Commit c5ae1954 authored by Yishai Hadas's avatar Yishai Hadas Committed by Jason Gunthorpe
Browse files

IB/mlx5: Use mlx5 core to create/destroy a DEVX DCT 



To prevent a hardware memory leak when a DEVX DCT object is destroyed
without calling DRAIN DCT before, (e.g. under cleanup flow), need to
manage its creation and destruction via mlx5 core.

In that case the DRAIN DCT command will be called and only once that it
will be completed the DESTROY DCT command will be called.  Otherwise, the
DESTROY DCT may fail and a hardware leak may occur.

As of that change the DRAIN DCT command should not be exposed any more
from DEVX, it's managed internally by the driver to work as expected by
the device specification.

Fixes: 7efce369 ("IB/mlx5: Add obj create and destroy functionality")
Signed-off-by: default avatarYishai Hadas <yishaih@mellanox.com>
Reviewed-by: default avatarArtemy Kovalyov <artemyko@mellanox.com>
Signed-off-by: default avatarLeon Romanovsky <leonro@mellanox.com>
Signed-off-by: default avatarJason Gunthorpe <jgg@mellanox.com>
parent f84b66b9

drivers/infiniband/hw/mlx5/devx.c

+26 −8
Original line number Diff line number Diff line
@@ -20,6 +20,7 @@ 


enum devx_obj_flags {

	DEVX_OBJ_FLAGS_INDIRECT_MKEY = 1 << 0,

	DEVX_OBJ_FLAGS_DCT = 1 << 1,

};



struct devx_async_data {
@@ -39,7 +40,10 @@ struct devx_obj { 
	u32			dinlen; /* destroy inbox length */

	u32			dinbox[MLX5_MAX_DESTROY_INBOX_SIZE_DW];

	u32			flags;

	union {

		struct mlx5_ib_devx_mr	devx_mr;

		struct mlx5_core_dct	core_dct;

	};

};



struct devx_umem {
@@ -347,7 +351,6 @@ static u64 devx_get_obj_id(const void *in) 
		obj_id = get_enc_obj_id(MLX5_CMD_OP_CREATE_RQ,

					MLX5_GET(arm_rq_in, in, srq_number));

		break;

	case MLX5_CMD_OP_DRAIN_DCT:

	case MLX5_CMD_OP_ARM_DCT_FOR_KEY_VIOLATION:

		obj_id = get_enc_obj_id(MLX5_CMD_OP_CREATE_DCT,

					MLX5_GET(drain_dct_in, in, dctn));
@@ -618,7 +621,6 @@ static bool devx_is_obj_modify_cmd(const void *in) 
	case MLX5_CMD_OP_2RST_QP:

	case MLX5_CMD_OP_ARM_XRC_SRQ:

	case MLX5_CMD_OP_ARM_RQ:

	case MLX5_CMD_OP_DRAIN_DCT:

	case MLX5_CMD_OP_ARM_DCT_FOR_KEY_VIOLATION:

	case MLX5_CMD_OP_ARM_XRQ:

	case MLX5_CMD_OP_SET_XRQ_DC_PARAMS_ENTRY:
@@ -1124,7 +1126,11 @@ static int devx_obj_cleanup(struct ib_uobject *uobject, 
	if (obj->flags & DEVX_OBJ_FLAGS_INDIRECT_MKEY)

		devx_cleanup_mkey(obj);



	ret = mlx5_cmd_exec(obj->mdev, obj->dinbox, obj->dinlen, out, sizeof(out));

	if (obj->flags & DEVX_OBJ_FLAGS_DCT)

		ret = mlx5_core_destroy_dct(obj->mdev, &obj->core_dct);

	else

		ret = mlx5_cmd_exec(obj->mdev, obj->dinbox, obj->dinlen, out,

				    sizeof(out));

	if (ib_is_destroy_retryable(ret, why, uobject))

		return ret;
@@ -1185,9 +1191,17 @@ static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OBJ_CREATE)( 
		devx_set_umem_valid(cmd_in);

	}



	if (opcode == MLX5_CMD_OP_CREATE_DCT) {

		obj->flags |= DEVX_OBJ_FLAGS_DCT;

		err = mlx5_core_create_dct(dev->mdev, &obj->core_dct,

					   cmd_in, cmd_in_len,

					   cmd_out, cmd_out_len);

	} else {

		err = mlx5_cmd_exec(dev->mdev, cmd_in,

				    cmd_in_len,

				    cmd_out, cmd_out_len);

	}



	if (err)

		goto obj_free;
@@ -1214,7 +1228,11 @@ err_copy: 
	if (obj->flags & DEVX_OBJ_FLAGS_INDIRECT_MKEY)

		devx_cleanup_mkey(obj);

obj_destroy:

	mlx5_cmd_exec(obj->mdev, obj->dinbox, obj->dinlen, out, sizeof(out));

	if (obj->flags & DEVX_OBJ_FLAGS_DCT)

		mlx5_core_destroy_dct(obj->mdev, &obj->core_dct);

	else

		mlx5_cmd_exec(obj->mdev, obj->dinbox, obj->dinlen, out,

			      sizeof(out));

obj_free:

	kfree(obj);

	return err;

drivers/infiniband/hw/mlx5/qp.c

+3 −1
Original line number Diff line number Diff line
@@ -3729,6 +3729,7 @@ static int mlx5_ib_modify_dct(struct ib_qp *ibqp, struct ib_qp_attr *attr, 


	} else if (cur_state == IB_QPS_INIT && new_state == IB_QPS_RTR) {

		struct mlx5_ib_modify_qp_resp resp = {};

		u32 out[MLX5_ST_SZ_DW(create_dct_out)] = {0};

		u32 min_resp_len = offsetof(typeof(resp), dctn) +

				   sizeof(resp.dctn);
@@ -3747,7 +3748,8 @@ static int mlx5_ib_modify_dct(struct ib_qp *ibqp, struct ib_qp_attr *attr, 
		MLX5_SET(dctc, dctc, hop_limit, attr->ah_attr.grh.hop_limit);



		err = mlx5_core_create_dct(dev->mdev, &qp->dct.mdct, qp->dct.in,

					   MLX5_ST_SZ_BYTES(create_dct_in));

					   MLX5_ST_SZ_BYTES(create_dct_in), out,

					   sizeof(out));

		if (err)

			return err;

		resp.dctn = qp->dct.mdct.mqp.qpn;

drivers/net/ethernet/mellanox/mlx5/core/qp.c

+3 −3
Original line number Diff line number Diff line
@@ -263,16 +263,16 @@ destroy: 


int mlx5_core_create_dct(struct mlx5_core_dev *dev,

			 struct mlx5_core_dct *dct,

			 u32 *in, int inlen)

			 u32 *in, int inlen,

			 u32 *out, int outlen)

{

	u32 out[MLX5_ST_SZ_DW(create_dct_out)]   = {0};

	struct mlx5_core_qp *qp = &dct->mqp;

	int err;



	init_completion(&dct->drained);

	MLX5_SET(create_dct_in, in, opcode, MLX5_CMD_OP_CREATE_DCT);



	err = mlx5_cmd_exec(dev, in, inlen, &out, sizeof(out));

	err = mlx5_cmd_exec(dev, in, inlen, out, outlen);

	if (err) {

		mlx5_core_warn(dev, "create DCT failed, ret %d\n", err);

		return err;

include/linux/mlx5/qp.h

+2 −1
Original line number Diff line number Diff line
@@ -557,7 +557,8 @@ static inline struct mlx5_core_mkey *__mlx5_mr_lookup(struct mlx5_core_dev *dev, 


int mlx5_core_create_dct(struct mlx5_core_dev *dev,

			 struct mlx5_core_dct *qp,

			 u32 *in, int inlen);

			 u32 *in, int inlen,

			 u32 *out, int outlen);

int mlx5_core_create_qp(struct mlx5_core_dev *dev,

			struct mlx5_core_qp *qp,

			u32 *in,

CRA Git | Maintained and supported by SUSTech CRA and CCSE