Merge branch 'txq_max_rate'

Or Gerlitz says:

====================
Add max rate TXQ attribute

Add the ability to set a max-rate limitation for TX queues.
The attribute name is maxrate and the units are Mbs, to make
it similar to the existing max-rate limitation knobs (ETS and
SRIOV ndo calls).

changes from V2:
  - added Documentation (thanks Florian and Tom)
  - rebased to latest net-next to comply with the swdev ndo removal
  - addressed more feedback from Dave on the comments style

changes from V1:
  - addressed feedback from Dave

changes from V0:
  - addressed feedback from Sergei

John Fastabend (1):
  net: Add max rate tx queue attribute
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Documentation/ABI/testing/sysfs-class-net-queues

@@ -24,6 +24,14 @@ Description:
 		Indicates the number of transmit timeout events seen by this
 		network interface transmit queue.
 
+What:		/sys/class/<iface>/queues/tx-<queue>/tx_maxrate
+Date:		March 2015
+KernelVersion:	4.1
+Contact:	netdev@vger.kernel.org
+Description:
+		A Mbps max-rate set for the queue, a value of zero means disabled,
+		default is disabled.
+
 What:		/sys/class/<iface>/queues/tx-<queue>/xps_cpus
 Date:		November 2010
 KernelVersion:	2.6.38

Documentation/networking/scaling.txt

@@ -421,6 +421,15 @@ best CPUs to share a given queue are probably those that share the cache
 with the CPU that processes transmit completions for that queue
 (transmit interrupts).
 
+Per TX Queue rate limitation:
+=============================
+
+These are rate-limitation mechanisms implemented by HW, where currently
+a max-rate attribute is supported, by setting a Mbps value to
+
+/sys/class/net/<dev>/queues/tx-<n>/tx_maxrate
+
+A value of zero means disabled, and this is the default.
 
 Further Information
 ===================

drivers/net/ethernet/mellanox/mlx4/en_netdev.c

@@ -2379,6 +2379,33 @@ static netdev_features_t mlx4_en_features_check(struct sk_buff *skb,
 }
 #endif
 
+int mlx4_en_set_tx_maxrate(struct net_device *dev, int queue_index, u32 maxrate)
+{
+	struct mlx4_en_priv *priv = netdev_priv(dev);
+	struct mlx4_en_tx_ring *tx_ring = priv->tx_ring[queue_index];
+	struct mlx4_update_qp_params params;
+	int err;
+
+	if (!(priv->mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_QP_RATE_LIMIT))
+		return -EOPNOTSUPP;
+
+	/* rate provided to us in Mbs, check if it fits into 12 bits, if not use Gbs */
+	if (maxrate >> 12) {
+		params.rate_unit = MLX4_QP_RATE_LIMIT_GBS;
+		params.rate_val  = maxrate / 1000;
+	} else if (maxrate) {
+		params.rate_unit = MLX4_QP_RATE_LIMIT_MBS;
+		params.rate_val  = maxrate;
+	} else { /* zero serves to revoke the QP rate-limitation */
+		params.rate_unit = 0;
+		params.rate_val  = 0;
+	}
+
+	err = mlx4_update_qp(priv->mdev->dev, tx_ring->qpn, MLX4_UPDATE_QP_RATE_LIMIT,
+			     &params);
+	return err;
+}
+
 static const struct net_device_ops mlx4_netdev_ops = {
 	.ndo_open		= mlx4_en_open,
 	.ndo_stop		= mlx4_en_close,
@@ -2410,6 +2437,7 @@ static const struct net_device_ops mlx4_netdev_ops = {
 	.ndo_del_vxlan_port	= mlx4_en_del_vxlan_port,
 	.ndo_features_check	= mlx4_en_features_check,
 #endif
+	.ndo_set_tx_maxrate	= mlx4_en_set_tx_maxrate,
 };
 
 static const struct net_device_ops mlx4_netdev_ops_master = {
@@ -2444,6 +2472,7 @@ static const struct net_device_ops mlx4_netdev_ops_master = {
 	.ndo_del_vxlan_port	= mlx4_en_del_vxlan_port,
 	.ndo_features_check	= mlx4_en_features_check,
 #endif
+	.ndo_set_tx_maxrate	= mlx4_en_set_tx_maxrate,
 };
 
 struct mlx4_en_bond {

drivers/net/ethernet/mellanox/mlx4/fw.c

@@ -144,7 +144,8 @@ static void dump_dev_cap_flags2(struct mlx4_dev *dev, u64 flags)
 		[19] = "Performance optimized for limited rule configuration flow steering support",
 		[20] = "Recoverable error events support",
 		[21] = "Port Remap support",
-		[22] = "QCN support"
+		[22] = "QCN support",
+		[23] = "QP rate limiting support"
 	};
 	int i;
 
@@ -697,6 +698,10 @@ int mlx4_QUERY_DEV_CAP(struct mlx4_dev *dev, struct mlx4_dev_cap *dev_cap)
 #define QUERY_DEV_CAP_MAD_DEMUX_OFFSET		0xb0
 #define QUERY_DEV_CAP_DMFS_HIGH_RATE_QPN_BASE_OFFSET	0xa8
 #define QUERY_DEV_CAP_DMFS_HIGH_RATE_QPN_RANGE_OFFSET	0xac
+#define QUERY_DEV_CAP_QP_RATE_LIMIT_NUM_OFFSET	0xcc
+#define QUERY_DEV_CAP_QP_RATE_LIMIT_MAX_OFFSET	0xd0
+#define QUERY_DEV_CAP_QP_RATE_LIMIT_MIN_OFFSET	0xd2
+
 
 	dev_cap->flags2 = 0;
 	mailbox = mlx4_alloc_cmd_mailbox(dev);
@@ -904,6 +909,18 @@ int mlx4_QUERY_DEV_CAP(struct mlx4_dev *dev, struct mlx4_dev_cap *dev_cap)
 		 QUERY_DEV_CAP_DMFS_HIGH_RATE_QPN_RANGE_OFFSET);
 	dev_cap->dmfs_high_rate_qpn_range &= MGM_QPN_MASK;
 
+	MLX4_GET(size, outbox, QUERY_DEV_CAP_QP_RATE_LIMIT_NUM_OFFSET);
+	dev_cap->rl_caps.num_rates = size;
+	if (dev_cap->rl_caps.num_rates) {
+		dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_QP_RATE_LIMIT;
+		MLX4_GET(size, outbox, QUERY_DEV_CAP_QP_RATE_LIMIT_MAX_OFFSET);
+		dev_cap->rl_caps.max_val  = size & 0xfff;
+		dev_cap->rl_caps.max_unit = size >> 14;
+		MLX4_GET(size, outbox, QUERY_DEV_CAP_QP_RATE_LIMIT_MIN_OFFSET);
+		dev_cap->rl_caps.min_val  = size & 0xfff;
+		dev_cap->rl_caps.min_unit = size >> 14;
+	}
+
 	MLX4_GET(field32, outbox, QUERY_DEV_CAP_EXT_2_FLAGS_OFFSET);
 	if (field32 & (1 << 16))
 		dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_UPDATE_QP;
@@ -979,6 +996,15 @@ void mlx4_dev_cap_dump(struct mlx4_dev *dev, struct mlx4_dev_cap *dev_cap)
 		 dev_cap->dmfs_high_rate_qpn_base);
 	mlx4_dbg(dev, "DMFS high rate steer QPn range: %d\n",
 		 dev_cap->dmfs_high_rate_qpn_range);
+
+	if (dev_cap->flags2 & MLX4_DEV_CAP_FLAG2_QP_RATE_LIMIT) {
+		struct mlx4_rate_limit_caps *rl_caps = &dev_cap->rl_caps;
+
+		mlx4_dbg(dev, "QP Rate-Limit: #rates %d, unit/val max %d/%d, min %d/%d\n",
+			 rl_caps->num_rates, rl_caps->max_unit, rl_caps->max_val,
+			 rl_caps->min_unit, rl_caps->min_val);
+	}
+
 	dump_dev_cap_flags(dev, dev_cap->flags);
 	dump_dev_cap_flags2(dev, dev_cap->flags2);
 }
@@ -1075,6 +1101,7 @@ int mlx4_QUERY_DEV_CAP_wrapper(struct mlx4_dev *dev, int slave,
 	u64	flags;
 	int	err = 0;
 	u8	field;
+	u16	field16;
 	u32	bmme_flags, field32;
 	int	real_port;
 	int	slave_port;
@@ -1158,6 +1185,10 @@ int mlx4_QUERY_DEV_CAP_wrapper(struct mlx4_dev *dev, int slave,
 	field &= 0xfe;
 	MLX4_PUT(outbox->buf, field, QUERY_DEV_CAP_ECN_QCN_VER_OFFSET);
 
+	/* turn off QP max-rate limiting for guests */
+	field16 = 0;
+	MLX4_PUT(outbox->buf, field16, QUERY_DEV_CAP_QP_RATE_LIMIT_NUM_OFFSET);
+
 	return 0;
 }
 

drivers/net/ethernet/mellanox/mlx4/fw.h

@@ -127,6 +127,7 @@ struct mlx4_dev_cap {
 	u32 max_counters;
 	u32 dmfs_high_rate_qpn_base;
 	u32 dmfs_high_rate_qpn_range;
+	struct mlx4_rate_limit_caps rl_caps;
 	struct mlx4_port_cap port_cap[MLX4_MAX_PORTS + 1];
 };
 

drivers/net/ethernet/mellanox/mlx4/main.c

@@ -489,6 +489,8 @@ static int mlx4_dev_cap(struct mlx4_dev *dev, struct mlx4_dev_cap *dev_cap)
 		dev->caps.dmfs_high_rate_qpn_range = MLX4_A0_STEERING_TABLE_SIZE;
 	}
 
+	dev->caps.rl_caps = dev_cap->rl_caps;
+
 	dev->caps.reserved_qps_cnt[MLX4_QP_REGION_RSS_RAW_ETH] =
 		dev->caps.dmfs_high_rate_qpn_range;
 

drivers/net/ethernet/mellanox/mlx4/qp.c

@@ -442,6 +442,11 @@ int mlx4_update_qp(struct mlx4_dev *dev, u32 qpn,
 			cmd->qp_context.param3 |= cpu_to_be32(MLX4_STRIP_VLAN);
 	}
 
+	if (attr & MLX4_UPDATE_QP_RATE_LIMIT) {
+		qp_mask |= 1ULL << MLX4_UPD_QP_MASK_RATE_LIMIT;
+		cmd->qp_context.rate_limit_params = cpu_to_be16((params->rate_unit << 14) | params->rate_val);
+	}
+
 	cmd->primary_addr_path_mask = cpu_to_be64(pri_addr_path_mask);
 	cmd->qp_mask = cpu_to_be64(qp_mask);
 

drivers/net/ethernet/mellanox/mlx4/resource_tracker.c

@@ -2947,8 +2947,12 @@ static int verify_qp_parameters(struct mlx4_dev *dev,
 	qp_type	= (be32_to_cpu(qp_ctx->flags) >> 16) & 0xff;
 	optpar	= be32_to_cpu(*(__be32 *) inbox->buf);
 
-	if (slave != mlx4_master_func_num(dev))
+	if (slave != mlx4_master_func_num(dev)) {
 		qp_ctx->params2 &= ~MLX4_QP_BIT_FPP;
+		/* setting QP rate-limit is disallowed for VFs */
+		if (qp_ctx->rate_limit_params)
+			return -EPERM;
+	}
 
 	switch (qp_type) {
 	case MLX4_QP_ST_RC:

include/linux/mlx4/device.h

@@ -205,6 +205,7 @@ enum {
 	MLX4_DEV_CAP_FLAG2_RECOVERABLE_ERROR_EVENT = 1LL << 20,
 	MLX4_DEV_CAP_FLAG2_PORT_REMAP		= 1LL <<  21,
 	MLX4_DEV_CAP_FLAG2_QCN			= 1LL <<  22,
+	MLX4_DEV_CAP_FLAG2_QP_RATE_LIMIT	= 1LL <<  23
 };
 
 enum {
@@ -450,6 +451,21 @@ enum mlx4_module_id {
 	MLX4_MODULE_ID_QSFP28           = 0x11,
 };
 
+enum { /* rl */
+	MLX4_QP_RATE_LIMIT_NONE		= 0,
+	MLX4_QP_RATE_LIMIT_KBS		= 1,
+	MLX4_QP_RATE_LIMIT_MBS		= 2,
+	MLX4_QP_RATE_LIMIT_GBS		= 3
+};
+
+struct mlx4_rate_limit_caps {
+	u16	num_rates; /* Number of different rates */
+	u8	min_unit;
+	u16	min_val;
+	u8	max_unit;
+	u16	max_val;
+};
+
 static inline u64 mlx4_fw_ver(u64 major, u64 minor, u64 subminor)
 {
 	return (major << 32) | (minor << 16) | subminor;
@@ -565,6 +581,7 @@ struct mlx4_caps {
 	u32			dmfs_high_rate_qpn_base;
 	u32			dmfs_high_rate_qpn_range;
 	u32			vf_caps;
+	struct mlx4_rate_limit_caps rl_caps;
 };
 
 struct mlx4_buf_list {

include/linux/mlx4/qp.h

@@ -207,14 +207,16 @@ struct mlx4_qp_context {
 	__be32			msn;
 	__be16			rq_wqe_counter;
 	__be16			sq_wqe_counter;
-	u32			reserved3[2];
+	u32			reserved3;
+	__be16			rate_limit_params;
+	__be16			reserved4;
 	__be32			param3;
 	__be32			nummmcpeers_basemkey;
 	u8			log_page_size;
-	u8			reserved4[2];
+	u8			reserved5[2];
 	u8			mtt_base_addr_h;
 	__be32			mtt_base_addr_l;
-	u32			reserved5[10];
+	u32			reserved6[10];
 };
 
 struct mlx4_update_qp_context {
@@ -229,6 +231,7 @@ struct mlx4_update_qp_context {
 enum {
 	MLX4_UPD_QP_MASK_PM_STATE	= 32,
 	MLX4_UPD_QP_MASK_VSD		= 33,
+	MLX4_UPD_QP_MASK_RATE_LIMIT	= 35,
 };
 
 enum {
@@ -428,7 +431,8 @@ struct mlx4_wqe_inline_seg {
 enum mlx4_update_qp_attr {
 	MLX4_UPDATE_QP_SMAC		= 1 << 0,
 	MLX4_UPDATE_QP_VSD		= 1 << 1,
-	MLX4_UPDATE_QP_SUPPORTED_ATTRS	= (1 << 2) - 1
+	MLX4_UPDATE_QP_RATE_LIMIT	= 1 << 2,
+	MLX4_UPDATE_QP_SUPPORTED_ATTRS	= (1 << 3) - 1
 };
 
 enum mlx4_update_qp_params_flags {
@@ -438,6 +442,8 @@ enum mlx4_update_qp_params_flags {
 struct mlx4_update_qp_params {
 	u8	smac_index;
 	u32	flags;
+	u16	rate_unit;
+	u16	rate_val;
 };
 
 int mlx4_update_qp(struct mlx4_dev *dev, u32 qpn,

include/linux/netdevice.h

@@ -587,6 +587,7 @@ struct netdev_queue {
 #ifdef CONFIG_BQL
 	struct dql		dql;
 #endif
+	unsigned long		tx_maxrate;
 } ____cacheline_aligned_in_smp;
 
 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
@@ -1022,6 +1023,10 @@ typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
  *	be otherwise expressed by feature flags. The check is called with
  *	the set of features that the stack has calculated and it returns
  *	those the driver believes to be appropriate.
+ * int (*ndo_set_tx_maxrate)(struct net_device *dev,
+ *			     int queue_index, u32 maxrate);
+ *	Called when a user wants to set a max-rate limitation of specific
+ *	TX queue.
  */
 struct net_device_ops {
 	int			(*ndo_init)(struct net_device *dev);
@@ -1178,6 +1183,9 @@ struct net_device_ops {
 	netdev_features_t	(*ndo_features_check) (struct sk_buff *skb,
 						       struct net_device *dev,
 						       netdev_features_t features);
+	int			(*ndo_set_tx_maxrate)(struct net_device *dev,
+						      int queue_index,
+						      u32 maxrate);
 };
 
 /**

net/core/net-sysfs.c

@@ -951,6 +951,60 @@ static ssize_t show_trans_timeout(struct netdev_queue *queue,
 	return sprintf(buf, "%lu", trans_timeout);
 }
 
+#ifdef CONFIG_XPS
+static inline unsigned int get_netdev_queue_index(struct netdev_queue *queue)
+{
+	struct net_device *dev = queue->dev;
+	int i;
+
+	for (i = 0; i < dev->num_tx_queues; i++)
+		if (queue == &dev->_tx[i])
+			break;
+
+	BUG_ON(i >= dev->num_tx_queues);
+
+	return i;
+}
+
+static ssize_t show_tx_maxrate(struct netdev_queue *queue,
+			       struct netdev_queue_attribute *attribute,
+			       char *buf)
+{
+	return sprintf(buf, "%lu\n", queue->tx_maxrate);
+}
+
+static ssize_t set_tx_maxrate(struct netdev_queue *queue,
+			      struct netdev_queue_attribute *attribute,
+			      const char *buf, size_t len)
+{
+	struct net_device *dev = queue->dev;
+	int err, index = get_netdev_queue_index(queue);
+	u32 rate = 0;
+
+	err = kstrtou32(buf, 10, &rate);
+	if (err < 0)
+		return err;
+
+	if (!rtnl_trylock())
+		return restart_syscall();
+
+	err = -EOPNOTSUPP;
+	if (dev->netdev_ops->ndo_set_tx_maxrate)
+		err = dev->netdev_ops->ndo_set_tx_maxrate(dev, index, rate);
+
+	rtnl_unlock();
+	if (!err) {
+		queue->tx_maxrate = rate;
+		return len;
+	}
+	return err;
+}
+
+static struct netdev_queue_attribute queue_tx_maxrate =
+	__ATTR(tx_maxrate, S_IRUGO | S_IWUSR,
+	       show_tx_maxrate, set_tx_maxrate);
+#endif
+
 static struct netdev_queue_attribute queue_trans_timeout =
 	__ATTR(tx_timeout, S_IRUGO, show_trans_timeout, NULL);
 
@@ -1065,18 +1119,6 @@ static struct attribute_group dql_group = {
 #endif /* CONFIG_BQL */
 
 #ifdef CONFIG_XPS
-static unsigned int get_netdev_queue_index(struct netdev_queue *queue)
-{
-	struct net_device *dev = queue->dev;
-	unsigned int i;
-
-	i = queue - dev->_tx;
-	BUG_ON(i >= dev->num_tx_queues);
-
-	return i;
-}
-
-
 static ssize_t show_xps_map(struct netdev_queue *queue,
 			    struct netdev_queue_attribute *attribute, char *buf)
 {
@@ -1153,6 +1195,7 @@ static struct attribute *netdev_queue_default_attrs[] = {
 	&queue_trans_timeout.attr,
 #ifdef CONFIG_XPS
 	&xps_cpus_attribute.attr,
+	&queue_tx_maxrate.attr,
 #endif
 	NULL
 };