Merge branch 'xen-netback-control-ring'

Paul Durrant says:

====================
xen-netback: support for control ring

My recent patch to import an up-to-date include/xen/interface/io/netif.h
from the Xen Project brought in the necessary definitions to support the
new control shared ring and protocol. This patch series updates xen-netback
to support the new ring.

Patch #1 adds the necessary boilerplate to map the control ring and handle
messages. No implementation of the new protocol is included in this patch
so that it can be kept to a reasonable size.

Patch #2 adds the protocol implementation.

Patch #3 adds support for passing has values calculated by xen-netback to
capable frontends.

Patch #4 adds support for accepting hash values calculated by capable
frontends and using them the set the socket buffer hash.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

drivers/net/xen-netback/Makefile

 obj-$(CONFIG_XEN_NETDEV_BACKEND) := xen-netback.o
 
-xen-netback-y := netback.o xenbus.o interface.o
+xen-netback-y := netback.o xenbus.o interface.o hash.o

drivers/net/xen-netback/common.h

@@ -220,6 +220,35 @@ struct xenvif_mcast_addr {
 
 #define XEN_NETBK_MCAST_MAX 64
 
+#define XEN_NETBK_MAX_HASH_KEY_SIZE 40
+#define XEN_NETBK_MAX_HASH_MAPPING_SIZE 128
+#define XEN_NETBK_HASH_TAG_SIZE 40
+
+struct xenvif_hash_cache_entry {
+	struct list_head link;
+	struct rcu_head rcu;
+	u8 tag[XEN_NETBK_HASH_TAG_SIZE];
+	unsigned int len;
+	u32 val;
+	int seq;
+};
+
+struct xenvif_hash_cache {
+	spinlock_t lock;
+	struct list_head list;
+	unsigned int count;
+	atomic_t seq;
+};
+
+struct xenvif_hash {
+	unsigned int alg;
+	u32 flags;
+	u8 key[XEN_NETBK_MAX_HASH_KEY_SIZE];
+	u32 mapping[XEN_NETBK_MAX_HASH_MAPPING_SIZE];
+	unsigned int size;
+	struct xenvif_hash_cache cache;
+};
+
 struct xenvif {
 	/* Unique identifier for this interface. */
 	domid_t          domid;
@@ -251,6 +280,8 @@ struct xenvif {
 	unsigned int num_queues; /* active queues, resource allocated */
 	unsigned int stalled_queues;
 
+	struct xenvif_hash hash;
+
 	struct xenbus_watch credit_watch;
 	struct xenbus_watch mcast_ctrl_watch;
 
@@ -260,6 +291,11 @@ struct xenvif {
 	struct dentry *xenvif_dbg_root;
 #endif
 
+	struct xen_netif_ctrl_back_ring ctrl;
+	struct task_struct *ctrl_task;
+	wait_queue_head_t ctrl_wq;
+	unsigned int ctrl_irq;
+
 	/* Miscellaneous private stuff. */
 	struct net_device *dev;
 };
@@ -285,10 +321,15 @@ struct xenvif *xenvif_alloc(struct device *parent,
 int xenvif_init_queue(struct xenvif_queue *queue);
 void xenvif_deinit_queue(struct xenvif_queue *queue);
 
-int xenvif_connect(struct xenvif_queue *queue, unsigned long tx_ring_ref,
-		   unsigned long rx_ring_ref, unsigned int tx_evtchn,
-		   unsigned int rx_evtchn);
-void xenvif_disconnect(struct xenvif *vif);
+int xenvif_connect_data(struct xenvif_queue *queue,
+			unsigned long tx_ring_ref,
+			unsigned long rx_ring_ref,
+			unsigned int tx_evtchn,
+			unsigned int rx_evtchn);
+void xenvif_disconnect_data(struct xenvif *vif);
+int xenvif_connect_ctrl(struct xenvif *vif, grant_ref_t ring_ref,
+			unsigned int evtchn);
+void xenvif_disconnect_ctrl(struct xenvif *vif);
 void xenvif_free(struct xenvif *vif);
 
 int xenvif_xenbus_init(void);
@@ -300,10 +341,10 @@ int xenvif_queue_stopped(struct xenvif_queue *queue);
 void xenvif_wake_queue(struct xenvif_queue *queue);
 
 /* (Un)Map communication rings. */
-void xenvif_unmap_frontend_rings(struct xenvif_queue *queue);
-int xenvif_map_frontend_rings(struct xenvif_queue *queue,
-			      grant_ref_t tx_ring_ref,
-			      grant_ref_t rx_ring_ref);
+void xenvif_unmap_frontend_data_rings(struct xenvif_queue *queue);
+int xenvif_map_frontend_data_rings(struct xenvif_queue *queue,
+				   grant_ref_t tx_ring_ref,
+				   grant_ref_t rx_ring_ref);
 
 /* Check for SKBs from frontend and schedule backend processing */
 void xenvif_napi_schedule_or_enable_events(struct xenvif_queue *queue);
@@ -318,6 +359,8 @@ void xenvif_kick_thread(struct xenvif_queue *queue);
 
 int xenvif_dealloc_kthread(void *data);
 
+int xenvif_ctrl_kthread(void *data);
+
 void xenvif_rx_queue_tail(struct xenvif_queue *queue, struct sk_buff *skb);
 
 void xenvif_carrier_on(struct xenvif *vif);
@@ -341,6 +384,7 @@ extern bool separate_tx_rx_irq;
 extern unsigned int rx_drain_timeout_msecs;
 extern unsigned int rx_stall_timeout_msecs;
 extern unsigned int xenvif_max_queues;
+extern unsigned int xenvif_hash_cache_size;
 
 #ifdef CONFIG_DEBUG_FS
 extern struct dentry *xen_netback_dbg_root;
@@ -354,4 +398,18 @@ void xenvif_skb_zerocopy_complete(struct xenvif_queue *queue);
 bool xenvif_mcast_match(struct xenvif *vif, const u8 *addr);
 void xenvif_mcast_addr_list_free(struct xenvif *vif);
 
+/* Hash */
+void xenvif_init_hash(struct xenvif *vif);
+void xenvif_deinit_hash(struct xenvif *vif);
+
+u32 xenvif_set_hash_alg(struct xenvif *vif, u32 alg);
+u32 xenvif_get_hash_flags(struct xenvif *vif, u32 *flags);
+u32 xenvif_set_hash_flags(struct xenvif *vif, u32 flags);
+u32 xenvif_set_hash_key(struct xenvif *vif, u32 gref, u32 len);
+u32 xenvif_set_hash_mapping_size(struct xenvif *vif, u32 size);
+u32 xenvif_set_hash_mapping(struct xenvif *vif, u32 gref, u32 len,
+			    u32 off);
+
+void xenvif_set_skb_hash(struct xenvif *vif, struct sk_buff *skb);
+
 #endif /* __XEN_NETBACK__COMMON_H__ */

drivers/net/xen-netback/hash.c

+/*
+ * Copyright (c) 2016 Citrix Systems Inc.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License version 2
+ * as published by the Free Softare Foundation; or, when distributed
+ * separately from the Linux kernel or incorporated into other
+ * software packages, subject to the following license:
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this source file (the "Software"), to deal in the Software without
+ * restriction, including without limitation the rights to use, copy, modify,
+ * merge, publish, distribute, sublicense, and/or sell copies of the Software,
+ * and to permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#define XEN_NETIF_DEFINE_TOEPLITZ
+
+#include "common.h"
+#include <linux/vmalloc.h>
+#include <linux/rculist.h>
+
+static void xenvif_del_hash(struct rcu_head *rcu)
+{
+	struct xenvif_hash_cache_entry *entry;
+
+	entry = container_of(rcu, struct xenvif_hash_cache_entry, rcu);
+
+	kfree(entry);
+}
+
+static void xenvif_add_hash(struct xenvif *vif, const u8 *tag,
+			    unsigned int len, u32 val)
+{
+	struct xenvif_hash_cache_entry *new, *entry, *oldest;
+	unsigned long flags;
+	bool found;
+
+	new = kmalloc(sizeof(*entry), GFP_KERNEL);
+	if (!new)
+		return;
+
+	memcpy(new->tag, tag, len);
+	new->len = len;
+	new->val = val;
+
+	spin_lock_irqsave(&vif->hash.cache.lock, flags);
+
+	found = false;
+	oldest = NULL;
+	list_for_each_entry_rcu(entry, &vif->hash.cache.list, link) {
+		/* Make sure we don't add duplicate entries */
+		if (entry->len == len &&
+		    memcmp(entry->tag, tag, len) == 0)
+			found = true;
+		if (!oldest || entry->seq < oldest->seq)
+			oldest = entry;
+	}
+
+	if (!found) {
+		new->seq = atomic_inc_return(&vif->hash.cache.seq);
+		list_add_rcu(&new->link, &vif->hash.cache.list);
+
+		if (++vif->hash.cache.count > xenvif_hash_cache_size) {
+			list_del_rcu(&oldest->link);
+			vif->hash.cache.count--;
+			call_rcu(&oldest->rcu, xenvif_del_hash);
+		}
+	}
+
+	spin_unlock_irqrestore(&vif->hash.cache.lock, flags);
+
+	if (found)
+		kfree(new);
+}
+
+static u32 xenvif_new_hash(struct xenvif *vif, const u8 *data,
+			   unsigned int len)
+{
+	u32 val;
+
+	val = xen_netif_toeplitz_hash(vif->hash.key,
+				      sizeof(vif->hash.key),
+				      data, len);
+
+	if (xenvif_hash_cache_size != 0)
+		xenvif_add_hash(vif, data, len, val);
+
+	return val;
+}
+
+static void xenvif_flush_hash(struct xenvif *vif)
+{
+	struct xenvif_hash_cache_entry *entry;
+	unsigned long flags;
+
+	if (xenvif_hash_cache_size == 0)
+		return;
+
+	spin_lock_irqsave(&vif->hash.cache.lock, flags);
+
+	list_for_each_entry_rcu(entry, &vif->hash.cache.list, link) {
+		list_del_rcu(&entry->link);
+		vif->hash.cache.count--;
+		call_rcu(&entry->rcu, xenvif_del_hash);
+	}
+
+	spin_unlock_irqrestore(&vif->hash.cache.lock, flags);
+}
+
+static u32 xenvif_find_hash(struct xenvif *vif, const u8 *data,
+			    unsigned int len)
+{
+	struct xenvif_hash_cache_entry *entry;
+	u32 val;
+	bool found;
+
+	if (len >= XEN_NETBK_HASH_TAG_SIZE)
+		return 0;
+
+	if (xenvif_hash_cache_size == 0)
+		return xenvif_new_hash(vif, data, len);
+
+	rcu_read_lock();
+
+	found = false;
+
+	list_for_each_entry_rcu(entry, &vif->hash.cache.list, link) {
+		if (entry->len == len &&
+		    memcmp(entry->tag, data, len) == 0) {
+			val = entry->val;
+			entry->seq = atomic_inc_return(&vif->hash.cache.seq);
+			found = true;
+			break;
+		}
+	}
+
+	rcu_read_unlock();
+
+	if (!found)
+		val = xenvif_new_hash(vif, data, len);
+
+	return val;
+}
+
+void xenvif_set_skb_hash(struct xenvif *vif, struct sk_buff *skb)
+{
+	struct flow_keys flow;
+	u32 hash = 0;
+	enum pkt_hash_types type = PKT_HASH_TYPE_NONE;
+	u32 flags = vif->hash.flags;
+	bool has_tcp_hdr;
+
+	/* Quick rejection test: If the network protocol doesn't
+	 * correspond to any enabled hash type then there's no point
+	 * in parsing the packet header.
+	 */
+	switch (skb->protocol) {
+	case htons(ETH_P_IP):
+		if (flags & (XEN_NETIF_CTRL_HASH_TYPE_IPV4_TCP |
+			     XEN_NETIF_CTRL_HASH_TYPE_IPV4))
+			break;
+
+		goto done;
+
+	case htons(ETH_P_IPV6):
+		if (flags & (XEN_NETIF_CTRL_HASH_TYPE_IPV6_TCP |
+			     XEN_NETIF_CTRL_HASH_TYPE_IPV6))
+			break;
+
+		goto done;
+
+	default:
+		goto done;
+	}
+
+	memset(&flow, 0, sizeof(flow));
+	if (!skb_flow_dissect_flow_keys(skb, &flow, 0))
+		goto done;
+
+	has_tcp_hdr = (flow.basic.ip_proto == IPPROTO_TCP) &&
+		      !(flow.control.flags & FLOW_DIS_IS_FRAGMENT);
+
+	switch (skb->protocol) {
+	case htons(ETH_P_IP):
+		if (has_tcp_hdr &&
+		    (flags & XEN_NETIF_CTRL_HASH_TYPE_IPV4_TCP)) {
+			u8 data[12];
+
+			memcpy(&data[0], &flow.addrs.v4addrs.src, 4);
+			memcpy(&data[4], &flow.addrs.v4addrs.dst, 4);
+			memcpy(&data[8], &flow.ports.src, 2);
+			memcpy(&data[10], &flow.ports.dst, 2);
+
+			hash = xenvif_find_hash(vif, data, sizeof(data));
+			type = PKT_HASH_TYPE_L4;
+		} else if (flags & XEN_NETIF_CTRL_HASH_TYPE_IPV4) {
+			u8 data[8];
+
+			memcpy(&data[0], &flow.addrs.v4addrs.src, 4);
+			memcpy(&data[4], &flow.addrs.v4addrs.dst, 4);
+
+			hash = xenvif_find_hash(vif, data, sizeof(data));
+			type = PKT_HASH_TYPE_L3;
+		}
+
+		break;
+
+	case htons(ETH_P_IPV6):
+		if (has_tcp_hdr &&
+		    (flags & XEN_NETIF_CTRL_HASH_TYPE_IPV6_TCP)) {
+			u8 data[36];
+
+			memcpy(&data[0], &flow.addrs.v6addrs.src, 16);
+			memcpy(&data[16], &flow.addrs.v6addrs.dst, 16);
+			memcpy(&data[32], &flow.ports.src, 2);
+			memcpy(&data[34], &flow.ports.dst, 2);
+
+			hash = xenvif_find_hash(vif, data, sizeof(data));
+			type = PKT_HASH_TYPE_L4;
+		} else if (flags & XEN_NETIF_CTRL_HASH_TYPE_IPV6) {
+			u8 data[32];
+
+			memcpy(&data[0], &flow.addrs.v6addrs.src, 16);
+			memcpy(&data[16], &flow.addrs.v6addrs.dst, 16);
+
+			hash = xenvif_find_hash(vif, data, sizeof(data));
+			type = PKT_HASH_TYPE_L3;
+		}
+
+		break;
+	}
+
+done:
+	if (type == PKT_HASH_TYPE_NONE)
+		skb_clear_hash(skb);
+	else
+		__skb_set_sw_hash(skb, hash, type == PKT_HASH_TYPE_L4);
+}
+
+u32 xenvif_set_hash_alg(struct xenvif *vif, u32 alg)
+{
+	switch (alg) {
+	case XEN_NETIF_CTRL_HASH_ALGORITHM_NONE:
+	case XEN_NETIF_CTRL_HASH_ALGORITHM_TOEPLITZ:
+		break;
+
+	default:
+		return XEN_NETIF_CTRL_STATUS_INVALID_PARAMETER;
+	}
+
+	vif->hash.alg = alg;
+
+	return XEN_NETIF_CTRL_STATUS_SUCCESS;
+}
+
+u32 xenvif_get_hash_flags(struct xenvif *vif, u32 *flags)
+{
+	if (vif->hash.alg == XEN_NETIF_CTRL_HASH_ALGORITHM_NONE)
+		return XEN_NETIF_CTRL_STATUS_NOT_SUPPORTED;
+
+	*flags = XEN_NETIF_CTRL_HASH_TYPE_IPV4 |
+		 XEN_NETIF_CTRL_HASH_TYPE_IPV4_TCP |
+		 XEN_NETIF_CTRL_HASH_TYPE_IPV6 |
+		 XEN_NETIF_CTRL_HASH_TYPE_IPV6_TCP;
+
+	return XEN_NETIF_CTRL_STATUS_SUCCESS;
+}
+
+u32 xenvif_set_hash_flags(struct xenvif *vif, u32 flags)
+{
+	if (flags & ~(XEN_NETIF_CTRL_HASH_TYPE_IPV4 |
+		      XEN_NETIF_CTRL_HASH_TYPE_IPV4_TCP |
+		      XEN_NETIF_CTRL_HASH_TYPE_IPV6 |
+		      XEN_NETIF_CTRL_HASH_TYPE_IPV6_TCP))
+		return XEN_NETIF_CTRL_STATUS_INVALID_PARAMETER;
+
+	if (vif->hash.alg == XEN_NETIF_CTRL_HASH_ALGORITHM_NONE)
+		return XEN_NETIF_CTRL_STATUS_INVALID_PARAMETER;
+
+	vif->hash.flags = flags;
+
+	return XEN_NETIF_CTRL_STATUS_SUCCESS;
+}
+
+u32 xenvif_set_hash_key(struct xenvif *vif, u32 gref, u32 len)
+{
+	u8 *key = vif->hash.key;
+	struct gnttab_copy copy_op = {
+		.source.u.ref = gref,
+		.source.domid = vif->domid,
+		.dest.u.gmfn = virt_to_gfn(key),
+		.dest.domid = DOMID_SELF,
+		.dest.offset = xen_offset_in_page(key),
+		.len = len,
+		.flags = GNTCOPY_source_gref
+	};
+
+	if (len > XEN_NETBK_MAX_HASH_KEY_SIZE)
+		return XEN_NETIF_CTRL_STATUS_INVALID_PARAMETER;
+
+	if (len != 0) {
+		gnttab_batch_copy(&copy_op, 1);
+
+		if (copy_op.status != GNTST_okay)
+			return XEN_NETIF_CTRL_STATUS_INVALID_PARAMETER;
+	}
+
+	/* Clear any remaining key octets */
+	if (len < XEN_NETBK_MAX_HASH_KEY_SIZE)
+		memset(key + len, 0, XEN_NETBK_MAX_HASH_KEY_SIZE - len);
+
+	xenvif_flush_hash(vif);
+
+	return XEN_NETIF_CTRL_STATUS_SUCCESS;
+}
+
+u32 xenvif_set_hash_mapping_size(struct xenvif *vif, u32 size)
+{
+	if (size > XEN_NETBK_MAX_HASH_MAPPING_SIZE)
+		return XEN_NETIF_CTRL_STATUS_INVALID_PARAMETER;
+
+	vif->hash.size = size;
+	memset(vif->hash.mapping, 0, sizeof(u32) * size);
+
+	return XEN_NETIF_CTRL_STATUS_SUCCESS;
+}
+
+u32 xenvif_set_hash_mapping(struct xenvif *vif, u32 gref, u32 len,
+			    u32 off)
+{
+	u32 *mapping = &vif->hash.mapping[off];
+	struct gnttab_copy copy_op = {
+		.source.u.ref = gref,
+		.source.domid = vif->domid,
+		.dest.u.gmfn = virt_to_gfn(mapping),
+		.dest.domid = DOMID_SELF,
+		.dest.offset = xen_offset_in_page(mapping),
+		.len = len * sizeof(u32),
+		.flags = GNTCOPY_source_gref
+	};
+
+	if ((off + len > vif->hash.size) || copy_op.len > XEN_PAGE_SIZE)
+		return XEN_NETIF_CTRL_STATUS_INVALID_PARAMETER;
+
+	while (len-- != 0)
+		if (mapping[off++] >= vif->num_queues)
+			return XEN_NETIF_CTRL_STATUS_INVALID_PARAMETER;
+
+	if (len != 0) {
+		gnttab_batch_copy(&copy_op, 1);
+
+		if (copy_op.status != GNTST_okay)
+			return XEN_NETIF_CTRL_STATUS_INVALID_PARAMETER;
+	}
+
+	return XEN_NETIF_CTRL_STATUS_SUCCESS;
+}
+
+void xenvif_init_hash(struct xenvif *vif)
+{
+	if (xenvif_hash_cache_size == 0)
+		return;
+
+	spin_lock_init(&vif->hash.cache.lock);
+	INIT_LIST_HEAD(&vif->hash.cache.list);
+}
+
+void xenvif_deinit_hash(struct xenvif *vif)
+{
+	xenvif_flush_hash(vif);
+}

drivers/net/xen-netback/interface.c

@@ -128,6 +128,15 @@ irqreturn_t xenvif_interrupt(int irq, void *dev_id)
 	return IRQ_HANDLED;
 }
 
+irqreturn_t xenvif_ctrl_interrupt(int irq, void *dev_id)
+{
+	struct xenvif *vif = dev_id;
+
+	wake_up(&vif->ctrl_wq);
+
+	return IRQ_HANDLED;
+}
+
 int xenvif_queue_stopped(struct xenvif_queue *queue)
 {
 	struct net_device *dev = queue->vif->dev;
@@ -142,6 +151,33 @@ void xenvif_wake_queue(struct xenvif_queue *queue)
 	netif_tx_wake_queue(netdev_get_tx_queue(dev, id));
 }
 
+static u16 xenvif_select_queue(struct net_device *dev, struct sk_buff *skb,
+			       void *accel_priv,
+			       select_queue_fallback_t fallback)
+{
+	struct xenvif *vif = netdev_priv(dev);
+	unsigned int size = vif->hash.size;
+
+	if (vif->hash.alg == XEN_NETIF_CTRL_HASH_ALGORITHM_NONE) {
+		u16 index = fallback(dev, skb) % dev->real_num_tx_queues;
+
+		/* Make sure there is no hash information in the socket
+		 * buffer otherwise it would be incorrectly forwarded
+		 * to the frontend.
+		 */
+		skb_clear_hash(skb);
+
+		return index;
+	}
+
+	xenvif_set_skb_hash(vif, skb);
+
+	if (size == 0)
+		return skb_get_hash_raw(skb) % dev->real_num_tx_queues;
+
+	return vif->hash.mapping[skb_get_hash_raw(skb) % size];
+}
+
 static int xenvif_start_xmit(struct sk_buff *skb, struct net_device *dev)
 {
 	struct xenvif *vif = netdev_priv(dev);
@@ -386,6 +422,7 @@ static const struct ethtool_ops xenvif_ethtool_ops = {
 };
 
 static const struct net_device_ops xenvif_netdev_ops = {
+	.ndo_select_queue = xenvif_select_queue,
 	.ndo_start_xmit	= xenvif_start_xmit,
 	.ndo_get_stats	= xenvif_get_stats,
 	.ndo_open	= xenvif_open,
@@ -527,9 +564,69 @@ void xenvif_carrier_on(struct xenvif *vif)
 	rtnl_unlock();
 }
 
-int xenvif_connect(struct xenvif_queue *queue, unsigned long tx_ring_ref,
-		   unsigned long rx_ring_ref, unsigned int tx_evtchn,
-		   unsigned int rx_evtchn)
+int xenvif_connect_ctrl(struct xenvif *vif, grant_ref_t ring_ref,
+			unsigned int evtchn)
+{
+	struct net_device *dev = vif->dev;
+	void *addr;
+	struct xen_netif_ctrl_sring *shared;
+	struct task_struct *task;
+	int err = -ENOMEM;
+
+	err = xenbus_map_ring_valloc(xenvif_to_xenbus_device(vif),
+				     &ring_ref, 1, &addr);
+	if (err)
+		goto err;
+
+	shared = (struct xen_netif_ctrl_sring *)addr;
+	BACK_RING_INIT(&vif->ctrl, shared, XEN_PAGE_SIZE);
+
+	init_waitqueue_head(&vif->ctrl_wq);
+
+	err = bind_interdomain_evtchn_to_irqhandler(vif->domid, evtchn,
+						    xenvif_ctrl_interrupt,
+						    0, dev->name, vif);
+	if (err < 0)
+		goto err_unmap;
+
+	vif->ctrl_irq = err;
+
+	xenvif_init_hash(vif);
+
+	task = kthread_create(xenvif_ctrl_kthread, (void *)vif,
+			      "%s-control", dev->name);
+	if (IS_ERR(task)) {
+		pr_warn("Could not allocate kthread for %s\n", dev->name);
+		err = PTR_ERR(task);
+		goto err_deinit;
+	}
+
+	get_task_struct(task);
+	vif->ctrl_task = task;
+
+	wake_up_process(vif->ctrl_task);
+
+	return 0;
+
+err_deinit:
+	xenvif_deinit_hash(vif);
+	unbind_from_irqhandler(vif->ctrl_irq, vif);
+	vif->ctrl_irq = 0;
+
+err_unmap:
+	xenbus_unmap_ring_vfree(xenvif_to_xenbus_device(vif),
+				vif->ctrl.sring);
+	vif->ctrl.sring = NULL;
+
+err:
+	return err;
+}
+
+int xenvif_connect_data(struct xenvif_queue *queue,
+			unsigned long tx_ring_ref,
+			unsigned long rx_ring_ref,
+			unsigned int tx_evtchn,
+			unsigned int rx_evtchn)
 {
 	struct task_struct *task;
 	int err = -ENOMEM;
@@ -538,7 +635,8 @@ int xenvif_connect(struct xenvif_queue *queue, unsigned long tx_ring_ref,
 	BUG_ON(queue->task);
 	BUG_ON(queue->dealloc_task);
 
-	err = xenvif_map_frontend_rings(queue, tx_ring_ref, rx_ring_ref);
+	err = xenvif_map_frontend_data_rings(queue, tx_ring_ref,
+					     rx_ring_ref);
 	if (err < 0)
 		goto err;
 
@@ -614,7 +712,7 @@ int xenvif_connect(struct xenvif_queue *queue, unsigned long tx_ring_ref,
 	unbind_from_irqhandler(queue->tx_irq, queue);
 	queue->tx_irq = 0;
 err_unmap:
-	xenvif_unmap_frontend_rings(queue);
+	xenvif_unmap_frontend_data_rings(queue);
 	netif_napi_del(&queue->napi);
 err:
 	module_put(THIS_MODULE);
@@ -634,7 +732,7 @@ void xenvif_carrier_off(struct xenvif *vif)
 	rtnl_unlock();
 }
 
-void xenvif_disconnect(struct xenvif *vif)
+void xenvif_disconnect_data(struct xenvif *vif)
 {
 	struct xenvif_queue *queue = NULL;
 	unsigned int num_queues = vif->num_queues;
@@ -668,12 +766,34 @@ void xenvif_disconnect(struct xenvif *vif)
 			queue->tx_irq = 0;
 		}
 
-		xenvif_unmap_frontend_rings(queue);
+		xenvif_unmap_frontend_data_rings(queue);
 	}
 
 	xenvif_mcast_addr_list_free(vif);
 }
 
+void xenvif_disconnect_ctrl(struct xenvif *vif)
+{
+	if (vif->ctrl_task) {
+		kthread_stop(vif->ctrl_task);
+		put_task_struct(vif->ctrl_task);
+		vif->ctrl_task = NULL;
+	}
+
+	xenvif_deinit_hash(vif);
+
+	if (vif->ctrl_irq) {
+		unbind_from_irqhandler(vif->ctrl_irq, vif);
+		vif->ctrl_irq = 0;
+	}
+
+	if (vif->ctrl.sring) {
+		xenbus_unmap_ring_vfree(xenvif_to_xenbus_device(vif),
+					vif->ctrl.sring);
+		vif->ctrl.sring = NULL;
+	}
+}
+
 /* Reverse the relevant parts of xenvif_init_queue().
  * Used for queue teardown from xenvif_free(), and on the
  * error handling paths in xenbus.c:connect().

drivers/net/xen-netback/netback.c

@@ -89,6 +89,11 @@ module_param(fatal_skb_slots, uint, 0444);
  */
 #define XEN_NETBACK_TX_COPY_LEN 128
 
+/* This is the maximum number of flows in the hash cache. */
+#define XENVIF_HASH_CACHE_SIZE_DEFAULT 64
+unsigned int xenvif_hash_cache_size = XENVIF_HASH_CACHE_SIZE_DEFAULT;
+module_param_named(hash_cache_size, xenvif_hash_cache_size, uint, 0644);
+MODULE_PARM_DESC(hash_cache_size, "Number of flows in the hash cache");
 
 static void xenvif_idx_release(struct xenvif_queue *queue, u16 pending_idx,
 			       u8 status);
@@ -163,6 +168,8 @@ static bool xenvif_rx_ring_slots_available(struct xenvif_queue *queue)
 	needed = DIV_ROUND_UP(skb->len, XEN_PAGE_SIZE);
 	if (skb_is_gso(skb))
 		needed++;
+	if (skb->sw_hash)
+		needed++;
 
 	do {
 		prod = queue->rx.sring->req_prod;
@@ -280,6 +287,8 @@ struct gop_frag_copy {
 	struct xenvif_rx_meta *meta;
 	int head;
 	int gso_type;
+	int protocol;
+	int hash_present;
 
 	struct page *page;
 };
@@ -326,8 +335,15 @@ static void xenvif_setup_copy_gop(unsigned long gfn,
 	npo->copy_off += *len;
 	info->meta->size += *len;
 
+	if (!info->head)
+		return;
+
 	/* Leave a gap for the GSO descriptor. */
-	if (info->head && ((1 << info->gso_type) & queue->vif->gso_mask))
+	if ((1 << info->gso_type) & queue->vif->gso_mask)
+		queue->rx.req_cons++;
+
+	/* Leave a gap for the hash extra segment. */
+	if (info->hash_present)
 		queue->rx.req_cons++;
 
 	info->head = 0; /* There must be something in this buffer now */
@@ -362,6 +378,11 @@ static void xenvif_gop_frag_copy(struct xenvif_queue *queue, struct sk_buff *skb
 		.npo = npo,
 		.head = *head,
 		.gso_type = XEN_NETIF_GSO_TYPE_NONE,
+		/* xenvif_set_skb_hash() will have either set a s/w
+		 * hash or cleared the hash depending on
+		 * whether the the frontend wants a hash for this skb.
+		 */
+		.hash_present = skb->sw_hash,
 	};
 	unsigned long bytes;
 
@@ -550,6 +571,7 @@ void xenvif_kick_thread(struct xenvif_queue *queue)
 
 static void xenvif_rx_action(struct xenvif_queue *queue)
 {
+	struct xenvif *vif = queue->vif;
 	s8 status;
 	u16 flags;
 	struct xen_netif_rx_response *resp;
@@ -585,9 +607,10 @@ static void xenvif_rx_action(struct xenvif_queue *queue)
 	gnttab_batch_copy(queue->grant_copy_op, npo.copy_prod);
 
 	while ((skb = __skb_dequeue(&rxq)) != NULL) {
+		struct xen_netif_extra_info *extra = NULL;
 
 		if ((1 << queue->meta[npo.meta_cons].gso_type) &
-		    queue->vif->gso_prefix_mask) {
+		    vif->gso_prefix_mask) {
 			resp = RING_GET_RESPONSE(&queue->rx,
 						 queue->rx.rsp_prod_pvt++);
 
@@ -605,7 +628,7 @@ static void xenvif_rx_action(struct xenvif_queue *queue)
 		queue->stats.tx_bytes += skb->len;
 		queue->stats.tx_packets++;
 
-		status = xenvif_check_gop(queue->vif,
+		status = xenvif_check_gop(vif,
 					  XENVIF_RX_CB(skb)->meta_slots_used,
 					  &npo);
 
@@ -627,21 +650,57 @@ static void xenvif_rx_action(struct xenvif_queue *queue)
 					flags);
 
 		if ((1 << queue->meta[npo.meta_cons].gso_type) &
-		    queue->vif->gso_mask) {
-			struct xen_netif_extra_info *gso =
-				(struct xen_netif_extra_info *)
+		    vif->gso_mask) {
+			extra = (struct xen_netif_extra_info *)
 				RING_GET_RESPONSE(&queue->rx,
 						  queue->rx.rsp_prod_pvt++);
 
 			resp->flags |= XEN_NETRXF_extra_info;
 
-			gso->u.gso.type = queue->meta[npo.meta_cons].gso_type;
-			gso->u.gso.size = queue->meta[npo.meta_cons].gso_size;
-			gso->u.gso.pad = 0;
-			gso->u.gso.features = 0;
+			extra->u.gso.type = queue->meta[npo.meta_cons].gso_type;
+			extra->u.gso.size = queue->meta[npo.meta_cons].gso_size;
+			extra->u.gso.pad = 0;
+			extra->u.gso.features = 0;
 
-			gso->type = XEN_NETIF_EXTRA_TYPE_GSO;
-			gso->flags = 0;
+			extra->type = XEN_NETIF_EXTRA_TYPE_GSO;
+			extra->flags = 0;
+		}
+
+		if (skb->sw_hash) {
+			/* Since the skb got here via xenvif_select_queue()
+			 * we know that the hash has been re-calculated
+			 * according to a configuration set by the frontend
+			 * and therefore we know that it is legitimate to
+			 * pass it to the frontend.
+			 */
+			if (resp->flags & XEN_NETRXF_extra_info)
+				extra->flags |= XEN_NETIF_EXTRA_FLAG_MORE;
+			else
+				resp->flags |= XEN_NETRXF_extra_info;
+
+			extra = (struct xen_netif_extra_info *)
+				RING_GET_RESPONSE(&queue->rx,
+						  queue->rx.rsp_prod_pvt++);
+
+			extra->u.hash.algorithm =
+				XEN_NETIF_CTRL_HASH_ALGORITHM_TOEPLITZ;
+
+			if (skb->l4_hash)
+				extra->u.hash.type =
+					skb->protocol == htons(ETH_P_IP) ?
+					_XEN_NETIF_CTRL_HASH_TYPE_IPV4_TCP :
+					_XEN_NETIF_CTRL_HASH_TYPE_IPV6_TCP;
+			else
+				extra->u.hash.type =
+					skb->protocol == htons(ETH_P_IP) ?
+					_XEN_NETIF_CTRL_HASH_TYPE_IPV4 :
+					_XEN_NETIF_CTRL_HASH_TYPE_IPV6;
+
+			*(uint32_t *)extra->u.hash.value =
+				skb_get_hash_raw(skb);
+
+			extra->type = XEN_NETIF_EXTRA_TYPE_HASH;
+			extra->flags = 0;
 		}
 
 		xenvif_add_frag_responses(queue, status,
@@ -1451,6 +1510,33 @@ static void xenvif_tx_build_gops(struct xenvif_queue *queue,
 			}
 		}
 
+		if (extras[XEN_NETIF_EXTRA_TYPE_HASH - 1].type) {
+			struct xen_netif_extra_info *extra;
+			enum pkt_hash_types type = PKT_HASH_TYPE_NONE;
+
+			extra = &extras[XEN_NETIF_EXTRA_TYPE_HASH - 1];
+
+			switch (extra->u.hash.type) {
+			case _XEN_NETIF_CTRL_HASH_TYPE_IPV4:
+			case _XEN_NETIF_CTRL_HASH_TYPE_IPV6:
+				type = PKT_HASH_TYPE_L3;
+				break;
+
+			case _XEN_NETIF_CTRL_HASH_TYPE_IPV4_TCP:
+			case _XEN_NETIF_CTRL_HASH_TYPE_IPV6_TCP:
+				type = PKT_HASH_TYPE_L4;
+				break;
+
+			default:
+				break;
+			}
+
+			if (type != PKT_HASH_TYPE_NONE)
+				skb_set_hash(skb,
+					     *(u32 *)extra->u.hash.value,
+					     type);
+		}
+
 		XENVIF_TX_CB(skb)->pending_idx = pending_idx;
 
 		__skb_put(skb, data_len);
@@ -1926,7 +2012,7 @@ static inline bool tx_dealloc_work_todo(struct xenvif_queue *queue)
 	return queue->dealloc_cons != queue->dealloc_prod;
 }
 
-void xenvif_unmap_frontend_rings(struct xenvif_queue *queue)
+void xenvif_unmap_frontend_data_rings(struct xenvif_queue *queue)
 {
 	if (queue->tx.sring)
 		xenbus_unmap_ring_vfree(xenvif_to_xenbus_device(queue->vif),
@@ -1936,9 +2022,9 @@ void xenvif_unmap_frontend_rings(struct xenvif_queue *queue)
 					queue->rx.sring);
 }
 
-int xenvif_map_frontend_rings(struct xenvif_queue *queue,
-			      grant_ref_t tx_ring_ref,
-			      grant_ref_t rx_ring_ref)
+int xenvif_map_frontend_data_rings(struct xenvif_queue *queue,
+				   grant_ref_t tx_ring_ref,
+				   grant_ref_t rx_ring_ref)
 {
 	void *addr;
 	struct xen_netif_tx_sring *txs;
@@ -1965,7 +2051,7 @@ int xenvif_map_frontend_rings(struct xenvif_queue *queue,
 	return 0;
 
 err:
-	xenvif_unmap_frontend_rings(queue);
+	xenvif_unmap_frontend_data_rings(queue);
 	return err;
 }
 
@@ -2164,6 +2250,135 @@ int xenvif_dealloc_kthread(void *data)
 	return 0;
 }
 
+static void make_ctrl_response(struct xenvif *vif,
+			       const struct xen_netif_ctrl_request *req,
+			       u32 status, u32 data)
+{
+	RING_IDX idx = vif->ctrl.rsp_prod_pvt;
+	struct xen_netif_ctrl_response rsp = {
+		.id = req->id,
+		.type = req->type,
+		.status = status,
+		.data = data,
+	};
+
+	*RING_GET_RESPONSE(&vif->ctrl, idx) = rsp;
+	vif->ctrl.rsp_prod_pvt = ++idx;
+}
+
+static void push_ctrl_response(struct xenvif *vif)
+{
+	int notify;
+
+	RING_PUSH_RESPONSES_AND_CHECK_NOTIFY(&vif->ctrl, notify);
+	if (notify)
+		notify_remote_via_irq(vif->ctrl_irq);
+}
+
+static void process_ctrl_request(struct xenvif *vif,
+				 const struct xen_netif_ctrl_request *req)
+{
+	u32 status = XEN_NETIF_CTRL_STATUS_NOT_SUPPORTED;
+	u32 data = 0;
+
+	switch (req->type) {
+	case XEN_NETIF_CTRL_TYPE_SET_HASH_ALGORITHM:
+		status = xenvif_set_hash_alg(vif, req->data[0]);
+		break;
+
+	case XEN_NETIF_CTRL_TYPE_GET_HASH_FLAGS:
+		status = xenvif_get_hash_flags(vif, &data);
+		break;
+
+	case XEN_NETIF_CTRL_TYPE_SET_HASH_FLAGS:
+		status = xenvif_set_hash_flags(vif, req->data[0]);
+		break;
+
+	case XEN_NETIF_CTRL_TYPE_SET_HASH_KEY:
+		status = xenvif_set_hash_key(vif, req->data[0],
+					     req->data[1]);
+		break;
+
+	case XEN_NETIF_CTRL_TYPE_GET_HASH_MAPPING_SIZE:
+		status = XEN_NETIF_CTRL_STATUS_SUCCESS;
+		data = XEN_NETBK_MAX_HASH_MAPPING_SIZE;
+		break;
+
+	case XEN_NETIF_CTRL_TYPE_SET_HASH_MAPPING_SIZE:
+		status = xenvif_set_hash_mapping_size(vif,
+						      req->data[0]);
+		break;
+
+	case XEN_NETIF_CTRL_TYPE_SET_HASH_MAPPING:
+		status = xenvif_set_hash_mapping(vif, req->data[0],
+						 req->data[1],
+						 req->data[2]);
+		break;
+
+	default:
+		break;
+	}
+
+	make_ctrl_response(vif, req, status, data);
+	push_ctrl_response(vif);
+}
+
+static void xenvif_ctrl_action(struct xenvif *vif)
+{
+	for (;;) {
+		RING_IDX req_prod, req_cons;
+
+		req_prod = vif->ctrl.sring->req_prod;
+		req_cons = vif->ctrl.req_cons;
+
+		/* Make sure we can see requests before we process them. */
+		rmb();
+
+		if (req_cons == req_prod)
+			break;
+
+		while (req_cons != req_prod) {
+			struct xen_netif_ctrl_request req;
+
+			RING_COPY_REQUEST(&vif->ctrl, req_cons, &req);
+			req_cons++;
+
+			process_ctrl_request(vif, &req);
+		}
+
+		vif->ctrl.req_cons = req_cons;
+		vif->ctrl.sring->req_event = req_cons + 1;
+	}
+}
+
+static bool xenvif_ctrl_work_todo(struct xenvif *vif)
+{
+	if (likely(RING_HAS_UNCONSUMED_REQUESTS(&vif->ctrl)))
+		return 1;
+
+	return 0;
+}
+
+int xenvif_ctrl_kthread(void *data)
+{
+	struct xenvif *vif = data;
+
+	for (;;) {
+		wait_event_interruptible(vif->ctrl_wq,
+					 xenvif_ctrl_work_todo(vif) ||
+					 kthread_should_stop());
+		if (kthread_should_stop())
+			break;
+
+		while (xenvif_ctrl_work_todo(vif))
+			xenvif_ctrl_action(vif);
+
+		cond_resched();
+	}
+
+	return 0;
+}
+
 static int __init netback_init(void)
 {
 	int rc = 0;

drivers/net/xen-netback/xenbus.c

@@ -38,7 +38,8 @@ struct backend_info {
 	const char *hotplug_script;
 };
 
-static int connect_rings(struct backend_info *be, struct xenvif_queue *queue);
+static int connect_data_rings(struct backend_info *be,
+			      struct xenvif_queue *queue);
 static void connect(struct backend_info *be);
 static int read_xenbus_vif_flags(struct backend_info *be);
 static int backend_create_xenvif(struct backend_info *be);
@@ -367,6 +368,12 @@ static int netback_probe(struct xenbus_device *dev,
 	if (err)
 		pr_debug("Error writing multi-queue-max-queues\n");
 
+	err = xenbus_printf(XBT_NIL, dev->nodename,
+			    "feature-ctrl-ring",
+			    "%u", true);
+	if (err)
+		pr_debug("Error writing feature-ctrl-ring\n");
+
 	script = xenbus_read(XBT_NIL, dev->nodename, "script", NULL);
 	if (IS_ERR(script)) {
 		err = PTR_ERR(script);
@@ -457,7 +464,8 @@ static void backend_disconnect(struct backend_info *be)
 #ifdef CONFIG_DEBUG_FS
 		xenvif_debugfs_delif(be->vif);
 #endif /* CONFIG_DEBUG_FS */
-		xenvif_disconnect(be->vif);
+		xenvif_disconnect_data(be->vif);
+		xenvif_disconnect_ctrl(be->vif);
 	}
 }
 
@@ -825,6 +833,48 @@ static void hotplug_status_changed(struct xenbus_watch *watch,
 	kfree(str);
 }
 
+static int connect_ctrl_ring(struct backend_info *be)
+{
+	struct xenbus_device *dev = be->dev;
+	struct xenvif *vif = be->vif;
+	unsigned int val;
+	grant_ref_t ring_ref;
+	unsigned int evtchn;
+	int err;
+
+	err = xenbus_gather(XBT_NIL, dev->otherend,
+			    "ctrl-ring-ref", "%u", &val, NULL);
+	if (err)
+		goto done; /* The frontend does not have a control ring */
+
+	ring_ref = val;
+
+	err = xenbus_gather(XBT_NIL, dev->otherend,
+			    "event-channel-ctrl", "%u", &val, NULL);
+	if (err) {
+		xenbus_dev_fatal(dev, err,
+				 "reading %s/event-channel-ctrl",
+				 dev->otherend);
+		goto fail;
+	}
+
+	evtchn = val;
+
+	err = xenvif_connect_ctrl(vif, ring_ref, evtchn);
+	if (err) {
+		xenbus_dev_fatal(dev, err,
+				 "mapping shared-frame %u port %u",
+				 ring_ref, evtchn);
+		goto fail;
+	}
+
+done:
+	return 0;
+
+fail:
+	return err;
+}
+
 static void connect(struct backend_info *be)
 {
 	int err;
@@ -861,6 +911,12 @@ static void connect(struct backend_info *be)
 	xen_register_watchers(dev, be->vif);
 	read_xenbus_vif_flags(be);
 
+	err = connect_ctrl_ring(be);
+	if (err) {
+		xenbus_dev_fatal(dev, err, "connecting control ring");
+		return;
+	}
+
 	/* Use the number of queues requested by the frontend */
 	be->vif->queues = vzalloc(requested_num_queues *
 				  sizeof(struct xenvif_queue));
@@ -896,11 +952,12 @@ static void connect(struct backend_info *be)
 		queue->remaining_credit = credit_bytes;
 		queue->credit_usec = credit_usec;
 
-		err = connect_rings(be, queue);
+		err = connect_data_rings(be, queue);
 		if (err) {
-			/* connect_rings() cleans up after itself on failure,
-			 * but we need to clean up after xenvif_init_queue() here,
-			 * and also clean up any previously initialised queues.
+			/* connect_data_rings() cleans up after itself on
+			 * failure, but we need to clean up after
+			 * xenvif_init_queue() here, and also clean up any
+			 * previously initialised queues.
 			 */
 			xenvif_deinit_queue(queue);
 			be->vif->num_queues = queue_index;
@@ -935,15 +992,17 @@ static void connect(struct backend_info *be)
 
 err:
 	if (be->vif->num_queues > 0)
-		xenvif_disconnect(be->vif); /* Clean up existing queues */
+		xenvif_disconnect_data(be->vif); /* Clean up existing queues */
 	vfree(be->vif->queues);
 	be->vif->queues = NULL;
 	be->vif->num_queues = 0;
+	xenvif_disconnect_ctrl(be->vif);
 	return;
 }
 
 
-static int connect_rings(struct backend_info *be, struct xenvif_queue *queue)
+static int connect_data_rings(struct backend_info *be,
+			      struct xenvif_queue *queue)
 {
 	struct xenbus_device *dev = be->dev;
 	unsigned int num_queues = queue->vif->num_queues;
@@ -1007,8 +1066,8 @@ static int connect_rings(struct backend_info *be, struct xenvif_queue *queue)
 	}
 
 	/* Map the shared frame, irq etc. */
-	err = xenvif_connect(queue, tx_ring_ref, rx_ring_ref,
-			     tx_evtchn, rx_evtchn);
+	err = xenvif_connect_data(queue, tx_ring_ref, rx_ring_ref,
+				  tx_evtchn, rx_evtchn);
 	if (err) {
 		xenbus_dev_fatal(dev, err,
 				 "mapping shared-frames %lu/%lu port tx %u rx %u",