channel = device->channel;
-	int num_outstanding_sends;
 	u16 q_idx = 0;
 	int queue_sends;
 
@@ -630,13 +623,10 @@ static void netvsc_send_tx_complete(struct netvsc_device *net_device,
 		dev_consume_skb_any(skb);
 	}
 
-	num_outstanding_sends =
-		atomic_dec_return(&net_device->num_outstanding_sends);
-
 	queue_sends =
 		atomic_dec_return(&net_device->chan_table[q_idx].queue_sends);
 
-	if (net_device->destroy && num_outstanding_sends == 0)
+	if (net_device->destroy && queue_sends == 0)
 		wake_up(&net_device->wait_drain);
 
 	if (netif_tx_queue_stopped(netdev_get_tx_queue(ndev, q_idx)) &&
@@ -823,15 +813,10 @@ static inline int netvsc_send_pkt(
 	}
 
 	if (ret == 0) {
-		atomic_inc(&net_device->num_outstanding_sends);
 		atomic_inc_return(&nvchan->queue_sends);
 
-		if (ring_avail < RING_AVAIL_PERCENT_LOWATER) {
+		if (ring_avail < RING_AVAIL_PERCENT_LOWATER)
 			netif_tx_stop_queue(txq);
-
-			if (atomic_read(&nvchan->queue_sends) < 1)
-				netif_tx_wake_queue(txq);
-		}
 	} else if (ret == -EAGAIN) {
 		netif_tx_stop_queue(txq);
 		if (atomic_read(&nvchan->queue_sends) < 1) {
@@ -1259,11 +1244,14 @@ void netvsc_channel_cb(void *context)
 	else
 		device = channel->device_obj;
 
-	net_device = get_inbound_net_device(device);
-	if (!net_device)
+	ndev = hv_get_drvdata(device);
+	if (unlikely(!ndev))
 		return;
 
-	ndev = hv_get_drvdata(device);
+	net_device = net_device_to_netvsc_device(ndev);
+	if (unlikely(net_device->destroy) &&
+	    netvsc_channel_idle(net_device, q_idx))
+		return;
 
 	while ((desc = get_next_pkt_raw(channel)) != NULL) {
 		netvsc_process_raw_pkt(device, channel, net_device,
diff --git a/drivers/net/hyperv/rndis_filter.c b/drivers/net/hyperv/rndis_filter.c
index 70b099a731a9..19356f56b7b1 100644
--- a/drivers/net/hyperv/rndis_filter.c
+++ b/drivers/net/hyperv/rndis_filter.c
@@ -903,6 +903,23 @@ cleanup:
 	return ret;
 }
 
+static bool netvsc_device_idle(const struct netvsc_device *nvdev)
+{
+	int i;
+
+	if (atomic_read(&nvdev->num_outstanding_recvs) > 0)
+		return false;
+
+	for (i = 0; i < nvdev->num_chn; i++) {
+		const struct netvsc_channel *nvchan = &nvdev->chan_table[i];
+
+		if (atomic_read(&nvchan->queue_sends) > 0)
+			return false;
+	}
+
+	return true;
+}
+
 static void rndis_filter_halt_device(struct rndis_device *dev)
 {
 	struct rndis_request *request;
@@ -933,9 +950,7 @@ cleanup:
 	spin_unlock_irqrestore(&hdev->channel->inbound_lock, flags);
 
 	/* Wait for all send completions */
-	wait_event(nvdev->wait_drain,
-		   atomic_read(&nvdev->num_outstanding_sends) == 0 &&
-		   atomic_read(&nvdev->num_outstanding_recvs) == 0);
+	wait_event(nvdev->wait_drain, netvsc_device_idle(nvdev));
 
 	if (request)
 		put_rndis_request(dev, request);
-- 
cgit 


From 793e39555511bccd73308c41205b72448d0077db Mon Sep 17 00:00:00 2001
From: stephen hemminger <stephen@networkplumber.org>
Date: Tue, 24 Jan 2017 13:06:12 -0800
Subject: netvsc: account for packets/bytes transmitted after completion

Most drivers do not increment transmit statistics until after the
transmit is completed. This will also be necessary for BQL support.

Slight additional complexity because the netvsc driver aggregates
multiple packets into one transmit.

Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
 drivers/net/hyperv/hyperv_net.h |  4 +++-
 drivers/net/hyperv/netvsc.c     | 19 ++++++++++++++++---
 drivers/net/hyperv/netvsc_drv.c | 13 +++----------
 3 files changed, 22 insertions(+), 14 deletions(-)

(limited to 'drivers/net/hyperv/netvsc.c')

diff --git a/drivers/net/hyperv/hyperv_net.h b/drivers/net/hyperv/hyperv_net.h
index fec365241f37..340f64233e2a 100644
--- a/drivers/net/hyperv/hyperv_net.h
+++ b/drivers/net/hyperv/hyperv_net.h
@@ -137,8 +137,10 @@ struct hv_netvsc_packet {
 	u8 page_buf_cnt;
 
 	u16 q_idx;
-	u32 send_buf_index;
+	u16 total_packets;
 
+	u32 total_bytes;
+	u32 send_buf_index;
 	u32 total_data_buflen;
 };
 
diff --git a/drivers/net/hyperv/netvsc.c b/drivers/net/hyperv/netvsc.c
index bd055146f098..397aa1d88a6f 100644
--- a/drivers/net/hyperv/netvsc.c
+++ b/drivers/net/hyperv/netvsc.c
@@ -611,15 +611,23 @@ static void netvsc_send_tx_complete(struct netvsc_device *net_device,
 
 	/* Notify the layer above us */
 	if (likely(skb)) {
-		struct hv_netvsc_packet *nvsc_packet
+		const struct hv_netvsc_packet *packet
 			= (struct hv_netvsc_packet *)skb->cb;
-		u32 send_index = nvsc_packet->send_buf_index;
+		u32 send_index = packet->send_buf_index;
+		struct netvsc_stats *tx_stats;
 
 		if (send_index != NETVSC_INVALID_INDEX)
 			netvsc_free_send_slot(net_device, send_index);
-		q_idx = nvsc_packet->q_idx;
+		q_idx = packet->q_idx;
 		channel = incoming_channel;
 
+		tx_stats = this_cpu_ptr(net_device_ctx->tx_stats);
+
+		u64_stats_update_begin(&tx_stats->syncp);
+		tx_stats->packets += packet->total_packets;
+		tx_stats->bytes += packet->total_bytes;
+		u64_stats_update_end(&tx_stats->syncp);
+
 		dev_consume_skb_any(skb);
 	}
 
@@ -924,6 +932,11 @@ int netvsc_send(struct hv_device *device,
 			packet->total_data_buflen += msd_len;
 		}
 
+		if (msdp->pkt) {
+			packet->total_packets += msdp->pkt->total_packets;
+			packet->total_bytes += msdp->pkt->total_bytes;
+		}
+
 		if (msdp->skb)
 			dev_consume_skb_any(msdp->skb);
 
diff --git a/drivers/net/hyperv/netvsc_drv.c b/drivers/net/hyperv/netvsc_drv.c
index 11755783c2f6..1dd13da79f02 100644
--- a/drivers/net/hyperv/netvsc_drv.c
+++ b/drivers/net/hyperv/netvsc_drv.c
@@ -364,7 +364,6 @@ static int netvsc_start_xmit(struct sk_buff *skb, struct net_device *net)
 	u32 rndis_msg_size;
 	struct rndis_per_packet_info *ppi;
 	u32 hash;
-	u32 skb_length;
 	struct hv_page_buffer page_buf[MAX_PAGE_BUFFER_COUNT];
 	struct hv_page_buffer *pb = page_buf;
 
@@ -374,7 +373,6 @@ static int netvsc_start_xmit(struct sk_buff *skb, struct net_device *net)
 	 * more pages we try linearizing it.
 	 */
 
-	skb_length = skb->len;
 	num_data_pgs = netvsc_get_slots(skb) + 2;
 
 	if (unlikely(num_data_pgs > MAX_PAGE_BUFFER_COUNT)) {
@@ -407,6 +405,8 @@ static int netvsc_start_xmit(struct sk_buff *skb, struct net_device *net)
 	packet->q_idx = skb_get_queue_mapping(skb);
 
 	packet->total_data_buflen = skb->len;
+	packet->total_bytes = skb->len;
+	packet->total_packets = 1;
 
 	rndis_msg = (struct rndis_message *)skb->head;
 
@@ -517,15 +517,8 @@ static int netvsc_start_xmit(struct sk_buff *skb, struct net_device *net)
 	skb_tx_timestamp(skb);
 	ret = netvsc_send(net_device_ctx->device_ctx, packet,
 			  rndis_msg, &pb, skb);
-	if (likely(ret == 0)) {
-		struct netvsc_stats *tx_stats = this_cpu_ptr(net_device_ctx->tx_stats);
-
-		u64_stats_update_begin(&tx_stats->syncp);
-		tx_stats->packets++;
-		tx_stats->bytes += skb_length;
-		u64_stats_update_end(&tx_stats->syncp);
+	if (likely(ret == 0))
 		return NETDEV_TX_OK;
-	}
 
 	if (ret == -EAGAIN) {
 		++net_device_ctx->eth_stats.tx_busy;
-- 
cgit 


From 6c80f3fc2398aef22798e8ac4258454b1062f3fb Mon Sep 17 00:00:00 2001
From: Simon Xiao <sixiao@microsoft.com>
Date: Tue, 24 Jan 2017 13:06:13 -0800
Subject: netvsc: report per-channel stats in ethtool statistics

Report packets and bytes transferred through a vmbus channel via ethtool.
This supersedes need for per-cpu statistics.

Example:
$ ethtool -S eth0
NIC statistics:
...
     tx_queue_0_packets: 3523179
     tx_queue_0_bytes: 505370920
     rx_queue_0_packets: 41430490
     rx_queue_0_bytes: 62714661254
     tx_queue_1_packets: 0
     tx_queue_1_bytes: 0
     rx_queue_1_packets: 0
     rx_queue_1_bytes: 0
...

Reviewed-by: Long Li <longli@microsoft.com>
Reviewed-by: K. Y. Srinivasan <kys@microsoft.com>
Reviewed-by: Haiyang Zhang <haiyangz@microsoft.com>
Signed-off-by: Simon Xiao <sixiao@microsoft.com>
Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
 drivers/net/hyperv/hyperv_net.h |   5 +-
 drivers/net/hyperv/netvsc.c     |   2 +-
 drivers/net/hyperv/netvsc_drv.c | 143 +++++++++++++++++++++++++---------------
 3 files changed, 93 insertions(+), 57 deletions(-)

(limited to 'drivers/net/hyperv/netvsc.c')

diff --git a/drivers/net/hyperv/hyperv_net.h b/drivers/net/hyperv/hyperv_net.h
index 340f64233e2a..d3e73ac158ae 100644
--- a/drivers/net/hyperv/hyperv_net.h
+++ b/drivers/net/hyperv/hyperv_net.h
@@ -699,8 +699,6 @@ struct net_device_context {
 	u32 msg_enable; /* debug level */
 
 	u32 tx_checksum_mask;
-	struct netvsc_stats __percpu *tx_stats;
-	struct netvsc_stats __percpu *rx_stats;
 
 	/* Ethtool settings */
 	u8 duplex;
@@ -725,6 +723,9 @@ struct netvsc_channel {
 	struct multi_send_data msd;
 	struct multi_recv_comp mrc;
 	atomic_t queue_sends;
+
+	struct netvsc_stats tx_stats;
+	struct netvsc_stats rx_stats;
 };
 
 /* Per netvsc device */
diff --git a/drivers/net/hyperv/netvsc.c b/drivers/net/hyperv/netvsc.c
index 397aa1d88a6f..c70c4ac77b74 100644
--- a/drivers/net/hyperv/netvsc.c
+++ b/drivers/net/hyperv/netvsc.c
@@ -621,7 +621,7 @@ static void netvsc_send_tx_complete(struct netvsc_device *net_device,
 		q_idx = packet->q_idx;
 		channel = incoming_channel;
 
-		tx_stats = this_cpu_ptr(net_device_ctx->tx_stats);
+		tx_stats = &net_device->chan_table[q_idx].tx_stats;
 
 		u64_stats_update_begin(&tx_stats->syncp);
 		tx_stats->packets += packet->total_packets;
diff --git a/drivers/net/hyperv/netvsc_drv.c b/drivers/net/hyperv/netvsc_drv.c
index 1dd13da79f02..fe0df72532a3 100644
--- a/drivers/net/hyperv/netvsc_drv.c
+++ b/drivers/net/hyperv/netvsc_drv.c
@@ -641,9 +641,12 @@ int netvsc_recv_callback(struct net_device *net,
 			 const struct ndis_pkt_8021q_info *vlan)
 {
 	struct net_device_context *net_device_ctx = netdev_priv(net);
+	struct netvsc_device *net_device = net_device_ctx->nvdev;
 	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;
@@ -669,15 +672,14 @@ int netvsc_recv_callback(struct net_device *net,
 	}
 
 	if (net != vf_netdev)
-		skb_record_rx_queue(skb,
-				    channel->offermsg.offer.sub_channel_index);
+		skb_record_rx_queue(skb, q_idx);
 
 	/*
 	 * Even if injecting the packet, record the statistics
 	 * on the synthetic device because modifying the VF device
 	 * statistics will not work correctly.
 	 */
-	rx_stats = this_cpu_ptr(net_device_ctx->rx_stats);
+	rx_stats = &net_device->chan_table[q_idx].rx_stats;
 	u64_stats_update_begin(&rx_stats->syncp);
 	rx_stats->packets++;
 	rx_stats->bytes += len;
@@ -882,34 +884,39 @@ static void netvsc_get_stats64(struct net_device *net,
 			       struct rtnl_link_stats64 *t)
 {
 	struct net_device_context *ndev_ctx = netdev_priv(net);
-	int cpu;
-
-	for_each_possible_cpu(cpu) {
-		struct netvsc_stats *tx_stats = per_cpu_ptr(ndev_ctx->tx_stats,
-							    cpu);
-		struct netvsc_stats *rx_stats = per_cpu_ptr(ndev_ctx->rx_stats,
-							    cpu);
-		u64 tx_packets, tx_bytes, rx_packets, rx_bytes, rx_multicast;
+	struct netvsc_device *nvdev = ndev_ctx->nvdev;
+	int i;
+
+	if (!nvdev)
+		return;
+
+	for (i = 0; i < nvdev->num_chn; i++) {
+		const struct netvsc_channel *nvchan = &nvdev->chan_table[i];
+		const struct netvsc_stats *stats;
+		u64 packets, bytes, multicast;
 		unsigned int start;
 
+		stats = &nvchan->tx_stats;
 		do {
-			start = u64_stats_fetch_begin_irq(&tx_stats->syncp);
-			tx_packets = tx_stats->packets;
-			tx_bytes = tx_stats->bytes;
-		} while (u64_stats_fetch_retry_irq(&tx_stats->syncp, start));
+			start = u64_stats_fetch_begin_irq(&stats->syncp);
+			packets = stats->packets;
+			bytes = stats->bytes;
+		} while (u64_stats_fetch_retry_irq(&stats->syncp, start));
+
+		t->tx_bytes	+= bytes;
+		t->tx_packets	+= packets;
 
+		stats = &nvchan->rx_stats;
 		do {
-			start = u64_stats_fetch_begin_irq(&rx_stats->syncp);
-			rx_packets = rx_stats->packets;
-			rx_bytes = rx_stats->bytes;
-			rx_multicast = rx_stats->multicast + rx_stats->broadcast;
-		} while (u64_stats_fetch_retry_irq(&rx_stats->syncp, start));
-
-		t->tx_bytes	+= tx_bytes;
-		t->tx_packets	+= tx_packets;
-		t->rx_bytes	+= rx_bytes;
-		t->rx_packets	+= rx_packets;
-		t->multicast	+= rx_multicast;
+			start = u64_stats_fetch_begin_irq(&stats->syncp);
+			packets = stats->packets;
+			bytes = stats->bytes;
+			multicast = stats->multicast + stats->broadcast;
+		} while (u64_stats_fetch_retry_irq(&stats->syncp, start));
+
+		t->rx_bytes	+= bytes;
+		t->rx_packets	+= packets;
+		t->multicast	+= multicast;
 	}
 
 	t->tx_dropped	= net->stats.tx_dropped;
@@ -954,11 +961,19 @@ static const struct {
 	{ "tx_busy",	  offsetof(struct netvsc_ethtool_stats, tx_busy) },
 };
 
+#define NETVSC_GLOBAL_STATS_LEN	ARRAY_SIZE(netvsc_stats)
+
+/* 4 statistics per queue (rx/tx packets/bytes) */
+#define NETVSC_QUEUE_STATS_LEN(dev) ((dev)->num_chn * 4)
+
 static int netvsc_get_sset_count(struct net_device *dev, int string_set)
 {
+	struct net_device_context *ndc = netdev_priv(dev);
+	struct netvsc_device *nvdev = ndc->nvdev;
+
 	switch (string_set) {
 	case ETH_SS_STATS:
-		return ARRAY_SIZE(netvsc_stats);
+		return NETVSC_GLOBAL_STATS_LEN + NETVSC_QUEUE_STATS_LEN(nvdev);
 	default:
 		return -EINVAL;
 	}
@@ -968,22 +983,63 @@ static void netvsc_get_ethtool_stats(struct net_device *dev,
 				     struct ethtool_stats *stats, u64 *data)
 {
 	struct net_device_context *ndc = netdev_priv(dev);
+	struct netvsc_device *nvdev = ndc->nvdev;
 	const void *nds = &ndc->eth_stats;
-	int i;
+	const struct netvsc_stats *qstats;
+	unsigned int start;
+	u64 packets, bytes;
+	int i, j;
 
-	for (i = 0; i < ARRAY_SIZE(netvsc_stats); i++)
+	for (i = 0; i < NETVSC_GLOBAL_STATS_LEN; i++)
 		data[i] = *(unsigned long *)(nds + netvsc_stats[i].offset);
+
+	for (j = 0; j < nvdev->num_chn; j++) {
+		qstats = &nvdev->chan_table[j].tx_stats;
+
+		do {
+			start = u64_stats_fetch_begin_irq(&qstats->syncp);
+			packets = qstats->packets;
+			bytes = qstats->bytes;
+		} while (u64_stats_fetch_retry_irq(&qstats->syncp, start));
+		data[i++] = packets;
+		data[i++] = bytes;
+
+		qstats = &nvdev->chan_table[j].rx_stats;
+		do {
+			start = u64_stats_fetch_begin_irq(&qstats->syncp);
+			packets = qstats->packets;
+			bytes = qstats->bytes;
+		} while (u64_stats_fetch_retry_irq(&qstats->syncp, start));
+		data[i++] = packets;
+		data[i++] = bytes;
+	}
 }
 
 static void netvsc_get_strings(struct net_device *dev, u32 stringset, u8 *data)
 {
+	struct net_device_context *ndc = netdev_priv(dev);
+	struct netvsc_device *nvdev = ndc->nvdev;
+	u8 *p = data;
 	int i;
 
 	switch (stringset) {
 	case ETH_SS_STATS:
 		for (i = 0; i < ARRAY_SIZE(netvsc_stats); i++)
-			memcpy(data + i * ETH_GSTRING_LEN,
+			memcpy(p + i * ETH_GSTRING_LEN,
 			       netvsc_stats[i].name, ETH_GSTRING_LEN);
+
+		p += i * ETH_GSTRING_LEN;
+		for (i = 0; i < nvdev->num_chn; i++) {
+			sprintf(p, "tx_queue_%u_packets", i);
+			p += ETH_GSTRING_LEN;
+			sprintf(p, "tx_queue_%u_bytes", i);
+			p += ETH_GSTRING_LEN;
+			sprintf(p, "rx_queue_%u_packets", i);
+			p += ETH_GSTRING_LEN;
+			sprintf(p, "rx_queue_%u_bytes", i);
+			p += ETH_GSTRING_LEN;
+		}
+
 		break;
 	}
 }
@@ -1237,15 +1293,6 @@ out_unlock:
 	rtnl_unlock();
 }
 
-static void netvsc_free_netdev(struct net_device *netdev)
-{
-	struct net_device_context *net_device_ctx = netdev_priv(netdev);
-
-	free_percpu(net_device_ctx->tx_stats);
-	free_percpu(net_device_ctx->rx_stats);
-	free_netdev(netdev);
-}
-
 static struct net_device *get_netvsc_bymac(const u8 *mac)
 {
 	struct net_device *dev;
@@ -1423,18 +1470,6 @@ static int netvsc_probe(struct hv_device *dev,
 		netdev_dbg(net, "netvsc msg_enable: %d\n",
 			   net_device_ctx->msg_enable);
 
-	net_device_ctx->tx_stats = netdev_alloc_pcpu_stats(struct netvsc_stats);
-	if (!net_device_ctx->tx_stats) {
-		free_netdev(net);
-		return -ENOMEM;
-	}
-	net_device_ctx->rx_stats = netdev_alloc_pcpu_stats(struct netvsc_stats);
-	if (!net_device_ctx->rx_stats) {
-		free_percpu(net_device_ctx->tx_stats);
-		free_netdev(net);
-		return -ENOMEM;
-	}
-
 	hv_set_drvdata(dev, net);
 
 	net_device_ctx->start_remove = false;
@@ -1460,7 +1495,7 @@ static int netvsc_probe(struct hv_device *dev,
 	ret = rndis_filter_device_add(dev, &device_info);
 	if (ret != 0) {
 		netdev_err(net, "unable to add netvsc device (ret %d)\n", ret);
-		netvsc_free_netdev(net);
+		free_netdev(net);
 		hv_set_drvdata(dev, NULL);
 		return ret;
 	}
@@ -1487,7 +1522,7 @@ static int netvsc_probe(struct hv_device *dev,
 	if (ret != 0) {
 		pr_err("Unable to register netdev.\n");
 		rndis_filter_device_remove(dev, nvdev);
-		netvsc_free_netdev(net);
+		free_netdev(net);
 	}
 
 	return ret;
@@ -1530,7 +1565,7 @@ static int netvsc_remove(struct hv_device *dev)
 
 	hv_set_drvdata(dev, NULL);
 
-	netvsc_free_netdev(net);
+	free_netdev(net);
 	return 0;
 }
 
-- 
cgit 


From b58a185801dab4eefccad868c466296795c5bae7 Mon Sep 17 00:00:00 2001
From: stephen hemminger <stephen@networkplumber.org>
Date: Tue, 24 Jan 2017 13:06:14 -0800
Subject: netvsc: simplify get next send section

Use kernel for_each_clear_bit macro to simplify finding next
available send section.

Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
 drivers/net/hyperv/netvsc.c | 28 ++++++++--------------------
 1 file changed, 8 insertions(+), 20 deletions(-)

(limited to 'drivers/net/hyperv/netvsc.c')

diff --git a/drivers/net/hyperv/netvsc.c b/drivers/net/hyperv/netvsc.c
index c70c4ac77b74..5cfdb1a1b4c1 100644
--- a/drivers/net/hyperv/netvsc.c
+++ b/drivers/net/hyperv/netvsc.c
@@ -680,27 +680,15 @@ static void netvsc_send_completion(struct netvsc_device *net_device,
 
 static u32 netvsc_get_next_send_section(struct netvsc_device *net_device)
 {
-	unsigned long index;
-	u32 max_words = net_device->map_words;
-	unsigned long *map_addr = (unsigned long *)net_device->send_section_map;
-	u32 section_cnt = net_device->send_section_cnt;
-	int ret_val = NETVSC_INVALID_INDEX;
-	int i;
-	int prev_val;
-
-	for (i = 0; i < max_words; i++) {
-		if (!~(map_addr[i]))
-			continue;
-		index = ffz(map_addr[i]);
-		prev_val = sync_test_and_set_bit(index, &map_addr[i]);
-		if (prev_val)
-			continue;
-		if ((index + (i * BITS_PER_LONG)) >= section_cnt)
-			break;
-		ret_val = (index + (i * BITS_PER_LONG));
-		break;
+	unsigned long *map_addr = net_device->send_section_map;
+	unsigned int i;
+
+	for_each_clear_bit(i, map_addr, net_device->map_words) {
+		if (sync_test_and_set_bit(i, map_addr) == 0)
+			return i;
 	}
-	return ret_val;
+
+	return NETVSC_INVALID_INDEX;
 }
 
 static u32 netvsc_copy_to_send_buf(struct netvsc_device *net_device,
-- 
cgit 


From 3454323c954775098871559b5c23d877c3e23f77 Mon Sep 17 00:00:00 2001
From: Stephen Hemminger <stephen@networkplumber.org>
Date: Sun, 5 Feb 2017 17:20:34 -0700
Subject: vmbus: remove unused kickq argument to sendpacket

Since sendpacket no longer uses kickq argument remove it.
Remove it no longer used xmit_more in sendpacket in netvsc as well.

Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
---
 drivers/hv/channel.c        | 19 +++++++++----------
 drivers/net/hyperv/netvsc.c | 21 +++------------------
 include/linux/hyperv.h      |  6 ++----
 3 files changed, 14 insertions(+), 32 deletions(-)

(limited to 'drivers/net/hyperv/netvsc.c')

diff --git a/drivers/hv/channel.c b/drivers/hv/channel.c
index e26285cde8e0..789c75f6df26 100644
--- a/drivers/hv/channel.c
+++ b/drivers/hv/channel.c
@@ -643,8 +643,8 @@ void vmbus_close(struct vmbus_channel *channel)
 EXPORT_SYMBOL_GPL(vmbus_close);
 
 int vmbus_sendpacket_ctl(struct vmbus_channel *channel, void *buffer,
-			   u32 bufferlen, u64 requestid,
-			   enum vmbus_packet_type type, u32 flags, bool kick_q)
+			 u32 bufferlen, u64 requestid,
+			 enum vmbus_packet_type type, u32 flags)
 {
 	struct vmpacket_descriptor desc;
 	u32 packetlen = sizeof(struct vmpacket_descriptor) + bufferlen;
@@ -693,7 +693,7 @@ int vmbus_sendpacket(struct vmbus_channel *channel, void *buffer,
 			   enum vmbus_packet_type type, u32 flags)
 {
 	return vmbus_sendpacket_ctl(channel, buffer, bufferlen, requestid,
-				    type, flags, true);
+				    type, flags);
 }
 EXPORT_SYMBOL(vmbus_sendpacket);
 
@@ -705,11 +705,9 @@ EXPORT_SYMBOL(vmbus_sendpacket);
  * explicitly.
  */
 int vmbus_sendpacket_pagebuffer_ctl(struct vmbus_channel *channel,
-				     struct hv_page_buffer pagebuffers[],
-				     u32 pagecount, void *buffer, u32 bufferlen,
-				     u64 requestid,
-				     u32 flags,
-				     bool kick_q)
+				    struct hv_page_buffer pagebuffers[],
+				    u32 pagecount, void *buffer, u32 bufferlen,
+				    u64 requestid, u32 flags)
 {
 	int i;
 	struct vmbus_channel_packet_page_buffer desc;
@@ -769,9 +767,10 @@ int vmbus_sendpacket_pagebuffer(struct vmbus_channel *channel,
 				     u64 requestid)
 {
 	u32 flags = VMBUS_DATA_PACKET_FLAG_COMPLETION_REQUESTED;
+
 	return vmbus_sendpacket_pagebuffer_ctl(channel, pagebuffers, pagecount,
-					       buffer, bufferlen, requestid,
-					       flags, true);
+					       buffer, bufferlen,
+					       requestid, flags);
 
 }
 EXPORT_SYMBOL_GPL(vmbus_sendpacket_pagebuffer);
diff --git a/drivers/net/hyperv/netvsc.c b/drivers/net/hyperv/netvsc.c
index 86e5749226ef..372603023053 100644
--- a/drivers/net/hyperv/netvsc.c
+++ b/drivers/net/hyperv/netvsc.c
@@ -723,8 +723,6 @@ static u32 netvsc_copy_to_send_buf(struct netvsc_device *net_device,
 	char *dest = start + (section_index * net_device->send_section_size)
 		     + pend_size;
 	int i;
-	bool is_data_pkt = (skb != NULL) ? true : false;
-	bool xmit_more = (skb != NULL) ? skb->xmit_more : false;
 	u32 msg_size = 0;
 	u32 padding = 0;
 	u32 remain = packet->total_data_buflen % net_device->pkt_align;
@@ -732,7 +730,7 @@ static u32 netvsc_copy_to_send_buf(struct netvsc_device *net_device,
 		packet->page_buf_cnt;
 
 	/* Add padding */
-	if (is_data_pkt && xmit_more && remain &&
+	if (skb && skb->xmit_more && remain &&
 	    !packet->cp_partial) {
 		padding = net_device->pkt_align - remain;
 		rndis_msg->msg_len += padding;
@@ -772,7 +770,6 @@ static inline int netvsc_send_pkt(
 	int ret;
 	struct hv_page_buffer *pgbuf;
 	u32 ring_avail = hv_ringbuf_avail_percent(&out_channel->outbound);
-	bool xmit_more = (skb != NULL) ? skb->xmit_more : false;
 
 	nvmsg.hdr.msg_type = NVSP_MSG1_TYPE_SEND_RNDIS_PKT;
 	if (skb != NULL) {
@@ -796,16 +793,6 @@ static inline int netvsc_send_pkt(
 	if (out_channel->rescind)
 		return -ENODEV;
 
-	/*
-	 * It is possible that once we successfully place this packet
-	 * on the ringbuffer, we may stop the queue. In that case, we want
-	 * to notify the host independent of the xmit_more flag. We don't
-	 * need to be precise here; in the worst case we may signal the host
-	 * unnecessarily.
-	 */
-	if (ring_avail < (RING_AVAIL_PERCENT_LOWATER + 1))
-		xmit_more = false;
-
 	if (packet->page_buf_cnt) {
 		pgbuf = packet->cp_partial ? (*pb) +
 			packet->rmsg_pgcnt : (*pb);
@@ -815,15 +802,13 @@ static inline int netvsc_send_pkt(
 						      &nvmsg,
 						      sizeof(struct nvsp_message),
 						      req_id,
-						      VMBUS_DATA_PACKET_FLAG_COMPLETION_REQUESTED,
-						      !xmit_more);
+						      VMBUS_DATA_PACKET_FLAG_COMPLETION_REQUESTED);
 	} else {
 		ret = vmbus_sendpacket_ctl(out_channel, &nvmsg,
 					   sizeof(struct nvsp_message),
 					   req_id,
 					   VM_PKT_DATA_INBAND,
-					   VMBUS_DATA_PACKET_FLAG_COMPLETION_REQUESTED,
-					   !xmit_more);
+					   VMBUS_DATA_PACKET_FLAG_COMPLETION_REQUESTED);
 	}
 
 	if (ret == 0) {
diff --git a/include/linux/hyperv.h b/include/linux/hyperv.h
index 7795966af0f9..e208e6437f5b 100644
--- a/include/linux/hyperv.h
+++ b/include/linux/hyperv.h
@@ -1037,8 +1037,7 @@ extern int vmbus_sendpacket_ctl(struct vmbus_channel *channel,
 				  u32 bufferLen,
 				  u64 requestid,
 				  enum vmbus_packet_type type,
-				  u32 flags,
-				  bool kick_q);
+				  u32 flags);
 
 extern int vmbus_sendpacket_pagebuffer(struct vmbus_channel *channel,
 					    struct hv_page_buffer pagebuffers[],
@@ -1053,8 +1052,7 @@ extern int vmbus_sendpacket_pagebuffer_ctl(struct vmbus_channel *channel,
 					   void *buffer,
 					   u32 bufferlen,
 					   u64 requestid,
-					   u32 flags,
-					   bool kick_q);
+					   u32 flags);
 
 extern int vmbus_sendpacket_multipagebuffer(struct vmbus_channel *channel,
 					struct hv_multipage_buffer *mpb,
-- 
cgit 


From 7ce101246655935b014b11d81f815342921f5654 Mon Sep 17 00:00:00 2001
From: stephen hemminger <stephen@networkplumber.org>
Date: Thu, 9 Mar 2017 14:58:29 -0800
Subject: netvsc: handle select_queue when device is being removed

Move the send indirection table from the inner device (netvsc)
to the network device context.

It is possible that netvsc_device is not present (remove in progress).
This solves potential use after free issues when packet is being
created during MTU change, shutdown, or queue count changes.

Fixes: d8e18ee0fa96 ("netvsc: enhance transmit select_queue")
Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
 drivers/net/hyperv/hyperv_net.h |  3 ++-
 drivers/net/hyperv/netvsc.c     |  8 ++------
 drivers/net/hyperv/netvsc_drv.c | 11 +++--------
 3 files changed, 7 insertions(+), 15 deletions(-)

(limited to 'drivers/net/hyperv/netvsc.c')

diff --git a/drivers/net/hyperv/hyperv_net.h b/drivers/net/hyperv/hyperv_net.h
index d3e73ac158ae..f9f3dba7a588 100644
--- a/drivers/net/hyperv/hyperv_net.h
+++ b/drivers/net/hyperv/hyperv_net.h
@@ -700,6 +700,8 @@ struct net_device_context {
 
 	u32 tx_checksum_mask;
 
+	u32 tx_send_table[VRSS_SEND_TAB_SIZE];
+
 	/* Ethtool settings */
 	u8 duplex;
 	u32 speed;
@@ -757,7 +759,6 @@ struct netvsc_device {
 
 	struct nvsp_message revoke_packet;
 
-	u32 send_table[VRSS_SEND_TAB_SIZE];
 	u32 max_chn;
 	u32 num_chn;
 	spinlock_t sc_lock; /* Protects num_sc_offered variable */
diff --git a/drivers/net/hyperv/netvsc.c b/drivers/net/hyperv/netvsc.c
index d35ebd993b38..4c1d8cca247b 100644
--- a/drivers/net/hyperv/netvsc.c
+++ b/drivers/net/hyperv/netvsc.c
@@ -1136,15 +1136,11 @@ static void netvsc_receive(struct net_device *ndev,
 static void netvsc_send_table(struct hv_device *hdev,
 			      struct nvsp_message *nvmsg)
 {
-	struct netvsc_device *nvscdev;
 	struct net_device *ndev = hv_get_drvdata(hdev);
+	struct net_device_context *net_device_ctx = netdev_priv(ndev);
 	int i;
 	u32 count, *tab;
 
-	nvscdev = get_outbound_net_device(hdev);
-	if (!nvscdev)
-		return;
-
 	count = nvmsg->msg.v5_msg.send_table.count;
 	if (count != VRSS_SEND_TAB_SIZE) {
 		netdev_err(ndev, "Received wrong send-table size:%u\n", count);
@@ -1155,7 +1151,7 @@ static void netvsc_send_table(struct hv_device *hdev,
 		      nvmsg->msg.v5_msg.send_table.offset);
 
 	for (i = 0; i < count; i++)
-		nvscdev->send_table[i] = tab[i];
+		net_device_ctx->tx_send_table[i] = tab[i];
 }
 
 static void netvsc_send_vf(struct net_device_context *net_device_ctx,
diff --git a/drivers/net/hyperv/netvsc_drv.c b/drivers/net/hyperv/netvsc_drv.c
index bc05c895d958..5ede87f30463 100644
--- a/drivers/net/hyperv/netvsc_drv.c
+++ b/drivers/net/hyperv/netvsc_drv.c
@@ -206,17 +206,15 @@ static u16 netvsc_select_queue(struct net_device *ndev, struct sk_buff *skb,
 			void *accel_priv, select_queue_fallback_t fallback)
 {
 	struct net_device_context *net_device_ctx = netdev_priv(ndev);
-	struct netvsc_device *nvsc_dev = net_device_ctx->nvdev;
+	unsigned int num_tx_queues = ndev->real_num_tx_queues;
 	struct sock *sk = skb->sk;
 	int q_idx = sk_tx_queue_get(sk);
 
-	if (q_idx < 0 || skb->ooo_okay ||
-	    q_idx >= ndev->real_num_tx_queues) {
+	if (q_idx < 0 || skb->ooo_okay || q_idx >= num_tx_queues) {
 		u16 hash = __skb_tx_hash(ndev, skb, VRSS_SEND_TAB_SIZE);
 		int new_idx;
 
-		new_idx = nvsc_dev->send_table[hash]
-			% nvsc_dev->num_chn;
+		new_idx = net_device_ctx->tx_send_table[hash] % num_tx_queues;
 
 		if (q_idx != new_idx && sk &&
 		    sk_fullsock(sk) && rcu_access_pointer(sk->sk_dst_cache))
@@ -225,9 +223,6 @@ static u16 netvsc_select_queue(struct net_device *ndev, struct sk_buff *skb,
 		q_idx = new_idx;
 	}
 
-	if (unlikely(!nvsc_dev->chan_table[q_idx].channel))
-		q_idx = 0;
-
 	return q_idx;
 }
 
-- 
cgit