Merge branch 'netvsc-bug-fixes-and-cleanups'

Stephen Hemminger says:

====================
netvsc: bug fixes and cleanups

These fix NAPI issues and bugs found during testing of shutdown
testing.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

drivers/net/hyperv/hyperv_net.h

@@ -686,7 +686,7 @@ struct net_device_context {
 	/* point back to our device context */
 	struct hv_device *device_ctx;
 	/* netvsc_device */
-	struct netvsc_device *nvdev;
+	struct netvsc_device __rcu *nvdev;
 	/* reconfigure work */
 	struct delayed_work dwork;
 	/* last reconfig time */
@@ -708,9 +708,6 @@ struct net_device_context {
 	u32 speed;
 	struct netvsc_ethtool_stats eth_stats;
 
-	/* the device is going away */
-	bool start_remove;
-
 	/* State to manage the associated VF interface. */
 	struct net_device __rcu *vf_netdev;
 
@@ -723,6 +720,7 @@ struct net_device_context {
 /* Per channel data */
 struct netvsc_channel {
 	struct vmbus_channel *channel;
+	const struct vmpacket_descriptor *desc;
 	struct napi_struct napi;
 	struct multi_send_data msd;
 	struct multi_recv_comp mrc;
@@ -763,8 +761,8 @@ struct netvsc_device {
 
 	u32 max_chn;
 	u32 num_chn;
-	spinlock_t sc_lock; /* Protects num_sc_offered variable */
-	u32 num_sc_offered;
+
+	refcount_t sc_offered;
 
 	/* Holds rndis device info */
 	void *extension;
@@ -779,6 +777,8 @@ struct netvsc_device {
 	atomic_t open_cnt;
 
 	struct netvsc_channel chan_table[VRSS_CHANNEL_MAX];
+
+	struct rcu_head rcu;
 };
 
 static inline struct netvsc_device *

drivers/net/hyperv/netvsc.c

@@ -80,8 +80,10 @@ static struct netvsc_device *alloc_net_device(void)
 	return net_device;
 }
 
-static void free_netvsc_device(struct netvsc_device *nvdev)
+static void free_netvsc_device(struct rcu_head *head)
 {
+	struct netvsc_device *nvdev
+		= container_of(head, struct netvsc_device, rcu);
 	int i;
 
 	for (i = 0; i < VRSS_CHANNEL_MAX; i++)
@@ -90,6 +92,10 @@ static void free_netvsc_device(struct netvsc_device *nvdev)
 	kfree(nvdev);
 }
 
+static void free_netvsc_device_rcu(struct netvsc_device *nvdev)
+{
+	call_rcu(&nvdev->rcu, free_netvsc_device);
+}
 
 static struct netvsc_device *get_outbound_net_device(struct hv_device *device)
 {
@@ -551,7 +557,7 @@ void netvsc_device_remove(struct hv_device *device)
 
 	netvsc_disconnect_vsp(device);
 
-	net_device_ctx->nvdev = NULL;
+	RCU_INIT_POINTER(net_device_ctx->nvdev, NULL);
 
 	/*
 	 * At this point, no one should be accessing net_device
@@ -566,7 +572,7 @@ void netvsc_device_remove(struct hv_device *device)
 		napi_disable(&net_device->chan_table[i].napi);
 
 	/* Release all resources */
-	free_netvsc_device(net_device);
+	free_netvsc_device_rcu(net_device);
 }
 
 #define RING_AVAIL_PERCENT_HIWATER 20
@@ -599,7 +605,6 @@ static void netvsc_send_tx_complete(struct netvsc_device *net_device,
 {
 	struct sk_buff *skb = (struct sk_buff *)(unsigned long)desc->trans_id;
 	struct net_device *ndev = hv_get_drvdata(device);
-	struct net_device_context *net_device_ctx = netdev_priv(ndev);
 	struct vmbus_channel *channel = device->channel;
 	u16 q_idx = 0;
 	int queue_sends;
@@ -633,7 +638,6 @@ static void netvsc_send_tx_complete(struct netvsc_device *net_device,
 		wake_up(&net_device->wait_drain);
 
 	if (netif_tx_queue_stopped(netdev_get_tx_queue(ndev, q_idx)) &&
-	    !net_device_ctx->start_remove &&
 	    (hv_ringbuf_avail_percent(&channel->outbound) > RING_AVAIL_PERCENT_HIWATER ||
 	     queue_sends < 1))
 		netif_tx_wake_queue(netdev_get_tx_queue(ndev, q_idx));
@@ -702,8 +706,7 @@ static u32 netvsc_copy_to_send_buf(struct netvsc_device *net_device,
 		packet->page_buf_cnt;
 
 	/* Add padding */
-	if (skb && skb->xmit_more && remain &&
-	    !packet->cp_partial) {
+	if (skb->xmit_more && remain && !packet->cp_partial) {
 		padding = net_device->pkt_align - remain;
 		rndis_msg->msg_len += padding;
 		packet->total_data_buflen += padding;
@@ -861,9 +864,7 @@ int netvsc_send(struct hv_device *device,
 	if (msdp->pkt)
 		msd_len = msdp->pkt->total_data_buflen;
 
-	try_batch = (skb != NULL) && msd_len > 0 && msdp->count <
-		    net_device->max_pkt;
-
+	try_batch =  msd_len > 0 && msdp->count < net_device->max_pkt;
 	if (try_batch && msd_len + pktlen + net_device->pkt_align <
 	    net_device->send_section_size) {
 		section_index = msdp->pkt->send_buf_index;
@@ -873,7 +874,7 @@ int netvsc_send(struct hv_device *device,
 		section_index = msdp->pkt->send_buf_index;
 		packet->cp_partial = true;
 
-	} else if ((skb != NULL) && pktlen + net_device->pkt_align <
+	} else if (pktlen + net_device->pkt_align <
 		   net_device->send_section_size) {
 		section_index = netvsc_get_next_send_section(net_device);
 		if (section_index != NETVSC_INVALID_INDEX) {
@@ -1173,7 +1174,6 @@ static int netvsc_process_raw_pkt(struct hv_device *device,
 				  struct vmbus_channel *channel,
 				  struct netvsc_device *net_device,
 				  struct net_device *ndev,
-				  u64 request_id,
 				  const struct vmpacket_descriptor *desc)
 {
 	struct net_device_context *net_device_ctx = netdev_priv(ndev);
@@ -1195,7 +1195,7 @@ static int netvsc_process_raw_pkt(struct hv_device *device,
 
 	default:
 		netdev_err(ndev, "unhandled packet type %d, tid %llx\n",
-			   desc->type, request_id);
+			   desc->type, desc->trans_id);
 		break;
 	}
 
@@ -1222,28 +1222,20 @@ int netvsc_poll(struct napi_struct *napi, int budget)
 	u16 q_idx = channel->offermsg.offer.sub_channel_index;
 	struct net_device *ndev = hv_get_drvdata(device);
 	struct netvsc_device *net_device = net_device_to_netvsc_device(ndev);
-	const struct vmpacket_descriptor *desc;
 	int work_done = 0;
 
-	desc = hv_pkt_iter_first(channel);
-	while (desc) {
-		int count;
-
-		count = netvsc_process_raw_pkt(device, channel, net_device,
-					       ndev, desc->trans_id, desc);
-		work_done += count;
-		desc = __hv_pkt_iter_next(channel, desc);
+	/* If starting a new interval */
+	if (!nvchan->desc)
+		nvchan->desc = hv_pkt_iter_first(channel);
 
-		/* If receive packet budget is exhausted, reschedule */
-		if (work_done >= budget) {
-			work_done = budget;
-			break;
-		}
+	while (nvchan->desc && work_done < budget) {
+		work_done += netvsc_process_raw_pkt(device, channel, net_device,
+						    ndev, nvchan->desc);
+		nvchan->desc = hv_pkt_iter_next(channel, nvchan->desc);
 	}
-	hv_pkt_iter_close(channel);
 
-	/* If budget was not exhausted and
-	 * not doing busy poll
+	/* If receive ring was exhausted
+	 * and not doing busy poll
 	 * then re-enable host interrupts
 	 *  and reschedule if ring is not empty.
 	 */
@@ -1253,7 +1245,9 @@ int netvsc_poll(struct napi_struct *napi, int budget)
 		napi_reschedule(napi);
 
 	netvsc_chk_recv_comp(net_device, channel, q_idx);
-	return work_done;
+
+	/* Driver may overshoot since multiple packets per descriptor */
+	return min(work_done, budget);
 }
 
 /* Call back when data is available in host ring buffer.
@@ -1263,10 +1257,12 @@ void netvsc_channel_cb(void *context)
 {
 	struct netvsc_channel *nvchan = context;
 
+	if (napi_schedule_prep(&nvchan->napi)) {
 		/* disable interupts from host */
 		hv_begin_read(&nvchan->channel->inbound);
 
-	napi_schedule(&nvchan->napi);
+		__napi_schedule(&nvchan->napi);
+	}
 }
 
 /*
@@ -1325,9 +1321,7 @@ int netvsc_device_add(struct hv_device *device,
 	/* Writing nvdev pointer unlocks netvsc_send(), make sure chn_table is
 	 * populated.
 	 */
-	wmb();
-
-	net_device_ctx->nvdev = net_device;
+	rcu_assign_pointer(net_device_ctx->nvdev, net_device);
 
 	/* Connect with the NetVsp */
 	ret = netvsc_connect_vsp(device);
@@ -1346,7 +1340,7 @@ int netvsc_device_add(struct hv_device *device,
 	vmbus_close(device->channel);
 
 cleanup:
-	free_netvsc_device(net_device);
+	free_netvsc_device(&net_device->rcu);
 
 	return ret;
 }

drivers/net/hyperv/rndis_filter.c

@@ -819,16 +819,14 @@ int rndis_filter_set_packet_filter(struct rndis_device *dev, u32 new_filter)
 {
 	struct rndis_request *request;
 	struct rndis_set_request *set;
-	struct rndis_set_complete *set_complete;
 	int ret;
 
 	request = get_rndis_request(dev, RNDIS_MSG_SET,
 			RNDIS_MESSAGE_SIZE(struct rndis_set_request) +
 			sizeof(u32));
-	if (!request) {
-		ret = -ENOMEM;
-		goto cleanup;
-	}
+	if (!request)
+		return -ENOMEM;
+
 
 	/* Setup the rndis set */
 	set = &request->request_msg.msg.set_req;
@@ -840,15 +838,11 @@ int rndis_filter_set_packet_filter(struct rndis_device *dev, u32 new_filter)
 	       &new_filter, sizeof(u32));
 
 	ret = rndis_filter_send_request(dev, request);
-	if (ret != 0)
-		goto cleanup;
-
+	if (ret == 0)
 		wait_for_completion(&request->wait_event);
 
-	set_complete = &request->response_msg.msg.set_complete;
-cleanup:
-	if (request)
 	put_rndis_request(dev, request);
+
 	return ret;
 }
 
@@ -926,8 +920,6 @@ static void rndis_filter_halt_device(struct rndis_device *dev)
 	struct rndis_halt_request *halt;
 	struct net_device_context *net_device_ctx = netdev_priv(dev->ndev);
 	struct netvsc_device *nvdev = net_device_ctx->nvdev;
-	struct hv_device *hdev = net_device_ctx->device_ctx;
-	ulong flags;
 
 	/* Attempt to do a rndis device halt */
 	request = get_rndis_request(dev, RNDIS_MSG_HALT,
@@ -945,9 +937,10 @@ static void rndis_filter_halt_device(struct rndis_device *dev)
 	dev->state = RNDIS_DEV_UNINITIALIZED;
 
 cleanup:
-	spin_lock_irqsave(&hdev->channel->inbound_lock, flags);
 	nvdev->destroy = true;
-	spin_unlock_irqrestore(&hdev->channel->inbound_lock, flags);
+
+	/* Force flag to be ordered before waiting */
+	wmb();
 
 	/* Wait for all send completions */
 	wait_event(nvdev->wait_drain, netvsc_device_idle(nvdev));
@@ -997,7 +990,6 @@ static void netvsc_sc_open(struct vmbus_channel *new_sc)
 	struct netvsc_device *nvscdev = net_device_to_netvsc_device(ndev);
 	u16 chn_index = new_sc->offermsg.offer.sub_channel_index;
 	struct netvsc_channel *nvchan;
-	unsigned long flags;
 	int ret;
 
 	if (chn_index >= nvscdev->num_chn)
@@ -1019,10 +1011,7 @@ static void netvsc_sc_open(struct vmbus_channel *new_sc)
 
 	napi_enable(&nvchan->napi);
 
-	spin_lock_irqsave(&nvscdev->sc_lock, flags);
-	nvscdev->num_sc_offered--;
-	spin_unlock_irqrestore(&nvscdev->sc_lock, flags);
-	if (nvscdev->num_sc_offered == 0)
+	if (refcount_dec_and_test(&nvscdev->sc_offered))
 		complete(&nvscdev->channel_init_wait);
 }
 
@@ -1039,12 +1028,9 @@ int rndis_filter_device_add(struct hv_device *dev,
 	struct ndis_recv_scale_cap rsscap;
 	u32 rsscap_size = sizeof(struct ndis_recv_scale_cap);
 	unsigned int gso_max_size = GSO_MAX_SIZE;
-	u32 mtu, size;
-	u32 num_rss_qs;
-	u32 sc_delta;
+	u32 mtu, size, num_rss_qs;
 	const struct cpumask *node_cpu_mask;
 	u32 num_possible_rss_qs;
-	unsigned long flags;
 	int i, ret;
 
 	rndis_device = get_rndis_device();
@@ -1067,7 +1053,7 @@ int rndis_filter_device_add(struct hv_device *dev,
 	net_device->max_chn = 1;
 	net_device->num_chn = 1;
 
-	spin_lock_init(&net_device->sc_lock);
+	refcount_set(&net_device->sc_offered, 0);
 
 	net_device->extension = rndis_device;
 	rndis_device->ndev = net;
@@ -1181,34 +1167,30 @@ int rndis_filter_device_add(struct hv_device *dev,
 	if (ret || rsscap.num_recv_que < 2)
 		goto out;
 
-	net_device->max_chn = min_t(u32, VRSS_CHANNEL_MAX, rsscap.num_recv_que);
-
-	num_rss_qs = min(device_info->max_num_vrss_chns, net_device->max_chn);
-
 	/*
 	 * We will limit the VRSS channels to the number CPUs in the NUMA node
 	 * the primary channel is currently bound to.
+	 *
+	 * This also guarantees that num_possible_rss_qs <= num_online_cpus
 	 */
 	node_cpu_mask = cpumask_of_node(cpu_to_node(dev->channel->target_cpu));
-	num_possible_rss_qs = cpumask_weight(node_cpu_mask);
+	num_possible_rss_qs = min_t(u32, cpumask_weight(node_cpu_mask),
+				    rsscap.num_recv_que);
 
-	/* We will use the given number of channels if available. */
-	if (device_info->num_chn && device_info->num_chn < net_device->max_chn)
-		net_device->num_chn = device_info->num_chn;
-	else
-		net_device->num_chn = min(num_possible_rss_qs, num_rss_qs);
+	net_device->max_chn = min_t(u32, VRSS_CHANNEL_MAX, num_possible_rss_qs);
 
-	num_rss_qs = net_device->num_chn - 1;
+	/* We will use the given number of channels if available. */
+	net_device->num_chn = min(net_device->max_chn, device_info->num_chn);
 
 	for (i = 0; i < ITAB_NUM; i++)
 		rndis_device->ind_table[i] = ethtool_rxfh_indir_default(i,
 							net_device->num_chn);
 
-	net_device->num_sc_offered = num_rss_qs;
-
-	if (net_device->num_chn == 1)
-		goto out;
+	num_rss_qs = net_device->num_chn - 1;
+	if (num_rss_qs == 0)
+		return 0;
 
+	refcount_set(&net_device->sc_offered, num_rss_qs);
 	vmbus_set_sc_create_callback(dev->channel, netvsc_sc_open);
 
 	init_packet = &net_device->channel_init_pkt;
@@ -1224,32 +1206,23 @@ int rndis_filter_device_add(struct hv_device *dev,
 			       VMBUS_DATA_PACKET_FLAG_COMPLETION_REQUESTED);
 	if (ret)
 		goto out;
-	wait_for_completion(&net_device->channel_init_wait);
 
-	if (init_packet->msg.v5_msg.subchn_comp.status !=
-	    NVSP_STAT_SUCCESS) {
+	if (init_packet->msg.v5_msg.subchn_comp.status != NVSP_STAT_SUCCESS) {
 		ret = -ENODEV;
 		goto out;
 	}
+	wait_for_completion(&net_device->channel_init_wait);
+
 	net_device->num_chn = 1 +
 		init_packet->msg.v5_msg.subchn_comp.num_subchannels;
 
-	ret = rndis_filter_set_rss_param(rndis_device, netvsc_hash_key,
+	/* ignore failues from setting rss parameters, still have channels */
+	rndis_filter_set_rss_param(rndis_device, netvsc_hash_key,
 				   net_device->num_chn);
-
-	/*
-	 * Set the number of sub-channels to be received.
-	 */
-	spin_lock_irqsave(&net_device->sc_lock, flags);
-	sc_delta = num_rss_qs - (net_device->num_chn - 1);
-	net_device->num_sc_offered -= sc_delta;
-	spin_unlock_irqrestore(&net_device->sc_lock, flags);
-
 out:
 	if (ret) {
 		net_device->max_chn = 1;
 		net_device->num_chn = 1;
-		net_device->num_sc_offered = 0;
 	}
 
 	return 0; /* return 0 because primary channel can be used alone */
@@ -1264,12 +1237,6 @@ void rndis_filter_device_remove(struct hv_device *dev,
 {
 	struct rndis_device *rndis_dev = net_dev->extension;
 
-	/* If not all subchannel offers are complete, wait for them until
-	 * completion to avoid race.
-	 */
-	if (net_dev->num_sc_offered > 0)
-		wait_for_completion(&net_dev->channel_init_wait);
-
 	/* Halt and release the rndis device */
 	rndis_filter_halt_device(rndis_dev);