net/mlx4_core: Flexible (asymmetric) allocation of EQs and MSI-X vectors for PF/VFs

Previously, the driver queried the firmware in order to get the number
of supported EQs. Under SRIOV, since this was done before the driver
notified the firmware how many VFs it actually needs, the firmware had
to take into account a worst case scenario and always allocated four EQs
per VF, where one was used for events while the others were used for completions.

Now, when the firmware supports the asymmetric allocation scheme, denoted
by exposing num_sys_eqs > 0 (--> MLX4_DEV_CAP_FLAG2_SYS_EQS), we use the
QUERY_FUNC command to query the firmware before enabling SRIOV. Thus we
can get more EQs and MSI-X vectors per function.

Moreover, when running in the new firmware/driver mode, the limitation
that the number of EQs should be a power of two is lifted.

Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

1 files changed, 125 insertions, 19 deletions

diff --git a/drivers/net/ethernet/mellanox/mlx4/main.c b/drivers/net/ethernet/mellanox/mlx4/main.c
index 43047b2a2aac..ebb279060a25 100644
--- a/drivers/net/ethernet/mellanox/mlx4/main.c
+++ b/drivers/net/ethernet/mellanox/mlx4/main.c
@@ -197,6 +197,29 @@ static void mlx4_set_port_mask(struct mlx4_dev *dev)
 		dev->caps.port_mask[i] = dev->caps.port_type[i];
 }
 
+enum {
+	MLX4_QUERY_FUNC_NUM_SYS_EQS = 1 << 0,
+};
+
+static int mlx4_query_func(struct mlx4_dev *dev, struct mlx4_dev_cap *dev_cap)
+{
+	int err = 0;
+	struct mlx4_func func;
+
+	if (dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_SYS_EQS) {
+		err = mlx4_QUERY_FUNC(dev, &func, 0);
+		if (err) {
+			mlx4_err(dev, "QUERY_DEV_CAP command failed, aborting.\n");
+			return err;
+		}
+		dev_cap->max_eqs = func.max_eq;
+		dev_cap->reserved_eqs = func.rsvd_eqs;
+		dev_cap->reserved_uars = func.rsvd_uars;
+		err |= MLX4_QUERY_FUNC_NUM_SYS_EQS;
+	}
+	return err;
+}
+
 static void mlx4_enable_cqe_eqe_stride(struct mlx4_dev *dev)
 {
 	struct mlx4_caps *dev_cap = &dev->caps;
@@ -261,7 +284,10 @@ static int mlx4_dev_cap(struct mlx4_dev *dev, struct mlx4_dev_cap *dev_cap)
 	}
 
 	dev->caps.num_ports	     = dev_cap->num_ports;
-	dev->phys_caps.num_phys_eqs  = MLX4_MAX_EQ_NUM;
+	dev->caps.num_sys_eqs = dev_cap->num_sys_eqs;
+	dev->phys_caps.num_phys_eqs = dev_cap->flags2 & MLX4_DEV_CAP_FLAG2_SYS_EQS ?
+				      dev->caps.num_sys_eqs :
+				      MLX4_MAX_EQ_NUM;
 	for (i = 1; i <= dev->caps.num_ports; ++i) {
 		dev->caps.vl_cap[i]	    = dev_cap->max_vl[i];
 		dev->caps.ib_mtu_cap[i]	    = dev_cap->ib_mtu[i];
@@ -1130,8 +1156,7 @@ static int mlx4_init_cmpt_table(struct mlx4_dev *dev, u64 cmpt_base,
 	if (err)
 		goto err_srq;
 
-	num_eqs = (mlx4_is_master(dev)) ? dev->phys_caps.num_phys_eqs :
-		  dev->caps.num_eqs;
+	num_eqs = dev->phys_caps.num_phys_eqs;
 	err = mlx4_init_icm_table(dev, &priv->eq_table.cmpt_table,
 				  cmpt_base +
 				  ((u64) (MLX4_CMPT_TYPE_EQ *
@@ -1193,8 +1218,7 @@ static int mlx4_init_icm(struct mlx4_dev *dev, struct mlx4_dev_cap *dev_cap,
 	}
 
 
-	num_eqs = (mlx4_is_master(dev)) ? dev->phys_caps.num_phys_eqs :
-		   dev->caps.num_eqs;
+	num_eqs = dev->phys_caps.num_phys_eqs;
 	err = mlx4_init_icm_table(dev, &priv->eq_table.table,
 				  init_hca->eqc_base, dev_cap->eqc_entry_sz,
 				  num_eqs, num_eqs, 0, 0);
@@ -1719,6 +1743,19 @@ static int mlx4_init_hca(struct mlx4_dev *dev)
 			mlx4_err(dev, "INIT_HCA command failed, aborting\n");
 			goto err_free_icm;
 		}
+
+		if (dev_cap.flags2 & MLX4_DEV_CAP_FLAG2_SYS_EQS) {
+			err = mlx4_query_func(dev, &dev_cap);
+			if (err < 0) {
+				mlx4_err(dev, "QUERY_FUNC command failed, aborting.\n");
+				goto err_stop_fw;
+			} else if (err & MLX4_QUERY_FUNC_NUM_SYS_EQS) {
+				dev->caps.num_eqs = dev_cap.max_eqs;
+				dev->caps.reserved_eqs = dev_cap.reserved_eqs;
+				dev->caps.reserved_uars = dev_cap.reserved_uars;
+			}
+		}
+
 		/*
 		 * If TS is supported by FW
 		 * read HCA frequency by QUERY_HCA command
@@ -2085,12 +2122,11 @@ static void mlx4_enable_msi_x(struct mlx4_dev *dev)
 {
 	struct mlx4_priv *priv = mlx4_priv(dev);
 	struct msix_entry *entries;
-	int nreq = min_t(int, dev->caps.num_ports *
-			 min_t(int, num_online_cpus() + 1,
-			       MAX_MSIX_P_PORT) + MSIX_LEGACY_SZ, MAX_MSIX);
 	int i;
 
 	if (msi_x) {
+		int nreq = dev->caps.num_ports * num_online_cpus() + MSIX_LEGACY_SZ;
+
 		nreq = min_t(int, dev->caps.num_eqs - dev->caps.reserved_eqs,
 			     nreq);
 
@@ -2345,6 +2381,7 @@ static int mlx4_load_one(struct pci_dev *pdev, int pci_dev_data,
 	int err;
 	int port;
 	int i;
+	struct mlx4_dev_cap *dev_cap = NULL;
 	int existing_vfs = 0;
 
 	dev = &priv->dev;
@@ -2381,15 +2418,6 @@ static int mlx4_load_one(struct pci_dev *pdev, int pci_dev_data,
 			}
 		}
 
-		if (total_vfs) {
-			existing_vfs = pci_num_vf(pdev);
-			dev->flags = MLX4_FLAG_MASTER;
-			dev->flags = mlx4_enable_sriov(dev, pdev, total_vfs,
-						       existing_vfs);
-			if (!SRIOV_VALID_STATE(dev->flags))
-				goto err_sriov;
-		}
-
 		atomic_set(&priv->opreq_count, 0);
 		INIT_WORK(&priv->opreq_task, mlx4_opreq_action);
 
@@ -2403,6 +2431,12 @@ static int mlx4_load_one(struct pci_dev *pdev, int pci_dev_data,
 			mlx4_err(dev, "Failed to reset HCA, aborting\n");
 			goto err_sriov;
 		}
+
+		if (total_vfs) {
+			existing_vfs = pci_num_vf(pdev);
+			dev->flags = MLX4_FLAG_MASTER;
+			dev->num_vfs = total_vfs;
+		}
 	}
 
 slave_start:
@@ -2416,9 +2450,10 @@ slave_start:
 	 * before posting commands. Also, init num_slaves before calling
 	 * mlx4_init_hca */
 	if (mlx4_is_mfunc(dev)) {
-		if (mlx4_is_master(dev))
+		if (mlx4_is_master(dev)) {
 			dev->num_slaves = MLX4_MAX_NUM_SLAVES;
-		else {
+
+		} else {
 			dev->num_slaves = 0;
 			err = mlx4_multi_func_init(dev);
 			if (err) {
@@ -2434,6 +2469,52 @@ slave_start:
 		goto err_mfunc;
 	}
 
+	if (mlx4_is_master(dev)) {
+		if (!dev_cap) {
+			dev_cap = kzalloc(sizeof(*dev_cap), GFP_KERNEL);
+
+			if (!dev_cap) {
+				err = -ENOMEM;
+				goto err_fw;
+			}
+
+			err = mlx4_QUERY_DEV_CAP(dev, dev_cap);
+			if (err) {
+				mlx4_err(dev, "QUERY_DEV_CAP command failed, aborting.\n");
+				goto err_fw;
+			}
+
+			if (!(dev_cap->flags2 & MLX4_DEV_CAP_FLAG2_SYS_EQS)) {
+				u64 dev_flags = mlx4_enable_sriov(dev, pdev, total_vfs,
+								  existing_vfs);
+
+				mlx4_cmd_cleanup(dev, MLX4_CMD_CLEANUP_ALL);
+				dev->flags = dev_flags;
+				if (!SRIOV_VALID_STATE(dev->flags)) {
+					mlx4_err(dev, "Invalid SRIOV state\n");
+					goto err_sriov;
+				}
+				err = mlx4_reset(dev);
+				if (err) {
+					mlx4_err(dev, "Failed to reset HCA, aborting.\n");
+					goto err_sriov;
+				}
+				goto slave_start;
+			}
+		} else {
+			/* Legacy mode FW requires SRIOV to be enabled before
+			 * doing QUERY_DEV_CAP, since max_eq's value is different if
+			 * SRIOV is enabled.
+			 */
+			memset(dev_cap, 0, sizeof(*dev_cap));
+			err = mlx4_QUERY_DEV_CAP(dev, dev_cap);
+			if (err) {
+				mlx4_err(dev, "QUERY_DEV_CAP command failed, aborting.\n");
+				goto err_fw;
+			}
+		}
+	}
+
 	err = mlx4_init_hca(dev);
 	if (err) {
 		if (err == -EACCES) {
@@ -2457,6 +2538,30 @@ slave_start:
 			goto err_fw;
 	}
 
+	if (mlx4_is_master(dev) && (dev_cap->flags2 & MLX4_DEV_CAP_FLAG2_SYS_EQS)) {
+		u64 dev_flags = mlx4_enable_sriov(dev, pdev, total_vfs, existing_vfs);
+
+		if ((dev->flags ^ dev_flags) & (MLX4_FLAG_MASTER | MLX4_FLAG_SLAVE)) {
+			mlx4_cmd_cleanup(dev, MLX4_CMD_CLEANUP_VHCR);
+			dev->flags = dev_flags;
+			err = mlx4_cmd_init(dev);
+			if (err) {
+				/* Only VHCR is cleaned up, so could still
+				 * send FW commands
+				 */
+				mlx4_err(dev, "Failed to init VHCR command interface, aborting\n");
+				goto err_close;
+			}
+		} else {
+			dev->flags = dev_flags;
+		}
+
+		if (!SRIOV_VALID_STATE(dev->flags)) {
+			mlx4_err(dev, "Invalid SRIOV state\n");
+			goto err_close;
+		}
+	}
+
 	/* check if the device is functioning at its maximum possible speed.
 	 * No return code for this call, just warn the user in case of PCI
 	 * express device capabilities are under-satisfied by the bus.
@@ -2631,6 +2736,7 @@ err_sriov:
 	if (!mlx4_is_slave(dev))
 		mlx4_free_ownership(dev);
 
+	kfree(dev_cap);
 	return err;
 }