IB/mlx5: Test write combining support

Linux can run in all sorts of physical machines and VMs where write
combining may or may not be supported. Currently there is no way to
reliably tell if the system supports WC, or not. The driver uses WC to
optimize posting work to the HCA, and getting this wrong in either
direction can cause a significant performance loss.

Add a test in mlx5_ib initialization process to test whether
write-combining is supported on the machine.  The test will run as part of
the enable_driver callback to ensure that the test runs after the device
is setup and can create and modify the QP needed, but runs before the
device is exposed to the users.

The test opens UD QP and posts NOP WQEs, the WQE written to the BlueFlame
is different from the WQE in memory, requesting CQE only on the BlueFlame
WQE. By checking whether we received a completion on one of these WQEs we
can know if BlueFlame succeeded and this write-combining must be
supported.

Change reporting of BlueFlame support to be dependent on write-combining
support instead of the FW's guess as to what the machine can do.

Link: https://lore.kernel.org/r/20191027062234.10993-1-leon@kernel.org
Signed-off-by: Michael Guralnik <michaelgur@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

1 files changed, 199 insertions, 0 deletions

diff --git a/drivers/infiniband/hw/mlx5/mem.c b/drivers/infiniband/hw/mlx5/mem.c
index b5aece786b36..048f4e974a61 100644
--- a/drivers/infiniband/hw/mlx5/mem.c
+++ b/drivers/infiniband/hw/mlx5/mem.c
@@ -34,6 +34,7 @@
 #include <rdma/ib_umem.h>
 #include <rdma/ib_umem_odp.h>
 #include "mlx5_ib.h"
+#include <linux/jiffies.h>
 
 /* @umem: umem object to scan
  * @addr: ib virtual address requested by the user
@@ -216,3 +217,201 @@ int mlx5_ib_get_buf_offset(u64 addr, int page_shift, u32 *offset)
 	*offset = buf_off >> ilog2(off_size);
 	return 0;
 }
+
+#define WR_ID_BF 0xBF
+#define WR_ID_END 0xBAD
+#define TEST_WC_NUM_WQES 255
+#define TEST_WC_POLLING_MAX_TIME_JIFFIES msecs_to_jiffies(100)
+static int post_send_nop(struct mlx5_ib_dev *dev, struct ib_qp *ibqp, u64 wr_id,
+			 bool signaled)
+{
+	struct mlx5_ib_qp *qp = to_mqp(ibqp);
+	struct mlx5_wqe_ctrl_seg *ctrl;
+	struct mlx5_bf *bf = &qp->bf;
+	__be32 mmio_wqe[16] = {};
+	unsigned long flags;
+	unsigned int idx;
+	int i;
+
+	if (unlikely(dev->mdev->state == MLX5_DEVICE_STATE_INTERNAL_ERROR))
+		return -EIO;
+
+	spin_lock_irqsave(&qp->sq.lock, flags);
+
+	idx = qp->sq.cur_post & (qp->sq.wqe_cnt - 1);
+	ctrl = mlx5_frag_buf_get_wqe(&qp->sq.fbc, idx);
+
+	memset(ctrl, 0, sizeof(struct mlx5_wqe_ctrl_seg));
+	ctrl->fm_ce_se = signaled ? MLX5_WQE_CTRL_CQ_UPDATE : 0;
+	ctrl->opmod_idx_opcode =
+		cpu_to_be32(((u32)(qp->sq.cur_post) << 8) | MLX5_OPCODE_NOP);
+	ctrl->qpn_ds = cpu_to_be32((sizeof(struct mlx5_wqe_ctrl_seg) / 16) |
+				   (qp->trans_qp.base.mqp.qpn << 8));
+
+	qp->sq.wrid[idx] = wr_id;
+	qp->sq.w_list[idx].opcode = MLX5_OPCODE_NOP;
+	qp->sq.wqe_head[idx] = qp->sq.head + 1;
+	qp->sq.cur_post += DIV_ROUND_UP(sizeof(struct mlx5_wqe_ctrl_seg),
+					MLX5_SEND_WQE_BB);
+	qp->sq.w_list[idx].next = qp->sq.cur_post;
+	qp->sq.head++;
+
+	memcpy(mmio_wqe, ctrl, sizeof(*ctrl));
+	((struct mlx5_wqe_ctrl_seg *)&mmio_wqe)->fm_ce_se |=
+		MLX5_WQE_CTRL_CQ_UPDATE;
+
+	/* Make sure that descriptors are written before
+	 * updating doorbell record and ringing the doorbell
+	 */
+	wmb();
+
+	qp->db.db[MLX5_SND_DBR] = cpu_to_be32(qp->sq.cur_post);
+
+	/* Make sure doorbell record is visible to the HCA before
+	 * we hit doorbell
+	 */
+	wmb();
+	for (i = 0; i < 8; i++)
+		mlx5_write64(&mmio_wqe[i * 2],
+			     bf->bfreg->map + bf->offset + i * 8);
+
+	bf->offset ^= bf->buf_size;
+
+	spin_unlock_irqrestore(&qp->sq.lock, flags);
+
+	return 0;
+}
+
+static int test_wc_poll_cq_result(struct mlx5_ib_dev *dev, struct ib_cq *cq)
+{
+	int ret;
+	struct ib_wc wc = {};
+	unsigned long end = jiffies + TEST_WC_POLLING_MAX_TIME_JIFFIES;
+
+	do {
+		ret = ib_poll_cq(cq, 1, &wc);
+		if (ret < 0 || wc.status)
+			return ret < 0 ? ret : -EINVAL;
+		if (ret)
+			break;
+	} while (!time_after(jiffies, end));
+
+	if (!ret)
+		return -ETIMEDOUT;
+
+	if (wc.wr_id != WR_ID_BF)
+		ret = 0;
+
+	return ret;
+}
+
+static int test_wc_do_send(struct mlx5_ib_dev *dev, struct ib_qp *qp)
+{
+	int err, i;
+
+	for (i = 0; i < TEST_WC_NUM_WQES; i++) {
+		err = post_send_nop(dev, qp, WR_ID_BF, false);
+		if (err)
+			return err;
+	}
+
+	return post_send_nop(dev, qp, WR_ID_END, true);
+}
+
+int mlx5_ib_test_wc(struct mlx5_ib_dev *dev)
+{
+	struct ib_cq_init_attr cq_attr = { .cqe = TEST_WC_NUM_WQES + 1 };
+	int port_type_cap = MLX5_CAP_GEN(dev->mdev, port_type);
+	struct ib_qp_init_attr qp_init_attr = {
+		.cap = { .max_send_wr = TEST_WC_NUM_WQES },
+		.qp_type = IB_QPT_UD,
+		.sq_sig_type = IB_SIGNAL_REQ_WR,
+		.create_flags = MLX5_IB_QP_CREATE_WC_TEST,
+	};
+	struct ib_qp_attr qp_attr = { .port_num = 1 };
+	struct ib_device *ibdev = &dev->ib_dev;
+	struct ib_qp *qp;
+	struct ib_cq *cq;
+	struct ib_pd *pd;
+	int ret;
+
+	if (!MLX5_CAP_GEN(dev->mdev, bf))
+		return 0;
+
+	if (!dev->mdev->roce.roce_en &&
+	    port_type_cap == MLX5_CAP_PORT_TYPE_ETH) {
+		if (mlx5_core_is_pf(dev->mdev))
+			dev->wc_support = true;
+		return 0;
+	}
+
+	ret = mlx5_alloc_bfreg(dev->mdev, &dev->wc_bfreg, true, false);
+	if (ret)
+		goto print_err;
+
+	if (!dev->wc_bfreg.wc)
+		goto out1;
+
+	pd = ib_alloc_pd(ibdev, 0);
+	if (IS_ERR(pd)) {
+		ret = PTR_ERR(pd);
+		goto out1;
+	}
+
+	cq = ib_create_cq(ibdev, NULL, NULL, NULL, &cq_attr);
+	if (IS_ERR(cq)) {
+		ret = PTR_ERR(cq);
+		goto out2;
+	}
+
+	qp_init_attr.recv_cq = cq;
+	qp_init_attr.send_cq = cq;
+	qp = ib_create_qp(pd, &qp_init_attr);
+	if (IS_ERR(qp)) {
+		ret = PTR_ERR(qp);
+		goto out3;
+	}
+
+	qp_attr.qp_state = IB_QPS_INIT;
+	ret = ib_modify_qp(qp, &qp_attr,
+			   IB_QP_STATE | IB_QP_PORT | IB_QP_PKEY_INDEX |
+				   IB_QP_QKEY);
+	if (ret)
+		goto out4;
+
+	qp_attr.qp_state = IB_QPS_RTR;
+	ret = ib_modify_qp(qp, &qp_attr, IB_QP_STATE);
+	if (ret)
+		goto out4;
+
+	qp_attr.qp_state = IB_QPS_RTS;
+	ret = ib_modify_qp(qp, &qp_attr, IB_QP_STATE | IB_QP_SQ_PSN);
+	if (ret)
+		goto out4;
+
+	ret = test_wc_do_send(dev, qp);
+	if (ret < 0)
+		goto out4;
+
+	ret = test_wc_poll_cq_result(dev, cq);
+	if (ret > 0) {
+		dev->wc_support = true;
+		ret = 0;
+	}
+
+out4:
+	ib_destroy_qp(qp);
+out3:
+	ib_destroy_cq(cq);
+out2:
+	ib_dealloc_pd(pd);
+out1:
+	mlx5_free_bfreg(dev->mdev, &dev->wc_bfreg);
+print_err:
+	if (ret)
+		mlx5_ib_err(
+			dev,
+			"Error %d while trying to test write-combining support\n",
+			ret);
+	return ret;
+}