[net-next PATCH v2 0/4] Add support for XPS when using DCB

[net-next PATCH v2 0/4] Add support for XPS when using DCB

[Alexander Duyck]

This patch series enables proper isolation between traffic classes when
using XPS while DCB is enabled.  Previously enabling XPS would cause the
traffic to be potentially pulled from one traffic class into another on
egress.  This change essentially multiplies the XPS map by the number of
traffic classes and allows us to do a lookup per traffic class for a given
CPU.

To guarantee the isolation I invalidate the XPS map for any queues that are
moved from one traffic class to another, or if we change the number of
traffic classes.

v2: Added sysfs to display traffic class
    Replaced do/while with for loop
    Cleaned up several other for for loops throughout the patch

---

Alexander Duyck (4):
      net: Move functions for configuring traffic classes out of inline headers
      net: Add sysfs value to determine queue traffic class
      net: Refactor removal of queues from XPS map and apply on num_tc changes
      net: Add support for XPS with QoS via traffic classes


 include/linux/netdevice.h |   36 +------
 net/core/dev.c            |  226 +++++++++++++++++++++++++++++++++++----------
 net/core/net-sysfs.c      |   51 ++++++++--
 3 files changed, 219 insertions(+), 94 deletions(-)

[net-next PATCH v2 1/4] net: Move functions for configuring traffic classes out of inline headers

[Alexander Duyck]

The functions for configuring the traffic class to queue mappings have
other effects that need to be addressed.  Instead of trying to export a
bunch of new functions just relocate the functions so that we can
instrument them directly with the functionality they will need.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
---
 include/linux/netdevice.h |   31 +++----------------------------
 net/core/dev.c            |   29 +++++++++++++++++++++++++++++
 2 files changed, 32 insertions(+), 28 deletions(-)

diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index 458c876..d045432 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -1920,34 +1920,9 @@ int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
 	return 0;
 }
 
-static inline
-void netdev_reset_tc(struct net_device *dev)
-{
-	dev->num_tc = 0;
-	memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
-	memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
-}
-
-static inline
-int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
-{
-	if (tc >= dev->num_tc)
-		return -EINVAL;
-
-	dev->tc_to_txq[tc].count = count;
-	dev->tc_to_txq[tc].offset = offset;
-	return 0;
-}
-
-static inline
-int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
-{
-	if (num_tc > TC_MAX_QUEUE)
-		return -EINVAL;
-
-	dev->num_tc = num_tc;
-	return 0;
-}
+void netdev_reset_tc(struct net_device *dev);
+int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
+int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
 
 static inline
 int netdev_get_num_tc(struct net_device *dev)
diff --git a/net/core/dev.c b/net/core/dev.c
index f55fb45..d4d45bf 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -2173,6 +2173,35 @@ int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
 EXPORT_SYMBOL(netif_set_xps_queue);
 
 #endif
+void netdev_reset_tc(struct net_device *dev)
+{
+	dev->num_tc = 0;
+	memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
+	memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
+}
+EXPORT_SYMBOL(netdev_reset_tc);
+
+int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
+{
+	if (tc >= dev->num_tc)
+		return -EINVAL;
+
+	dev->tc_to_txq[tc].count = count;
+	dev->tc_to_txq[tc].offset = offset;
+	return 0;
+}
+EXPORT_SYMBOL(netdev_set_tc_queue);
+
+int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
+{
+	if (num_tc > TC_MAX_QUEUE)
+		return -EINVAL;
+
+	dev->num_tc = num_tc;
+	return 0;
+}
+EXPORT_SYMBOL(netdev_set_num_tc);
+
 /*
  * Routine to help set real_num_tx_queues. To avoid skbs mapped to queues
  * greater then real_num_tx_queues stale skbs on the qdisc must be flushed.

[net-next PATCH v2 2/4] net: Add sysfs value to determine queue traffic class

[Alexander Duyck]

Add a sysfs attribute for a Tx queue that allows us to determine the
traffic class for a given queue.  This will allow us to more easily
determine this in the future.  It is needed as XPS will take the traffic
class for a group of queues into account in order to avoid pulling traffic
from one traffic class into another.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
---

v2: Added this patch to the series.

 include/linux/netdevice.h |    1 +
 net/core/dev.c            |   17 +++++++++++++++++
 net/core/net-sysfs.c      |   20 +++++++++++++++++++-
 3 files changed, 37 insertions(+), 1 deletion(-)

diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index d045432..b60a156 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -1920,6 +1920,7 @@ int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
 	return 0;
 }
 
+int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
 void netdev_reset_tc(struct net_device *dev);
 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
diff --git a/net/core/dev.c b/net/core/dev.c
index d4d45bf..08ed625 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -1948,6 +1948,23 @@ static void netif_setup_tc(struct net_device *dev, unsigned int txq)
 	}
 }
 
+int netdev_txq_to_tc(struct net_device *dev, unsigned int txq)
+{
+	if (dev->num_tc) {
+		struct netdev_tc_txq *tc = &dev->tc_to_txq[0];
+		int i;
+
+		for (i = 0; i < TC_MAX_QUEUE; i++, tc++) {
+			if ((txq - tc->offset) < tc->count)
+				return i;
+		}
+
+		return -1;
+	}
+
+	return 0;
+}
+
 #ifdef CONFIG_XPS
 static DEFINE_MUTEX(xps_map_mutex);
 #define xmap_dereference(P)		\
diff --git a/net/core/net-sysfs.c b/net/core/net-sysfs.c
index 6e4f347..c4e4e7d 100644
--- a/net/core/net-sysfs.c
+++ b/net/core/net-sysfs.c
@@ -1021,7 +1021,6 @@ static ssize_t show_trans_timeout(struct netdev_queue *queue,
 	return sprintf(buf, "%lu", trans_timeout);
 }
 
-#ifdef CONFIG_XPS
 static unsigned int get_netdev_queue_index(struct netdev_queue *queue)
 {
 	struct net_device *dev = queue->dev;
@@ -1033,6 +1032,21 @@ static unsigned int get_netdev_queue_index(struct netdev_queue *queue)
 	return i;
 }
 
+static ssize_t show_traffic_class(struct netdev_queue *queue,
+				  struct netdev_queue_attribute *attribute,
+				  char *buf)
+{
+	struct net_device *dev = queue->dev;
+	int index = get_netdev_queue_index(queue);
+	int tc = netdev_txq_to_tc(dev, index);
+
+	if (tc < 0)
+		return -EINVAL;
+
+	return sprintf(buf, "%u\n", tc);
+}
+
+#ifdef CONFIG_XPS
 static ssize_t show_tx_maxrate(struct netdev_queue *queue,
 			       struct netdev_queue_attribute *attribute,
 			       char *buf)
@@ -1075,6 +1089,9 @@ static ssize_t set_tx_maxrate(struct netdev_queue *queue,
 static struct netdev_queue_attribute queue_trans_timeout =
 	__ATTR(tx_timeout, S_IRUGO, show_trans_timeout, NULL);
 
+static struct netdev_queue_attribute queue_traffic_class =
+	__ATTR(traffic_class, S_IRUGO, show_traffic_class, NULL);
+
 #ifdef CONFIG_BQL
 /*
  * Byte queue limits sysfs structures and functions.
@@ -1260,6 +1277,7 @@ static ssize_t store_xps_map(struct netdev_queue *queue,
 
 static struct attribute *netdev_queue_default_attrs[] = {
 	&queue_trans_timeout.attr,
+	&queue_traffic_class.attr,
 #ifdef CONFIG_XPS
 	&xps_cpus_attribute.attr,
 	&queue_tx_maxrate.attr,

[net-next PATCH v2 3/4] net: Refactor removal of queues from XPS map and apply on num_tc changes

[Alexander Duyck]

This patch updates the code for removing queues from the XPS map and makes
it so that we can apply the code any time we change either the number of
traffic classes or the mapping of a given block of queues.  This way we
avoid having queues pulling traffic from a foreign traffic class.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
---

v2: Replaced do/while with for loop
    Changed order for remove_xps_queue_cpu loop
    Updated i to reflect count, j to be temporary offset
    Updated test to check for i < 0 instead of i < offset

 net/core/dev.c |   73 ++++++++++++++++++++++++++++++++++++++------------------
 1 file changed, 50 insertions(+), 23 deletions(-)

diff --git a/net/core/dev.c b/net/core/dev.c
index 08ed625..fea7061 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -1970,32 +1970,50 @@ int netdev_txq_to_tc(struct net_device *dev, unsigned int txq)
 #define xmap_dereference(P)		\
 	rcu_dereference_protected((P), lockdep_is_held(&xps_map_mutex))
 
-static struct xps_map *remove_xps_queue(struct xps_dev_maps *dev_maps,
-					int cpu, u16 index)
+static bool remove_xps_queue(struct xps_dev_maps *dev_maps,
+			     int tci, u16 index)
 {
 	struct xps_map *map = NULL;
 	int pos;
 
 	if (dev_maps)
-		map = xmap_dereference(dev_maps->cpu_map[cpu]);
+		map = xmap_dereference(dev_maps->cpu_map[tci]);
+	if (!map)
+		return false;
 
-	for (pos = 0; map && pos < map->len; pos++) {
-		if (map->queues[pos] == index) {
-			if (map->len > 1) {
-				map->queues[pos] = map->queues[--map->len];
-			} else {
-				RCU_INIT_POINTER(dev_maps->cpu_map[cpu], NULL);
-				kfree_rcu(map, rcu);
-				map = NULL;
-			}
+	for (pos = map->len; pos--;) {
+		if (map->queues[pos] != index)
+			continue;
+
+		if (map->len > 1) {
+			map->queues[pos] = map->queues[--map->len];
 			break;
 		}
+
+		RCU_INIT_POINTER(dev_maps->cpu_map[tci], NULL);
+		kfree_rcu(map, rcu);
+		return false;
 	}
 
-	return map;
+	return true;
 }
 
-static void netif_reset_xps_queues_gt(struct net_device *dev, u16 index)
+static bool remove_xps_queue_cpu(struct net_device *dev,
+				 struct xps_dev_maps *dev_maps,
+				 int cpu, u16 offset, u16 count)
+{
+	int i, j;
+
+	for (i = count, j = offset; i--; j++) {
+		if (!remove_xps_queue(dev_maps, cpu, j))
+			break;
+	}
+
+	return i < 0;
+}
+
+static void netif_reset_xps_queues(struct net_device *dev, u16 offset,
+				   u16 count)
 {
 	struct xps_dev_maps *dev_maps;
 	int cpu, i;
@@ -2007,21 +2025,16 @@ static void netif_reset_xps_queues_gt(struct net_device *dev, u16 index)
 	if (!dev_maps)
 		goto out_no_maps;
 
-	for_each_possible_cpu(cpu) {
-		for (i = index; i < dev->num_tx_queues; i++) {
-			if (!remove_xps_queue(dev_maps, cpu, i))
-				break;
-		}
-		if (i == dev->num_tx_queues)
-			active = true;
-	}
+	for_each_possible_cpu(cpu)
+		active |= remove_xps_queue_cpu(dev, dev_maps, cpu,
+					       offset, count);
 
 	if (!active) {
 		RCU_INIT_POINTER(dev->xps_maps, NULL);
 		kfree_rcu(dev_maps, rcu);
 	}
 
-	for (i = index; i < dev->num_tx_queues; i++)
+	for (i = offset + (count - 1); count--; i--)
 		netdev_queue_numa_node_write(netdev_get_tx_queue(dev, i),
 					     NUMA_NO_NODE);
 
@@ -2029,6 +2042,11 @@ static void netif_reset_xps_queues_gt(struct net_device *dev, u16 index)
 	mutex_unlock(&xps_map_mutex);
 }
 
+static void netif_reset_xps_queues_gt(struct net_device *dev, u16 index)
+{
+	netif_reset_xps_queues(dev, index, dev->num_tx_queues - index);
+}
+
 static struct xps_map *expand_xps_map(struct xps_map *map,
 				      int cpu, u16 index)
 {
@@ -2192,6 +2210,9 @@ int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
 #endif
 void netdev_reset_tc(struct net_device *dev)
 {
+#ifdef CONFIG_XPS
+	netif_reset_xps_queues_gt(dev, 0);
+#endif
 	dev->num_tc = 0;
 	memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
 	memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
@@ -2203,6 +2224,9 @@ int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
 	if (tc >= dev->num_tc)
 		return -EINVAL;
 
+#ifdef CONFIG_XPS
+	netif_reset_xps_queues(dev, offset, count);
+#endif
 	dev->tc_to_txq[tc].count = count;
 	dev->tc_to_txq[tc].offset = offset;
 	return 0;
@@ -2214,6 +2238,9 @@ int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
 	if (num_tc > TC_MAX_QUEUE)
 		return -EINVAL;
 
+#ifdef CONFIG_XPS
+	netif_reset_xps_queues_gt(dev, 0);
+#endif
 	dev->num_tc = num_tc;
 	return 0;
 }

[net-next PATCH v2 4/4] net: Add support for XPS with QoS via traffic classes

[Alexander Duyck]

This patch adds support for setting and using XPS when QoS via traffic
classes is enabled.  With this change we will factor in the priority and
traffic class mapping of the packet and use that information to correctly
select the queue.

This allows us to define a set of queues for a given traffic class via
mqprio and then configure the XPS mapping for those queues so that the
traffic flows can avoid head-of-line blocking between the individual CPUs
if so desired.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
---

v2: Mostly for loop updates based on changes in patch 3
    Replaced "i = 0; i < X; i++" with "i = X; i--;" as i was just a counter

 include/linux/netdevice.h |    4 +-
 net/core/dev.c            |  117 ++++++++++++++++++++++++++++++++-------------
 net/core/net-sysfs.c      |   31 ++++++++----
 3 files changed, 105 insertions(+), 47 deletions(-)

diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index b60a156..56f90f7 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -732,8 +732,8 @@ struct xps_dev_maps {
 	struct rcu_head rcu;
 	struct xps_map __rcu *cpu_map[0];
 };
-#define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) +		\
-    (nr_cpu_ids * sizeof(struct xps_map *)))
+#define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) +		\
+	(nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
 #endif /* CONFIG_XPS */
 
 #define TC_MAX_QUEUE	16
diff --git a/net/core/dev.c b/net/core/dev.c
index fea7061..0ebb035 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -2002,14 +2002,22 @@ static bool remove_xps_queue_cpu(struct net_device *dev,
 				 struct xps_dev_maps *dev_maps,
 				 int cpu, u16 offset, u16 count)
 {
-	int i, j;
+	int num_tc = dev->num_tc ? : 1;
+	bool active = false;
+	int tci;
 
-	for (i = count, j = offset; i--; j++) {
-		if (!remove_xps_queue(dev_maps, cpu, j))
-			break;
+	for (tci = cpu * num_tc; num_tc--; tci++) {
+		int i, j;
+
+		for (i = count, j = offset; i--; j++) {
+			if (!remove_xps_queue(dev_maps, cpu, j))
+				break;
+		}
+
+		active |= i < 0;
 	}
 
-	return i < 0;
+	return active;
 }
 
 static void netif_reset_xps_queues(struct net_device *dev, u16 offset,
@@ -2086,20 +2094,28 @@ int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
 			u16 index)
 {
 	struct xps_dev_maps *dev_maps, *new_dev_maps = NULL;
+	int i, cpu, tci, numa_node_id = -2;
+	int maps_sz, num_tc = 1, tc = 0;
 	struct xps_map *map, *new_map;
-	int maps_sz = max_t(unsigned int, XPS_DEV_MAPS_SIZE, L1_CACHE_BYTES);
-	int cpu, numa_node_id = -2;
 	bool active = false;
 
+	if (dev->num_tc) {
+		num_tc = dev->num_tc;
+		tc = netdev_txq_to_tc(dev, index);
+		if (tc < 0)
+			return -EINVAL;
+	}
+
+	maps_sz = XPS_DEV_MAPS_SIZE(num_tc);
+	if (maps_sz < L1_CACHE_BYTES)
+		maps_sz = L1_CACHE_BYTES;
+
 	mutex_lock(&xps_map_mutex);
 
 	dev_maps = xmap_dereference(dev->xps_maps);
 
 	/* allocate memory for queue storage */
-	for_each_online_cpu(cpu) {
-		if (!cpumask_test_cpu(cpu, mask))
-			continue;
-
+	for_each_cpu_and(cpu, cpu_online_mask, mask) {
 		if (!new_dev_maps)
 			new_dev_maps = kzalloc(maps_sz, GFP_KERNEL);
 		if (!new_dev_maps) {
@@ -2107,25 +2123,38 @@ int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
 			return -ENOMEM;
 		}
 
-		map = dev_maps ? xmap_dereference(dev_maps->cpu_map[cpu]) :
+		tci = cpu * num_tc + tc;
+		map = dev_maps ? xmap_dereference(dev_maps->cpu_map[tci]) :
 				 NULL;
 
 		map = expand_xps_map(map, cpu, index);
 		if (!map)
 			goto error;
 
-		RCU_INIT_POINTER(new_dev_maps->cpu_map[cpu], map);
+		RCU_INIT_POINTER(new_dev_maps->cpu_map[tci], map);
 	}
 
 	if (!new_dev_maps)
 		goto out_no_new_maps;
 
 	for_each_possible_cpu(cpu) {
+		/* copy maps belonging to foreign traffic classes */
+		for (i = tc, tci = cpu * num_tc; dev_maps && i--; tci++) {
+			/* fill in the new device map from the old device map */
+			map = xmap_dereference(dev_maps->cpu_map[tci]);
+			RCU_INIT_POINTER(new_dev_maps->cpu_map[tci], map);
+		}
+
+		/* We need to explicitly update tci as prevous loop
+		 * could break out early if dev_maps is NULL.
+		 */
+		tci = cpu * num_tc + tc;
+
 		if (cpumask_test_cpu(cpu, mask) && cpu_online(cpu)) {
 			/* add queue to CPU maps */
 			int pos = 0;
 
-			map = xmap_dereference(new_dev_maps->cpu_map[cpu]);
+			map = xmap_dereference(new_dev_maps->cpu_map[tci]);
 			while ((pos < map->len) && (map->queues[pos] != index))
 				pos++;
 
@@ -2139,26 +2168,36 @@ int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
 #endif
 		} else if (dev_maps) {
 			/* fill in the new device map from the old device map */
-			map = xmap_dereference(dev_maps->cpu_map[cpu]);
-			RCU_INIT_POINTER(new_dev_maps->cpu_map[cpu], map);
+			map = xmap_dereference(dev_maps->cpu_map[tci]);
+			RCU_INIT_POINTER(new_dev_maps->cpu_map[tci], map);
 		}
 
+		/* copy maps belonging to foreign traffic classes */
+		for (i = num_tc - tc, tci++; dev_maps && --i; tci++) {
+			/* fill in the new device map from the old device map */
+			map = xmap_dereference(dev_maps->cpu_map[tci]);
+			RCU_INIT_POINTER(new_dev_maps->cpu_map[tci], map);
+		}
 	}
 
 	rcu_assign_pointer(dev->xps_maps, new_dev_maps);
 
 	/* Cleanup old maps */
-	if (dev_maps) {
-		for_each_possible_cpu(cpu) {
-			new_map = xmap_dereference(new_dev_maps->cpu_map[cpu]);
-			map = xmap_dereference(dev_maps->cpu_map[cpu]);
+	if (!dev_maps)
+		goto out_no_old_maps;
+
+	for_each_possible_cpu(cpu) {
+		for (i = num_tc, tci = cpu * num_tc; i--; tci++) {
+			new_map = xmap_dereference(new_dev_maps->cpu_map[tci]);
+			map = xmap_dereference(dev_maps->cpu_map[tci]);
 			if (map && map != new_map)
 				kfree_rcu(map, rcu);
 		}
-
-		kfree_rcu(dev_maps, rcu);
 	}
 
+	kfree_rcu(dev_maps, rcu);
+
+out_no_old_maps:
 	dev_maps = new_dev_maps;
 	active = true;
 
@@ -2173,11 +2212,12 @@ int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
 
 	/* removes queue from unused CPUs */
 	for_each_possible_cpu(cpu) {
-		if (cpumask_test_cpu(cpu, mask) && cpu_online(cpu))
-			continue;
-
-		if (remove_xps_queue(dev_maps, cpu, index))
-			active = true;
+		for (i = tc, tci = cpu * num_tc; i--; tci++)
+			active |= remove_xps_queue(dev_maps, tci, index);
+		if (!cpumask_test_cpu(cpu, mask) || !cpu_online(cpu))
+			active |= remove_xps_queue(dev_maps, tci, index);
+		for (i = num_tc - tc, tci++; --i; tci++)
+			active |= remove_xps_queue(dev_maps, tci, index);
 	}
 
 	/* free map if not active */
@@ -2193,11 +2233,14 @@ int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
 error:
 	/* remove any maps that we added */
 	for_each_possible_cpu(cpu) {
-		new_map = xmap_dereference(new_dev_maps->cpu_map[cpu]);
-		map = dev_maps ? xmap_dereference(dev_maps->cpu_map[cpu]) :
-				 NULL;
-		if (new_map && new_map != map)
-			kfree(new_map);
+		for (i = num_tc, tci = cpu * num_tc; i--; tci++) {
+			new_map = xmap_dereference(new_dev_maps->cpu_map[tci]);
+			map = dev_maps ?
+			      xmap_dereference(dev_maps->cpu_map[tci]) :
+			      NULL;
+			if (new_map && new_map != map)
+				kfree(new_map);
+		}
 	}
 
 	mutex_unlock(&xps_map_mutex);
@@ -3157,8 +3200,14 @@ static inline int get_xps_queue(struct net_device *dev, struct sk_buff *skb)
 	rcu_read_lock();
 	dev_maps = rcu_dereference(dev->xps_maps);
 	if (dev_maps) {
-		map = rcu_dereference(
-		    dev_maps->cpu_map[skb->sender_cpu - 1]);
+		unsigned int tci = skb->sender_cpu - 1;
+
+		if (dev->num_tc) {
+			tci *= dev->num_tc;
+			tci += netdev_get_prio_tc_map(dev, skb->priority);
+		}
+
+		map = rcu_dereference(dev_maps->cpu_map[tci]);
 		if (map) {
 			if (map->len == 1)
 				queue_index = map->queues[0];
diff --git a/net/core/net-sysfs.c b/net/core/net-sysfs.c
index c4e4e7d..3b2b935 100644
--- a/net/core/net-sysfs.c
+++ b/net/core/net-sysfs.c
@@ -1207,29 +1207,38 @@ static ssize_t show_xps_map(struct netdev_queue *queue,
 			    struct netdev_queue_attribute *attribute, char *buf)
 {
 	struct net_device *dev = queue->dev;
+	int cpu, len, num_tc = 1, tc = 0;
 	struct xps_dev_maps *dev_maps;
 	cpumask_var_t mask;
 	unsigned long index;
-	int i, len;
 
 	if (!zalloc_cpumask_var(&mask, GFP_KERNEL))
 		return -ENOMEM;
 
 	index = get_netdev_queue_index(queue);
 
+	if (dev->num_tc) {
+		num_tc = dev->num_tc;
+		tc = netdev_txq_to_tc(dev, index);
+		if (tc < 0)
+			return -EINVAL;
+	}
+
 	rcu_read_lock();
 	dev_maps = rcu_dereference(dev->xps_maps);
 	if (dev_maps) {
-		for_each_possible_cpu(i) {
-			struct xps_map *map =
-			    rcu_dereference(dev_maps->cpu_map[i]);
-			if (map) {
-				int j;
-				for (j = 0; j < map->len; j++) {
-					if (map->queues[j] == index) {
-						cpumask_set_cpu(i, mask);
-						break;
-					}
+		for_each_possible_cpu(cpu) {
+			int i, tci = cpu * num_tc + tc;
+			struct xps_map *map;
+
+			map = rcu_dereference(dev_maps->cpu_map[tci]);
+			if (!map)
+				continue;
+
+			for (i = map->len; i--;) {
+				if (map->queues[i] == index) {
+					cpumask_set_cpu(cpu, mask);
+					break;
 				}
 			}
 		}

Re: [net-next PATCH v2 0/4] Add support for XPS when using DCB

[David Miller]

From: Alexander Duyck <alexander.h.duyck@intel.com>
Date: Fri, 28 Oct 2016 11:42:45 -0400

> This patch series enables proper isolation between traffic classes when
> using XPS while DCB is enabled.  Previously enabling XPS would cause the
> traffic to be potentially pulled from one traffic class into another on
> egress.  This change essentially multiplies the XPS map by the number of
> traffic classes and allows us to do a lookup per traffic class for a given
> CPU.
> 
> To guarantee the isolation I invalidate the XPS map for any queues that are
> moved from one traffic class to another, or if we change the number of
> traffic classes.
> 
> v2: Added sysfs to display traffic class
>     Replaced do/while with for loop
>     Cleaned up several other for for loops throughout the patch

Series applied, thanks Alex.