netdev - [PATCH mlx5-next 2/2] IB/mlx5: Add DCT RoCE LAG support

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-Id: <20200818115245.700581-3-leon@kernel.org>
Date:   Tue, 18 Aug 2020 14:52:45 +0300
From:   Leon Romanovsky <leon@...nel.org>
To:     Doug Ledford <dledford@...hat.com>,
        Jason Gunthorpe <jgg@...dia.com>
Cc:     Mark Zhang <markz@...lanox.com>,
        Leon Romanovsky <leonro@...lanox.com>,
        linux-rdma@...r.kernel.org, Maor Gottlieb <maorg@...lanox.com>,
        netdev@...r.kernel.org, Saeed Mahameed <saeedm@...dia.com>
Subject: [PATCH mlx5-next 2/2] IB/mlx5: Add DCT RoCE LAG support

From: Mark Zhang <markz@...lanox.com>

When DCT QPs work in RoCE LAG mode:
1. DCT creation is allowed only when it is supported.
2. The "port" of a DCT QP is assigned in a round-robin way.

Signed-off-by: Mark Zhang <markz@...lanox.com>
Reviewed-by: Maor Gottlieb <maorg@...lanox.com>
Signed-off-by: Leon Romanovsky <leonro@...lanox.com>
---
 drivers/infiniband/hw/mlx5/qp.c | 9 ++++++++-
 include/linux/mlx5/mlx5_ifc.h   | 3 ++-
 2 files changed, 10 insertions(+), 2 deletions(-)

diff --git a/drivers/infiniband/hw/mlx5/qp.c b/drivers/infiniband/hw/mlx5/qp.c
index 0526d574cd9b..4eff325c4d2c 100644
--- a/drivers/infiniband/hw/mlx5/qp.c
+++ b/drivers/infiniband/hw/mlx5/qp.c
@@ -2409,6 +2409,9 @@ static int create_dct(struct mlx5_ib_dev *dev, struct ib_pd *pd,
 	u32 uidx = params->uidx;
 	void *dctc;

+	if (mlx5_lag_is_active(dev->mdev) && !MLX5_CAP_GEN(dev->mdev, lag_dct))
+		return -EOPNOTSUPP;
+
 	qp->dct.in = kzalloc(MLX5_ST_SZ_BYTES(create_dct_in), GFP_KERNEL);
 	if (!qp->dct.in)
 		return -ENOMEM;
@@ -4183,7 +4186,11 @@ static int mlx5_ib_modify_dct(struct ib_qp *ibqp, struct ib_qp_attr *attr,
 			MLX5_SET(dctc, dctc, rae, 1);
 		}
 		MLX5_SET(dctc, dctc, pkey_index, attr->pkey_index);
-		MLX5_SET(dctc, dctc, port, attr->port_num);
+		if (mlx5_lag_is_active(dev->mdev))
+			MLX5_SET(dctc, dctc, port,
+				 get_tx_affinity_rr(dev, udata));
+		else
+			MLX5_SET(dctc, dctc, port, attr->port_num);

 		set_id = mlx5_ib_get_counters_id(dev, attr->port_num - 1);
 		MLX5_SET(dctc, dctc, counter_set_id, set_id);
diff --git a/include/linux/mlx5/mlx5_ifc.h b/include/linux/mlx5/mlx5_ifc.h
index de1ffb4804d6..aee25e4fb2cc 100644
--- a/include/linux/mlx5/mlx5_ifc.h
+++ b/include/linux/mlx5/mlx5_ifc.h
@@ -1430,7 +1430,8 @@ struct mlx5_ifc_cmd_hca_cap_bits {

 	u8         log_bf_reg_size[0x5];

-	u8         reserved_at_270[0x8];
+	u8         reserved_at_270[0x6];
+	u8         lag_dct[0x2];
 	u8         lag_tx_port_affinity[0x1];
 	u8         reserved_at_279[0x2];
 	u8         lag_master[0x1];
--
2.26.2